diff options
Diffstat (limited to 'third_party/aom/av1/common/x86')
43 files changed, 24693 insertions, 0 deletions
diff --git a/third_party/aom/av1/common/x86/av1_convolve_horiz_rs_sse4.c b/third_party/aom/av1/common/x86/av1_convolve_horiz_rs_sse4.c new file mode 100644 index 0000000000..8aa14696f6 --- /dev/null +++ b/third_party/aom/av1/common/x86/av1_convolve_horiz_rs_sse4.c @@ -0,0 +1,228 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <assert.h> +#include <smmintrin.h> + +#include "config/av1_rtcd.h" + +#include "av1/common/convolve.h" +#include "av1/common/resize.h" +#include "aom_dsp/x86/synonyms.h" + +// Note: If the crop width is not a multiple of 4, then, unlike the C version, +// this function will overwrite some of the padding on the right hand side of +// the frame. This padding appears to be trashed anyway, so this should not +// affect the running of the decoder. +void av1_convolve_horiz_rs_sse4_1(const uint8_t *src, int src_stride, + uint8_t *dst, int dst_stride, int w, int h, + const int16_t *x_filters, int x0_qn, + int x_step_qn) { + assert(UPSCALE_NORMATIVE_TAPS == 8); + + src -= UPSCALE_NORMATIVE_TAPS / 2 - 1; + + const __m128i round_add = _mm_set1_epi32((1 << FILTER_BITS) >> 1); + const __m128i zero = _mm_setzero_si128(); + + const uint8_t *src_y; + uint8_t *dst_y; + int x_qn = x0_qn; + for (int x = 0; x < w; x += 4, x_qn += 4 * x_step_qn) { + const int x_filter_idx0 = + ((x_qn + 0 * x_step_qn) & RS_SCALE_SUBPEL_MASK) >> RS_SCALE_EXTRA_BITS; + const int x_filter_idx1 = + ((x_qn + 1 * x_step_qn) & RS_SCALE_SUBPEL_MASK) >> RS_SCALE_EXTRA_BITS; + const int x_filter_idx2 = + ((x_qn + 2 * x_step_qn) & RS_SCALE_SUBPEL_MASK) >> RS_SCALE_EXTRA_BITS; + const int x_filter_idx3 = + ((x_qn + 3 * x_step_qn) & RS_SCALE_SUBPEL_MASK) >> RS_SCALE_EXTRA_BITS; + + assert(x_filter_idx0 <= RS_SUBPEL_MASK); + assert(x_filter_idx1 <= RS_SUBPEL_MASK); + assert(x_filter_idx2 <= RS_SUBPEL_MASK); + assert(x_filter_idx3 <= RS_SUBPEL_MASK); + + const int16_t *const x_filter0 = + &x_filters[x_filter_idx0 * UPSCALE_NORMATIVE_TAPS]; + const int16_t *const x_filter1 = + &x_filters[x_filter_idx1 * UPSCALE_NORMATIVE_TAPS]; + const int16_t *const x_filter2 = + &x_filters[x_filter_idx2 * UPSCALE_NORMATIVE_TAPS]; + const int16_t *const x_filter3 = + &x_filters[x_filter_idx3 * UPSCALE_NORMATIVE_TAPS]; + + const __m128i fil0_16 = xx_loadu_128(x_filter0); + const __m128i fil1_16 = xx_loadu_128(x_filter1); + const __m128i fil2_16 = xx_loadu_128(x_filter2); + const __m128i fil3_16 = xx_loadu_128(x_filter3); + + src_y = src; + dst_y = dst; + for (int y = 0; y < h; y++, src_y += src_stride, dst_y += dst_stride) { + const uint8_t *const src_x0 = + &src_y[(x_qn + 0 * x_step_qn) >> RS_SCALE_SUBPEL_BITS]; + const uint8_t *const src_x1 = + &src_y[(x_qn + 1 * x_step_qn) >> RS_SCALE_SUBPEL_BITS]; + const uint8_t *const src_x2 = + &src_y[(x_qn + 2 * x_step_qn) >> RS_SCALE_SUBPEL_BITS]; + const uint8_t *const src_x3 = + &src_y[(x_qn + 3 * x_step_qn) >> RS_SCALE_SUBPEL_BITS]; + + // Load up the source data. This is 8-bit input data, so each load + // gets 8 pixels. + const __m128i src0_8 = xx_loadl_64(src_x0); + const __m128i src1_8 = xx_loadl_64(src_x1); + const __m128i src2_8 = xx_loadl_64(src_x2); + const __m128i src3_8 = xx_loadl_64(src_x3); + + // Now zero-extend up to 16-bit precision, i.e. + // [ 00 00 00 00 hg fe dc ba ] -> [ 0h 0g 0f 0e 0d 0c 0b 0a ] + const __m128i src0_16 = _mm_cvtepu8_epi16(src0_8); + const __m128i src1_16 = _mm_cvtepu8_epi16(src1_8); + const __m128i src2_16 = _mm_cvtepu8_epi16(src2_8); + const __m128i src3_16 = _mm_cvtepu8_epi16(src3_8); + + // Multiply by filter coefficients (results in a 32-bit value), + // and add adjacent pairs, i.e. + // ([ s7 s6 s5 s4 s3 s2 s1 s0], [ f7 f6 f5 f4 f3 f2 f1 f0 ]) + // -> [ {s7*f7+s6*f6} {s5*f5+s4*f4} {s3*f3+s2*f2} {s1*f1+s0*f0} ] + const __m128i conv0_32 = _mm_madd_epi16(src0_16, fil0_16); + const __m128i conv1_32 = _mm_madd_epi16(src1_16, fil1_16); + const __m128i conv2_32 = _mm_madd_epi16(src2_16, fil2_16); + const __m128i conv3_32 = _mm_madd_epi16(src3_16, fil3_16); + + // Reduce horizontally and add, i.e. + // ([ D C B A ], [ S R Q P ]) -> [ S+R Q+P D+C B+A ] + const __m128i conv01_32 = _mm_hadd_epi32(conv0_32, conv1_32); + const __m128i conv23_32 = _mm_hadd_epi32(conv2_32, conv3_32); + + const __m128i conv0123_32 = _mm_hadd_epi32(conv01_32, conv23_32); + + // Divide down by (1 << FILTER_BITS), rounding to nearest. + const __m128i shifted_32 = + _mm_srai_epi32(_mm_add_epi32(conv0123_32, round_add), FILTER_BITS); + + // Pack 32-bit values into 16-bit values, i.e. + // ([ D C B A ], [ 0 0 0 0 ]) -> [ 0 0 0 0 D C B A ] + const __m128i shifted_16 = _mm_packus_epi32(shifted_32, zero); + + // Pack 16-bit values into 8-bit values, i.e. + // ([ 0 0 0 0 D C B A ], [ 0 0 0 0 0 0 0 0 ]) + // -> [ 0 0 0 0 0 0 DC BA ] + const __m128i shifted_8 = _mm_packus_epi16(shifted_16, zero); + + // Write to the output + xx_storel_32(&dst_y[x], shifted_8); + } + } +} + +// Note: If the crop width is not a multiple of 4, then, unlike the C version, +// this function will overwrite some of the padding on the right hand side of +// the frame. This padding appears to be trashed anyway, so this should not +// affect the running of the decoder. +void av1_highbd_convolve_horiz_rs_sse4_1(const uint16_t *src, int src_stride, + uint16_t *dst, int dst_stride, int w, + int h, const int16_t *x_filters, + int x0_qn, int x_step_qn, int bd) { + assert(UPSCALE_NORMATIVE_TAPS == 8); + assert(bd == 8 || bd == 10 || bd == 12); + + src -= UPSCALE_NORMATIVE_TAPS / 2 - 1; + + const __m128i round_add = _mm_set1_epi32((1 << FILTER_BITS) >> 1); + const __m128i zero = _mm_setzero_si128(); + const __m128i clip_maximum = _mm_set1_epi16((1 << bd) - 1); + + const uint16_t *src_y; + uint16_t *dst_y; + int x_qn = x0_qn; + for (int x = 0; x < w; x += 4, x_qn += 4 * x_step_qn) { + const int x_filter_idx0 = + ((x_qn + 0 * x_step_qn) & RS_SCALE_SUBPEL_MASK) >> RS_SCALE_EXTRA_BITS; + const int x_filter_idx1 = + ((x_qn + 1 * x_step_qn) & RS_SCALE_SUBPEL_MASK) >> RS_SCALE_EXTRA_BITS; + const int x_filter_idx2 = + ((x_qn + 2 * x_step_qn) & RS_SCALE_SUBPEL_MASK) >> RS_SCALE_EXTRA_BITS; + const int x_filter_idx3 = + ((x_qn + 3 * x_step_qn) & RS_SCALE_SUBPEL_MASK) >> RS_SCALE_EXTRA_BITS; + + assert(x_filter_idx0 <= RS_SUBPEL_MASK); + assert(x_filter_idx1 <= RS_SUBPEL_MASK); + assert(x_filter_idx2 <= RS_SUBPEL_MASK); + assert(x_filter_idx3 <= RS_SUBPEL_MASK); + + const int16_t *const x_filter0 = + &x_filters[x_filter_idx0 * UPSCALE_NORMATIVE_TAPS]; + const int16_t *const x_filter1 = + &x_filters[x_filter_idx1 * UPSCALE_NORMATIVE_TAPS]; + const int16_t *const x_filter2 = + &x_filters[x_filter_idx2 * UPSCALE_NORMATIVE_TAPS]; + const int16_t *const x_filter3 = + &x_filters[x_filter_idx3 * UPSCALE_NORMATIVE_TAPS]; + + const __m128i fil0_16 = xx_loadu_128(x_filter0); + const __m128i fil1_16 = xx_loadu_128(x_filter1); + const __m128i fil2_16 = xx_loadu_128(x_filter2); + const __m128i fil3_16 = xx_loadu_128(x_filter3); + + src_y = src; + dst_y = dst; + for (int y = 0; y < h; y++, src_y += src_stride, dst_y += dst_stride) { + const uint16_t *const src_x0 = + &src_y[(x_qn + 0 * x_step_qn) >> RS_SCALE_SUBPEL_BITS]; + const uint16_t *const src_x1 = + &src_y[(x_qn + 1 * x_step_qn) >> RS_SCALE_SUBPEL_BITS]; + const uint16_t *const src_x2 = + &src_y[(x_qn + 2 * x_step_qn) >> RS_SCALE_SUBPEL_BITS]; + const uint16_t *const src_x3 = + &src_y[(x_qn + 3 * x_step_qn) >> RS_SCALE_SUBPEL_BITS]; + + // Load up the source data. This is 16-bit input data, so each load + // gets 8 pixels. + const __m128i src0_16 = xx_loadu_128(src_x0); + const __m128i src1_16 = xx_loadu_128(src_x1); + const __m128i src2_16 = xx_loadu_128(src_x2); + const __m128i src3_16 = xx_loadu_128(src_x3); + + // Multiply by filter coefficients (results in a 32-bit value), + // and add adjacent pairs, i.e. + // ([ s7 s6 s5 s4 s3 s2 s1 s0], [ f7 f6 f5 f4 f3 f2 f1 f0 ]) + // -> [ {s7*f7+s6*f6} {s5*f5+s4*f4} {s3*f3+s2*f2} {s1*f1+s0*f0} ] + const __m128i conv0_32 = _mm_madd_epi16(src0_16, fil0_16); + const __m128i conv1_32 = _mm_madd_epi16(src1_16, fil1_16); + const __m128i conv2_32 = _mm_madd_epi16(src2_16, fil2_16); + const __m128i conv3_32 = _mm_madd_epi16(src3_16, fil3_16); + + // Reduce horizontally and add, i.e. + // ([ D C B A ], [ S R Q P ]) -> [ S+R Q+P D+C B+A ] + const __m128i conv01_32 = _mm_hadd_epi32(conv0_32, conv1_32); + const __m128i conv23_32 = _mm_hadd_epi32(conv2_32, conv3_32); + + const __m128i conv0123_32 = _mm_hadd_epi32(conv01_32, conv23_32); + + // Divide down by (1 << FILTER_BITS), rounding to nearest. + const __m128i shifted_32 = + _mm_srai_epi32(_mm_add_epi32(conv0123_32, round_add), FILTER_BITS); + + // Pack 32-bit values into 16-bit values, i.e. + // ([ D C B A ], [ 0 0 0 0 ]) -> [ 0 0 0 0 D C B A ] + const __m128i shifted_16 = _mm_packus_epi32(shifted_32, zero); + + // Clip the values at (1 << bd) - 1 + const __m128i clipped_16 = _mm_min_epi16(shifted_16, clip_maximum); + + // Write to the output + xx_storel_64(&dst_y[x], clipped_16); + } + } +} diff --git a/third_party/aom/av1/common/x86/av1_convolve_scale_sse4.c b/third_party/aom/av1/common/x86/av1_convolve_scale_sse4.c new file mode 100644 index 0000000000..d9fb537856 --- /dev/null +++ b/third_party/aom/av1/common/x86/av1_convolve_scale_sse4.c @@ -0,0 +1,499 @@ +/* + * Copyright (c) 2017, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <assert.h> +#include <smmintrin.h> + +#include "config/aom_dsp_rtcd.h" + +#include "aom_dsp/aom_dsp_common.h" +#include "aom_dsp/aom_filter.h" +#include "av1/common/convolve.h" + +// A specialised version of hfilter, the horizontal filter for +// av1_convolve_2d_scale_sse4_1. This version only supports 8 tap filters. +static void hfilter8(const uint8_t *src, int src_stride, int16_t *dst, int w, + int h, int subpel_x_qn, int x_step_qn, + const InterpFilterParams *filter_params, unsigned round) { + const int bd = 8; + const int ntaps = 8; + + src -= ntaps / 2 - 1; + + int32_t round_add32 = (1 << round) / 2 + (1 << (bd + FILTER_BITS - 1)); + const __m128i round_add = _mm_set1_epi32(round_add32); + const __m128i round_shift = _mm_cvtsi32_si128(round); + + int x_qn = subpel_x_qn; + for (int x = 0; x < w; ++x, x_qn += x_step_qn) { + const uint8_t *const src_col = src + (x_qn >> SCALE_SUBPEL_BITS); + const int filter_idx = (x_qn & SCALE_SUBPEL_MASK) >> SCALE_EXTRA_BITS; + assert(filter_idx < SUBPEL_SHIFTS); + const int16_t *filter = + av1_get_interp_filter_subpel_kernel(filter_params, filter_idx); + + // Load the filter coefficients + const __m128i coefflo = _mm_loadu_si128((__m128i *)filter); + const __m128i zero = _mm_castps_si128(_mm_setzero_ps()); + + int y; + for (y = 0; y <= h - 4; y += 4) { + const uint8_t *const src0 = src_col + y * src_stride; + const uint8_t *const src1 = src0 + 1 * src_stride; + const uint8_t *const src2 = src0 + 2 * src_stride; + const uint8_t *const src3 = src0 + 3 * src_stride; + + // Load up source data. This is 8-bit input data; each load is just + // loading the lower half of the register and gets 8 pixels + const __m128i data08 = _mm_loadl_epi64((__m128i *)src0); + const __m128i data18 = _mm_loadl_epi64((__m128i *)src1); + const __m128i data28 = _mm_loadl_epi64((__m128i *)src2); + const __m128i data38 = _mm_loadl_epi64((__m128i *)src3); + + // Now zero-extend up to 16-bit precision by interleaving with + // zeros. Drop the upper half of each register (which just had zeros) + const __m128i data0lo = _mm_unpacklo_epi8(data08, zero); + const __m128i data1lo = _mm_unpacklo_epi8(data18, zero); + const __m128i data2lo = _mm_unpacklo_epi8(data28, zero); + const __m128i data3lo = _mm_unpacklo_epi8(data38, zero); + + // Multiply by coefficients + const __m128i conv0lo = _mm_madd_epi16(data0lo, coefflo); + const __m128i conv1lo = _mm_madd_epi16(data1lo, coefflo); + const __m128i conv2lo = _mm_madd_epi16(data2lo, coefflo); + const __m128i conv3lo = _mm_madd_epi16(data3lo, coefflo); + + // Reduce horizontally and add + const __m128i conv01lo = _mm_hadd_epi32(conv0lo, conv1lo); + const __m128i conv23lo = _mm_hadd_epi32(conv2lo, conv3lo); + const __m128i conv = _mm_hadd_epi32(conv01lo, conv23lo); + + // Divide down by (1 << round), rounding to nearest. + __m128i shifted = + _mm_sra_epi32(_mm_add_epi32(conv, round_add), round_shift); + + shifted = _mm_packus_epi32(shifted, shifted); + // Write transposed to the output + _mm_storel_epi64((__m128i *)(dst + y + x * h), shifted); + } + for (; y < h; ++y) { + const uint8_t *const src_row = src_col + y * src_stride; + + int32_t sum = (1 << (bd + FILTER_BITS - 1)); + for (int k = 0; k < ntaps; ++k) { + sum += filter[k] * src_row[k]; + } + + dst[y + x * h] = ROUND_POWER_OF_TWO(sum, round); + } + } +} + +static __m128i convolve_16_8(const int16_t *src, __m128i coeff) { + __m128i data = _mm_loadu_si128((__m128i *)src); + return _mm_madd_epi16(data, coeff); +} + +// A specialised version of vfilter, the vertical filter for +// av1_convolve_2d_scale_sse4_1. This version only supports 8 tap filters. +static void vfilter8(const int16_t *src, int src_stride, uint8_t *dst, + int dst_stride, int w, int h, int subpel_y_qn, + int y_step_qn, const InterpFilterParams *filter_params, + const ConvolveParams *conv_params, int bd) { + const int offset_bits = bd + 2 * FILTER_BITS - conv_params->round_0; + const int ntaps = 8; + + const __m128i round_shift = _mm_cvtsi32_si128(conv_params->round_1); + + const int32_t sub32 = ((1 << (offset_bits - conv_params->round_1)) + + (1 << (offset_bits - conv_params->round_1 - 1))); + const __m128i sub = _mm_set1_epi16(sub32); + + CONV_BUF_TYPE *dst16 = conv_params->dst; + const int dst16_stride = conv_params->dst_stride; + const int bits = + FILTER_BITS * 2 - conv_params->round_0 - conv_params->round_1; + const __m128i bits_shift = _mm_cvtsi32_si128(bits); + const __m128i bits_const = _mm_set1_epi16(((1 << bits) >> 1)); + const __m128i round_shift_add = + _mm_set1_epi32(((1 << conv_params->round_1) >> 1)); + const __m128i res_add_const = _mm_set1_epi32(1 << offset_bits); + + const int w0 = conv_params->fwd_offset; + const int w1 = conv_params->bck_offset; + const __m128i wt0 = _mm_set1_epi16(w0); + const __m128i wt1 = _mm_set1_epi16(w1); + const __m128i wt = _mm_unpacklo_epi16(wt0, wt1); + + int y_qn = subpel_y_qn; + for (int y = 0; y < h; ++y, y_qn += y_step_qn) { + const int16_t *src_y = src + (y_qn >> SCALE_SUBPEL_BITS); + const int filter_idx = (y_qn & SCALE_SUBPEL_MASK) >> SCALE_EXTRA_BITS; + assert(filter_idx < SUBPEL_SHIFTS); + const int16_t *filter = + av1_get_interp_filter_subpel_kernel(filter_params, filter_idx); + + const __m128i coeff0716 = _mm_loadu_si128((__m128i *)filter); + int x; + for (x = 0; x <= w - 4; x += 4) { + const int16_t *const src0 = src_y + x * src_stride; + const int16_t *const src1 = src0 + 1 * src_stride; + const int16_t *const src2 = src0 + 2 * src_stride; + const int16_t *const src3 = src0 + 3 * src_stride; + + // Load the source data for the three rows, adding the three registers of + // convolved products to one as we go (conv0..conv3) to avoid the + // register pressure getting too high. + const __m128i conv0 = convolve_16_8(src0, coeff0716); + const __m128i conv1 = convolve_16_8(src1, coeff0716); + const __m128i conv2 = convolve_16_8(src2, coeff0716); + const __m128i conv3 = convolve_16_8(src3, coeff0716); + + // Now reduce horizontally to get one lane for each result + const __m128i conv01 = _mm_hadd_epi32(conv0, conv1); + const __m128i conv23 = _mm_hadd_epi32(conv2, conv3); + __m128i conv = _mm_hadd_epi32(conv01, conv23); + + conv = _mm_add_epi32(conv, res_add_const); + // Divide down by (1 << round_1), rounding to nearest and subtract sub32. + __m128i shifted = + _mm_sra_epi32(_mm_add_epi32(conv, round_shift_add), round_shift); + + uint8_t *dst_x = dst + y * dst_stride + x; + CONV_BUF_TYPE *dst_16_x = dst16 + y * dst16_stride + x; + __m128i result; + __m128i shifted_16 = _mm_packus_epi32(shifted, shifted); + + if (conv_params->is_compound) { + if (conv_params->do_average) { + const __m128i p_16 = _mm_loadl_epi64((__m128i *)dst_16_x); + if (conv_params->use_jnt_comp_avg) { + const __m128i p_16_lo = _mm_unpacklo_epi16(p_16, shifted_16); + const __m128i wt_res_lo = _mm_madd_epi16(p_16_lo, wt); + const __m128i shifted_32 = + _mm_srai_epi32(wt_res_lo, DIST_PRECISION_BITS); + shifted_16 = _mm_packus_epi32(shifted_32, shifted_32); + } else { + shifted_16 = _mm_srai_epi16(_mm_add_epi16(p_16, shifted_16), 1); + } + const __m128i subbed = _mm_sub_epi16(shifted_16, sub); + result = _mm_sra_epi16(_mm_add_epi16(subbed, bits_const), bits_shift); + const __m128i result_8 = _mm_packus_epi16(result, result); + *(uint32_t *)dst_x = _mm_cvtsi128_si32(result_8); + } else { + _mm_storel_epi64((__m128i *)dst_16_x, shifted_16); + } + } else { + const __m128i subbed = _mm_sub_epi16(shifted_16, sub); + result = _mm_sra_epi16(_mm_add_epi16(subbed, bits_const), bits_shift); + const __m128i result_8 = _mm_packus_epi16(result, result); + *(uint32_t *)dst_x = _mm_cvtsi128_si32(result_8); + } + } + for (; x < w; ++x) { + const int16_t *src_x = src_y + x * src_stride; + int32_t sum = 1 << offset_bits; + for (int k = 0; k < ntaps; ++k) sum += filter[k] * src_x[k]; + CONV_BUF_TYPE res = ROUND_POWER_OF_TWO(sum, conv_params->round_1); + + if (conv_params->is_compound) { + if (conv_params->do_average) { + int32_t tmp = dst16[y * dst16_stride + x]; + if (conv_params->use_jnt_comp_avg) { + tmp = tmp * conv_params->fwd_offset + res * conv_params->bck_offset; + tmp = tmp >> DIST_PRECISION_BITS; + } else { + tmp += res; + tmp = tmp >> 1; + } + /* Subtract round offset and convolve round */ + tmp = tmp - sub32; + dst[y * dst_stride + x] = clip_pixel(ROUND_POWER_OF_TWO(tmp, bits)); + } else { + dst16[y * dst16_stride + x] = res; + } + } else { + /* Subtract round offset and convolve round */ + int32_t tmp = res - ((1 << (offset_bits - conv_params->round_1)) + + (1 << (offset_bits - conv_params->round_1 - 1))); + dst[y * dst_stride + x] = clip_pixel(ROUND_POWER_OF_TWO(tmp, bits)); + } + } + } +} +void av1_convolve_2d_scale_sse4_1(const uint8_t *src, int src_stride, + uint8_t *dst8, int dst8_stride, int w, int h, + const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, + const int subpel_x_qn, const int x_step_qn, + const int subpel_y_qn, const int y_step_qn, + ConvolveParams *conv_params) { + // TODO(yaowu): remove unnecessary initializations + int16_t tmp[(2 * MAX_SB_SIZE + MAX_FILTER_TAP) * MAX_SB_SIZE] = { 0 }; + int im_h = (((h - 1) * y_step_qn + subpel_y_qn) >> SCALE_SUBPEL_BITS) + + filter_params_y->taps; + + const int xtaps = filter_params_x->taps; + const int ytaps = filter_params_y->taps; + const int fo_vert = ytaps / 2 - 1; + assert((xtaps == 8) && (ytaps == 8)); + (void)xtaps; + + // horizontal filter + hfilter8(src - fo_vert * src_stride, src_stride, tmp, w, im_h, subpel_x_qn, + x_step_qn, filter_params_x, conv_params->round_0); + + // vertical filter (input is transposed) + vfilter8(tmp, im_h, dst8, dst8_stride, w, h, subpel_y_qn, y_step_qn, + filter_params_y, conv_params, 8); +} + +// A specialised version of hfilter, the horizontal filter for +// av1_highbd_convolve_2d_scale_sse4_1. This version only supports 8 tap +// filters. +static void highbd_hfilter8(const uint16_t *src, int src_stride, int16_t *dst, + int w, int h, int subpel_x_qn, int x_step_qn, + const InterpFilterParams *filter_params, + unsigned round, int bd) { + const int ntaps = 8; + + src -= ntaps / 2 - 1; + + int32_t round_add32 = (1 << round) / 2 + (1 << (bd + FILTER_BITS - 1)); + const __m128i round_add = _mm_set1_epi32(round_add32); + const __m128i round_shift = _mm_cvtsi32_si128(round); + + int x_qn = subpel_x_qn; + for (int x = 0; x < w; ++x, x_qn += x_step_qn) { + const uint16_t *const src_col = src + (x_qn >> SCALE_SUBPEL_BITS); + const int filter_idx = (x_qn & SCALE_SUBPEL_MASK) >> SCALE_EXTRA_BITS; + assert(filter_idx < SUBPEL_SHIFTS); + const int16_t *filter = + av1_get_interp_filter_subpel_kernel(filter_params, filter_idx); + + // Load the filter coefficients + const __m128i coefflo = _mm_loadu_si128((__m128i *)filter); + + int y; + for (y = 0; y <= h - 4; y += 4) { + const uint16_t *const src0 = src_col + y * src_stride; + const uint16_t *const src1 = src0 + 1 * src_stride; + const uint16_t *const src2 = src0 + 2 * src_stride; + const uint16_t *const src3 = src0 + 3 * src_stride; + + // Load up source data. This is 16-bit input data, so each load gets the 8 + // pixels we need. + const __m128i data0lo = _mm_loadu_si128((__m128i *)src0); + const __m128i data1lo = _mm_loadu_si128((__m128i *)src1); + const __m128i data2lo = _mm_loadu_si128((__m128i *)src2); + const __m128i data3lo = _mm_loadu_si128((__m128i *)src3); + + // Multiply by coefficients + const __m128i conv0lo = _mm_madd_epi16(data0lo, coefflo); + const __m128i conv1lo = _mm_madd_epi16(data1lo, coefflo); + const __m128i conv2lo = _mm_madd_epi16(data2lo, coefflo); + const __m128i conv3lo = _mm_madd_epi16(data3lo, coefflo); + + // Reduce horizontally and add + const __m128i conv01lo = _mm_hadd_epi32(conv0lo, conv1lo); + const __m128i conv23lo = _mm_hadd_epi32(conv2lo, conv3lo); + const __m128i conv = _mm_hadd_epi32(conv01lo, conv23lo); + + // Divide down by (1 << round), rounding to nearest. + __m128i shifted = + _mm_sra_epi32(_mm_add_epi32(conv, round_add), round_shift); + + shifted = _mm_packus_epi32(shifted, shifted); + // Write transposed to the output + _mm_storel_epi64((__m128i *)(dst + y + x * h), shifted); + } + for (; y < h; ++y) { + const uint16_t *const src_row = src_col + y * src_stride; + + int32_t sum = (1 << (bd + FILTER_BITS - 1)); + for (int k = 0; k < ntaps; ++k) { + sum += filter[k] * src_row[k]; + } + + dst[y + x * h] = ROUND_POWER_OF_TWO(sum, round); + } + } +} +// A specialised version of vfilter, the vertical filter for +// av1_highbd_convolve_2d_scale_sse4_1. This version only supports 8 tap +// filters. +static void highbd_vfilter8(const int16_t *src, int src_stride, uint16_t *dst, + int dst_stride, int w, int h, int subpel_y_qn, + int y_step_qn, + const InterpFilterParams *filter_params, + const ConvolveParams *conv_params, int bd) { + const int offset_bits = bd + 2 * FILTER_BITS - conv_params->round_0; + const int ntaps = 8; + + const __m128i round_shift = _mm_cvtsi32_si128(conv_params->round_1); + + const int32_t sub32 = ((1 << (offset_bits - conv_params->round_1)) + + (1 << (offset_bits - conv_params->round_1 - 1))); + const __m128i sub = _mm_set1_epi32(sub32); + + CONV_BUF_TYPE *dst16 = conv_params->dst; + const int dst16_stride = conv_params->dst_stride; + const __m128i clip_pixel_ = + _mm_set1_epi16(bd == 10 ? 1023 : (bd == 12 ? 4095 : 255)); + const int bits = + FILTER_BITS * 2 - conv_params->round_0 - conv_params->round_1; + const __m128i bits_shift = _mm_cvtsi32_si128(bits); + const __m128i bits_const = _mm_set1_epi32(((1 << bits) >> 1)); + const __m128i round_shift_add = + _mm_set1_epi32(((1 << conv_params->round_1) >> 1)); + const __m128i res_add_const = _mm_set1_epi32(1 << offset_bits); + const int round_bits = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + __m128i round_bits_shift = _mm_cvtsi32_si128(round_bits); + __m128i round_bits_const = _mm_set1_epi32(((1 << round_bits) >> 1)); + + const int w0 = conv_params->fwd_offset; + const int w1 = conv_params->bck_offset; + const __m128i wt0 = _mm_set1_epi32(w0); + const __m128i wt1 = _mm_set1_epi32(w1); + + int y_qn = subpel_y_qn; + for (int y = 0; y < h; ++y, y_qn += y_step_qn) { + const int16_t *src_y = src + (y_qn >> SCALE_SUBPEL_BITS); + const int filter_idx = (y_qn & SCALE_SUBPEL_MASK) >> SCALE_EXTRA_BITS; + assert(filter_idx < SUBPEL_SHIFTS); + const int16_t *filter = + av1_get_interp_filter_subpel_kernel(filter_params, filter_idx); + + const __m128i coeff0716 = _mm_loadu_si128((__m128i *)filter); + int x; + for (x = 0; x <= w - 4; x += 4) { + const int16_t *const src0 = src_y + x * src_stride; + const int16_t *const src1 = src0 + 1 * src_stride; + const int16_t *const src2 = src0 + 2 * src_stride; + const int16_t *const src3 = src0 + 3 * src_stride; + + // Load the source data for the three rows, adding the three registers of + // convolved products to one as we go (conv0..conv3) to avoid the + // register pressure getting too high. + const __m128i conv0 = convolve_16_8(src0, coeff0716); + const __m128i conv1 = convolve_16_8(src1, coeff0716); + const __m128i conv2 = convolve_16_8(src2, coeff0716); + const __m128i conv3 = convolve_16_8(src3, coeff0716); + + // Now reduce horizontally to get one lane for each result + const __m128i conv01 = _mm_hadd_epi32(conv0, conv1); + const __m128i conv23 = _mm_hadd_epi32(conv2, conv3); + __m128i conv = _mm_hadd_epi32(conv01, conv23); + conv = _mm_add_epi32(conv, res_add_const); + + // Divide down by (1 << round_1), rounding to nearest and subtract sub32. + __m128i shifted = + _mm_sra_epi32(_mm_add_epi32(conv, round_shift_add), round_shift); + + uint16_t *dst_x = dst + y * dst_stride + x; + CONV_BUF_TYPE *dst_16_x = dst16 + y * dst16_stride + x; + + __m128i result; + if (conv_params->is_compound) { + if (conv_params->do_average) { + __m128i p_32 = + _mm_cvtepu16_epi32(_mm_loadl_epi64((__m128i *)dst_16_x)); + + if (conv_params->use_jnt_comp_avg) { + shifted = _mm_add_epi32(_mm_mullo_epi32(p_32, wt0), + _mm_mullo_epi32(shifted, wt1)); + shifted = _mm_srai_epi32(shifted, DIST_PRECISION_BITS); + } else { + shifted = _mm_srai_epi32(_mm_add_epi32(p_32, shifted), 1); + } + __m128i res32 = _mm_sub_epi32(shifted, sub); + res32 = _mm_sra_epi32(_mm_add_epi32(res32, round_bits_const), + round_bits_shift); + + __m128i res16 = _mm_packus_epi32(res32, res32); + res16 = _mm_min_epi16(res16, clip_pixel_); + _mm_storel_epi64((__m128i *)dst_x, res16); + } else { + __m128i shifted_16 = _mm_packus_epi32(shifted, shifted); + _mm_storel_epi64((__m128i *)dst_16_x, shifted_16); + } + } else { + const __m128i subbed = _mm_sub_epi32(shifted, sub); + result = _mm_sra_epi16(_mm_add_epi32(subbed, bits_const), bits_shift); + result = _mm_packus_epi32(result, result); + result = _mm_min_epi16(result, clip_pixel_); + _mm_storel_epi64((__m128i *)dst_x, result); + } + } + + for (; x < w; ++x) { + const int16_t *src_x = src_y + x * src_stride; + int32_t sum = 1 << offset_bits; + for (int k = 0; k < ntaps; ++k) sum += filter[k] * src_x[k]; + CONV_BUF_TYPE res = ROUND_POWER_OF_TWO(sum, conv_params->round_1); + if (conv_params->is_compound) { + if (conv_params->do_average) { + int32_t tmp = dst16[y * dst16_stride + x]; + if (conv_params->use_jnt_comp_avg) { + tmp = tmp * conv_params->fwd_offset + res * conv_params->bck_offset; + tmp = tmp >> DIST_PRECISION_BITS; + } else { + tmp += res; + tmp = tmp >> 1; + } + /* Subtract round offset and convolve round */ + tmp = tmp - ((1 << (offset_bits - conv_params->round_1)) + + (1 << (offset_bits - conv_params->round_1 - 1))); + dst[y * dst_stride + x] = + clip_pixel_highbd(ROUND_POWER_OF_TWO(tmp, bits), bd); + } else { + dst16[y * dst16_stride + x] = res; + } + } else { + /* Subtract round offset and convolve round */ + int32_t tmp = res - ((1 << (offset_bits - conv_params->round_1)) + + (1 << (offset_bits - conv_params->round_1 - 1))); + dst[y * dst_stride + x] = + clip_pixel_highbd(ROUND_POWER_OF_TWO(tmp, bits), bd); + } + } + } +} + +void av1_highbd_convolve_2d_scale_sse4_1( + const uint16_t *src, int src_stride, uint16_t *dst, int dst_stride, int w, + int h, const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, const int subpel_x_qn, + const int x_step_qn, const int subpel_y_qn, const int y_step_qn, + ConvolveParams *conv_params, int bd) { + // TODO(yaowu): Move this out of stack + DECLARE_ALIGNED(16, int16_t, + tmp[(2 * MAX_SB_SIZE + MAX_FILTER_TAP) * MAX_SB_SIZE]); + int im_h = (((h - 1) * y_step_qn + subpel_y_qn) >> SCALE_SUBPEL_BITS) + + filter_params_y->taps; + const int xtaps = filter_params_x->taps; + const int ytaps = filter_params_y->taps; + const int fo_vert = ytaps / 2 - 1; + + memset(tmp, 0, sizeof(tmp)); + assert((xtaps == 8) && (ytaps == 8)); + (void)xtaps; + + // horizontal filter + highbd_hfilter8(src - fo_vert * src_stride, src_stride, tmp, w, im_h, + subpel_x_qn, x_step_qn, filter_params_x, conv_params->round_0, + bd); + + // vertical filter (input is transposed) + highbd_vfilter8(tmp, im_h, dst, dst_stride, w, h, subpel_y_qn, y_step_qn, + filter_params_y, conv_params, bd); +} diff --git a/third_party/aom/av1/common/x86/av1_highbd_convolve_sse4.c b/third_party/aom/av1/common/x86/av1_highbd_convolve_sse4.c new file mode 100644 index 0000000000..212d3bd723 --- /dev/null +++ b/third_party/aom/av1/common/x86/av1_highbd_convolve_sse4.c @@ -0,0 +1,205 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <assert.h> +#include <smmintrin.h> + +#include "config/av1_rtcd.h" + +#include "av1/common/filter.h" + +typedef void (*TransposeSave)(int width, int pixelsNum, uint32_t *src, + int src_stride, uint16_t *dst, int dst_stride, + int bd); + +// pixelsNum 0: write all 4 pixels +// 1/2/3: residual pixels 1/2/3 +static void writePixel(__m128i *u, int width, int pixelsNum, uint16_t *dst, + int dst_stride) { + if (2 == width) { + if (0 == pixelsNum) { + *(int *)dst = _mm_cvtsi128_si32(u[0]); + *(int *)(dst + dst_stride) = _mm_cvtsi128_si32(u[1]); + *(int *)(dst + 2 * dst_stride) = _mm_cvtsi128_si32(u[2]); + *(int *)(dst + 3 * dst_stride) = _mm_cvtsi128_si32(u[3]); + } else if (1 == pixelsNum) { + *(int *)dst = _mm_cvtsi128_si32(u[0]); + } else if (2 == pixelsNum) { + *(int *)dst = _mm_cvtsi128_si32(u[0]); + *(int *)(dst + dst_stride) = _mm_cvtsi128_si32(u[1]); + } else if (3 == pixelsNum) { + *(int *)dst = _mm_cvtsi128_si32(u[0]); + *(int *)(dst + dst_stride) = _mm_cvtsi128_si32(u[1]); + *(int *)(dst + 2 * dst_stride) = _mm_cvtsi128_si32(u[2]); + } + } else { + if (0 == pixelsNum) { + _mm_storel_epi64((__m128i *)dst, u[0]); + _mm_storel_epi64((__m128i *)(dst + dst_stride), u[1]); + _mm_storel_epi64((__m128i *)(dst + 2 * dst_stride), u[2]); + _mm_storel_epi64((__m128i *)(dst + 3 * dst_stride), u[3]); + } else if (1 == pixelsNum) { + _mm_storel_epi64((__m128i *)dst, u[0]); + } else if (2 == pixelsNum) { + _mm_storel_epi64((__m128i *)dst, u[0]); + _mm_storel_epi64((__m128i *)(dst + dst_stride), u[1]); + } else if (3 == pixelsNum) { + _mm_storel_epi64((__m128i *)dst, u[0]); + _mm_storel_epi64((__m128i *)(dst + dst_stride), u[1]); + _mm_storel_epi64((__m128i *)(dst + 2 * dst_stride), u[2]); + } + } +} + +// 16-bit pixels clip with bd (10/12) +static void highbd_clip(__m128i *p, int numVecs, int bd) { + const __m128i zero = _mm_setzero_si128(); + const __m128i one = _mm_set1_epi16(1); + const __m128i max = _mm_sub_epi16(_mm_slli_epi16(one, bd), one); + __m128i clamped, mask; + int i; + + for (i = 0; i < numVecs; i++) { + mask = _mm_cmpgt_epi16(p[i], max); + clamped = _mm_andnot_si128(mask, p[i]); + mask = _mm_and_si128(mask, max); + clamped = _mm_or_si128(mask, clamped); + mask = _mm_cmpgt_epi16(clamped, zero); + p[i] = _mm_and_si128(clamped, mask); + } +} + +static void transClipPixel(uint32_t *src, int src_stride, __m128i *u, int bd) { + __m128i v0, v1; + __m128i rnd = _mm_set1_epi32(1 << (FILTER_BITS - 1)); + + u[0] = _mm_loadu_si128((__m128i const *)src); + u[1] = _mm_loadu_si128((__m128i const *)(src + src_stride)); + u[2] = _mm_loadu_si128((__m128i const *)(src + 2 * src_stride)); + u[3] = _mm_loadu_si128((__m128i const *)(src + 3 * src_stride)); + + u[0] = _mm_add_epi32(u[0], rnd); + u[1] = _mm_add_epi32(u[1], rnd); + u[2] = _mm_add_epi32(u[2], rnd); + u[3] = _mm_add_epi32(u[3], rnd); + + u[0] = _mm_srai_epi32(u[0], FILTER_BITS); + u[1] = _mm_srai_epi32(u[1], FILTER_BITS); + u[2] = _mm_srai_epi32(u[2], FILTER_BITS); + u[3] = _mm_srai_epi32(u[3], FILTER_BITS); + + u[0] = _mm_packus_epi32(u[0], u[1]); + u[1] = _mm_packus_epi32(u[2], u[3]); + + highbd_clip(u, 2, bd); + + v0 = _mm_unpacklo_epi16(u[0], u[1]); + v1 = _mm_unpackhi_epi16(u[0], u[1]); + + u[0] = _mm_unpacklo_epi16(v0, v1); + u[2] = _mm_unpackhi_epi16(v0, v1); + + u[1] = _mm_srli_si128(u[0], 8); + u[3] = _mm_srli_si128(u[2], 8); +} + +// pixelsNum = 0 : all 4 rows of pixels will be saved. +// pixelsNum = 1/2/3 : residual 1/2/4 rows of pixels will be saved. +void trans_save_4x4(int width, int pixelsNum, uint32_t *src, int src_stride, + uint16_t *dst, int dst_stride, int bd) { + __m128i u[4]; + transClipPixel(src, src_stride, u, bd); + writePixel(u, width, pixelsNum, dst, dst_stride); +} + +void trans_accum_save_4x4(int width, int pixelsNum, uint32_t *src, + int src_stride, uint16_t *dst, int dst_stride, + int bd) { + __m128i u[4], v[4]; + const __m128i ones = _mm_set1_epi16(1); + + transClipPixel(src, src_stride, u, bd); + + v[0] = _mm_loadl_epi64((__m128i const *)dst); + v[1] = _mm_loadl_epi64((__m128i const *)(dst + dst_stride)); + v[2] = _mm_loadl_epi64((__m128i const *)(dst + 2 * dst_stride)); + v[3] = _mm_loadl_epi64((__m128i const *)(dst + 3 * dst_stride)); + + u[0] = _mm_add_epi16(u[0], v[0]); + u[1] = _mm_add_epi16(u[1], v[1]); + u[2] = _mm_add_epi16(u[2], v[2]); + u[3] = _mm_add_epi16(u[3], v[3]); + + u[0] = _mm_add_epi16(u[0], ones); + u[1] = _mm_add_epi16(u[1], ones); + u[2] = _mm_add_epi16(u[2], ones); + u[3] = _mm_add_epi16(u[3], ones); + + u[0] = _mm_srai_epi16(u[0], 1); + u[1] = _mm_srai_epi16(u[1], 1); + u[2] = _mm_srai_epi16(u[2], 1); + u[3] = _mm_srai_epi16(u[3], 1); + + writePixel(u, width, pixelsNum, dst, dst_stride); +} + +// Vertical convolutional filter + +typedef void (*WritePixels)(__m128i *u, int bd, uint16_t *dst); + +static void highbdRndingPacks(__m128i *u) { + __m128i rnd = _mm_set1_epi32(1 << (FILTER_BITS - 1)); + u[0] = _mm_add_epi32(u[0], rnd); + u[0] = _mm_srai_epi32(u[0], FILTER_BITS); + u[0] = _mm_packus_epi32(u[0], u[0]); +} + +static void write2pixelsOnly(__m128i *u, int bd, uint16_t *dst) { + highbdRndingPacks(u); + highbd_clip(u, 1, bd); + *(uint32_t *)dst = _mm_cvtsi128_si32(u[0]); +} + +static void write2pixelsAccum(__m128i *u, int bd, uint16_t *dst) { + __m128i v = _mm_loadl_epi64((__m128i const *)dst); + const __m128i ones = _mm_set1_epi16(1); + + highbdRndingPacks(u); + highbd_clip(u, 1, bd); + + v = _mm_add_epi16(v, u[0]); + v = _mm_add_epi16(v, ones); + v = _mm_srai_epi16(v, 1); + *(uint32_t *)dst = _mm_cvtsi128_si32(v); +} + +WritePixels write2pixelsTab[2] = { write2pixelsOnly, write2pixelsAccum }; + +static void write4pixelsOnly(__m128i *u, int bd, uint16_t *dst) { + highbdRndingPacks(u); + highbd_clip(u, 1, bd); + _mm_storel_epi64((__m128i *)dst, u[0]); +} + +static void write4pixelsAccum(__m128i *u, int bd, uint16_t *dst) { + __m128i v = _mm_loadl_epi64((__m128i const *)dst); + const __m128i ones = _mm_set1_epi16(1); + + highbdRndingPacks(u); + highbd_clip(u, 1, bd); + + v = _mm_add_epi16(v, u[0]); + v = _mm_add_epi16(v, ones); + v = _mm_srai_epi16(v, 1); + _mm_storel_epi64((__m128i *)dst, v); +} + +WritePixels write4pixelsTab[2] = { write4pixelsOnly, write4pixelsAccum }; diff --git a/third_party/aom/av1/common/x86/av1_inv_txfm_avx2.c b/third_party/aom/av1/common/x86/av1_inv_txfm_avx2.c new file mode 100644 index 0000000000..5db2ccf6c3 --- /dev/null +++ b/third_party/aom/av1/common/x86/av1_inv_txfm_avx2.c @@ -0,0 +1,1945 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include "config/aom_config.h" + +#include "config/av1_rtcd.h" + +#include "av1/common/av1_inv_txfm1d_cfg.h" +#include "av1/common/x86/av1_txfm_sse2.h" +#include "av1/common/x86/av1_inv_txfm_avx2.h" +#include "av1/common/x86/av1_inv_txfm_ssse3.h" + +// TODO(venkatsanampudi@ittiam.com): move this to header file + +// Sqrt2, Sqrt2^2, Sqrt2^3, Sqrt2^4, Sqrt2^5 +static int32_t NewSqrt2list[TX_SIZES] = { 5793, 2 * 4096, 2 * 5793, 4 * 4096, + 4 * 5793 }; + +static INLINE void idct16_stage5_avx2(__m256i *x1, const int32_t *cospi, + const __m256i _r, int8_t cos_bit) { + const __m256i cospi_m32_p32 = pair_set_w16_epi16(-cospi[32], cospi[32]); + const __m256i cospi_p32_p32 = pair_set_w16_epi16(cospi[32], cospi[32]); + btf_16_adds_subs_avx2(&x1[0], &x1[3]); + btf_16_adds_subs_avx2(&x1[1], &x1[2]); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x1[5], &x1[6], _r, cos_bit); + + btf_16_adds_subs_avx2(&x1[8], &x1[11]); + btf_16_adds_subs_avx2(&x1[9], &x1[10]); + btf_16_adds_subs_avx2(&x1[15], &x1[12]); + btf_16_adds_subs_avx2(&x1[14], &x1[13]); +} + +static INLINE void idct16_stage6_avx2(__m256i *x, const int32_t *cospi, + const __m256i _r, int8_t cos_bit) { + const __m256i cospi_m32_p32 = pair_set_w16_epi16(-cospi[32], cospi[32]); + const __m256i cospi_p32_p32 = pair_set_w16_epi16(cospi[32], cospi[32]); + btf_16_adds_subs_avx2(&x[0], &x[7]); + btf_16_adds_subs_avx2(&x[1], &x[6]); + btf_16_adds_subs_avx2(&x[2], &x[5]); + btf_16_adds_subs_avx2(&x[3], &x[4]); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[10], &x[13], _r, cos_bit); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[11], &x[12], _r, cos_bit); +} + +static INLINE void idct16_stage7_avx2(__m256i *output, __m256i *x1) { + btf_16_adds_subs_out_avx2(&output[0], &output[15], x1[0], x1[15]); + btf_16_adds_subs_out_avx2(&output[1], &output[14], x1[1], x1[14]); + btf_16_adds_subs_out_avx2(&output[2], &output[13], x1[2], x1[13]); + btf_16_adds_subs_out_avx2(&output[3], &output[12], x1[3], x1[12]); + btf_16_adds_subs_out_avx2(&output[4], &output[11], x1[4], x1[11]); + btf_16_adds_subs_out_avx2(&output[5], &output[10], x1[5], x1[10]); + btf_16_adds_subs_out_avx2(&output[6], &output[9], x1[6], x1[9]); + btf_16_adds_subs_out_avx2(&output[7], &output[8], x1[7], x1[8]); +} + +static void idct16_new_avx2(const __m256i *input, __m256i *output, + int8_t cos_bit) { + (void)(cos_bit); + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m256i _r = _mm256_set1_epi32(1 << (INV_COS_BIT - 1)); + + __m256i cospi_p60_m04 = pair_set_w16_epi16(cospi[60], -cospi[4]); + __m256i cospi_p04_p60 = pair_set_w16_epi16(cospi[4], cospi[60]); + __m256i cospi_p28_m36 = pair_set_w16_epi16(cospi[28], -cospi[36]); + __m256i cospi_p36_p28 = pair_set_w16_epi16(cospi[36], cospi[28]); + __m256i cospi_p44_m20 = pair_set_w16_epi16(cospi[44], -cospi[20]); + __m256i cospi_p20_p44 = pair_set_w16_epi16(cospi[20], cospi[44]); + __m256i cospi_p12_m52 = pair_set_w16_epi16(cospi[12], -cospi[52]); + __m256i cospi_p52_p12 = pair_set_w16_epi16(cospi[52], cospi[12]); + __m256i cospi_p56_m08 = pair_set_w16_epi16(cospi[56], -cospi[8]); + __m256i cospi_p08_p56 = pair_set_w16_epi16(cospi[8], cospi[56]); + __m256i cospi_p24_m40 = pair_set_w16_epi16(cospi[24], -cospi[40]); + __m256i cospi_p40_p24 = pair_set_w16_epi16(cospi[40], cospi[24]); + __m256i cospi_p32_p32 = pair_set_w16_epi16(cospi[32], cospi[32]); + __m256i cospi_p32_m32 = pair_set_w16_epi16(cospi[32], -cospi[32]); + __m256i cospi_p48_m16 = pair_set_w16_epi16(cospi[48], -cospi[16]); + __m256i cospi_p16_p48 = pair_set_w16_epi16(cospi[16], cospi[48]); + __m256i cospi_m16_p48 = pair_set_w16_epi16(-cospi[16], cospi[48]); + __m256i cospi_p48_p16 = pair_set_w16_epi16(cospi[48], cospi[16]); + __m256i cospi_m48_m16 = pair_set_w16_epi16(-cospi[48], -cospi[16]); + + // stage 1 + __m256i x1[16]; + x1[0] = input[0]; + x1[1] = input[8]; + x1[2] = input[4]; + x1[3] = input[12]; + x1[4] = input[2]; + x1[5] = input[10]; + x1[6] = input[6]; + x1[7] = input[14]; + x1[8] = input[1]; + x1[9] = input[9]; + x1[10] = input[5]; + x1[11] = input[13]; + x1[12] = input[3]; + x1[13] = input[11]; + x1[14] = input[7]; + x1[15] = input[15]; + + // stage 2 + btf_16_w16_avx2(cospi_p60_m04, cospi_p04_p60, &x1[8], &x1[15], _r, cos_bit); + btf_16_w16_avx2(cospi_p28_m36, cospi_p36_p28, &x1[9], &x1[14], _r, cos_bit); + btf_16_w16_avx2(cospi_p44_m20, cospi_p20_p44, &x1[10], &x1[13], _r, cos_bit); + btf_16_w16_avx2(cospi_p12_m52, cospi_p52_p12, &x1[11], &x1[12], _r, cos_bit); + + // stage 3 + btf_16_w16_avx2(cospi_p56_m08, cospi_p08_p56, &x1[4], &x1[7], _r, cos_bit); + btf_16_w16_avx2(cospi_p24_m40, cospi_p40_p24, &x1[5], &x1[6], _r, cos_bit); + btf_16_adds_subs_avx2(&x1[8], &x1[9]); + btf_16_adds_subs_avx2(&x1[11], &x1[10]); + btf_16_adds_subs_avx2(&x1[12], &x1[13]); + btf_16_adds_subs_avx2(&x1[15], &x1[14]); + + // stage 4 + btf_16_w16_avx2(cospi_p32_p32, cospi_p32_m32, &x1[0], &x1[1], _r, cos_bit); + btf_16_w16_avx2(cospi_p48_m16, cospi_p16_p48, &x1[2], &x1[3], _r, cos_bit); + btf_16_adds_subs_avx2(&x1[4], &x1[5]); + btf_16_adds_subs_avx2(&x1[7], &x1[6]); + btf_16_w16_avx2(cospi_m16_p48, cospi_p48_p16, &x1[9], &x1[14], _r, cos_bit); + btf_16_w16_avx2(cospi_m48_m16, cospi_m16_p48, &x1[10], &x1[13], _r, cos_bit); + + idct16_stage5_avx2(x1, cospi, _r, cos_bit); + idct16_stage6_avx2(x1, cospi, _r, cos_bit); + idct16_stage7_avx2(output, x1); +} + +static void idct16_low8_new_avx2(const __m256i *input, __m256i *output, + int8_t cos_bit) { + (void)(cos_bit); + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m256i _r = _mm256_set1_epi32(1 << (INV_COS_BIT - 1)); + + const __m256i cospi_m16_p48 = pair_set_w16_epi16(-cospi[16], cospi[48]); + const __m256i cospi_p48_p16 = pair_set_w16_epi16(cospi[48], cospi[16]); + const __m256i cospi_m48_m16 = pair_set_w16_epi16(-cospi[48], -cospi[16]); + + // stage 1 + __m256i x1[16]; + x1[0] = input[0]; + x1[2] = input[4]; + x1[4] = input[2]; + x1[6] = input[6]; + x1[8] = input[1]; + x1[10] = input[5]; + x1[12] = input[3]; + x1[14] = input[7]; + + // stage 2 + btf_16_w16_0_avx2(cospi[60], cospi[4], x1[8], x1[8], x1[15]); + btf_16_w16_0_avx2(-cospi[36], cospi[28], x1[14], x1[9], x1[14]); + btf_16_w16_0_avx2(cospi[44], cospi[20], x1[10], x1[10], x1[13]); + btf_16_w16_0_avx2(-cospi[52], cospi[12], x1[12], x1[11], x1[12]); + + // stage 3 + btf_16_w16_0_avx2(cospi[56], cospi[8], x1[4], x1[4], x1[7]); + btf_16_w16_0_avx2(-cospi[40], cospi[24], x1[6], x1[5], x1[6]); + btf_16_adds_subs_avx2(&x1[8], &x1[9]); + btf_16_adds_subs_avx2(&x1[11], &x1[10]); + btf_16_adds_subs_avx2(&x1[12], &x1[13]); + btf_16_adds_subs_avx2(&x1[15], &x1[14]); + + // stage 4 + btf_16_w16_0_avx2(cospi[32], cospi[32], x1[0], x1[0], x1[1]); + btf_16_w16_0_avx2(cospi[48], cospi[16], x1[2], x1[2], x1[3]); + btf_16_adds_subs_avx2(&x1[4], &x1[5]); + btf_16_adds_subs_avx2(&x1[7], &x1[6]); + btf_16_w16_avx2(cospi_m16_p48, cospi_p48_p16, &x1[9], &x1[14], _r, cos_bit); + btf_16_w16_avx2(cospi_m48_m16, cospi_m16_p48, &x1[10], &x1[13], _r, cos_bit); + + idct16_stage5_avx2(x1, cospi, _r, cos_bit); + idct16_stage6_avx2(x1, cospi, _r, cos_bit); + idct16_stage7_avx2(output, x1); +} + +static void idct16_low1_new_avx2(const __m256i *input, __m256i *output, + int8_t cos_bit) { + (void)(cos_bit); + const int32_t *cospi = cospi_arr(INV_COS_BIT); + + // stage 1 + __m256i x1[2]; + x1[0] = input[0]; + + // stage 2 + // stage 3 + // stage 4 + btf_16_w16_0_avx2(cospi[32], cospi[32], x1[0], x1[0], x1[1]); + + // stage 5 + // stage 6 + output[0] = x1[0]; + output[1] = x1[1]; + output[2] = x1[1]; + output[3] = x1[0]; + output[4] = x1[0]; + output[5] = x1[1]; + output[6] = x1[1]; + output[7] = x1[0]; + output[8] = x1[0]; + output[9] = x1[1]; + output[10] = x1[1]; + output[11] = x1[0]; + output[12] = x1[0]; + output[13] = x1[1]; + output[14] = x1[1]; + output[15] = x1[0]; +} + +static INLINE void iadst16_stage3_avx2(__m256i *x) { + btf_16_adds_subs_avx2(&x[0], &x[8]); + btf_16_adds_subs_avx2(&x[1], &x[9]); + btf_16_adds_subs_avx2(&x[2], &x[10]); + btf_16_adds_subs_avx2(&x[3], &x[11]); + btf_16_adds_subs_avx2(&x[4], &x[12]); + btf_16_adds_subs_avx2(&x[5], &x[13]); + btf_16_adds_subs_avx2(&x[6], &x[14]); + btf_16_adds_subs_avx2(&x[7], &x[15]); +} + +static INLINE void iadst16_stage4_avx2(__m256i *x, const int32_t *cospi, + const __m256i _r, int8_t cos_bit) { + const __m256i cospi_p08_p56 = pair_set_w16_epi16(cospi[8], cospi[56]); + const __m256i cospi_p56_m08 = pair_set_w16_epi16(cospi[56], -cospi[8]); + const __m256i cospi_p40_p24 = pair_set_w16_epi16(cospi[40], cospi[24]); + const __m256i cospi_p24_m40 = pair_set_w16_epi16(cospi[24], -cospi[40]); + const __m256i cospi_m56_p08 = pair_set_w16_epi16(-cospi[56], cospi[8]); + const __m256i cospi_m24_p40 = pair_set_w16_epi16(-cospi[24], cospi[40]); + btf_16_w16_avx2(cospi_p08_p56, cospi_p56_m08, &x[8], &x[9], _r, cos_bit); + btf_16_w16_avx2(cospi_p40_p24, cospi_p24_m40, &x[10], &x[11], _r, cos_bit); + btf_16_w16_avx2(cospi_m56_p08, cospi_p08_p56, &x[12], &x[13], _r, cos_bit); + btf_16_w16_avx2(cospi_m24_p40, cospi_p40_p24, &x[14], &x[15], _r, cos_bit); +} + +static INLINE void iadst16_stage5_avx2(__m256i *x) { + btf_16_adds_subs_avx2(&x[0], &x[4]); + btf_16_adds_subs_avx2(&x[1], &x[5]); + btf_16_adds_subs_avx2(&x[2], &x[6]); + btf_16_adds_subs_avx2(&x[3], &x[7]); + btf_16_adds_subs_avx2(&x[8], &x[12]); + btf_16_adds_subs_avx2(&x[9], &x[13]); + btf_16_adds_subs_avx2(&x[10], &x[14]); + btf_16_adds_subs_avx2(&x[11], &x[15]); +} + +static INLINE void iadst16_stage6_avx2(__m256i *x, const int32_t *cospi, + const __m256i _r, int8_t cos_bit) { + const __m256i cospi_p16_p48 = pair_set_w16_epi16(cospi[16], cospi[48]); + const __m256i cospi_p48_m16 = pair_set_w16_epi16(cospi[48], -cospi[16]); + const __m256i cospi_m48_p16 = pair_set_w16_epi16(-cospi[48], cospi[16]); + btf_16_w16_avx2(cospi_p16_p48, cospi_p48_m16, &x[4], &x[5], _r, cos_bit); + btf_16_w16_avx2(cospi_m48_p16, cospi_p16_p48, &x[6], &x[7], _r, cos_bit); + btf_16_w16_avx2(cospi_p16_p48, cospi_p48_m16, &x[12], &x[13], _r, cos_bit); + btf_16_w16_avx2(cospi_m48_p16, cospi_p16_p48, &x[14], &x[15], _r, cos_bit); +} + +static INLINE void iadst16_stage7_avx2(__m256i *x) { + btf_16_adds_subs_avx2(&x[0], &x[2]); + btf_16_adds_subs_avx2(&x[1], &x[3]); + btf_16_adds_subs_avx2(&x[4], &x[6]); + btf_16_adds_subs_avx2(&x[5], &x[7]); + btf_16_adds_subs_avx2(&x[8], &x[10]); + btf_16_adds_subs_avx2(&x[9], &x[11]); + btf_16_adds_subs_avx2(&x[12], &x[14]); + btf_16_adds_subs_avx2(&x[13], &x[15]); +} + +static INLINE void iadst16_stage8_avx2(__m256i *x1, const int32_t *cospi, + const __m256i _r, int8_t cos_bit) { + const __m256i cospi_p32_p32 = pair_set_w16_epi16(cospi[32], cospi[32]); + const __m256i cospi_p32_m32 = pair_set_w16_epi16(cospi[32], -cospi[32]); + btf_16_w16_avx2(cospi_p32_p32, cospi_p32_m32, &x1[2], &x1[3], _r, cos_bit); + btf_16_w16_avx2(cospi_p32_p32, cospi_p32_m32, &x1[6], &x1[7], _r, cos_bit); + btf_16_w16_avx2(cospi_p32_p32, cospi_p32_m32, &x1[10], &x1[11], _r, cos_bit); + btf_16_w16_avx2(cospi_p32_p32, cospi_p32_m32, &x1[14], &x1[15], _r, cos_bit); +} + +static INLINE void iadst16_stage9_avx2(__m256i *output, __m256i *x1) { + const __m256i __zero = _mm256_setzero_si256(); + output[0] = x1[0]; + output[1] = _mm256_subs_epi16(__zero, x1[8]); + output[2] = x1[12]; + output[3] = _mm256_subs_epi16(__zero, x1[4]); + output[4] = x1[6]; + output[5] = _mm256_subs_epi16(__zero, x1[14]); + output[6] = x1[10]; + output[7] = _mm256_subs_epi16(__zero, x1[2]); + output[8] = x1[3]; + output[9] = _mm256_subs_epi16(__zero, x1[11]); + output[10] = x1[15]; + output[11] = _mm256_subs_epi16(__zero, x1[7]); + output[12] = x1[5]; + output[13] = _mm256_subs_epi16(__zero, x1[13]); + output[14] = x1[9]; + output[15] = _mm256_subs_epi16(__zero, x1[1]); +} + +static void iadst16_new_avx2(const __m256i *input, __m256i *output, + int8_t cos_bit) { + (void)(cos_bit); + const int32_t *cospi = cospi_arr(INV_COS_BIT); + + const __m256i _r = _mm256_set1_epi32(1 << (INV_COS_BIT - 1)); + + __m256i cospi_p02_p62 = pair_set_w16_epi16(cospi[2], cospi[62]); + __m256i cospi_p62_m02 = pair_set_w16_epi16(cospi[62], -cospi[2]); + __m256i cospi_p10_p54 = pair_set_w16_epi16(cospi[10], cospi[54]); + __m256i cospi_p54_m10 = pair_set_w16_epi16(cospi[54], -cospi[10]); + __m256i cospi_p18_p46 = pair_set_w16_epi16(cospi[18], cospi[46]); + __m256i cospi_p46_m18 = pair_set_w16_epi16(cospi[46], -cospi[18]); + __m256i cospi_p26_p38 = pair_set_w16_epi16(cospi[26], cospi[38]); + __m256i cospi_p38_m26 = pair_set_w16_epi16(cospi[38], -cospi[26]); + __m256i cospi_p34_p30 = pair_set_w16_epi16(cospi[34], cospi[30]); + __m256i cospi_p30_m34 = pair_set_w16_epi16(cospi[30], -cospi[34]); + __m256i cospi_p42_p22 = pair_set_w16_epi16(cospi[42], cospi[22]); + __m256i cospi_p22_m42 = pair_set_w16_epi16(cospi[22], -cospi[42]); + __m256i cospi_p50_p14 = pair_set_w16_epi16(cospi[50], cospi[14]); + __m256i cospi_p14_m50 = pair_set_w16_epi16(cospi[14], -cospi[50]); + __m256i cospi_p58_p06 = pair_set_w16_epi16(cospi[58], cospi[6]); + __m256i cospi_p06_m58 = pair_set_w16_epi16(cospi[6], -cospi[58]); + + // stage 1 + __m256i x1[16]; + x1[0] = input[15]; + x1[1] = input[0]; + x1[2] = input[13]; + x1[3] = input[2]; + x1[4] = input[11]; + x1[5] = input[4]; + x1[6] = input[9]; + x1[7] = input[6]; + x1[8] = input[7]; + x1[9] = input[8]; + x1[10] = input[5]; + x1[11] = input[10]; + x1[12] = input[3]; + x1[13] = input[12]; + x1[14] = input[1]; + x1[15] = input[14]; + + // stage 2 + btf_16_w16_avx2(cospi_p02_p62, cospi_p62_m02, &x1[0], &x1[1], _r, cos_bit); + btf_16_w16_avx2(cospi_p10_p54, cospi_p54_m10, &x1[2], &x1[3], _r, cos_bit); + btf_16_w16_avx2(cospi_p18_p46, cospi_p46_m18, &x1[4], &x1[5], _r, cos_bit); + btf_16_w16_avx2(cospi_p26_p38, cospi_p38_m26, &x1[6], &x1[7], _r, cos_bit); + btf_16_w16_avx2(cospi_p34_p30, cospi_p30_m34, &x1[8], &x1[9], _r, cos_bit); + btf_16_w16_avx2(cospi_p42_p22, cospi_p22_m42, &x1[10], &x1[11], _r, cos_bit); + btf_16_w16_avx2(cospi_p50_p14, cospi_p14_m50, &x1[12], &x1[13], _r, cos_bit); + btf_16_w16_avx2(cospi_p58_p06, cospi_p06_m58, &x1[14], &x1[15], _r, cos_bit); + + iadst16_stage3_avx2(x1); + iadst16_stage4_avx2(x1, cospi, _r, cos_bit); + iadst16_stage5_avx2(x1); + iadst16_stage6_avx2(x1, cospi, _r, cos_bit); + iadst16_stage7_avx2(x1); + iadst16_stage8_avx2(x1, cospi, _r, cos_bit); + iadst16_stage9_avx2(output, x1); +} + +static void iadst16_low8_new_avx2(const __m256i *input, __m256i *output, + int8_t cos_bit) { + (void)(cos_bit); + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m256i _r = _mm256_set1_epi32(1 << (INV_COS_BIT - 1)); + + // stage 1 + __m256i x1[16]; + x1[1] = input[0]; + x1[3] = input[2]; + x1[5] = input[4]; + x1[7] = input[6]; + x1[8] = input[7]; + x1[10] = input[5]; + x1[12] = input[3]; + x1[14] = input[1]; + + // stage 2 + btf_16_w16_0_avx2(cospi[62], -cospi[2], x1[1], x1[0], x1[1]); + btf_16_w16_0_avx2(cospi[54], -cospi[10], x1[3], x1[2], x1[3]); + btf_16_w16_0_avx2(cospi[46], -cospi[18], x1[5], x1[4], x1[5]); + btf_16_w16_0_avx2(cospi[38], -cospi[26], x1[7], x1[6], x1[7]); + btf_16_w16_0_avx2(cospi[34], cospi[30], x1[8], x1[8], x1[9]); + btf_16_w16_0_avx2(cospi[42], cospi[22], x1[10], x1[10], x1[11]); + btf_16_w16_0_avx2(cospi[50], cospi[14], x1[12], x1[12], x1[13]); + btf_16_w16_0_avx2(cospi[58], cospi[06], x1[14], x1[14], x1[15]); + + iadst16_stage3_avx2(x1); + iadst16_stage4_avx2(x1, cospi, _r, cos_bit); + iadst16_stage5_avx2(x1); + iadst16_stage6_avx2(x1, cospi, _r, cos_bit); + iadst16_stage7_avx2(x1); + iadst16_stage8_avx2(x1, cospi, _r, cos_bit); + iadst16_stage9_avx2(output, x1); +} + +static void iadst16_low1_new_avx2(const __m256i *input, __m256i *output, + int8_t cos_bit) { + (void)(cos_bit); + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m256i _r = _mm256_set1_epi32(1 << (INV_COS_BIT - 1)); + + const __m256i cospi_p08_p56 = pair_set_w16_epi16(cospi[8], cospi[56]); + const __m256i cospi_p56_m08 = pair_set_w16_epi16(cospi[56], -cospi[8]); + const __m256i cospi_p16_p48 = pair_set_w16_epi16(cospi[16], cospi[48]); + const __m256i cospi_p48_m16 = pair_set_w16_epi16(cospi[48], -cospi[16]); + + // stage 1 + __m256i x1[16]; + x1[1] = input[0]; + + // stage 2 + btf_16_w16_0_avx2(cospi[62], -cospi[2], x1[1], x1[0], x1[1]); + + // stage 3 + x1[8] = x1[0]; + x1[9] = x1[1]; + + // stage 4 + btf_16_w16_avx2(cospi_p08_p56, cospi_p56_m08, &x1[8], &x1[9], _r, cos_bit); + + // stage 5 + x1[4] = x1[0]; + x1[5] = x1[1]; + + x1[12] = x1[8]; + x1[13] = x1[9]; + + // stage 6 + btf_16_w16_avx2(cospi_p16_p48, cospi_p48_m16, &x1[4], &x1[5], _r, cos_bit); + btf_16_w16_avx2(cospi_p16_p48, cospi_p48_m16, &x1[12], &x1[13], _r, cos_bit); + + // stage 7 + x1[2] = x1[0]; + x1[3] = x1[1]; + x1[6] = x1[4]; + x1[7] = x1[5]; + x1[10] = x1[8]; + x1[11] = x1[9]; + x1[14] = x1[12]; + x1[15] = x1[13]; + + iadst16_stage8_avx2(x1, cospi, _r, cos_bit); + iadst16_stage9_avx2(output, x1); +} + +static INLINE void idct32_high16_stage3_avx2(__m256i *x) { + btf_16_adds_subs_avx2(&x[16], &x[17]); + btf_16_adds_subs_avx2(&x[19], &x[18]); + btf_16_adds_subs_avx2(&x[20], &x[21]); + btf_16_adds_subs_avx2(&x[23], &x[22]); + btf_16_adds_subs_avx2(&x[24], &x[25]); + btf_16_adds_subs_avx2(&x[27], &x[26]); + btf_16_adds_subs_avx2(&x[28], &x[29]); + btf_16_adds_subs_avx2(&x[31], &x[30]); +} + +static INLINE void idct32_high16_stage4_avx2(__m256i *x, const int32_t *cospi, + const __m256i _r, int8_t cos_bit) { + const __m256i cospi_m08_p56 = pair_set_w16_epi16(-cospi[8], cospi[56]); + const __m256i cospi_p56_p08 = pair_set_w16_epi16(cospi[56], cospi[8]); + const __m256i cospi_m56_m08 = pair_set_w16_epi16(-cospi[56], -cospi[8]); + const __m256i cospi_m40_p24 = pair_set_w16_epi16(-cospi[40], cospi[24]); + const __m256i cospi_p24_p40 = pair_set_w16_epi16(cospi[24], cospi[40]); + const __m256i cospi_m24_m40 = pair_set_w16_epi16(-cospi[24], -cospi[40]); + btf_16_w16_avx2(cospi_m08_p56, cospi_p56_p08, &x[17], &x[30], _r, cos_bit); + btf_16_w16_avx2(cospi_m56_m08, cospi_m08_p56, &x[18], &x[29], _r, cos_bit); + btf_16_w16_avx2(cospi_m40_p24, cospi_p24_p40, &x[21], &x[26], _r, cos_bit); + btf_16_w16_avx2(cospi_m24_m40, cospi_m40_p24, &x[22], &x[25], _r, cos_bit); +} + +static INLINE void idct32_high24_stage5_avx2(__m256i *x, const int32_t *cospi, + const __m256i _r, int8_t cos_bit) { + const __m256i cospi_m16_p48 = pair_set_w16_epi16(-cospi[16], cospi[48]); + const __m256i cospi_p48_p16 = pair_set_w16_epi16(cospi[48], cospi[16]); + const __m256i cospi_m48_m16 = pair_set_w16_epi16(-cospi[48], -cospi[16]); + btf_16_w16_avx2(cospi_m16_p48, cospi_p48_p16, &x[9], &x[14], _r, cos_bit); + btf_16_w16_avx2(cospi_m48_m16, cospi_m16_p48, &x[10], &x[13], _r, cos_bit); + btf_16_adds_subs_avx2(&x[16], &x[19]); + btf_16_adds_subs_avx2(&x[17], &x[18]); + btf_16_adds_subs_avx2(&x[23], &x[20]); + btf_16_adds_subs_avx2(&x[22], &x[21]); + btf_16_adds_subs_avx2(&x[24], &x[27]); + btf_16_adds_subs_avx2(&x[25], &x[26]); + btf_16_adds_subs_avx2(&x[31], &x[28]); + btf_16_adds_subs_avx2(&x[30], &x[29]); +} + +static INLINE void idct32_high28_stage6_avx2(__m256i *x, const int32_t *cospi, + const __m256i _r, int8_t cos_bit) { + const __m256i cospi_m32_p32 = pair_set_w16_epi16(-cospi[32], cospi[32]); + const __m256i cospi_p32_p32 = pair_set_w16_epi16(cospi[32], cospi[32]); + const __m256i cospi_m16_p48 = pair_set_w16_epi16(-cospi[16], cospi[48]); + const __m256i cospi_p48_p16 = pair_set_w16_epi16(cospi[48], cospi[16]); + const __m256i cospi_m48_m16 = pair_set_w16_epi16(-cospi[48], -cospi[16]); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[5], &x[6], _r, cos_bit); + btf_16_adds_subs_avx2(&x[8], &x[11]); + btf_16_adds_subs_avx2(&x[9], &x[10]); + btf_16_adds_subs_avx2(&x[15], &x[12]); + btf_16_adds_subs_avx2(&x[14], &x[13]); + btf_16_w16_avx2(cospi_m16_p48, cospi_p48_p16, &x[18], &x[29], _r, cos_bit); + btf_16_w16_avx2(cospi_m16_p48, cospi_p48_p16, &x[19], &x[28], _r, cos_bit); + btf_16_w16_avx2(cospi_m48_m16, cospi_m16_p48, &x[20], &x[27], _r, cos_bit); + btf_16_w16_avx2(cospi_m48_m16, cospi_m16_p48, &x[21], &x[26], _r, cos_bit); +} + +static INLINE void idct32_stage7_avx2(__m256i *x, const int32_t *cospi, + const __m256i _r, int8_t cos_bit) { + const __m256i cospi_m32_p32 = pair_set_w16_epi16(-cospi[32], cospi[32]); + const __m256i cospi_p32_p32 = pair_set_w16_epi16(cospi[32], cospi[32]); + btf_16_adds_subs_avx2(&x[0], &x[7]); + btf_16_adds_subs_avx2(&x[1], &x[6]); + btf_16_adds_subs_avx2(&x[2], &x[5]); + btf_16_adds_subs_avx2(&x[3], &x[4]); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[10], &x[13], _r, cos_bit); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[11], &x[12], _r, cos_bit); + btf_16_adds_subs_avx2(&x[16], &x[23]); + btf_16_adds_subs_avx2(&x[17], &x[22]); + btf_16_adds_subs_avx2(&x[18], &x[21]); + btf_16_adds_subs_avx2(&x[19], &x[20]); + btf_16_adds_subs_avx2(&x[31], &x[24]); + btf_16_adds_subs_avx2(&x[30], &x[25]); + btf_16_adds_subs_avx2(&x[29], &x[26]); + btf_16_adds_subs_avx2(&x[28], &x[27]); +} + +static INLINE void idct32_stage8_avx2(__m256i *x, const int32_t *cospi, + const __m256i _r, int8_t cos_bit) { + const __m256i cospi_m32_p32 = pair_set_w16_epi16(-cospi[32], cospi[32]); + const __m256i cospi_p32_p32 = pair_set_w16_epi16(cospi[32], cospi[32]); + btf_16_adds_subs_avx2(&x[0], &x[15]); + btf_16_adds_subs_avx2(&x[1], &x[14]); + btf_16_adds_subs_avx2(&x[2], &x[13]); + btf_16_adds_subs_avx2(&x[3], &x[12]); + btf_16_adds_subs_avx2(&x[4], &x[11]); + btf_16_adds_subs_avx2(&x[5], &x[10]); + btf_16_adds_subs_avx2(&x[6], &x[9]); + btf_16_adds_subs_avx2(&x[7], &x[8]); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[20], &x[27], _r, cos_bit); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[21], &x[26], _r, cos_bit); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[22], &x[25], _r, cos_bit); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[23], &x[24], _r, cos_bit); +} + +static INLINE void idct32_stage9_avx2(__m256i *output, __m256i *x) { + btf_16_adds_subs_out_avx2(&output[0], &output[31], x[0], x[31]); + btf_16_adds_subs_out_avx2(&output[1], &output[30], x[1], x[30]); + btf_16_adds_subs_out_avx2(&output[2], &output[29], x[2], x[29]); + btf_16_adds_subs_out_avx2(&output[3], &output[28], x[3], x[28]); + btf_16_adds_subs_out_avx2(&output[4], &output[27], x[4], x[27]); + btf_16_adds_subs_out_avx2(&output[5], &output[26], x[5], x[26]); + btf_16_adds_subs_out_avx2(&output[6], &output[25], x[6], x[25]); + btf_16_adds_subs_out_avx2(&output[7], &output[24], x[7], x[24]); + btf_16_adds_subs_out_avx2(&output[8], &output[23], x[8], x[23]); + btf_16_adds_subs_out_avx2(&output[9], &output[22], x[9], x[22]); + btf_16_adds_subs_out_avx2(&output[10], &output[21], x[10], x[21]); + btf_16_adds_subs_out_avx2(&output[11], &output[20], x[11], x[20]); + btf_16_adds_subs_out_avx2(&output[12], &output[19], x[12], x[19]); + btf_16_adds_subs_out_avx2(&output[13], &output[18], x[13], x[18]); + btf_16_adds_subs_out_avx2(&output[14], &output[17], x[14], x[17]); + btf_16_adds_subs_out_avx2(&output[15], &output[16], x[15], x[16]); +} + +static void idct32_low1_new_avx2(const __m256i *input, __m256i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + + // stage 1 + __m256i x[2]; + x[0] = input[0]; + + // stage 2 + // stage 3 + // stage 4 + // stage 5 + btf_16_w16_0_avx2(cospi[32], cospi[32], x[0], x[0], x[1]); + + // stage 6 + // stage 7 + // stage 8 + // stage 9 + output[0] = x[0]; + output[31] = x[0]; + output[1] = x[1]; + output[30] = x[1]; + output[2] = x[1]; + output[29] = x[1]; + output[3] = x[0]; + output[28] = x[0]; + output[4] = x[0]; + output[27] = x[0]; + output[5] = x[1]; + output[26] = x[1]; + output[6] = x[1]; + output[25] = x[1]; + output[7] = x[0]; + output[24] = x[0]; + output[8] = x[0]; + output[23] = x[0]; + output[9] = x[1]; + output[22] = x[1]; + output[10] = x[1]; + output[21] = x[1]; + output[11] = x[0]; + output[20] = x[0]; + output[12] = x[0]; + output[19] = x[0]; + output[13] = x[1]; + output[18] = x[1]; + output[14] = x[1]; + output[17] = x[1]; + output[15] = x[0]; + output[16] = x[0]; +} + +static void idct32_low8_new_avx2(const __m256i *input, __m256i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m256i _r = _mm256_set1_epi32(1 << (INV_COS_BIT - 1)); + + // stage 1 + __m256i x[32]; + x[0] = input[0]; + x[4] = input[4]; + x[8] = input[2]; + x[12] = input[6]; + x[16] = input[1]; + x[20] = input[5]; + x[24] = input[3]; + x[28] = input[7]; + + // stage 2 + btf_16_w16_0_avx2(cospi[62], cospi[2], x[16], x[16], x[31]); + btf_16_w16_0_avx2(-cospi[50], cospi[14], x[28], x[19], x[28]); + btf_16_w16_0_avx2(cospi[54], cospi[10], x[20], x[20], x[27]); + btf_16_w16_0_avx2(-cospi[58], cospi[6], x[24], x[23], x[24]); + + // stage 3 + btf_16_w16_0_avx2(cospi[60], cospi[4], x[8], x[8], x[15]); + btf_16_w16_0_avx2(-cospi[52], cospi[12], x[12], x[11], x[12]); + x[17] = x[16]; + x[18] = x[19]; + x[21] = x[20]; + x[22] = x[23]; + x[25] = x[24]; + x[26] = x[27]; + x[29] = x[28]; + x[30] = x[31]; + + // stage 4 + btf_16_w16_0_avx2(cospi[56], cospi[8], x[4], x[4], x[7]); + x[9] = x[8]; + x[10] = x[11]; + x[13] = x[12]; + x[14] = x[15]; + idct32_high16_stage4_avx2(x, cospi, _r, cos_bit); + + // stage 5 + btf_16_w16_0_avx2(cospi[32], cospi[32], x[0], x[0], x[1]); + x[5] = x[4]; + x[6] = x[7]; + idct32_high24_stage5_avx2(x, cospi, _r, cos_bit); + // stage 6 + x[3] = x[0]; + x[2] = x[1]; + idct32_high28_stage6_avx2(x, cospi, _r, cos_bit); + + idct32_stage7_avx2(x, cospi, _r, cos_bit); + idct32_stage8_avx2(x, cospi, _r, cos_bit); + idct32_stage9_avx2(output, x); +} + +static void idct32_low16_new_avx2(const __m256i *input, __m256i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m256i _r = _mm256_set1_epi32(1 << (INV_COS_BIT - 1)); + + // stage 1 + __m256i x[32]; + x[0] = input[0]; + x[2] = input[8]; + x[4] = input[4]; + x[6] = input[12]; + x[8] = input[2]; + x[10] = input[10]; + x[12] = input[6]; + x[14] = input[14]; + x[16] = input[1]; + x[18] = input[9]; + x[20] = input[5]; + x[22] = input[13]; + x[24] = input[3]; + x[26] = input[11]; + x[28] = input[7]; + x[30] = input[15]; + + // stage 2 + btf_16_w16_0_avx2(cospi[62], cospi[2], x[16], x[16], x[31]); + btf_16_w16_0_avx2(-cospi[34], cospi[30], x[30], x[17], x[30]); + btf_16_w16_0_avx2(cospi[46], cospi[18], x[18], x[18], x[29]); + btf_16_w16_0_avx2(-cospi[50], cospi[14], x[28], x[19], x[28]); + btf_16_w16_0_avx2(cospi[54], cospi[10], x[20], x[20], x[27]); + btf_16_w16_0_avx2(-cospi[42], cospi[22], x[26], x[21], x[26]); + btf_16_w16_0_avx2(cospi[38], cospi[26], x[22], x[22], x[25]); + btf_16_w16_0_avx2(-cospi[58], cospi[6], x[24], x[23], x[24]); + + // stage 3 + btf_16_w16_0_avx2(cospi[60], cospi[4], x[8], x[8], x[15]); + btf_16_w16_0_avx2(-cospi[36], cospi[28], x[14], x[9], x[14]); + btf_16_w16_0_avx2(cospi[44], cospi[20], x[10], x[10], x[13]); + btf_16_w16_0_avx2(-cospi[52], cospi[12], x[12], x[11], x[12]); + idct32_high16_stage3_avx2(x); + + // stage 4 + btf_16_w16_0_avx2(cospi[56], cospi[8], x[4], x[4], x[7]); + btf_16_w16_0_avx2(-cospi[40], cospi[24], x[6], x[5], x[6]); + btf_16_adds_subs_avx2(&x[8], &x[9]); + btf_16_adds_subs_avx2(&x[11], &x[10]); + btf_16_adds_subs_avx2(&x[12], &x[13]); + btf_16_adds_subs_avx2(&x[15], &x[14]); + idct32_high16_stage4_avx2(x, cospi, _r, cos_bit); + + // stage 5 + btf_16_w16_0_avx2(cospi[32], cospi[32], x[0], x[0], x[1]); + btf_16_w16_0_avx2(cospi[48], cospi[16], x[2], x[2], x[3]); + btf_16_adds_subs_avx2(&x[4], &x[5]); + btf_16_adds_subs_avx2(&x[7], &x[6]); + idct32_high24_stage5_avx2(x, cospi, _r, cos_bit); + + btf_16_adds_subs_avx2(&x[0], &x[3]); + btf_16_adds_subs_avx2(&x[1], &x[2]); + idct32_high28_stage6_avx2(x, cospi, _r, cos_bit); + + idct32_stage7_avx2(x, cospi, _r, cos_bit); + idct32_stage8_avx2(x, cospi, _r, cos_bit); + idct32_stage9_avx2(output, x); +} + +static void idct32_new_avx2(const __m256i *input, __m256i *output, + int8_t cos_bit) { + (void)(cos_bit); + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m256i _r = _mm256_set1_epi32(1 << (INV_COS_BIT - 1)); + + __m256i cospi_p62_m02 = pair_set_w16_epi16(cospi[62], -cospi[2]); + __m256i cospi_p02_p62 = pair_set_w16_epi16(cospi[2], cospi[62]); + __m256i cospi_p30_m34 = pair_set_w16_epi16(cospi[30], -cospi[34]); + __m256i cospi_p34_p30 = pair_set_w16_epi16(cospi[34], cospi[30]); + __m256i cospi_p46_m18 = pair_set_w16_epi16(cospi[46], -cospi[18]); + __m256i cospi_p18_p46 = pair_set_w16_epi16(cospi[18], cospi[46]); + __m256i cospi_p14_m50 = pair_set_w16_epi16(cospi[14], -cospi[50]); + __m256i cospi_p50_p14 = pair_set_w16_epi16(cospi[50], cospi[14]); + __m256i cospi_p54_m10 = pair_set_w16_epi16(cospi[54], -cospi[10]); + __m256i cospi_p10_p54 = pair_set_w16_epi16(cospi[10], cospi[54]); + __m256i cospi_p22_m42 = pair_set_w16_epi16(cospi[22], -cospi[42]); + __m256i cospi_p42_p22 = pair_set_w16_epi16(cospi[42], cospi[22]); + __m256i cospi_p38_m26 = pair_set_w16_epi16(cospi[38], -cospi[26]); + __m256i cospi_p26_p38 = pair_set_w16_epi16(cospi[26], cospi[38]); + __m256i cospi_p06_m58 = pair_set_w16_epi16(cospi[6], -cospi[58]); + __m256i cospi_p58_p06 = pair_set_w16_epi16(cospi[58], cospi[6]); + __m256i cospi_p60_m04 = pair_set_w16_epi16(cospi[60], -cospi[4]); + __m256i cospi_p04_p60 = pair_set_w16_epi16(cospi[4], cospi[60]); + __m256i cospi_p28_m36 = pair_set_w16_epi16(cospi[28], -cospi[36]); + __m256i cospi_p36_p28 = pair_set_w16_epi16(cospi[36], cospi[28]); + __m256i cospi_p44_m20 = pair_set_w16_epi16(cospi[44], -cospi[20]); + __m256i cospi_p20_p44 = pair_set_w16_epi16(cospi[20], cospi[44]); + __m256i cospi_p12_m52 = pair_set_w16_epi16(cospi[12], -cospi[52]); + __m256i cospi_p52_p12 = pair_set_w16_epi16(cospi[52], cospi[12]); + __m256i cospi_p56_m08 = pair_set_w16_epi16(cospi[56], -cospi[8]); + __m256i cospi_p08_p56 = pair_set_w16_epi16(cospi[8], cospi[56]); + __m256i cospi_p24_m40 = pair_set_w16_epi16(cospi[24], -cospi[40]); + __m256i cospi_p40_p24 = pair_set_w16_epi16(cospi[40], cospi[24]); + __m256i cospi_p32_p32 = pair_set_w16_epi16(cospi[32], cospi[32]); + __m256i cospi_p32_m32 = pair_set_w16_epi16(cospi[32], -cospi[32]); + __m256i cospi_p48_m16 = pair_set_w16_epi16(cospi[48], -cospi[16]); + __m256i cospi_p16_p48 = pair_set_w16_epi16(cospi[16], cospi[48]); + + // stage 1 + __m256i x1[32]; + x1[0] = input[0]; + x1[1] = input[16]; + x1[2] = input[8]; + x1[3] = input[24]; + x1[4] = input[4]; + x1[5] = input[20]; + x1[6] = input[12]; + x1[7] = input[28]; + x1[8] = input[2]; + x1[9] = input[18]; + x1[10] = input[10]; + x1[11] = input[26]; + x1[12] = input[6]; + x1[13] = input[22]; + x1[14] = input[14]; + x1[15] = input[30]; + x1[16] = input[1]; + x1[17] = input[17]; + x1[18] = input[9]; + x1[19] = input[25]; + x1[20] = input[5]; + x1[21] = input[21]; + x1[22] = input[13]; + x1[23] = input[29]; + x1[24] = input[3]; + x1[25] = input[19]; + x1[26] = input[11]; + x1[27] = input[27]; + x1[28] = input[7]; + x1[29] = input[23]; + x1[30] = input[15]; + x1[31] = input[31]; + + // stage 2 + btf_16_w16_avx2(cospi_p62_m02, cospi_p02_p62, &x1[16], &x1[31], _r, cos_bit); + btf_16_w16_avx2(cospi_p30_m34, cospi_p34_p30, &x1[17], &x1[30], _r, cos_bit); + btf_16_w16_avx2(cospi_p46_m18, cospi_p18_p46, &x1[18], &x1[29], _r, cos_bit); + btf_16_w16_avx2(cospi_p14_m50, cospi_p50_p14, &x1[19], &x1[28], _r, cos_bit); + btf_16_w16_avx2(cospi_p54_m10, cospi_p10_p54, &x1[20], &x1[27], _r, cos_bit); + btf_16_w16_avx2(cospi_p22_m42, cospi_p42_p22, &x1[21], &x1[26], _r, cos_bit); + btf_16_w16_avx2(cospi_p38_m26, cospi_p26_p38, &x1[22], &x1[25], _r, cos_bit); + btf_16_w16_avx2(cospi_p06_m58, cospi_p58_p06, &x1[23], &x1[24], _r, cos_bit); + + // stage 3 + btf_16_w16_avx2(cospi_p60_m04, cospi_p04_p60, &x1[8], &x1[15], _r, cos_bit); + btf_16_w16_avx2(cospi_p28_m36, cospi_p36_p28, &x1[9], &x1[14], _r, cos_bit); + btf_16_w16_avx2(cospi_p44_m20, cospi_p20_p44, &x1[10], &x1[13], _r, cos_bit); + btf_16_w16_avx2(cospi_p12_m52, cospi_p52_p12, &x1[11], &x1[12], _r, cos_bit); + idct32_high16_stage3_avx2(x1); + + // stage 4 + btf_16_w16_avx2(cospi_p56_m08, cospi_p08_p56, &x1[4], &x1[7], _r, cos_bit); + btf_16_w16_avx2(cospi_p24_m40, cospi_p40_p24, &x1[5], &x1[6], _r, cos_bit); + btf_16_adds_subs_avx2(&x1[8], &x1[9]); + btf_16_adds_subs_avx2(&x1[11], &x1[10]); + btf_16_adds_subs_avx2(&x1[12], &x1[13]); + btf_16_adds_subs_avx2(&x1[15], &x1[14]); + idct32_high16_stage4_avx2(x1, cospi, _r, cos_bit); + + // stage 5 + btf_16_w16_avx2(cospi_p32_p32, cospi_p32_m32, &x1[0], &x1[1], _r, cos_bit); + btf_16_w16_avx2(cospi_p48_m16, cospi_p16_p48, &x1[2], &x1[3], _r, cos_bit); + btf_16_adds_subs_avx2(&x1[4], &x1[5]); + btf_16_adds_subs_avx2(&x1[7], &x1[6]); + idct32_high24_stage5_avx2(x1, cospi, _r, cos_bit); + + // stage 6 + btf_16_adds_subs_avx2(&x1[0], &x1[3]); + btf_16_adds_subs_avx2(&x1[1], &x1[2]); + idct32_high28_stage6_avx2(x1, cospi, _r, cos_bit); + + idct32_stage7_avx2(x1, cospi, _r, cos_bit); + idct32_stage8_avx2(x1, cospi, _r, cos_bit); + idct32_stage9_avx2(output, x1); +} + +static INLINE void idct64_stage4_high32_avx2(__m256i *x, const int32_t *cospi, + const __m256i _r, int8_t cos_bit) { + (void)cos_bit; + const __m256i cospi_m04_p60 = pair_set_w16_epi16(-cospi[4], cospi[60]); + const __m256i cospi_p60_p04 = pair_set_w16_epi16(cospi[60], cospi[4]); + const __m256i cospi_m60_m04 = pair_set_w16_epi16(-cospi[60], -cospi[4]); + const __m256i cospi_m36_p28 = pair_set_w16_epi16(-cospi[36], cospi[28]); + const __m256i cospi_p28_p36 = pair_set_w16_epi16(cospi[28], cospi[36]); + const __m256i cospi_m28_m36 = pair_set_w16_epi16(-cospi[28], -cospi[36]); + const __m256i cospi_m20_p44 = pair_set_w16_epi16(-cospi[20], cospi[44]); + const __m256i cospi_p44_p20 = pair_set_w16_epi16(cospi[44], cospi[20]); + const __m256i cospi_m44_m20 = pair_set_w16_epi16(-cospi[44], -cospi[20]); + const __m256i cospi_m52_p12 = pair_set_w16_epi16(-cospi[52], cospi[12]); + const __m256i cospi_p12_p52 = pair_set_w16_epi16(cospi[12], cospi[52]); + const __m256i cospi_m12_m52 = pair_set_w16_epi16(-cospi[12], -cospi[52]); + btf_16_w16_avx2(cospi_m04_p60, cospi_p60_p04, &x[33], &x[62], _r, cos_bit); + btf_16_w16_avx2(cospi_m60_m04, cospi_m04_p60, &x[34], &x[61], _r, cos_bit); + btf_16_w16_avx2(cospi_m36_p28, cospi_p28_p36, &x[37], &x[58], _r, cos_bit); + btf_16_w16_avx2(cospi_m28_m36, cospi_m36_p28, &x[38], &x[57], _r, cos_bit); + btf_16_w16_avx2(cospi_m20_p44, cospi_p44_p20, &x[41], &x[54], _r, cos_bit); + btf_16_w16_avx2(cospi_m44_m20, cospi_m20_p44, &x[42], &x[53], _r, cos_bit); + btf_16_w16_avx2(cospi_m52_p12, cospi_p12_p52, &x[45], &x[50], _r, cos_bit); + btf_16_w16_avx2(cospi_m12_m52, cospi_m52_p12, &x[46], &x[49], _r, cos_bit); +} + +static INLINE void idct64_stage5_high48_avx2(__m256i *x, const int32_t *cospi, + const __m256i _r, int8_t cos_bit) { + (void)cos_bit; + const __m256i cospi_m08_p56 = pair_set_w16_epi16(-cospi[8], cospi[56]); + const __m256i cospi_p56_p08 = pair_set_w16_epi16(cospi[56], cospi[8]); + const __m256i cospi_m56_m08 = pair_set_w16_epi16(-cospi[56], -cospi[8]); + const __m256i cospi_m40_p24 = pair_set_w16_epi16(-cospi[40], cospi[24]); + const __m256i cospi_p24_p40 = pair_set_w16_epi16(cospi[24], cospi[40]); + const __m256i cospi_m24_m40 = pair_set_w16_epi16(-cospi[24], -cospi[40]); + btf_16_w16_avx2(cospi_m08_p56, cospi_p56_p08, &x[17], &x[30], _r, cos_bit); + btf_16_w16_avx2(cospi_m56_m08, cospi_m08_p56, &x[18], &x[29], _r, cos_bit); + btf_16_w16_avx2(cospi_m40_p24, cospi_p24_p40, &x[21], &x[26], _r, cos_bit); + btf_16_w16_avx2(cospi_m24_m40, cospi_m40_p24, &x[22], &x[25], _r, cos_bit); + btf_16_adds_subs_avx2(&x[32], &x[35]); + btf_16_adds_subs_avx2(&x[33], &x[34]); + btf_16_adds_subs_avx2(&x[39], &x[36]); + btf_16_adds_subs_avx2(&x[38], &x[37]); + btf_16_adds_subs_avx2(&x[40], &x[43]); + btf_16_adds_subs_avx2(&x[41], &x[42]); + btf_16_adds_subs_avx2(&x[47], &x[44]); + btf_16_adds_subs_avx2(&x[46], &x[45]); + btf_16_adds_subs_avx2(&x[48], &x[51]); + btf_16_adds_subs_avx2(&x[49], &x[50]); + btf_16_adds_subs_avx2(&x[55], &x[52]); + btf_16_adds_subs_avx2(&x[54], &x[53]); + btf_16_adds_subs_avx2(&x[56], &x[59]); + btf_16_adds_subs_avx2(&x[57], &x[58]); + btf_16_adds_subs_avx2(&x[63], &x[60]); + btf_16_adds_subs_avx2(&x[62], &x[61]); +} + +static INLINE void idct64_stage6_high32_avx2(__m256i *x, const int32_t *cospi, + const __m256i _r, int8_t cos_bit) { + (void)cos_bit; + const __m256i cospi_m08_p56 = pair_set_w16_epi16(-cospi[8], cospi[56]); + const __m256i cospi_p56_p08 = pair_set_w16_epi16(cospi[56], cospi[8]); + const __m256i cospi_m56_m08 = pair_set_w16_epi16(-cospi[56], -cospi[8]); + const __m256i cospi_m40_p24 = pair_set_w16_epi16(-cospi[40], cospi[24]); + const __m256i cospi_p24_p40 = pair_set_w16_epi16(cospi[24], cospi[40]); + const __m256i cospi_m24_m40 = pair_set_w16_epi16(-cospi[24], -cospi[40]); + btf_16_w16_avx2(cospi_m08_p56, cospi_p56_p08, &x[34], &x[61], _r, cos_bit); + btf_16_w16_avx2(cospi_m08_p56, cospi_p56_p08, &x[35], &x[60], _r, cos_bit); + btf_16_w16_avx2(cospi_m56_m08, cospi_m08_p56, &x[36], &x[59], _r, cos_bit); + btf_16_w16_avx2(cospi_m56_m08, cospi_m08_p56, &x[37], &x[58], _r, cos_bit); + btf_16_w16_avx2(cospi_m40_p24, cospi_p24_p40, &x[42], &x[53], _r, cos_bit); + btf_16_w16_avx2(cospi_m40_p24, cospi_p24_p40, &x[43], &x[52], _r, cos_bit); + btf_16_w16_avx2(cospi_m24_m40, cospi_m40_p24, &x[44], &x[51], _r, cos_bit); + btf_16_w16_avx2(cospi_m24_m40, cospi_m40_p24, &x[45], &x[50], _r, cos_bit); +} + +static INLINE void idct64_stage6_high48_avx2(__m256i *x, const int32_t *cospi, + const __m256i _r, int8_t cos_bit) { + btf_16_adds_subs_avx2(&x[16], &x[19]); + btf_16_adds_subs_avx2(&x[17], &x[18]); + btf_16_adds_subs_avx2(&x[23], &x[20]); + btf_16_adds_subs_avx2(&x[22], &x[21]); + btf_16_adds_subs_avx2(&x[24], &x[27]); + btf_16_adds_subs_avx2(&x[25], &x[26]); + btf_16_adds_subs_avx2(&x[31], &x[28]); + btf_16_adds_subs_avx2(&x[30], &x[29]); + idct64_stage6_high32_avx2(x, cospi, _r, cos_bit); +} + +static INLINE void idct64_stage7_high48_avx2(__m256i *x, const int32_t *cospi, + const __m256i _r, int8_t cos_bit) { + (void)cos_bit; + const __m256i cospi_m16_p48 = pair_set_w16_epi16(-cospi[16], cospi[48]); + const __m256i cospi_p48_p16 = pair_set_w16_epi16(cospi[48], cospi[16]); + const __m256i cospi_m48_m16 = pair_set_w16_epi16(-cospi[48], -cospi[16]); + btf_16_w16_avx2(cospi_m16_p48, cospi_p48_p16, &x[18], &x[29], _r, cos_bit); + btf_16_w16_avx2(cospi_m16_p48, cospi_p48_p16, &x[19], &x[28], _r, cos_bit); + btf_16_w16_avx2(cospi_m48_m16, cospi_m16_p48, &x[20], &x[27], _r, cos_bit); + btf_16_w16_avx2(cospi_m48_m16, cospi_m16_p48, &x[21], &x[26], _r, cos_bit); + btf_16_adds_subs_avx2(&x[32], &x[39]); + btf_16_adds_subs_avx2(&x[33], &x[38]); + btf_16_adds_subs_avx2(&x[34], &x[37]); + btf_16_adds_subs_avx2(&x[35], &x[36]); + btf_16_adds_subs_avx2(&x[47], &x[40]); + btf_16_adds_subs_avx2(&x[46], &x[41]); + btf_16_adds_subs_avx2(&x[45], &x[42]); + btf_16_adds_subs_avx2(&x[44], &x[43]); + btf_16_adds_subs_avx2(&x[48], &x[55]); + btf_16_adds_subs_avx2(&x[49], &x[54]); + btf_16_adds_subs_avx2(&x[50], &x[53]); + btf_16_adds_subs_avx2(&x[51], &x[52]); + btf_16_adds_subs_avx2(&x[63], &x[56]); + btf_16_adds_subs_avx2(&x[62], &x[57]); + btf_16_adds_subs_avx2(&x[61], &x[58]); + btf_16_adds_subs_avx2(&x[60], &x[59]); +} + +static INLINE void idct64_stage8_high48_avx2(__m256i *x, const int32_t *cospi, + const __m256i _r, int8_t cos_bit) { + (void)cos_bit; + const __m256i cospi_m16_p48 = pair_set_w16_epi16(-cospi[16], cospi[48]); + const __m256i cospi_p48_p16 = pair_set_w16_epi16(cospi[48], cospi[16]); + const __m256i cospi_m48_m16 = pair_set_w16_epi16(-cospi[48], -cospi[16]); + btf_16_adds_subs_avx2(&x[16], &x[23]); + btf_16_adds_subs_avx2(&x[17], &x[22]); + btf_16_adds_subs_avx2(&x[18], &x[21]); + btf_16_adds_subs_avx2(&x[19], &x[20]); + btf_16_adds_subs_avx2(&x[31], &x[24]); + btf_16_adds_subs_avx2(&x[30], &x[25]); + btf_16_adds_subs_avx2(&x[29], &x[26]); + btf_16_adds_subs_avx2(&x[28], &x[27]); + btf_16_w16_avx2(cospi_m16_p48, cospi_p48_p16, &x[36], &x[59], _r, cos_bit); + btf_16_w16_avx2(cospi_m16_p48, cospi_p48_p16, &x[37], &x[58], _r, cos_bit); + btf_16_w16_avx2(cospi_m16_p48, cospi_p48_p16, &x[38], &x[57], _r, cos_bit); + btf_16_w16_avx2(cospi_m16_p48, cospi_p48_p16, &x[39], &x[56], _r, cos_bit); + btf_16_w16_avx2(cospi_m48_m16, cospi_m16_p48, &x[40], &x[55], _r, cos_bit); + btf_16_w16_avx2(cospi_m48_m16, cospi_m16_p48, &x[41], &x[54], _r, cos_bit); + btf_16_w16_avx2(cospi_m48_m16, cospi_m16_p48, &x[42], &x[53], _r, cos_bit); + btf_16_w16_avx2(cospi_m48_m16, cospi_m16_p48, &x[43], &x[52], _r, cos_bit); +} + +static INLINE void idct64_stage9_avx2(__m256i *x, const int32_t *cospi, + const __m256i _r, int8_t cos_bit) { + (void)cos_bit; + const __m256i cospi_m32_p32 = pair_set_w16_epi16(-cospi[32], cospi[32]); + const __m256i cospi_p32_p32 = pair_set_w16_epi16(cospi[32], cospi[32]); + btf_16_adds_subs_avx2(&x[0], &x[15]); + btf_16_adds_subs_avx2(&x[1], &x[14]); + btf_16_adds_subs_avx2(&x[2], &x[13]); + btf_16_adds_subs_avx2(&x[3], &x[12]); + btf_16_adds_subs_avx2(&x[4], &x[11]); + btf_16_adds_subs_avx2(&x[5], &x[10]); + btf_16_adds_subs_avx2(&x[6], &x[9]); + btf_16_adds_subs_avx2(&x[7], &x[8]); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[20], &x[27], _r, cos_bit); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[21], &x[26], _r, cos_bit); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[22], &x[25], _r, cos_bit); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[23], &x[24], _r, cos_bit); + btf_16_adds_subs_avx2(&x[32], &x[47]); + btf_16_adds_subs_avx2(&x[33], &x[46]); + btf_16_adds_subs_avx2(&x[34], &x[45]); + btf_16_adds_subs_avx2(&x[35], &x[44]); + btf_16_adds_subs_avx2(&x[36], &x[43]); + btf_16_adds_subs_avx2(&x[37], &x[42]); + btf_16_adds_subs_avx2(&x[38], &x[41]); + btf_16_adds_subs_avx2(&x[39], &x[40]); + btf_16_adds_subs_avx2(&x[63], &x[48]); + btf_16_adds_subs_avx2(&x[62], &x[49]); + btf_16_adds_subs_avx2(&x[61], &x[50]); + btf_16_adds_subs_avx2(&x[60], &x[51]); + btf_16_adds_subs_avx2(&x[59], &x[52]); + btf_16_adds_subs_avx2(&x[58], &x[53]); + btf_16_adds_subs_avx2(&x[57], &x[54]); + btf_16_adds_subs_avx2(&x[56], &x[55]); +} + +static INLINE void idct64_stage10_avx2(__m256i *x, const int32_t *cospi, + const __m256i _r, int8_t cos_bit) { + (void)cos_bit; + const __m256i cospi_m32_p32 = pair_set_w16_epi16(-cospi[32], cospi[32]); + const __m256i cospi_p32_p32 = pair_set_w16_epi16(cospi[32], cospi[32]); + btf_16_adds_subs_avx2(&x[0], &x[31]); + btf_16_adds_subs_avx2(&x[1], &x[30]); + btf_16_adds_subs_avx2(&x[2], &x[29]); + btf_16_adds_subs_avx2(&x[3], &x[28]); + btf_16_adds_subs_avx2(&x[4], &x[27]); + btf_16_adds_subs_avx2(&x[5], &x[26]); + btf_16_adds_subs_avx2(&x[6], &x[25]); + btf_16_adds_subs_avx2(&x[7], &x[24]); + btf_16_adds_subs_avx2(&x[8], &x[23]); + btf_16_adds_subs_avx2(&x[9], &x[22]); + btf_16_adds_subs_avx2(&x[10], &x[21]); + btf_16_adds_subs_avx2(&x[11], &x[20]); + btf_16_adds_subs_avx2(&x[12], &x[19]); + btf_16_adds_subs_avx2(&x[13], &x[18]); + btf_16_adds_subs_avx2(&x[14], &x[17]); + btf_16_adds_subs_avx2(&x[15], &x[16]); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[40], &x[55], _r, cos_bit); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[41], &x[54], _r, cos_bit); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[42], &x[53], _r, cos_bit); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[43], &x[52], _r, cos_bit); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[44], &x[51], _r, cos_bit); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[45], &x[50], _r, cos_bit); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[46], &x[49], _r, cos_bit); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[47], &x[48], _r, cos_bit); +} + +static INLINE void idct64_stage11_avx2(__m256i *output, __m256i *x) { + btf_16_adds_subs_out_avx2(&output[0], &output[63], x[0], x[63]); + btf_16_adds_subs_out_avx2(&output[1], &output[62], x[1], x[62]); + btf_16_adds_subs_out_avx2(&output[2], &output[61], x[2], x[61]); + btf_16_adds_subs_out_avx2(&output[3], &output[60], x[3], x[60]); + btf_16_adds_subs_out_avx2(&output[4], &output[59], x[4], x[59]); + btf_16_adds_subs_out_avx2(&output[5], &output[58], x[5], x[58]); + btf_16_adds_subs_out_avx2(&output[6], &output[57], x[6], x[57]); + btf_16_adds_subs_out_avx2(&output[7], &output[56], x[7], x[56]); + btf_16_adds_subs_out_avx2(&output[8], &output[55], x[8], x[55]); + btf_16_adds_subs_out_avx2(&output[9], &output[54], x[9], x[54]); + btf_16_adds_subs_out_avx2(&output[10], &output[53], x[10], x[53]); + btf_16_adds_subs_out_avx2(&output[11], &output[52], x[11], x[52]); + btf_16_adds_subs_out_avx2(&output[12], &output[51], x[12], x[51]); + btf_16_adds_subs_out_avx2(&output[13], &output[50], x[13], x[50]); + btf_16_adds_subs_out_avx2(&output[14], &output[49], x[14], x[49]); + btf_16_adds_subs_out_avx2(&output[15], &output[48], x[15], x[48]); + btf_16_adds_subs_out_avx2(&output[16], &output[47], x[16], x[47]); + btf_16_adds_subs_out_avx2(&output[17], &output[46], x[17], x[46]); + btf_16_adds_subs_out_avx2(&output[18], &output[45], x[18], x[45]); + btf_16_adds_subs_out_avx2(&output[19], &output[44], x[19], x[44]); + btf_16_adds_subs_out_avx2(&output[20], &output[43], x[20], x[43]); + btf_16_adds_subs_out_avx2(&output[21], &output[42], x[21], x[42]); + btf_16_adds_subs_out_avx2(&output[22], &output[41], x[22], x[41]); + btf_16_adds_subs_out_avx2(&output[23], &output[40], x[23], x[40]); + btf_16_adds_subs_out_avx2(&output[24], &output[39], x[24], x[39]); + btf_16_adds_subs_out_avx2(&output[25], &output[38], x[25], x[38]); + btf_16_adds_subs_out_avx2(&output[26], &output[37], x[26], x[37]); + btf_16_adds_subs_out_avx2(&output[27], &output[36], x[27], x[36]); + btf_16_adds_subs_out_avx2(&output[28], &output[35], x[28], x[35]); + btf_16_adds_subs_out_avx2(&output[29], &output[34], x[29], x[34]); + btf_16_adds_subs_out_avx2(&output[30], &output[33], x[30], x[33]); + btf_16_adds_subs_out_avx2(&output[31], &output[32], x[31], x[32]); +} + +static void idct64_low1_new_avx2(const __m256i *input, __m256i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + + // stage 1 + __m256i x[32]; + x[0] = input[0]; + + // stage 2 + // stage 3 + // stage 4 + // stage 5 + // stage 6 + btf_16_w16_0_avx2(cospi[32], cospi[32], x[0], x[0], x[1]); + + // stage 7 + // stage 8 + // stage 9 + // stage 10 + // stage 11 + output[0] = x[0]; + output[63] = x[0]; + output[1] = x[1]; + output[62] = x[1]; + output[2] = x[1]; + output[61] = x[1]; + output[3] = x[0]; + output[60] = x[0]; + output[4] = x[0]; + output[59] = x[0]; + output[5] = x[1]; + output[58] = x[1]; + output[6] = x[1]; + output[57] = x[1]; + output[7] = x[0]; + output[56] = x[0]; + output[8] = x[0]; + output[55] = x[0]; + output[9] = x[1]; + output[54] = x[1]; + output[10] = x[1]; + output[53] = x[1]; + output[11] = x[0]; + output[52] = x[0]; + output[12] = x[0]; + output[51] = x[0]; + output[13] = x[1]; + output[50] = x[1]; + output[14] = x[1]; + output[49] = x[1]; + output[15] = x[0]; + output[48] = x[0]; + output[16] = x[0]; + output[47] = x[0]; + output[17] = x[1]; + output[46] = x[1]; + output[18] = x[1]; + output[45] = x[1]; + output[19] = x[0]; + output[44] = x[0]; + output[20] = x[0]; + output[43] = x[0]; + output[21] = x[1]; + output[42] = x[1]; + output[22] = x[1]; + output[41] = x[1]; + output[23] = x[0]; + output[40] = x[0]; + output[24] = x[0]; + output[39] = x[0]; + output[25] = x[1]; + output[38] = x[1]; + output[26] = x[1]; + output[37] = x[1]; + output[27] = x[0]; + output[36] = x[0]; + output[28] = x[0]; + output[35] = x[0]; + output[29] = x[1]; + output[34] = x[1]; + output[30] = x[1]; + output[33] = x[1]; + output[31] = x[0]; + output[32] = x[0]; +} + +static void idct64_low8_new_avx2(const __m256i *input, __m256i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m256i _r = _mm256_set1_epi32(1 << (INV_COS_BIT - 1)); + const __m256i cospi_m04_p60 = pair_set_w16_epi16(-cospi[4], cospi[60]); + const __m256i cospi_p60_p04 = pair_set_w16_epi16(cospi[60], cospi[4]); + const __m256i cospi_m36_p28 = pair_set_w16_epi16(-cospi[36], cospi[28]); + const __m256i cospi_m28_m36 = pair_set_w16_epi16(-cospi[28], -cospi[36]); + const __m256i cospi_m20_p44 = pair_set_w16_epi16(-cospi[20], cospi[44]); + const __m256i cospi_p44_p20 = pair_set_w16_epi16(cospi[44], cospi[20]); + const __m256i cospi_m52_p12 = pair_set_w16_epi16(-cospi[52], cospi[12]); + const __m256i cospi_m12_m52 = pair_set_w16_epi16(-cospi[12], -cospi[52]); + const __m256i cospi_m08_p56 = pair_set_w16_epi16(-cospi[8], cospi[56]); + const __m256i cospi_p56_p08 = pair_set_w16_epi16(cospi[56], cospi[8]); + const __m256i cospi_m40_p24 = pair_set_w16_epi16(-cospi[40], cospi[24]); + const __m256i cospi_m24_m40 = pair_set_w16_epi16(-cospi[24], -cospi[40]); + const __m256i cospi_p32_p32 = pair_set_w16_epi16(cospi[32], cospi[32]); + const __m256i cospi_m16_p48 = pair_set_w16_epi16(-cospi[16], cospi[48]); + const __m256i cospi_p48_p16 = pair_set_w16_epi16(cospi[48], cospi[16]); + const __m256i cospi_m32_p32 = pair_set_w16_epi16(-cospi[32], cospi[32]); + + // stage 1 + __m256i x[64]; + x[0] = input[0]; + x[8] = input[4]; + x[16] = input[2]; + x[24] = input[6]; + x[32] = input[1]; + x[40] = input[5]; + x[48] = input[3]; + x[56] = input[7]; + + // stage 2 + btf_16_w16_0_avx2(cospi[63], cospi[1], x[32], x[32], x[63]); + btf_16_w16_0_avx2(-cospi[57], cospi[7], x[56], x[39], x[56]); + btf_16_w16_0_avx2(cospi[59], cospi[5], x[40], x[40], x[55]); + btf_16_w16_0_avx2(-cospi[61], cospi[3], x[48], x[47], x[48]); + + // stage 3 + btf_16_w16_0_avx2(cospi[62], cospi[2], x[16], x[16], x[31]); + btf_16_w16_0_avx2(-cospi[58], cospi[6], x[24], x[23], x[24]); + x[33] = x[32]; + x[38] = x[39]; + x[41] = x[40]; + x[46] = x[47]; + x[49] = x[48]; + x[54] = x[55]; + x[57] = x[56]; + x[62] = x[63]; + + // stage 4 + btf_16_w16_0_avx2(cospi[60], cospi[4], x[8], x[8], x[15]); + x[17] = x[16]; + x[22] = x[23]; + x[25] = x[24]; + x[30] = x[31]; + btf_16_w16_avx2(cospi_m04_p60, cospi_p60_p04, &x[33], &x[62], _r, cos_bit); + btf_16_w16_avx2(cospi_m28_m36, cospi_m36_p28, &x[38], &x[57], _r, cos_bit); + btf_16_w16_avx2(cospi_m20_p44, cospi_p44_p20, &x[41], &x[54], _r, cos_bit); + btf_16_w16_avx2(cospi_m12_m52, cospi_m52_p12, &x[46], &x[49], _r, cos_bit); + + // stage 5 + x[9] = x[8]; + x[14] = x[15]; + btf_16_w16_avx2(cospi_m08_p56, cospi_p56_p08, &x[17], &x[30], _r, cos_bit); + btf_16_w16_avx2(cospi_m24_m40, cospi_m40_p24, &x[22], &x[25], _r, cos_bit); + x[35] = x[32]; + x[34] = x[33]; + x[36] = x[39]; + x[37] = x[38]; + x[43] = x[40]; + x[42] = x[41]; + x[44] = x[47]; + x[45] = x[46]; + x[51] = x[48]; + x[50] = x[49]; + x[52] = x[55]; + x[53] = x[54]; + x[59] = x[56]; + x[58] = x[57]; + x[60] = x[63]; + x[61] = x[62]; + + // stage 6 + btf_16_w16_0_avx2(cospi[32], cospi[32], x[0], x[0], x[1]); + btf_16_w16_avx2(cospi_m16_p48, cospi_p48_p16, &x[9], &x[14], _r, cos_bit); + x[19] = x[16]; + x[18] = x[17]; + x[20] = x[23]; + x[21] = x[22]; + x[27] = x[24]; + x[26] = x[25]; + x[28] = x[31]; + x[29] = x[30]; + idct64_stage6_high32_avx2(x, cospi, _r, cos_bit); + + // stage 7 + x[3] = x[0]; + x[2] = x[1]; + x[11] = x[8]; + x[10] = x[9]; + x[12] = x[15]; + x[13] = x[14]; + idct64_stage7_high48_avx2(x, cospi, _r, cos_bit); + + // stage 8 + x[7] = x[0]; + x[6] = x[1]; + x[5] = x[2]; + x[4] = x[3]; + x[9] = x[9]; + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[10], &x[13], _r, cos_bit); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[11], &x[12], _r, cos_bit); + idct64_stage8_high48_avx2(x, cospi, _r, cos_bit); + + idct64_stage9_avx2(x, cospi, _r, cos_bit); + idct64_stage10_avx2(x, cospi, _r, cos_bit); + idct64_stage11_avx2(output, x); +} + +static void idct64_low16_new_avx2(const __m256i *input, __m256i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m256i _r = _mm256_set1_epi32(1 << (INV_COS_BIT - 1)); + + const __m256i cospi_p32_p32 = pair_set_w16_epi16(cospi[32], cospi[32]); + const __m256i cospi_m16_p48 = pair_set_w16_epi16(-cospi[16], cospi[48]); + const __m256i cospi_p48_p16 = pair_set_w16_epi16(cospi[48], cospi[16]); + const __m256i cospi_m48_m16 = pair_set_w16_epi16(-cospi[48], -cospi[16]); + const __m256i cospi_m32_p32 = pair_set_w16_epi16(-cospi[32], cospi[32]); + + // stage 1 + __m256i x[64]; + x[0] = input[0]; + x[4] = input[8]; + x[8] = input[4]; + x[12] = input[12]; + x[16] = input[2]; + x[20] = input[10]; + x[24] = input[6]; + x[28] = input[14]; + x[32] = input[1]; + x[36] = input[9]; + x[40] = input[5]; + x[44] = input[13]; + x[48] = input[3]; + x[52] = input[11]; + x[56] = input[7]; + x[60] = input[15]; + + // stage 2 + btf_16_w16_0_avx2(cospi[63], cospi[1], x[32], x[32], x[63]); + btf_16_w16_0_avx2(-cospi[49], cospi[15], x[60], x[35], x[60]); + btf_16_w16_0_avx2(cospi[55], cospi[9], x[36], x[36], x[59]); + btf_16_w16_0_avx2(-cospi[57], cospi[7], x[56], x[39], x[56]); + btf_16_w16_0_avx2(cospi[59], cospi[5], x[40], x[40], x[55]); + btf_16_w16_0_avx2(-cospi[53], cospi[11], x[52], x[43], x[52]); + btf_16_w16_0_avx2(cospi[51], cospi[13], x[44], x[44], x[51]); + btf_16_w16_0_avx2(-cospi[61], cospi[3], x[48], x[47], x[48]); + + // stage 3 + btf_16_w16_0_avx2(cospi[62], cospi[2], x[16], x[16], x[31]); + btf_16_w16_0_avx2(-cospi[50], cospi[14], x[28], x[19], x[28]); + btf_16_w16_0_avx2(cospi[54], cospi[10], x[20], x[20], x[27]); + btf_16_w16_0_avx2(-cospi[58], cospi[6], x[24], x[23], x[24]); + x[33] = x[32]; + x[34] = x[35]; + x[37] = x[36]; + x[38] = x[39]; + x[41] = x[40]; + x[42] = x[43]; + x[45] = x[44]; + x[46] = x[47]; + x[49] = x[48]; + x[50] = x[51]; + x[53] = x[52]; + x[54] = x[55]; + x[57] = x[56]; + x[58] = x[59]; + x[61] = x[60]; + x[62] = x[63]; + + // stage 4 + btf_16_w16_0_avx2(cospi[60], cospi[4], x[8], x[8], x[15]); + btf_16_w16_0_avx2(-cospi[52], cospi[12], x[12], x[11], x[12]); + x[17] = x[16]; + x[18] = x[19]; + x[21] = x[20]; + x[22] = x[23]; + x[25] = x[24]; + x[26] = x[27]; + x[29] = x[28]; + x[30] = x[31]; + idct64_stage4_high32_avx2(x, cospi, _r, cos_bit); + + // stage 5 + btf_16_w16_0_avx2(cospi[56], cospi[8], x[4], x[4], x[7]); + x[9] = x[8]; + x[10] = x[11]; + x[13] = x[12]; + x[14] = x[15]; + idct64_stage5_high48_avx2(x, cospi, _r, cos_bit); + + // stage 6 + btf_16_w16_0_avx2(cospi[32], cospi[32], x[0], x[0], x[1]); + x[5] = x[4]; + x[6] = x[7]; + btf_16_w16_avx2(cospi_m16_p48, cospi_p48_p16, &x[9], &x[14], _r, cos_bit); + btf_16_w16_avx2(cospi_m48_m16, cospi_m16_p48, &x[10], &x[13], _r, cos_bit); + idct64_stage6_high48_avx2(x, cospi, _r, cos_bit); + + // stage 7 + x[3] = x[0]; + x[2] = x[1]; + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[5], &x[6], _r, cos_bit); + btf_16_adds_subs_avx2(&x[8], &x[11]); + btf_16_adds_subs_avx2(&x[9], &x[10]); + btf_16_adds_subs_avx2(&x[15], &x[12]); + btf_16_adds_subs_avx2(&x[14], &x[13]); + idct64_stage7_high48_avx2(x, cospi, _r, cos_bit); + + // stage 8 + btf_16_adds_subs_avx2(&x[0], &x[7]); + btf_16_adds_subs_avx2(&x[1], &x[6]); + btf_16_adds_subs_avx2(&x[2], &x[5]); + btf_16_adds_subs_avx2(&x[3], &x[4]); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[10], &x[13], _r, cos_bit); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[11], &x[12], _r, cos_bit); + idct64_stage8_high48_avx2(x, cospi, _r, cos_bit); + + idct64_stage9_avx2(x, cospi, _r, cos_bit); + idct64_stage10_avx2(x, cospi, _r, cos_bit); + idct64_stage11_avx2(output, x); +} + +static void idct64_low32_new_avx2(const __m256i *input, __m256i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m256i _r = _mm256_set1_epi32(1 << (INV_COS_BIT - 1)); + + const __m256i cospi_p32_p32 = pair_set_w16_epi16(cospi[32], cospi[32]); + const __m256i cospi_m16_p48 = pair_set_w16_epi16(-cospi[16], cospi[48]); + const __m256i cospi_p48_p16 = pair_set_w16_epi16(cospi[48], cospi[16]); + const __m256i cospi_m48_m16 = pair_set_w16_epi16(-cospi[48], -cospi[16]); + const __m256i cospi_m32_p32 = pair_set_w16_epi16(-cospi[32], cospi[32]); + + // stage 1 + __m256i x[64]; + x[0] = input[0]; + x[2] = input[16]; + x[4] = input[8]; + x[6] = input[24]; + x[8] = input[4]; + x[10] = input[20]; + x[12] = input[12]; + x[14] = input[28]; + x[16] = input[2]; + x[18] = input[18]; + x[20] = input[10]; + x[22] = input[26]; + x[24] = input[6]; + x[26] = input[22]; + x[28] = input[14]; + x[30] = input[30]; + x[32] = input[1]; + x[34] = input[17]; + x[36] = input[9]; + x[38] = input[25]; + x[40] = input[5]; + x[42] = input[21]; + x[44] = input[13]; + x[46] = input[29]; + x[48] = input[3]; + x[50] = input[19]; + x[52] = input[11]; + x[54] = input[27]; + x[56] = input[7]; + x[58] = input[23]; + x[60] = input[15]; + x[62] = input[31]; + + // stage 2 + btf_16_w16_0_avx2(cospi[63], cospi[1], x[32], x[32], x[63]); + btf_16_w16_0_avx2(-cospi[33], cospi[31], x[62], x[33], x[62]); + btf_16_w16_0_avx2(cospi[47], cospi[17], x[34], x[34], x[61]); + btf_16_w16_0_avx2(-cospi[49], cospi[15], x[60], x[35], x[60]); + btf_16_w16_0_avx2(cospi[55], cospi[9], x[36], x[36], x[59]); + btf_16_w16_0_avx2(-cospi[41], cospi[23], x[58], x[37], x[58]); + btf_16_w16_0_avx2(cospi[39], cospi[25], x[38], x[38], x[57]); + btf_16_w16_0_avx2(-cospi[57], cospi[7], x[56], x[39], x[56]); + btf_16_w16_0_avx2(cospi[59], cospi[5], x[40], x[40], x[55]); + btf_16_w16_0_avx2(-cospi[37], cospi[27], x[54], x[41], x[54]); + btf_16_w16_0_avx2(cospi[43], cospi[21], x[42], x[42], x[53]); + btf_16_w16_0_avx2(-cospi[53], cospi[11], x[52], x[43], x[52]); + btf_16_w16_0_avx2(cospi[51], cospi[13], x[44], x[44], x[51]); + btf_16_w16_0_avx2(-cospi[45], cospi[19], x[50], x[45], x[50]); + btf_16_w16_0_avx2(cospi[35], cospi[29], x[46], x[46], x[49]); + btf_16_w16_0_avx2(-cospi[61], cospi[3], x[48], x[47], x[48]); + + // stage 3 + btf_16_w16_0_avx2(cospi[62], cospi[2], x[16], x[16], x[31]); + btf_16_w16_0_avx2(-cospi[34], cospi[30], x[30], x[17], x[30]); + btf_16_w16_0_avx2(cospi[46], cospi[18], x[18], x[18], x[29]); + btf_16_w16_0_avx2(-cospi[50], cospi[14], x[28], x[19], x[28]); + btf_16_w16_0_avx2(cospi[54], cospi[10], x[20], x[20], x[27]); + btf_16_w16_0_avx2(-cospi[42], cospi[22], x[26], x[21], x[26]); + btf_16_w16_0_avx2(cospi[38], cospi[26], x[22], x[22], x[25]); + btf_16_w16_0_avx2(-cospi[58], cospi[6], x[24], x[23], x[24]); + btf_16_adds_subs_avx2(&x[32], &x[33]); + btf_16_adds_subs_avx2(&x[35], &x[34]); + btf_16_adds_subs_avx2(&x[36], &x[37]); + btf_16_adds_subs_avx2(&x[39], &x[38]); + btf_16_adds_subs_avx2(&x[40], &x[41]); + btf_16_adds_subs_avx2(&x[43], &x[42]); + btf_16_adds_subs_avx2(&x[44], &x[45]); + btf_16_adds_subs_avx2(&x[47], &x[46]); + btf_16_adds_subs_avx2(&x[48], &x[49]); + btf_16_adds_subs_avx2(&x[51], &x[50]); + btf_16_adds_subs_avx2(&x[52], &x[53]); + btf_16_adds_subs_avx2(&x[55], &x[54]); + btf_16_adds_subs_avx2(&x[56], &x[57]); + btf_16_adds_subs_avx2(&x[59], &x[58]); + btf_16_adds_subs_avx2(&x[60], &x[61]); + btf_16_adds_subs_avx2(&x[63], &x[62]); + + // stage 4 + btf_16_w16_0_avx2(cospi[60], cospi[4], x[8], x[8], x[15]); + btf_16_w16_0_avx2(-cospi[36], cospi[28], x[14], x[9], x[14]); + btf_16_w16_0_avx2(cospi[44], cospi[20], x[10], x[10], x[13]); + btf_16_w16_0_avx2(-cospi[52], cospi[12], x[12], x[11], x[12]); + btf_16_adds_subs_avx2(&x[16], &x[17]); + btf_16_adds_subs_avx2(&x[19], &x[18]); + btf_16_adds_subs_avx2(&x[20], &x[21]); + btf_16_adds_subs_avx2(&x[23], &x[22]); + btf_16_adds_subs_avx2(&x[24], &x[25]); + btf_16_adds_subs_avx2(&x[27], &x[26]); + btf_16_adds_subs_avx2(&x[28], &x[29]); + btf_16_adds_subs_avx2(&x[31], &x[30]); + idct64_stage4_high32_avx2(x, cospi, _r, cos_bit); + + // stage 5 + btf_16_w16_0_avx2(cospi[56], cospi[8], x[4], x[4], x[7]); + btf_16_w16_0_avx2(-cospi[40], cospi[24], x[6], x[5], x[6]); + btf_16_adds_subs_avx2(&x[8], &x[9]); + btf_16_adds_subs_avx2(&x[11], &x[10]); + btf_16_adds_subs_avx2(&x[12], &x[13]); + btf_16_adds_subs_avx2(&x[15], &x[14]); + idct64_stage5_high48_avx2(x, cospi, _r, cos_bit); + + // stage 6 + btf_16_w16_0_avx2(cospi[32], cospi[32], x[0], x[0], x[1]); + btf_16_w16_0_avx2(cospi[48], cospi[16], x[2], x[2], x[3]); + btf_16_adds_subs_avx2(&x[4], &x[5]); + btf_16_adds_subs_avx2(&x[7], &x[6]); + btf_16_w16_avx2(cospi_m16_p48, cospi_p48_p16, &x[9], &x[14], _r, cos_bit); + btf_16_w16_avx2(cospi_m48_m16, cospi_m16_p48, &x[10], &x[13], _r, cos_bit); + idct64_stage6_high48_avx2(x, cospi, _r, cos_bit); + + // stage 7 + btf_16_adds_subs_avx2(&x[0], &x[3]); + btf_16_adds_subs_avx2(&x[1], &x[2]); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[5], &x[6], _r, cos_bit); + btf_16_adds_subs_avx2(&x[8], &x[11]); + btf_16_adds_subs_avx2(&x[9], &x[10]); + btf_16_adds_subs_avx2(&x[15], &x[12]); + btf_16_adds_subs_avx2(&x[14], &x[13]); + idct64_stage7_high48_avx2(x, cospi, _r, cos_bit); + + // stage 8 + btf_16_adds_subs_avx2(&x[0], &x[7]); + btf_16_adds_subs_avx2(&x[1], &x[6]); + btf_16_adds_subs_avx2(&x[2], &x[5]); + btf_16_adds_subs_avx2(&x[3], &x[4]); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[10], &x[13], _r, cos_bit); + btf_16_w16_avx2(cospi_m32_p32, cospi_p32_p32, &x[11], &x[12], _r, cos_bit); + idct64_stage8_high48_avx2(x, cospi, _r, cos_bit); + + // stage 9~11 + idct64_stage9_avx2(x, cospi, _r, cos_bit); + idct64_stage10_avx2(x, cospi, _r, cos_bit); + idct64_stage11_avx2(output, x); +} + +// 1D functions process 16 pixels at one time. +static const transform_1d_avx2 + lowbd_txfm_all_1d_zeros_w16_arr[TX_SIZES][ITX_TYPES_1D][4] = { + { + { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL }, + }, + { { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL } }, + { + { idct16_low1_new_avx2, idct16_low8_new_avx2, idct16_new_avx2, NULL }, + { iadst16_low1_new_avx2, iadst16_low8_new_avx2, iadst16_new_avx2, + NULL }, + { NULL, NULL, NULL, NULL }, + }, + { { idct32_low1_new_avx2, idct32_low8_new_avx2, idct32_low16_new_avx2, + idct32_new_avx2 }, + { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL } }, + { { idct64_low1_new_avx2, idct64_low8_new_avx2, idct64_low16_new_avx2, + idct64_low32_new_avx2 }, + { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL } } + }; + +// only process w >= 16 h >= 16 +static INLINE void lowbd_inv_txfm2d_add_no_identity_avx2( + const int32_t *input, uint8_t *output, int stride, TX_TYPE tx_type, + TX_SIZE tx_size, int eob) { + __m256i buf1[64 * 16]; + int eobx, eoby; + get_eobx_eoby_scan_default(&eobx, &eoby, tx_size, eob); + const int8_t *shift = inv_txfm_shift_ls[tx_size]; + const int txw_idx = get_txw_idx(tx_size); + const int txh_idx = get_txh_idx(tx_size); + const int cos_bit_col = inv_cos_bit_col[txw_idx][txh_idx]; + const int cos_bit_row = inv_cos_bit_row[txw_idx][txh_idx]; + const int txfm_size_col = tx_size_wide[tx_size]; + const int txfm_size_row = tx_size_high[tx_size]; + const int buf_size_w_div16 = txfm_size_col >> 4; + const int buf_size_nonzero_w_div16 = (eobx + 16) >> 4; + const int buf_size_nonzero_h_div16 = (eoby + 16) >> 4; + const int input_stride = AOMMIN(32, txfm_size_col); + const int rect_type = get_rect_tx_log_ratio(txfm_size_col, txfm_size_row); + + const int fun_idx_x = lowbd_txfm_all_1d_zeros_idx[eobx]; + const int fun_idx_y = lowbd_txfm_all_1d_zeros_idx[eoby]; + const transform_1d_avx2 row_txfm = + lowbd_txfm_all_1d_zeros_w16_arr[txw_idx][hitx_1d_tab[tx_type]][fun_idx_x]; + const transform_1d_avx2 col_txfm = + lowbd_txfm_all_1d_zeros_w16_arr[txh_idx][vitx_1d_tab[tx_type]][fun_idx_y]; + + assert(col_txfm != NULL); + assert(row_txfm != NULL); + int ud_flip, lr_flip; + get_flip_cfg(tx_type, &ud_flip, &lr_flip); + for (int i = 0; i < buf_size_nonzero_h_div16; i++) { + __m256i buf0[64]; + const int32_t *input_row = input + (i << 4) * input_stride; + for (int j = 0; j < buf_size_nonzero_w_div16; ++j) { + __m256i *buf0_cur = buf0 + j * 16; + const int32_t *input_cur = input_row + j * 16; + load_buffer_32bit_to_16bit_w16_avx2(input_cur, input_stride, buf0_cur, + 16); + transpose_16bit_16x16_avx2(buf0_cur, buf0_cur); + } + if (rect_type == 1 || rect_type == -1) { + round_shift_avx2(buf0, buf0, input_stride); // rect special code + } + row_txfm(buf0, buf0, cos_bit_row); + round_shift_16bit_w16_avx2(buf0, txfm_size_col, shift[0]); + + __m256i *buf1_cur = buf1 + (i << 4); + if (lr_flip) { + for (int j = 0; j < buf_size_w_div16; ++j) { + __m256i temp[16]; + flip_buf_avx2(buf0 + 16 * j, temp, 16); + int offset = txfm_size_row * (buf_size_w_div16 - 1 - j); + transpose_16bit_16x16_avx2(temp, buf1_cur + offset); + } + } else { + for (int j = 0; j < buf_size_w_div16; ++j) { + transpose_16bit_16x16_avx2(buf0 + 16 * j, buf1_cur + txfm_size_row * j); + } + } + } + for (int i = 0; i < buf_size_w_div16; i++) { + __m256i *buf1_cur = buf1 + i * txfm_size_row; + col_txfm(buf1_cur, buf1_cur, cos_bit_col); + round_shift_16bit_w16_avx2(buf1_cur, txfm_size_row, shift[1]); + } + for (int i = 0; i < buf_size_w_div16; i++) { + lowbd_write_buffer_16xn_avx2(buf1 + i * txfm_size_row, output + 16 * i, + stride, ud_flip, txfm_size_row); + } +} + +static INLINE void iidentity_row_16xn_avx2(__m256i *out, const int32_t *input, + int stride, int shift, int height, + int txw_idx, int rect_type) { + const int32_t *input_row = input; + const __m256i scale = _mm256_set1_epi16(NewSqrt2list[txw_idx]); + const __m256i _r = _mm256_set1_epi16((1 << (NewSqrt2Bits - 1)) + + (1 << (NewSqrt2Bits - shift - 1))); + const __m256i one = _mm256_set1_epi16(1); + const __m256i scale__r = _mm256_unpacklo_epi16(scale, _r); + if (rect_type != 1 && rect_type != -1) { + for (int i = 0; i < height; ++i) { + const __m256i src = load_32bit_to_16bit_w16_avx2(input_row); + input_row += stride; + __m256i lo = _mm256_unpacklo_epi16(src, one); + __m256i hi = _mm256_unpackhi_epi16(src, one); + lo = _mm256_madd_epi16(lo, scale__r); + hi = _mm256_madd_epi16(hi, scale__r); + lo = _mm256_srai_epi32(lo, NewSqrt2Bits - shift); + hi = _mm256_srai_epi32(hi, NewSqrt2Bits - shift); + out[i] = _mm256_packs_epi32(lo, hi); + } + } else { + const __m256i rect_scale = + _mm256_set1_epi16(NewInvSqrt2 << (15 - NewSqrt2Bits)); + for (int i = 0; i < height; ++i) { + __m256i src = load_32bit_to_16bit_w16_avx2(input_row); + src = _mm256_mulhrs_epi16(src, rect_scale); + input_row += stride; + __m256i lo = _mm256_unpacklo_epi16(src, one); + __m256i hi = _mm256_unpackhi_epi16(src, one); + lo = _mm256_madd_epi16(lo, scale__r); + hi = _mm256_madd_epi16(hi, scale__r); + lo = _mm256_srai_epi32(lo, NewSqrt2Bits - shift); + hi = _mm256_srai_epi32(hi, NewSqrt2Bits - shift); + out[i] = _mm256_packs_epi32(lo, hi); + } + } +} + +static INLINE void iidentity_col_16xn_avx2(uint8_t *output, int stride, + __m256i *buf, int shift, int height, + int txh_idx) { + const __m256i scale = _mm256_set1_epi16(NewSqrt2list[txh_idx]); + const __m256i scale__r = _mm256_set1_epi16(1 << (NewSqrt2Bits - 1)); + const __m256i shift__r = _mm256_set1_epi32(1 << (-shift - 1)); + const __m256i one = _mm256_set1_epi16(1); + const __m256i scale_coeff = _mm256_unpacklo_epi16(scale, scale__r); + for (int h = 0; h < height; ++h) { + __m256i lo = _mm256_unpacklo_epi16(buf[h], one); + __m256i hi = _mm256_unpackhi_epi16(buf[h], one); + lo = _mm256_madd_epi16(lo, scale_coeff); + hi = _mm256_madd_epi16(hi, scale_coeff); + lo = _mm256_srai_epi32(lo, NewSqrt2Bits); + hi = _mm256_srai_epi32(hi, NewSqrt2Bits); + lo = _mm256_add_epi32(lo, shift__r); + hi = _mm256_add_epi32(hi, shift__r); + lo = _mm256_srai_epi32(lo, -shift); + hi = _mm256_srai_epi32(hi, -shift); + const __m256i x = _mm256_packs_epi32(lo, hi); + write_recon_w16_avx2(x, output); + output += stride; + } +} + +static INLINE void lowbd_inv_txfm2d_add_idtx_avx2(const int32_t *input, + uint8_t *output, int stride, + TX_SIZE tx_size, + int32_t eob) { + (void)eob; + const int8_t *shift = inv_txfm_shift_ls[tx_size]; + const int txw_idx = get_txw_idx(tx_size); + const int txh_idx = get_txh_idx(tx_size); + const int txfm_size_col = tx_size_wide[tx_size]; + const int txfm_size_row = tx_size_high[tx_size]; + const int input_stride = AOMMIN(32, txfm_size_col); + const int row_max = AOMMIN(32, txfm_size_row); + const int rect_type = get_rect_tx_log_ratio(txfm_size_col, txfm_size_row); + __m256i buf[32]; + for (int i = 0; i < input_stride; i += 16) { + iidentity_row_16xn_avx2(buf, input + i, input_stride, shift[0], row_max, + txw_idx, rect_type); + iidentity_col_16xn_avx2(output + i, stride, buf, shift[1], row_max, + txh_idx); + } +} + +static INLINE void lowbd_inv_txfm2d_add_h_identity_avx2( + const int32_t *input, uint8_t *output, int stride, TX_TYPE tx_type, + TX_SIZE tx_size, int eob) { + int eobx, eoby; + get_eobx_eoby_scan_h_identity(&eobx, &eoby, tx_size, eob); + const int8_t *shift = inv_txfm_shift_ls[tx_size]; + const int txw_idx = get_txw_idx(tx_size); + const int txh_idx = get_txh_idx(tx_size); + const int cos_bit_col = inv_cos_bit_col[txw_idx][txh_idx]; + const int txfm_size_col = tx_size_wide[tx_size]; + const int txfm_size_row = tx_size_high[tx_size]; + const int txfm_size_col_notzero = AOMMIN(32, txfm_size_col); + const int input_stride = txfm_size_col_notzero; + const int buf_size_w_div16 = (eobx + 16) >> 4; + const int rect_type = get_rect_tx_log_ratio(txfm_size_col, txfm_size_row); + + const int fun_idx_y = lowbd_txfm_all_1d_zeros_idx[eoby]; + const transform_1d_avx2 col_txfm = + lowbd_txfm_all_1d_zeros_w16_arr[txh_idx][vitx_1d_tab[tx_type]][fun_idx_y]; + + assert(col_txfm != NULL); + + int ud_flip, lr_flip; + get_flip_cfg(tx_type, &ud_flip, &lr_flip); + for (int i = 0; i < buf_size_w_div16; i++) { + __m256i buf0[64]; + iidentity_row_16xn_avx2(buf0, input + (i << 4), input_stride, shift[0], + eoby + 1, txw_idx, rect_type); + col_txfm(buf0, buf0, cos_bit_col); + __m256i mshift = _mm256_set1_epi16(1 << (15 + shift[1])); + int k = ud_flip ? (txfm_size_row - 1) : 0; + const int step = ud_flip ? -1 : 1; + for (int j = 0; j < txfm_size_row; ++j, k += step) { + __m256i res = _mm256_mulhrs_epi16(buf0[k], mshift); + write_recon_w16_avx2(res, output + (i << 4) + j * stride); + } + } +} + +static INLINE void lowbd_inv_txfm2d_add_v_identity_avx2( + const int32_t *input, uint8_t *output, int stride, TX_TYPE tx_type, + TX_SIZE tx_size, int eob) { + __m256i buf1[64]; + int eobx, eoby; + get_eobx_eoby_scan_v_identity(&eobx, &eoby, tx_size, eob); + const int8_t *shift = inv_txfm_shift_ls[tx_size]; + const int txw_idx = get_txw_idx(tx_size); + const int txh_idx = get_txh_idx(tx_size); + const int cos_bit_row = inv_cos_bit_row[txw_idx][txh_idx]; + const int txfm_size_col = tx_size_wide[tx_size]; + const int txfm_size_row = tx_size_high[tx_size]; + const int buf_size_w_div16 = txfm_size_col >> 4; + const int buf_size_h_div16 = (eoby + 16) >> 4; + const int input_stride = AOMMIN(32, txfm_size_col); + const int rect_type = get_rect_tx_log_ratio(txfm_size_col, txfm_size_row); + + const int fun_idx_x = lowbd_txfm_all_1d_zeros_idx[eobx]; + const transform_1d_avx2 row_txfm = + lowbd_txfm_all_1d_zeros_w16_arr[txw_idx][hitx_1d_tab[tx_type]][fun_idx_x]; + + assert(row_txfm != NULL); + + int ud_flip, lr_flip; + get_flip_cfg(tx_type, &ud_flip, &lr_flip); + for (int i = 0; i < buf_size_h_div16; i++) { + __m256i buf0[64]; + const int32_t *input_row = input + i * input_stride * 16; + for (int j = 0; j < AOMMIN(4, buf_size_w_div16); ++j) { + __m256i *buf0_cur = buf0 + j * 16; + load_buffer_32bit_to_16bit_w16_avx2(input_row + j * 16, input_stride, + buf0_cur, 16); + transpose_16bit_16x16_avx2(buf0_cur, buf0_cur); + } + if (rect_type == 1 || rect_type == -1) { + round_shift_avx2(buf0, buf0, input_stride); // rect special code + } + row_txfm(buf0, buf0, cos_bit_row); + round_shift_16bit_w16_avx2(buf0, txfm_size_col, shift[0]); + __m256i *_buf1 = buf1; + if (lr_flip) { + for (int j = 0; j < buf_size_w_div16; ++j) { + __m256i temp[16]; + flip_buf_avx2(buf0 + 16 * j, temp, 16); + transpose_16bit_16x16_avx2(temp, + _buf1 + 16 * (buf_size_w_div16 - 1 - j)); + } + } else { + for (int j = 0; j < buf_size_w_div16; ++j) { + transpose_16bit_16x16_avx2(buf0 + 16 * j, _buf1 + 16 * j); + } + } + for (int j = 0; j < buf_size_w_div16; ++j) { + iidentity_col_16xn_avx2(output + i * 16 * stride + j * 16, stride, + buf1 + j * 16, shift[1], 16, txh_idx); + } + } +} + +// for 32x32,32x64,64x32,64x64,16x32,32x16,64x16,16x64 +static INLINE void lowbd_inv_txfm2d_add_universe_avx2( + const int32_t *input, uint8_t *output, int stride, TX_TYPE tx_type, + TX_SIZE tx_size, int eob) { + (void)eob; + switch (tx_type) { + case DCT_DCT: + case ADST_DCT: // ADST in vertical, DCT in horizontal + case DCT_ADST: // DCT in vertical, ADST in horizontal + case ADST_ADST: // ADST in both directions + case FLIPADST_DCT: + case DCT_FLIPADST: + case FLIPADST_FLIPADST: + case ADST_FLIPADST: + case FLIPADST_ADST: + lowbd_inv_txfm2d_add_no_identity_avx2(input, output, stride, tx_type, + tx_size, eob); + break; + case IDTX: + lowbd_inv_txfm2d_add_idtx_avx2(input, output, stride, tx_size, eob); + break; + case V_DCT: + case V_ADST: + case V_FLIPADST: + lowbd_inv_txfm2d_add_h_identity_avx2(input, output, stride, tx_type, + tx_size, eob); + break; + case H_DCT: + case H_ADST: + case H_FLIPADST: + lowbd_inv_txfm2d_add_v_identity_avx2(input, output, stride, tx_type, + tx_size, eob); + break; + default: + av1_lowbd_inv_txfm2d_add_ssse3(input, output, stride, tx_type, tx_size, + eob); + break; + } +} + +void av1_lowbd_inv_txfm2d_add_avx2(const int32_t *input, uint8_t *output, + int stride, TX_TYPE tx_type, TX_SIZE tx_size, + int eob) { + switch (tx_size) { + case TX_4X4: + case TX_8X8: + case TX_4X8: + case TX_8X4: + case TX_8X16: + case TX_16X8: + case TX_4X16: + case TX_16X4: + case TX_8X32: + case TX_32X8: + av1_lowbd_inv_txfm2d_add_ssse3(input, output, stride, tx_type, tx_size, + eob); + break; + case TX_16X16: + case TX_32X32: + case TX_64X64: + case TX_16X32: + case TX_32X16: + case TX_32X64: + case TX_64X32: + case TX_16X64: + case TX_64X16: + default: + lowbd_inv_txfm2d_add_universe_avx2(input, output, stride, tx_type, + tx_size, eob); + break; + } +} + +void av1_inv_txfm_add_avx2(const tran_low_t *dqcoeff, uint8_t *dst, int stride, + const TxfmParam *txfm_param) { + const TX_TYPE tx_type = txfm_param->tx_type; + if (!txfm_param->lossless) { + av1_lowbd_inv_txfm2d_add_avx2(dqcoeff, dst, stride, tx_type, + txfm_param->tx_size, txfm_param->eob); + } else { + av1_inv_txfm_add_c(dqcoeff, dst, stride, txfm_param); + } +} diff --git a/third_party/aom/av1/common/x86/av1_inv_txfm_avx2.h b/third_party/aom/av1/common/x86/av1_inv_txfm_avx2.h new file mode 100644 index 0000000000..f74cbaeaa5 --- /dev/null +++ b/third_party/aom/av1/common/x86/av1_inv_txfm_avx2.h @@ -0,0 +1,71 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ +#ifndef AOM_AV1_COMMON_X86_AV1_INV_TXFM_AVX2_H_ +#define AOM_AV1_COMMON_X86_AV1_INV_TXFM_AVX2_H_ + +#include <immintrin.h> + +#include "config/aom_config.h" +#include "config/av1_rtcd.h" + +#include "aom/aom_integer.h" +#include "aom_dsp/x86/transpose_sse2.h" +#include "aom_dsp/x86/txfm_common_sse2.h" +#include "aom_dsp/x86/txfm_common_avx2.h" + +#ifdef __cplusplus +extern "C" { +#endif + +// half input is zero +#define btf_16_w16_0_avx2(w0, w1, in, out0, out1) \ + { \ + const __m256i _w0 = _mm256_set1_epi16(w0 * 8); \ + const __m256i _w1 = _mm256_set1_epi16(w1 * 8); \ + const __m256i _in = in; \ + out0 = _mm256_mulhrs_epi16(_in, _w0); \ + out1 = _mm256_mulhrs_epi16(_in, _w1); \ + } + +static INLINE void round_shift_avx2(const __m256i *input, __m256i *output, + int size) { + const __m256i scale = _mm256_set1_epi16(NewInvSqrt2 * 8); + for (int i = 0; i < size; ++i) { + output[i] = _mm256_mulhrs_epi16(input[i], scale); + } +} + +static INLINE void write_recon_w16_avx2(__m256i res, uint8_t *output) { + __m128i pred = _mm_loadu_si128((__m128i const *)(output)); + __m256i u = _mm256_adds_epi16(_mm256_cvtepu8_epi16(pred), res); + __m128i y = _mm256_castsi256_si128( + _mm256_permute4x64_epi64(_mm256_packus_epi16(u, u), 168)); + _mm_storeu_si128((__m128i *)(output), y); +} + +static INLINE void lowbd_write_buffer_16xn_avx2(__m256i *in, uint8_t *output, + int stride, int flipud, + int height) { + int j = flipud ? (height - 1) : 0; + const int step = flipud ? -1 : 1; + for (int i = 0; i < height; ++i, j += step) { + write_recon_w16_avx2(in[j], output + i * stride); + } +} + +void av1_lowbd_inv_txfm2d_add_avx2(const int32_t *input, uint8_t *output, + int stride, TX_TYPE tx_type, TX_SIZE tx_size, + int eob); +#ifdef __cplusplus +} +#endif + +#endif // AOM_AV1_COMMON_X86_AV1_INV_TXFM_AVX2_H_ diff --git a/third_party/aom/av1/common/x86/av1_inv_txfm_ssse3.c b/third_party/aom/av1/common/x86/av1_inv_txfm_ssse3.c new file mode 100644 index 0000000000..995bc3da44 --- /dev/null +++ b/third_party/aom/av1/common/x86/av1_inv_txfm_ssse3.c @@ -0,0 +1,2923 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include "config/aom_config.h" +#include "config/av1_rtcd.h" + +#include "av1/common/av1_inv_txfm1d_cfg.h" +#include "av1/common/x86/av1_inv_txfm_ssse3.h" +#include "av1/common/x86/av1_txfm_sse2.h" + +// TODO(venkatsanampudi@ittiam.com): move this to header file + +// Sqrt2, Sqrt2^2, Sqrt2^3, Sqrt2^4, Sqrt2^5 +static int32_t NewSqrt2list[TX_SIZES] = { 5793, 2 * 4096, 2 * 5793, 4 * 4096, + 4 * 5793 }; + +// TODO(binpengsmail@gmail.com): replace some for loop with do {} while + +static void idct4_new_sse2(const __m128i *input, __m128i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m128i __rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + const __m128i cospi_p32_m32 = pair_set_epi16(cospi[32], -cospi[32]); + const __m128i cospi_p48_m16 = pair_set_epi16(cospi[48], -cospi[16]); + const __m128i cospi_p16_p48 = pair_set_epi16(cospi[16], cospi[48]); + + // stage 1 + __m128i x[4]; + x[0] = input[0]; + x[1] = input[2]; + x[2] = input[1]; + x[3] = input[3]; + + // stage 2 + btf_16_sse2(cospi_p32_p32, cospi_p32_m32, x[0], x[1], x[0], x[1]); + btf_16_sse2(cospi_p48_m16, cospi_p16_p48, x[2], x[3], x[2], x[3]); + + // stage 3 + btf_16_adds_subs_out_sse2(output[0], output[3], x[0], x[3]); + btf_16_adds_subs_out_sse2(output[1], output[2], x[1], x[2]); +} + +void idct4_w4_new_sse2(const __m128i *input, __m128i *output, int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m128i __rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + const __m128i cospi_p32_m32 = pair_set_epi16(cospi[32], -cospi[32]); + const __m128i cospi_p48_m16 = pair_set_epi16(cospi[48], -cospi[16]); + const __m128i cospi_p16_p48 = pair_set_epi16(cospi[16], cospi[48]); + + // stage 1 + __m128i x[4]; + x[0] = input[0]; + x[1] = input[2]; + x[2] = input[1]; + x[3] = input[3]; + + // stage 2 + btf_16_4p_sse2(cospi_p32_p32, cospi_p32_m32, x[0], x[1], x[0], x[1]); + btf_16_4p_sse2(cospi_p48_m16, cospi_p16_p48, x[2], x[3], x[2], x[3]); + + // stage 3 + btf_16_adds_subs_out_sse2(output[0], output[3], x[0], x[3]); + btf_16_adds_subs_out_sse2(output[1], output[2], x[1], x[2]); +} + +void idct8_low1_new_ssse3(const __m128i *input, __m128i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + + // stage 1 + __m128i x[2]; + x[0] = input[0]; + + // stage 2 + // stage 3 + btf_16_ssse3(cospi[32], cospi[32], x[0], x[0], x[1]); + + // stage 4 + // stage 5 + output[0] = x[0]; + output[7] = x[0]; + output[1] = x[1]; + output[6] = x[1]; + output[2] = x[1]; + output[5] = x[1]; + output[3] = x[0]; + output[4] = x[0]; +} + +void idct8_new_sse2(const __m128i *input, __m128i *output, int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m128i __rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + + const __m128i cospi_p56_m08 = pair_set_epi16(cospi[56], -cospi[8]); + const __m128i cospi_p08_p56 = pair_set_epi16(cospi[8], cospi[56]); + const __m128i cospi_p24_m40 = pair_set_epi16(cospi[24], -cospi[40]); + const __m128i cospi_p40_p24 = pair_set_epi16(cospi[40], cospi[24]); + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + const __m128i cospi_p32_m32 = pair_set_epi16(cospi[32], -cospi[32]); + const __m128i cospi_p48_m16 = pair_set_epi16(cospi[48], -cospi[16]); + const __m128i cospi_p16_p48 = pair_set_epi16(cospi[16], cospi[48]); + const __m128i cospi_m32_p32 = pair_set_epi16(-cospi[32], cospi[32]); + + // stage 1 + __m128i x[8]; + x[0] = input[0]; + x[1] = input[4]; + x[2] = input[2]; + x[3] = input[6]; + x[4] = input[1]; + x[5] = input[5]; + x[6] = input[3]; + x[7] = input[7]; + + // stage 2 + btf_16_sse2(cospi_p56_m08, cospi_p08_p56, x[4], x[7], x[4], x[7]); + btf_16_sse2(cospi_p24_m40, cospi_p40_p24, x[5], x[6], x[5], x[6]); + + // stage 3 + btf_16_sse2(cospi_p32_p32, cospi_p32_m32, x[0], x[1], x[0], x[1]); + btf_16_sse2(cospi_p48_m16, cospi_p16_p48, x[2], x[3], x[2], x[3]); + btf_16_adds_subs_sse2(x[4], x[5]); + btf_16_subs_adds_sse2(x[7], x[6]); + + // stage 4 + btf_16_adds_subs_sse2(x[0], x[3]); + btf_16_adds_subs_sse2(x[1], x[2]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[5], x[6], x[5], x[6]); + + // stage 5 + btf_16_adds_subs_out_sse2(output[0], output[7], x[0], x[7]); + btf_16_adds_subs_out_sse2(output[1], output[6], x[1], x[6]); + btf_16_adds_subs_out_sse2(output[2], output[5], x[2], x[5]); + btf_16_adds_subs_out_sse2(output[3], output[4], x[3], x[4]); +} + +void idct8_w4_new_sse2(const __m128i *input, __m128i *output, int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m128i __rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + + const __m128i cospi_p56_m08 = pair_set_epi16(cospi[56], -cospi[8]); + const __m128i cospi_p08_p56 = pair_set_epi16(cospi[8], cospi[56]); + const __m128i cospi_p24_m40 = pair_set_epi16(cospi[24], -cospi[40]); + const __m128i cospi_p40_p24 = pair_set_epi16(cospi[40], cospi[24]); + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + const __m128i cospi_p32_m32 = pair_set_epi16(cospi[32], -cospi[32]); + const __m128i cospi_p48_m16 = pair_set_epi16(cospi[48], -cospi[16]); + const __m128i cospi_p16_p48 = pair_set_epi16(cospi[16], cospi[48]); + const __m128i cospi_m32_p32 = pair_set_epi16(-cospi[32], cospi[32]); + + // stage 1 + __m128i x[8]; + x[0] = input[0]; + x[1] = input[4]; + x[2] = input[2]; + x[3] = input[6]; + x[4] = input[1]; + x[5] = input[5]; + x[6] = input[3]; + x[7] = input[7]; + + // stage 2 + btf_16_4p_sse2(cospi_p56_m08, cospi_p08_p56, x[4], x[7], x[4], x[7]); + btf_16_4p_sse2(cospi_p24_m40, cospi_p40_p24, x[5], x[6], x[5], x[6]); + + // stage 3 + btf_16_4p_sse2(cospi_p32_p32, cospi_p32_m32, x[0], x[1], x[0], x[1]); + btf_16_4p_sse2(cospi_p48_m16, cospi_p16_p48, x[2], x[3], x[2], x[3]); + btf_16_adds_subs_sse2(x[4], x[5]); + btf_16_subs_adds_sse2(x[7], x[6]); + + // stage 4 + btf_16_adds_subs_sse2(x[0], x[3]); + btf_16_adds_subs_sse2(x[1], x[2]); + btf_16_4p_sse2(cospi_m32_p32, cospi_p32_p32, x[5], x[6], x[5], x[6]); + + // stage 5 + btf_16_adds_subs_out_sse2(output[0], output[7], x[0], x[7]); + btf_16_adds_subs_out_sse2(output[1], output[6], x[1], x[6]); + btf_16_adds_subs_out_sse2(output[2], output[5], x[2], x[5]); + btf_16_adds_subs_out_sse2(output[3], output[4], x[3], x[4]); +} + +static INLINE void idct16_stage5_sse2(__m128i *x, const int32_t *cospi, + const __m128i __rounding, + int8_t cos_bit) { + const __m128i cospi_m32_p32 = pair_set_epi16(-cospi[32], cospi[32]); + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + btf_16_adds_subs_sse2(x[0], x[3]); + btf_16_adds_subs_sse2(x[1], x[2]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[5], x[6], x[5], x[6]); + btf_16_adds_subs_sse2(x[8], x[11]); + btf_16_adds_subs_sse2(x[9], x[10]); + btf_16_subs_adds_sse2(x[15], x[12]); + btf_16_subs_adds_sse2(x[14], x[13]); +} + +static INLINE void idct16_stage6_sse2(__m128i *x, const int32_t *cospi, + const __m128i __rounding, + int8_t cos_bit) { + const __m128i cospi_m32_p32 = pair_set_epi16(-cospi[32], cospi[32]); + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + btf_16_adds_subs_sse2(x[0], x[7]); + btf_16_adds_subs_sse2(x[1], x[6]); + btf_16_adds_subs_sse2(x[2], x[5]); + btf_16_adds_subs_sse2(x[3], x[4]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[10], x[13], x[10], x[13]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[11], x[12], x[11], x[12]); +} + +static INLINE void idct16_stage7_sse2(__m128i *output, __m128i *x) { + btf_16_adds_subs_out_sse2(output[0], output[15], x[0], x[15]); + btf_16_adds_subs_out_sse2(output[1], output[14], x[1], x[14]); + btf_16_adds_subs_out_sse2(output[2], output[13], x[2], x[13]); + btf_16_adds_subs_out_sse2(output[3], output[12], x[3], x[12]); + btf_16_adds_subs_out_sse2(output[4], output[11], x[4], x[11]); + btf_16_adds_subs_out_sse2(output[5], output[10], x[5], x[10]); + btf_16_adds_subs_out_sse2(output[6], output[9], x[6], x[9]); + btf_16_adds_subs_out_sse2(output[7], output[8], x[7], x[8]); +} + +static void idct16_low1_new_ssse3(const __m128i *input, __m128i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + + // stage 1 + __m128i x[2]; + x[0] = input[0]; + + // stage 2 + // stage 3 + // stage 4 + btf_16_ssse3(cospi[32], cospi[32], x[0], x[0], x[1]); + + // stage 5 + // stage 6 + // stage 7 + output[0] = x[0]; + output[15] = x[0]; + output[1] = x[1]; + output[14] = x[1]; + output[2] = x[1]; + output[13] = x[1]; + output[3] = x[0]; + output[12] = x[0]; + output[4] = x[0]; + output[11] = x[0]; + output[5] = x[1]; + output[10] = x[1]; + output[6] = x[1]; + output[9] = x[1]; + output[7] = x[0]; + output[8] = x[0]; +} + +static void idct16_low8_new_ssse3(const __m128i *input, __m128i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m128i __rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + const __m128i cospi_m16_p48 = pair_set_epi16(-cospi[16], cospi[48]); + const __m128i cospi_p48_p16 = pair_set_epi16(cospi[48], cospi[16]); + const __m128i cospi_m48_m16 = pair_set_epi16(-cospi[48], -cospi[16]); + + // stage 1 + __m128i x[16]; + x[0] = input[0]; + x[2] = input[4]; + x[4] = input[2]; + x[6] = input[6]; + x[8] = input[1]; + x[10] = input[5]; + x[12] = input[3]; + x[14] = input[7]; + + // stage 2 + btf_16_ssse3(cospi[60], cospi[4], x[8], x[8], x[15]); + btf_16_ssse3(-cospi[36], cospi[28], x[14], x[9], x[14]); + btf_16_ssse3(cospi[44], cospi[20], x[10], x[10], x[13]); + btf_16_ssse3(-cospi[52], cospi[12], x[12], x[11], x[12]); + + // stage 3 + btf_16_ssse3(cospi[56], cospi[8], x[4], x[4], x[7]); + btf_16_ssse3(-cospi[40], cospi[24], x[6], x[5], x[6]); + btf_16_adds_subs_sse2(x[8], x[9]); + btf_16_subs_adds_sse2(x[11], x[10]); + btf_16_adds_subs_sse2(x[12], x[13]); + btf_16_subs_adds_sse2(x[15], x[14]); + + // stage 4 + btf_16_ssse3(cospi[32], cospi[32], x[0], x[0], x[1]); + btf_16_ssse3(cospi[48], cospi[16], x[2], x[2], x[3]); + btf_16_adds_subs_sse2(x[4], x[5]); + btf_16_subs_adds_sse2(x[7], x[6]); + btf_16_sse2(cospi_m16_p48, cospi_p48_p16, x[9], x[14], x[9], x[14]); + btf_16_sse2(cospi_m48_m16, cospi_m16_p48, x[10], x[13], x[10], x[13]); + + idct16_stage5_sse2(x, cospi, __rounding, cos_bit); + idct16_stage6_sse2(x, cospi, __rounding, cos_bit); + idct16_stage7_sse2(output, x); +} + +void idct16_new_sse2(const __m128i *input, __m128i *output, int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m128i __rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + + const __m128i cospi_p60_m04 = pair_set_epi16(cospi[60], -cospi[4]); + const __m128i cospi_p04_p60 = pair_set_epi16(cospi[4], cospi[60]); + const __m128i cospi_p28_m36 = pair_set_epi16(cospi[28], -cospi[36]); + const __m128i cospi_p36_p28 = pair_set_epi16(cospi[36], cospi[28]); + const __m128i cospi_p44_m20 = pair_set_epi16(cospi[44], -cospi[20]); + const __m128i cospi_p20_p44 = pair_set_epi16(cospi[20], cospi[44]); + const __m128i cospi_p12_m52 = pair_set_epi16(cospi[12], -cospi[52]); + const __m128i cospi_p52_p12 = pair_set_epi16(cospi[52], cospi[12]); + const __m128i cospi_p56_m08 = pair_set_epi16(cospi[56], -cospi[8]); + const __m128i cospi_p08_p56 = pair_set_epi16(cospi[8], cospi[56]); + const __m128i cospi_p24_m40 = pair_set_epi16(cospi[24], -cospi[40]); + const __m128i cospi_p40_p24 = pair_set_epi16(cospi[40], cospi[24]); + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + const __m128i cospi_p32_m32 = pair_set_epi16(cospi[32], -cospi[32]); + const __m128i cospi_p48_m16 = pair_set_epi16(cospi[48], -cospi[16]); + const __m128i cospi_p16_p48 = pair_set_epi16(cospi[16], cospi[48]); + const __m128i cospi_m16_p48 = pair_set_epi16(-cospi[16], cospi[48]); + const __m128i cospi_p48_p16 = pair_set_epi16(cospi[48], cospi[16]); + const __m128i cospi_m48_m16 = pair_set_epi16(-cospi[48], -cospi[16]); + + // stage 1 + __m128i x[16]; + x[0] = input[0]; + x[1] = input[8]; + x[2] = input[4]; + x[3] = input[12]; + x[4] = input[2]; + x[5] = input[10]; + x[6] = input[6]; + x[7] = input[14]; + x[8] = input[1]; + x[9] = input[9]; + x[10] = input[5]; + x[11] = input[13]; + x[12] = input[3]; + x[13] = input[11]; + x[14] = input[7]; + x[15] = input[15]; + + // stage 2 + btf_16_sse2(cospi_p60_m04, cospi_p04_p60, x[8], x[15], x[8], x[15]); + btf_16_sse2(cospi_p28_m36, cospi_p36_p28, x[9], x[14], x[9], x[14]); + btf_16_sse2(cospi_p44_m20, cospi_p20_p44, x[10], x[13], x[10], x[13]); + btf_16_sse2(cospi_p12_m52, cospi_p52_p12, x[11], x[12], x[11], x[12]); + + // stage 3 + btf_16_sse2(cospi_p56_m08, cospi_p08_p56, x[4], x[7], x[4], x[7]); + btf_16_sse2(cospi_p24_m40, cospi_p40_p24, x[5], x[6], x[5], x[6]); + btf_16_adds_subs_sse2(x[8], x[9]); + btf_16_subs_adds_sse2(x[11], x[10]); + btf_16_adds_subs_sse2(x[12], x[13]); + btf_16_subs_adds_sse2(x[15], x[14]); + + // stage 4 + btf_16_sse2(cospi_p32_p32, cospi_p32_m32, x[0], x[1], x[0], x[1]); + btf_16_sse2(cospi_p48_m16, cospi_p16_p48, x[2], x[3], x[2], x[3]); + btf_16_adds_subs_sse2(x[4], x[5]); + btf_16_subs_adds_sse2(x[7], x[6]); + btf_16_sse2(cospi_m16_p48, cospi_p48_p16, x[9], x[14], x[9], x[14]); + btf_16_sse2(cospi_m48_m16, cospi_m16_p48, x[10], x[13], x[10], x[13]); + + // stage 5~7 + idct16_stage5_sse2(x, cospi, __rounding, cos_bit); + idct16_stage6_sse2(x, cospi, __rounding, cos_bit); + idct16_stage7_sse2(output, x); +} + +void idct16_w4_new_sse2(const __m128i *input, __m128i *output, int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m128i __rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + + const __m128i cospi_p60_m04 = pair_set_epi16(cospi[60], -cospi[4]); + const __m128i cospi_p04_p60 = pair_set_epi16(cospi[4], cospi[60]); + const __m128i cospi_p28_m36 = pair_set_epi16(cospi[28], -cospi[36]); + const __m128i cospi_p36_p28 = pair_set_epi16(cospi[36], cospi[28]); + const __m128i cospi_p44_m20 = pair_set_epi16(cospi[44], -cospi[20]); + const __m128i cospi_p20_p44 = pair_set_epi16(cospi[20], cospi[44]); + const __m128i cospi_p12_m52 = pair_set_epi16(cospi[12], -cospi[52]); + const __m128i cospi_p52_p12 = pair_set_epi16(cospi[52], cospi[12]); + const __m128i cospi_p56_m08 = pair_set_epi16(cospi[56], -cospi[8]); + const __m128i cospi_p08_p56 = pair_set_epi16(cospi[8], cospi[56]); + const __m128i cospi_p24_m40 = pair_set_epi16(cospi[24], -cospi[40]); + const __m128i cospi_p40_p24 = pair_set_epi16(cospi[40], cospi[24]); + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + const __m128i cospi_p32_m32 = pair_set_epi16(cospi[32], -cospi[32]); + const __m128i cospi_p48_m16 = pair_set_epi16(cospi[48], -cospi[16]); + const __m128i cospi_p16_p48 = pair_set_epi16(cospi[16], cospi[48]); + const __m128i cospi_m16_p48 = pair_set_epi16(-cospi[16], cospi[48]); + const __m128i cospi_p48_p16 = pair_set_epi16(cospi[48], cospi[16]); + const __m128i cospi_m48_m16 = pair_set_epi16(-cospi[48], -cospi[16]); + const __m128i cospi_m32_p32 = pair_set_epi16(-cospi[32], cospi[32]); + + // stage 1 + __m128i x[16]; + x[0] = input[0]; + x[1] = input[8]; + x[2] = input[4]; + x[3] = input[12]; + x[4] = input[2]; + x[5] = input[10]; + x[6] = input[6]; + x[7] = input[14]; + x[8] = input[1]; + x[9] = input[9]; + x[10] = input[5]; + x[11] = input[13]; + x[12] = input[3]; + x[13] = input[11]; + x[14] = input[7]; + x[15] = input[15]; + + // stage 2 + btf_16_4p_sse2(cospi_p60_m04, cospi_p04_p60, x[8], x[15], x[8], x[15]); + btf_16_4p_sse2(cospi_p28_m36, cospi_p36_p28, x[9], x[14], x[9], x[14]); + btf_16_4p_sse2(cospi_p44_m20, cospi_p20_p44, x[10], x[13], x[10], x[13]); + btf_16_4p_sse2(cospi_p12_m52, cospi_p52_p12, x[11], x[12], x[11], x[12]); + + // stage 3 + btf_16_4p_sse2(cospi_p56_m08, cospi_p08_p56, x[4], x[7], x[4], x[7]); + btf_16_4p_sse2(cospi_p24_m40, cospi_p40_p24, x[5], x[6], x[5], x[6]); + btf_16_adds_subs_sse2(x[8], x[9]); + btf_16_subs_adds_sse2(x[11], x[10]); + btf_16_adds_subs_sse2(x[12], x[13]); + btf_16_subs_adds_sse2(x[15], x[14]); + + // stage 4 + btf_16_4p_sse2(cospi_p32_p32, cospi_p32_m32, x[0], x[1], x[0], x[1]); + btf_16_4p_sse2(cospi_p48_m16, cospi_p16_p48, x[2], x[3], x[2], x[3]); + btf_16_adds_subs_sse2(x[4], x[5]); + btf_16_subs_adds_sse2(x[7], x[6]); + btf_16_4p_sse2(cospi_m16_p48, cospi_p48_p16, x[9], x[14], x[9], x[14]); + btf_16_4p_sse2(cospi_m48_m16, cospi_m16_p48, x[10], x[13], x[10], x[13]); + + // stage 5 + btf_16_adds_subs_sse2(x[0], x[3]); + btf_16_adds_subs_sse2(x[1], x[2]); + btf_16_4p_sse2(cospi_m32_p32, cospi_p32_p32, x[5], x[6], x[5], x[6]); + btf_16_adds_subs_sse2(x[8], x[11]); + btf_16_adds_subs_sse2(x[9], x[10]); + btf_16_subs_adds_sse2(x[15], x[12]); + btf_16_subs_adds_sse2(x[14], x[13]); + + // stage 6 + btf_16_adds_subs_sse2(x[0], x[7]); + btf_16_adds_subs_sse2(x[1], x[6]); + btf_16_adds_subs_sse2(x[2], x[5]); + btf_16_adds_subs_sse2(x[3], x[4]); + btf_16_4p_sse2(cospi_m32_p32, cospi_p32_p32, x[10], x[13], x[10], x[13]); + btf_16_4p_sse2(cospi_m32_p32, cospi_p32_p32, x[11], x[12], x[11], x[12]); + + // stage 7 + idct16_stage7_sse2(output, x); +} + +static INLINE void idct32_high16_stage3_sse2(__m128i *x) { + btf_16_adds_subs_sse2(x[16], x[17]); + btf_16_subs_adds_sse2(x[19], x[18]); + btf_16_adds_subs_sse2(x[20], x[21]); + btf_16_subs_adds_sse2(x[23], x[22]); + btf_16_adds_subs_sse2(x[24], x[25]); + btf_16_subs_adds_sse2(x[27], x[26]); + btf_16_adds_subs_sse2(x[28], x[29]); + btf_16_subs_adds_sse2(x[31], x[30]); +} + +static INLINE void idct32_high16_stage4_sse2(__m128i *x, const int32_t *cospi, + const __m128i __rounding, + int8_t cos_bit) { + const __m128i cospi_m08_p56 = pair_set_epi16(-cospi[8], cospi[56]); + const __m128i cospi_p56_p08 = pair_set_epi16(cospi[56], cospi[8]); + const __m128i cospi_m56_m08 = pair_set_epi16(-cospi[56], -cospi[8]); + const __m128i cospi_m40_p24 = pair_set_epi16(-cospi[40], cospi[24]); + const __m128i cospi_p24_p40 = pair_set_epi16(cospi[24], cospi[40]); + const __m128i cospi_m24_m40 = pair_set_epi16(-cospi[24], -cospi[40]); + btf_16_sse2(cospi_m08_p56, cospi_p56_p08, x[17], x[30], x[17], x[30]); + btf_16_sse2(cospi_m56_m08, cospi_m08_p56, x[18], x[29], x[18], x[29]); + btf_16_sse2(cospi_m40_p24, cospi_p24_p40, x[21], x[26], x[21], x[26]); + btf_16_sse2(cospi_m24_m40, cospi_m40_p24, x[22], x[25], x[22], x[25]); +} + +static INLINE void idct32_high24_stage5_sse2(__m128i *x, const int32_t *cospi, + const __m128i __rounding, + int8_t cos_bit) { + const __m128i cospi_m16_p48 = pair_set_epi16(-cospi[16], cospi[48]); + const __m128i cospi_p48_p16 = pair_set_epi16(cospi[48], cospi[16]); + const __m128i cospi_m48_m16 = pair_set_epi16(-cospi[48], -cospi[16]); + btf_16_sse2(cospi_m16_p48, cospi_p48_p16, x[9], x[14], x[9], x[14]); + btf_16_sse2(cospi_m48_m16, cospi_m16_p48, x[10], x[13], x[10], x[13]); + btf_16_adds_subs_sse2(x[16], x[19]); + btf_16_adds_subs_sse2(x[17], x[18]); + btf_16_subs_adds_sse2(x[23], x[20]); + btf_16_subs_adds_sse2(x[22], x[21]); + btf_16_adds_subs_sse2(x[24], x[27]); + btf_16_adds_subs_sse2(x[25], x[26]); + btf_16_subs_adds_sse2(x[31], x[28]); + btf_16_subs_adds_sse2(x[30], x[29]); +} + +static INLINE void idct32_high28_stage6_sse2(__m128i *x, const int32_t *cospi, + const __m128i __rounding, + int8_t cos_bit) { + const __m128i cospi_m32_p32 = pair_set_epi16(-cospi[32], cospi[32]); + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + const __m128i cospi_m16_p48 = pair_set_epi16(-cospi[16], cospi[48]); + const __m128i cospi_p48_p16 = pair_set_epi16(cospi[48], cospi[16]); + const __m128i cospi_m48_m16 = pair_set_epi16(-cospi[48], -cospi[16]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[5], x[6], x[5], x[6]); + btf_16_adds_subs_sse2(x[8], x[11]); + btf_16_adds_subs_sse2(x[9], x[10]); + btf_16_subs_adds_sse2(x[15], x[12]); + btf_16_subs_adds_sse2(x[14], x[13]); + btf_16_sse2(cospi_m16_p48, cospi_p48_p16, x[18], x[29], x[18], x[29]); + btf_16_sse2(cospi_m16_p48, cospi_p48_p16, x[19], x[28], x[19], x[28]); + btf_16_sse2(cospi_m48_m16, cospi_m16_p48, x[20], x[27], x[20], x[27]); + btf_16_sse2(cospi_m48_m16, cospi_m16_p48, x[21], x[26], x[21], x[26]); +} + +static INLINE void idct32_stage7_sse2(__m128i *x, const int32_t *cospi, + const __m128i __rounding, + int8_t cos_bit) { + const __m128i cospi_m32_p32 = pair_set_epi16(-cospi[32], cospi[32]); + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + btf_16_adds_subs_sse2(x[0], x[7]); + btf_16_adds_subs_sse2(x[1], x[6]); + btf_16_adds_subs_sse2(x[2], x[5]); + btf_16_adds_subs_sse2(x[3], x[4]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[10], x[13], x[10], x[13]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[11], x[12], x[11], x[12]); + btf_16_adds_subs_sse2(x[16], x[23]); + btf_16_adds_subs_sse2(x[17], x[22]); + btf_16_adds_subs_sse2(x[18], x[21]); + btf_16_adds_subs_sse2(x[19], x[20]); + btf_16_subs_adds_sse2(x[31], x[24]); + btf_16_subs_adds_sse2(x[30], x[25]); + btf_16_subs_adds_sse2(x[29], x[26]); + btf_16_subs_adds_sse2(x[28], x[27]); +} + +static INLINE void idct32_stage8_sse2(__m128i *x, const int32_t *cospi, + const __m128i __rounding, + int8_t cos_bit) { + const __m128i cospi_m32_p32 = pair_set_epi16(-cospi[32], cospi[32]); + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + btf_16_adds_subs_sse2(x[0], x[15]); + btf_16_adds_subs_sse2(x[1], x[14]); + btf_16_adds_subs_sse2(x[2], x[13]); + btf_16_adds_subs_sse2(x[3], x[12]); + btf_16_adds_subs_sse2(x[4], x[11]); + btf_16_adds_subs_sse2(x[5], x[10]); + btf_16_adds_subs_sse2(x[6], x[9]); + btf_16_adds_subs_sse2(x[7], x[8]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[20], x[27], x[20], x[27]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[21], x[26], x[21], x[26]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[22], x[25], x[22], x[25]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[23], x[24], x[23], x[24]); +} + +static INLINE void idct32_stage9_sse2(__m128i *output, __m128i *x) { + btf_16_adds_subs_out_sse2(output[0], output[31], x[0], x[31]); + btf_16_adds_subs_out_sse2(output[1], output[30], x[1], x[30]); + btf_16_adds_subs_out_sse2(output[2], output[29], x[2], x[29]); + btf_16_adds_subs_out_sse2(output[3], output[28], x[3], x[28]); + btf_16_adds_subs_out_sse2(output[4], output[27], x[4], x[27]); + btf_16_adds_subs_out_sse2(output[5], output[26], x[5], x[26]); + btf_16_adds_subs_out_sse2(output[6], output[25], x[6], x[25]); + btf_16_adds_subs_out_sse2(output[7], output[24], x[7], x[24]); + btf_16_adds_subs_out_sse2(output[8], output[23], x[8], x[23]); + btf_16_adds_subs_out_sse2(output[9], output[22], x[9], x[22]); + btf_16_adds_subs_out_sse2(output[10], output[21], x[10], x[21]); + btf_16_adds_subs_out_sse2(output[11], output[20], x[11], x[20]); + btf_16_adds_subs_out_sse2(output[12], output[19], x[12], x[19]); + btf_16_adds_subs_out_sse2(output[13], output[18], x[13], x[18]); + btf_16_adds_subs_out_sse2(output[14], output[17], x[14], x[17]); + btf_16_adds_subs_out_sse2(output[15], output[16], x[15], x[16]); +} + +static void idct32_low1_new_ssse3(const __m128i *input, __m128i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + + // stage 1 + __m128i x[2]; + x[0] = input[0]; + + // stage 2 + // stage 3 + // stage 4 + // stage 5 + btf_16_ssse3(cospi[32], cospi[32], x[0], x[0], x[1]); + + // stage 6 + // stage 7 + // stage 8 + // stage 9 + output[0] = x[0]; + output[31] = x[0]; + output[1] = x[1]; + output[30] = x[1]; + output[2] = x[1]; + output[29] = x[1]; + output[3] = x[0]; + output[28] = x[0]; + output[4] = x[0]; + output[27] = x[0]; + output[5] = x[1]; + output[26] = x[1]; + output[6] = x[1]; + output[25] = x[1]; + output[7] = x[0]; + output[24] = x[0]; + output[8] = x[0]; + output[23] = x[0]; + output[9] = x[1]; + output[22] = x[1]; + output[10] = x[1]; + output[21] = x[1]; + output[11] = x[0]; + output[20] = x[0]; + output[12] = x[0]; + output[19] = x[0]; + output[13] = x[1]; + output[18] = x[1]; + output[14] = x[1]; + output[17] = x[1]; + output[15] = x[0]; + output[16] = x[0]; +} + +static void idct32_low8_new_ssse3(const __m128i *input, __m128i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m128i __rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + + // stage 1 + __m128i x[32]; + x[0] = input[0]; + x[4] = input[4]; + x[8] = input[2]; + x[12] = input[6]; + x[16] = input[1]; + x[20] = input[5]; + x[24] = input[3]; + x[28] = input[7]; + + // stage 2 + btf_16_ssse3(cospi[62], cospi[2], x[16], x[16], x[31]); + btf_16_ssse3(-cospi[50], cospi[14], x[28], x[19], x[28]); + btf_16_ssse3(cospi[54], cospi[10], x[20], x[20], x[27]); + btf_16_ssse3(-cospi[58], cospi[6], x[24], x[23], x[24]); + + // stage 3 + btf_16_ssse3(cospi[60], cospi[4], x[8], x[8], x[15]); + btf_16_ssse3(-cospi[52], cospi[12], x[12], x[11], x[12]); + x[17] = x[16]; + x[18] = x[19]; + x[21] = x[20]; + x[22] = x[23]; + x[25] = x[24]; + x[26] = x[27]; + x[29] = x[28]; + x[30] = x[31]; + + // stage 4 + btf_16_ssse3(cospi[56], cospi[8], x[4], x[4], x[7]); + x[9] = x[8]; + x[10] = x[11]; + x[13] = x[12]; + x[14] = x[15]; + idct32_high16_stage4_sse2(x, cospi, __rounding, cos_bit); + + // stage 5 + btf_16_ssse3(cospi[32], cospi[32], x[0], x[0], x[1]); + x[5] = x[4]; + x[6] = x[7]; + idct32_high24_stage5_sse2(x, cospi, __rounding, cos_bit); + // stage 6 + x[3] = x[0]; + x[2] = x[1]; + idct32_high28_stage6_sse2(x, cospi, __rounding, cos_bit); + + idct32_stage7_sse2(x, cospi, __rounding, cos_bit); + idct32_stage8_sse2(x, cospi, __rounding, cos_bit); + idct32_stage9_sse2(output, x); +} + +static void idct32_low16_new_ssse3(const __m128i *input, __m128i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m128i __rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + + // stage 1 + __m128i x[32]; + x[0] = input[0]; + x[2] = input[8]; + x[4] = input[4]; + x[6] = input[12]; + x[8] = input[2]; + x[10] = input[10]; + x[12] = input[6]; + x[14] = input[14]; + x[16] = input[1]; + x[18] = input[9]; + x[20] = input[5]; + x[22] = input[13]; + x[24] = input[3]; + x[26] = input[11]; + x[28] = input[7]; + x[30] = input[15]; + + // stage 2 + btf_16_ssse3(cospi[62], cospi[2], x[16], x[16], x[31]); + btf_16_ssse3(-cospi[34], cospi[30], x[30], x[17], x[30]); + btf_16_ssse3(cospi[46], cospi[18], x[18], x[18], x[29]); + btf_16_ssse3(-cospi[50], cospi[14], x[28], x[19], x[28]); + btf_16_ssse3(cospi[54], cospi[10], x[20], x[20], x[27]); + btf_16_ssse3(-cospi[42], cospi[22], x[26], x[21], x[26]); + btf_16_ssse3(cospi[38], cospi[26], x[22], x[22], x[25]); + btf_16_ssse3(-cospi[58], cospi[6], x[24], x[23], x[24]); + + // stage 3 + btf_16_ssse3(cospi[60], cospi[4], x[8], x[8], x[15]); + btf_16_ssse3(-cospi[36], cospi[28], x[14], x[9], x[14]); + btf_16_ssse3(cospi[44], cospi[20], x[10], x[10], x[13]); + btf_16_ssse3(-cospi[52], cospi[12], x[12], x[11], x[12]); + idct32_high16_stage3_sse2(x); + + // stage 4 + btf_16_ssse3(cospi[56], cospi[8], x[4], x[4], x[7]); + btf_16_ssse3(-cospi[40], cospi[24], x[6], x[5], x[6]); + btf_16_adds_subs_sse2(x[8], x[9]); + btf_16_subs_adds_sse2(x[11], x[10]); + btf_16_adds_subs_sse2(x[12], x[13]); + btf_16_subs_adds_sse2(x[15], x[14]); + idct32_high16_stage4_sse2(x, cospi, __rounding, cos_bit); + + // stage 5 + btf_16_ssse3(cospi[32], cospi[32], x[0], x[0], x[1]); + btf_16_ssse3(cospi[48], cospi[16], x[2], x[2], x[3]); + btf_16_adds_subs_sse2(x[4], x[5]); + btf_16_subs_adds_sse2(x[7], x[6]); + idct32_high24_stage5_sse2(x, cospi, __rounding, cos_bit); + + btf_16_adds_subs_sse2(x[0], x[3]); + btf_16_adds_subs_sse2(x[1], x[2]); + idct32_high28_stage6_sse2(x, cospi, __rounding, cos_bit); + + idct32_stage7_sse2(x, cospi, __rounding, cos_bit); + idct32_stage8_sse2(x, cospi, __rounding, cos_bit); + idct32_stage9_sse2(output, x); +} + +static void idct32_new_sse2(const __m128i *input, __m128i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m128i __rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + + const __m128i cospi_p62_m02 = pair_set_epi16(cospi[62], -cospi[2]); + const __m128i cospi_p02_p62 = pair_set_epi16(cospi[2], cospi[62]); + const __m128i cospi_p30_m34 = pair_set_epi16(cospi[30], -cospi[34]); + const __m128i cospi_p34_p30 = pair_set_epi16(cospi[34], cospi[30]); + const __m128i cospi_p46_m18 = pair_set_epi16(cospi[46], -cospi[18]); + const __m128i cospi_p18_p46 = pair_set_epi16(cospi[18], cospi[46]); + const __m128i cospi_p14_m50 = pair_set_epi16(cospi[14], -cospi[50]); + const __m128i cospi_p50_p14 = pair_set_epi16(cospi[50], cospi[14]); + const __m128i cospi_p54_m10 = pair_set_epi16(cospi[54], -cospi[10]); + const __m128i cospi_p10_p54 = pair_set_epi16(cospi[10], cospi[54]); + const __m128i cospi_p22_m42 = pair_set_epi16(cospi[22], -cospi[42]); + const __m128i cospi_p42_p22 = pair_set_epi16(cospi[42], cospi[22]); + const __m128i cospi_p38_m26 = pair_set_epi16(cospi[38], -cospi[26]); + const __m128i cospi_p26_p38 = pair_set_epi16(cospi[26], cospi[38]); + const __m128i cospi_p06_m58 = pair_set_epi16(cospi[6], -cospi[58]); + const __m128i cospi_p58_p06 = pair_set_epi16(cospi[58], cospi[6]); + const __m128i cospi_p60_m04 = pair_set_epi16(cospi[60], -cospi[4]); + const __m128i cospi_p04_p60 = pair_set_epi16(cospi[4], cospi[60]); + const __m128i cospi_p28_m36 = pair_set_epi16(cospi[28], -cospi[36]); + const __m128i cospi_p36_p28 = pair_set_epi16(cospi[36], cospi[28]); + const __m128i cospi_p44_m20 = pair_set_epi16(cospi[44], -cospi[20]); + const __m128i cospi_p20_p44 = pair_set_epi16(cospi[20], cospi[44]); + const __m128i cospi_p12_m52 = pair_set_epi16(cospi[12], -cospi[52]); + const __m128i cospi_p52_p12 = pair_set_epi16(cospi[52], cospi[12]); + const __m128i cospi_p56_m08 = pair_set_epi16(cospi[56], -cospi[8]); + const __m128i cospi_p08_p56 = pair_set_epi16(cospi[8], cospi[56]); + const __m128i cospi_p24_m40 = pair_set_epi16(cospi[24], -cospi[40]); + const __m128i cospi_p40_p24 = pair_set_epi16(cospi[40], cospi[24]); + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + const __m128i cospi_p32_m32 = pair_set_epi16(cospi[32], -cospi[32]); + const __m128i cospi_p48_m16 = pair_set_epi16(cospi[48], -cospi[16]); + const __m128i cospi_p16_p48 = pair_set_epi16(cospi[16], cospi[48]); + + // stage 1 + __m128i x[32]; + x[0] = input[0]; + x[1] = input[16]; + x[2] = input[8]; + x[3] = input[24]; + x[4] = input[4]; + x[5] = input[20]; + x[6] = input[12]; + x[7] = input[28]; + x[8] = input[2]; + x[9] = input[18]; + x[10] = input[10]; + x[11] = input[26]; + x[12] = input[6]; + x[13] = input[22]; + x[14] = input[14]; + x[15] = input[30]; + x[16] = input[1]; + x[17] = input[17]; + x[18] = input[9]; + x[19] = input[25]; + x[20] = input[5]; + x[21] = input[21]; + x[22] = input[13]; + x[23] = input[29]; + x[24] = input[3]; + x[25] = input[19]; + x[26] = input[11]; + x[27] = input[27]; + x[28] = input[7]; + x[29] = input[23]; + x[30] = input[15]; + x[31] = input[31]; + + // stage 2 + btf_16_sse2(cospi_p62_m02, cospi_p02_p62, x[16], x[31], x[16], x[31]); + btf_16_sse2(cospi_p30_m34, cospi_p34_p30, x[17], x[30], x[17], x[30]); + btf_16_sse2(cospi_p46_m18, cospi_p18_p46, x[18], x[29], x[18], x[29]); + btf_16_sse2(cospi_p14_m50, cospi_p50_p14, x[19], x[28], x[19], x[28]); + btf_16_sse2(cospi_p54_m10, cospi_p10_p54, x[20], x[27], x[20], x[27]); + btf_16_sse2(cospi_p22_m42, cospi_p42_p22, x[21], x[26], x[21], x[26]); + btf_16_sse2(cospi_p38_m26, cospi_p26_p38, x[22], x[25], x[22], x[25]); + btf_16_sse2(cospi_p06_m58, cospi_p58_p06, x[23], x[24], x[23], x[24]); + + // stage 3 + btf_16_sse2(cospi_p60_m04, cospi_p04_p60, x[8], x[15], x[8], x[15]); + btf_16_sse2(cospi_p28_m36, cospi_p36_p28, x[9], x[14], x[9], x[14]); + btf_16_sse2(cospi_p44_m20, cospi_p20_p44, x[10], x[13], x[10], x[13]); + btf_16_sse2(cospi_p12_m52, cospi_p52_p12, x[11], x[12], x[11], x[12]); + idct32_high16_stage3_sse2(x); + + // stage 4 + btf_16_sse2(cospi_p56_m08, cospi_p08_p56, x[4], x[7], x[4], x[7]); + btf_16_sse2(cospi_p24_m40, cospi_p40_p24, x[5], x[6], x[5], x[6]); + btf_16_adds_subs_sse2(x[8], x[9]); + btf_16_subs_adds_sse2(x[11], x[10]); + btf_16_adds_subs_sse2(x[12], x[13]); + btf_16_subs_adds_sse2(x[15], x[14]); + idct32_high16_stage4_sse2(x, cospi, __rounding, cos_bit); + + // stage 5 + btf_16_sse2(cospi_p32_p32, cospi_p32_m32, x[0], x[1], x[0], x[1]); + btf_16_sse2(cospi_p48_m16, cospi_p16_p48, x[2], x[3], x[2], x[3]); + btf_16_adds_subs_sse2(x[4], x[5]); + btf_16_adds_subs_sse2(x[7], x[6]); + idct32_high24_stage5_sse2(x, cospi, __rounding, cos_bit); + + // stage 6 + btf_16_adds_subs_sse2(x[0], x[3]); + btf_16_adds_subs_sse2(x[1], x[2]); + idct32_high28_stage6_sse2(x, cospi, __rounding, cos_bit); + + // stage 7~8 + idct32_stage7_sse2(x, cospi, __rounding, cos_bit); + idct32_stage8_sse2(x, cospi, __rounding, cos_bit); + idct32_stage9_sse2(output, x); +} + +static INLINE void idct64_stage4_high32_sse2(__m128i *x, const int32_t *cospi, + const __m128i __rounding, + int8_t cos_bit) { + const __m128i cospi_m04_p60 = pair_set_epi16(-cospi[4], cospi[60]); + const __m128i cospi_p60_p04 = pair_set_epi16(cospi[60], cospi[4]); + const __m128i cospi_m60_m04 = pair_set_epi16(-cospi[60], -cospi[4]); + const __m128i cospi_m36_p28 = pair_set_epi16(-cospi[36], cospi[28]); + const __m128i cospi_p28_p36 = pair_set_epi16(cospi[28], cospi[36]); + const __m128i cospi_m28_m36 = pair_set_epi16(-cospi[28], -cospi[36]); + const __m128i cospi_m20_p44 = pair_set_epi16(-cospi[20], cospi[44]); + const __m128i cospi_p44_p20 = pair_set_epi16(cospi[44], cospi[20]); + const __m128i cospi_m44_m20 = pair_set_epi16(-cospi[44], -cospi[20]); + const __m128i cospi_m52_p12 = pair_set_epi16(-cospi[52], cospi[12]); + const __m128i cospi_p12_p52 = pair_set_epi16(cospi[12], cospi[52]); + const __m128i cospi_m12_m52 = pair_set_epi16(-cospi[12], -cospi[52]); + btf_16_sse2(cospi_m04_p60, cospi_p60_p04, x[33], x[62], x[33], x[62]); + btf_16_sse2(cospi_m60_m04, cospi_m04_p60, x[34], x[61], x[34], x[61]); + btf_16_sse2(cospi_m36_p28, cospi_p28_p36, x[37], x[58], x[37], x[58]); + btf_16_sse2(cospi_m28_m36, cospi_m36_p28, x[38], x[57], x[38], x[57]); + btf_16_sse2(cospi_m20_p44, cospi_p44_p20, x[41], x[54], x[41], x[54]); + btf_16_sse2(cospi_m44_m20, cospi_m20_p44, x[42], x[53], x[42], x[53]); + btf_16_sse2(cospi_m52_p12, cospi_p12_p52, x[45], x[50], x[45], x[50]); + btf_16_sse2(cospi_m12_m52, cospi_m52_p12, x[46], x[49], x[46], x[49]); +} + +static INLINE void idct64_stage5_high48_sse2(__m128i *x, const int32_t *cospi, + const __m128i __rounding, + int8_t cos_bit) { + const __m128i cospi_m08_p56 = pair_set_epi16(-cospi[8], cospi[56]); + const __m128i cospi_p56_p08 = pair_set_epi16(cospi[56], cospi[8]); + const __m128i cospi_m56_m08 = pair_set_epi16(-cospi[56], -cospi[8]); + const __m128i cospi_m40_p24 = pair_set_epi16(-cospi[40], cospi[24]); + const __m128i cospi_p24_p40 = pair_set_epi16(cospi[24], cospi[40]); + const __m128i cospi_m24_m40 = pair_set_epi16(-cospi[24], -cospi[40]); + btf_16_sse2(cospi_m08_p56, cospi_p56_p08, x[17], x[30], x[17], x[30]); + btf_16_sse2(cospi_m56_m08, cospi_m08_p56, x[18], x[29], x[18], x[29]); + btf_16_sse2(cospi_m40_p24, cospi_p24_p40, x[21], x[26], x[21], x[26]); + btf_16_sse2(cospi_m24_m40, cospi_m40_p24, x[22], x[25], x[22], x[25]); + btf_16_adds_subs_sse2(x[32], x[35]); + btf_16_adds_subs_sse2(x[33], x[34]); + btf_16_subs_adds_sse2(x[39], x[36]); + btf_16_subs_adds_sse2(x[38], x[37]); + btf_16_adds_subs_sse2(x[40], x[43]); + btf_16_adds_subs_sse2(x[41], x[42]); + btf_16_subs_adds_sse2(x[47], x[44]); + btf_16_subs_adds_sse2(x[46], x[45]); + btf_16_adds_subs_sse2(x[48], x[51]); + btf_16_adds_subs_sse2(x[49], x[50]); + btf_16_subs_adds_sse2(x[55], x[52]); + btf_16_subs_adds_sse2(x[54], x[53]); + btf_16_adds_subs_sse2(x[56], x[59]); + btf_16_adds_subs_sse2(x[57], x[58]); + btf_16_subs_adds_sse2(x[63], x[60]); + btf_16_subs_adds_sse2(x[62], x[61]); +} + +static INLINE void idct64_stage6_high32_sse2(__m128i *x, const int32_t *cospi, + const __m128i __rounding, + int8_t cos_bit) { + const __m128i cospi_m08_p56 = pair_set_epi16(-cospi[8], cospi[56]); + const __m128i cospi_p56_p08 = pair_set_epi16(cospi[56], cospi[8]); + const __m128i cospi_m56_m08 = pair_set_epi16(-cospi[56], -cospi[8]); + const __m128i cospi_m40_p24 = pair_set_epi16(-cospi[40], cospi[24]); + const __m128i cospi_p24_p40 = pair_set_epi16(cospi[24], cospi[40]); + const __m128i cospi_m24_m40 = pair_set_epi16(-cospi[24], -cospi[40]); + btf_16_sse2(cospi_m08_p56, cospi_p56_p08, x[34], x[61], x[34], x[61]); + btf_16_sse2(cospi_m08_p56, cospi_p56_p08, x[35], x[60], x[35], x[60]); + btf_16_sse2(cospi_m56_m08, cospi_m08_p56, x[36], x[59], x[36], x[59]); + btf_16_sse2(cospi_m56_m08, cospi_m08_p56, x[37], x[58], x[37], x[58]); + btf_16_sse2(cospi_m40_p24, cospi_p24_p40, x[42], x[53], x[42], x[53]); + btf_16_sse2(cospi_m40_p24, cospi_p24_p40, x[43], x[52], x[43], x[52]); + btf_16_sse2(cospi_m24_m40, cospi_m40_p24, x[44], x[51], x[44], x[51]); + btf_16_sse2(cospi_m24_m40, cospi_m40_p24, x[45], x[50], x[45], x[50]); +} + +static INLINE void idct64_stage6_high48_sse2(__m128i *x, const int32_t *cospi, + const __m128i __rounding, + int8_t cos_bit) { + btf_16_adds_subs_sse2(x[16], x[19]); + btf_16_adds_subs_sse2(x[17], x[18]); + btf_16_subs_adds_sse2(x[23], x[20]); + btf_16_subs_adds_sse2(x[22], x[21]); + btf_16_adds_subs_sse2(x[24], x[27]); + btf_16_adds_subs_sse2(x[25], x[26]); + btf_16_subs_adds_sse2(x[31], x[28]); + btf_16_subs_adds_sse2(x[30], x[29]); + idct64_stage6_high32_sse2(x, cospi, __rounding, cos_bit); +} + +static INLINE void idct64_stage7_high48_sse2(__m128i *x, const int32_t *cospi, + const __m128i __rounding, + int8_t cos_bit) { + const __m128i cospi_m16_p48 = pair_set_epi16(-cospi[16], cospi[48]); + const __m128i cospi_p48_p16 = pair_set_epi16(cospi[48], cospi[16]); + const __m128i cospi_m48_m16 = pair_set_epi16(-cospi[48], -cospi[16]); + btf_16_sse2(cospi_m16_p48, cospi_p48_p16, x[18], x[29], x[18], x[29]); + btf_16_sse2(cospi_m16_p48, cospi_p48_p16, x[19], x[28], x[19], x[28]); + btf_16_sse2(cospi_m48_m16, cospi_m16_p48, x[20], x[27], x[20], x[27]); + btf_16_sse2(cospi_m48_m16, cospi_m16_p48, x[21], x[26], x[21], x[26]); + btf_16_adds_subs_sse2(x[32], x[39]); + btf_16_adds_subs_sse2(x[33], x[38]); + btf_16_adds_subs_sse2(x[34], x[37]); + btf_16_adds_subs_sse2(x[35], x[36]); + btf_16_subs_adds_sse2(x[47], x[40]); + btf_16_subs_adds_sse2(x[46], x[41]); + btf_16_subs_adds_sse2(x[45], x[42]); + btf_16_subs_adds_sse2(x[44], x[43]); + btf_16_adds_subs_sse2(x[48], x[55]); + btf_16_adds_subs_sse2(x[49], x[54]); + btf_16_adds_subs_sse2(x[50], x[53]); + btf_16_adds_subs_sse2(x[51], x[52]); + btf_16_subs_adds_sse2(x[63], x[56]); + btf_16_subs_adds_sse2(x[62], x[57]); + btf_16_subs_adds_sse2(x[61], x[58]); + btf_16_subs_adds_sse2(x[60], x[59]); +} + +static INLINE void idct64_stage8_high48_sse2(__m128i *x, const int32_t *cospi, + const __m128i __rounding, + int8_t cos_bit) { + const __m128i cospi_m16_p48 = pair_set_epi16(-cospi[16], cospi[48]); + const __m128i cospi_p48_p16 = pair_set_epi16(cospi[48], cospi[16]); + const __m128i cospi_m48_m16 = pair_set_epi16(-cospi[48], -cospi[16]); + btf_16_adds_subs_sse2(x[16], x[23]); + btf_16_adds_subs_sse2(x[17], x[22]); + btf_16_adds_subs_sse2(x[18], x[21]); + btf_16_adds_subs_sse2(x[19], x[20]); + btf_16_subs_adds_sse2(x[31], x[24]); + btf_16_subs_adds_sse2(x[30], x[25]); + btf_16_subs_adds_sse2(x[29], x[26]); + btf_16_subs_adds_sse2(x[28], x[27]); + btf_16_sse2(cospi_m16_p48, cospi_p48_p16, x[36], x[59], x[36], x[59]); + btf_16_sse2(cospi_m16_p48, cospi_p48_p16, x[37], x[58], x[37], x[58]); + btf_16_sse2(cospi_m16_p48, cospi_p48_p16, x[38], x[57], x[38], x[57]); + btf_16_sse2(cospi_m16_p48, cospi_p48_p16, x[39], x[56], x[39], x[56]); + btf_16_sse2(cospi_m48_m16, cospi_m16_p48, x[40], x[55], x[40], x[55]); + btf_16_sse2(cospi_m48_m16, cospi_m16_p48, x[41], x[54], x[41], x[54]); + btf_16_sse2(cospi_m48_m16, cospi_m16_p48, x[42], x[53], x[42], x[53]); + btf_16_sse2(cospi_m48_m16, cospi_m16_p48, x[43], x[52], x[43], x[52]); +} + +static INLINE void idct64_stage9_sse2(__m128i *x, const int32_t *cospi, + const __m128i __rounding, + int8_t cos_bit) { + const __m128i cospi_m32_p32 = pair_set_epi16(-cospi[32], cospi[32]); + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + btf_16_adds_subs_sse2(x[0], x[15]); + btf_16_adds_subs_sse2(x[1], x[14]); + btf_16_adds_subs_sse2(x[2], x[13]); + btf_16_adds_subs_sse2(x[3], x[12]); + btf_16_adds_subs_sse2(x[4], x[11]); + btf_16_adds_subs_sse2(x[5], x[10]); + btf_16_adds_subs_sse2(x[6], x[9]); + btf_16_adds_subs_sse2(x[7], x[8]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[20], x[27], x[20], x[27]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[21], x[26], x[21], x[26]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[22], x[25], x[22], x[25]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[23], x[24], x[23], x[24]); + btf_16_adds_subs_sse2(x[32], x[47]); + btf_16_adds_subs_sse2(x[33], x[46]); + btf_16_adds_subs_sse2(x[34], x[45]); + btf_16_adds_subs_sse2(x[35], x[44]); + btf_16_adds_subs_sse2(x[36], x[43]); + btf_16_adds_subs_sse2(x[37], x[42]); + btf_16_adds_subs_sse2(x[38], x[41]); + btf_16_adds_subs_sse2(x[39], x[40]); + btf_16_subs_adds_sse2(x[63], x[48]); + btf_16_subs_adds_sse2(x[62], x[49]); + btf_16_subs_adds_sse2(x[61], x[50]); + btf_16_subs_adds_sse2(x[60], x[51]); + btf_16_subs_adds_sse2(x[59], x[52]); + btf_16_subs_adds_sse2(x[58], x[53]); + btf_16_subs_adds_sse2(x[57], x[54]); + btf_16_subs_adds_sse2(x[56], x[55]); +} + +static INLINE void idct64_stage10_sse2(__m128i *x, const int32_t *cospi, + const __m128i __rounding, + int8_t cos_bit) { + const __m128i cospi_m32_p32 = pair_set_epi16(-cospi[32], cospi[32]); + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + btf_16_adds_subs_sse2(x[0], x[31]); + btf_16_adds_subs_sse2(x[1], x[30]); + btf_16_adds_subs_sse2(x[2], x[29]); + btf_16_adds_subs_sse2(x[3], x[28]); + btf_16_adds_subs_sse2(x[4], x[27]); + btf_16_adds_subs_sse2(x[5], x[26]); + btf_16_adds_subs_sse2(x[6], x[25]); + btf_16_adds_subs_sse2(x[7], x[24]); + btf_16_adds_subs_sse2(x[8], x[23]); + btf_16_adds_subs_sse2(x[9], x[22]); + btf_16_adds_subs_sse2(x[10], x[21]); + btf_16_adds_subs_sse2(x[11], x[20]); + btf_16_adds_subs_sse2(x[12], x[19]); + btf_16_adds_subs_sse2(x[13], x[18]); + btf_16_adds_subs_sse2(x[14], x[17]); + btf_16_adds_subs_sse2(x[15], x[16]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[40], x[55], x[40], x[55]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[41], x[54], x[41], x[54]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[42], x[53], x[42], x[53]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[43], x[52], x[43], x[52]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[44], x[51], x[44], x[51]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[45], x[50], x[45], x[50]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[46], x[49], x[46], x[49]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[47], x[48], x[47], x[48]); +} + +static INLINE void idct64_stage11_sse2(__m128i *output, __m128i *x) { + btf_16_adds_subs_out_sse2(output[0], output[63], x[0], x[63]); + btf_16_adds_subs_out_sse2(output[1], output[62], x[1], x[62]); + btf_16_adds_subs_out_sse2(output[2], output[61], x[2], x[61]); + btf_16_adds_subs_out_sse2(output[3], output[60], x[3], x[60]); + btf_16_adds_subs_out_sse2(output[4], output[59], x[4], x[59]); + btf_16_adds_subs_out_sse2(output[5], output[58], x[5], x[58]); + btf_16_adds_subs_out_sse2(output[6], output[57], x[6], x[57]); + btf_16_adds_subs_out_sse2(output[7], output[56], x[7], x[56]); + btf_16_adds_subs_out_sse2(output[8], output[55], x[8], x[55]); + btf_16_adds_subs_out_sse2(output[9], output[54], x[9], x[54]); + btf_16_adds_subs_out_sse2(output[10], output[53], x[10], x[53]); + btf_16_adds_subs_out_sse2(output[11], output[52], x[11], x[52]); + btf_16_adds_subs_out_sse2(output[12], output[51], x[12], x[51]); + btf_16_adds_subs_out_sse2(output[13], output[50], x[13], x[50]); + btf_16_adds_subs_out_sse2(output[14], output[49], x[14], x[49]); + btf_16_adds_subs_out_sse2(output[15], output[48], x[15], x[48]); + btf_16_adds_subs_out_sse2(output[16], output[47], x[16], x[47]); + btf_16_adds_subs_out_sse2(output[17], output[46], x[17], x[46]); + btf_16_adds_subs_out_sse2(output[18], output[45], x[18], x[45]); + btf_16_adds_subs_out_sse2(output[19], output[44], x[19], x[44]); + btf_16_adds_subs_out_sse2(output[20], output[43], x[20], x[43]); + btf_16_adds_subs_out_sse2(output[21], output[42], x[21], x[42]); + btf_16_adds_subs_out_sse2(output[22], output[41], x[22], x[41]); + btf_16_adds_subs_out_sse2(output[23], output[40], x[23], x[40]); + btf_16_adds_subs_out_sse2(output[24], output[39], x[24], x[39]); + btf_16_adds_subs_out_sse2(output[25], output[38], x[25], x[38]); + btf_16_adds_subs_out_sse2(output[26], output[37], x[26], x[37]); + btf_16_adds_subs_out_sse2(output[27], output[36], x[27], x[36]); + btf_16_adds_subs_out_sse2(output[28], output[35], x[28], x[35]); + btf_16_adds_subs_out_sse2(output[29], output[34], x[29], x[34]); + btf_16_adds_subs_out_sse2(output[30], output[33], x[30], x[33]); + btf_16_adds_subs_out_sse2(output[31], output[32], x[31], x[32]); +} + +static void idct64_low1_new_ssse3(const __m128i *input, __m128i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + + // stage 1 + __m128i x[32]; + x[0] = input[0]; + + // stage 2 + // stage 3 + // stage 4 + // stage 5 + // stage 6 + btf_16_ssse3(cospi[32], cospi[32], x[0], x[0], x[1]); + + // stage 7 + // stage 8 + // stage 9 + // stage 10 + // stage 11 + output[0] = x[0]; + output[63] = x[0]; + output[1] = x[1]; + output[62] = x[1]; + output[2] = x[1]; + output[61] = x[1]; + output[3] = x[0]; + output[60] = x[0]; + output[4] = x[0]; + output[59] = x[0]; + output[5] = x[1]; + output[58] = x[1]; + output[6] = x[1]; + output[57] = x[1]; + output[7] = x[0]; + output[56] = x[0]; + output[8] = x[0]; + output[55] = x[0]; + output[9] = x[1]; + output[54] = x[1]; + output[10] = x[1]; + output[53] = x[1]; + output[11] = x[0]; + output[52] = x[0]; + output[12] = x[0]; + output[51] = x[0]; + output[13] = x[1]; + output[50] = x[1]; + output[14] = x[1]; + output[49] = x[1]; + output[15] = x[0]; + output[48] = x[0]; + output[16] = x[0]; + output[47] = x[0]; + output[17] = x[1]; + output[46] = x[1]; + output[18] = x[1]; + output[45] = x[1]; + output[19] = x[0]; + output[44] = x[0]; + output[20] = x[0]; + output[43] = x[0]; + output[21] = x[1]; + output[42] = x[1]; + output[22] = x[1]; + output[41] = x[1]; + output[23] = x[0]; + output[40] = x[0]; + output[24] = x[0]; + output[39] = x[0]; + output[25] = x[1]; + output[38] = x[1]; + output[26] = x[1]; + output[37] = x[1]; + output[27] = x[0]; + output[36] = x[0]; + output[28] = x[0]; + output[35] = x[0]; + output[29] = x[1]; + output[34] = x[1]; + output[30] = x[1]; + output[33] = x[1]; + output[31] = x[0]; + output[32] = x[0]; +} + +static void idct64_low8_new_ssse3(const __m128i *input, __m128i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m128i __rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + const __m128i cospi_m04_p60 = pair_set_epi16(-cospi[4], cospi[60]); + const __m128i cospi_p60_p04 = pair_set_epi16(cospi[60], cospi[4]); + const __m128i cospi_m36_p28 = pair_set_epi16(-cospi[36], cospi[28]); + const __m128i cospi_m28_m36 = pair_set_epi16(-cospi[28], -cospi[36]); + const __m128i cospi_m20_p44 = pair_set_epi16(-cospi[20], cospi[44]); + const __m128i cospi_p44_p20 = pair_set_epi16(cospi[44], cospi[20]); + const __m128i cospi_m52_p12 = pair_set_epi16(-cospi[52], cospi[12]); + const __m128i cospi_m12_m52 = pair_set_epi16(-cospi[12], -cospi[52]); + const __m128i cospi_m08_p56 = pair_set_epi16(-cospi[8], cospi[56]); + const __m128i cospi_p56_p08 = pair_set_epi16(cospi[56], cospi[8]); + const __m128i cospi_m40_p24 = pair_set_epi16(-cospi[40], cospi[24]); + const __m128i cospi_m24_m40 = pair_set_epi16(-cospi[24], -cospi[40]); + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + const __m128i cospi_m16_p48 = pair_set_epi16(-cospi[16], cospi[48]); + const __m128i cospi_p48_p16 = pair_set_epi16(cospi[48], cospi[16]); + const __m128i cospi_m32_p32 = pair_set_epi16(-cospi[32], cospi[32]); + + // stage 1 + __m128i x[64]; + x[0] = input[0]; + x[8] = input[4]; + x[16] = input[2]; + x[24] = input[6]; + x[32] = input[1]; + x[40] = input[5]; + x[48] = input[3]; + x[56] = input[7]; + + // stage 2 + btf_16_ssse3(cospi[63], cospi[1], x[32], x[32], x[63]); + btf_16_ssse3(-cospi[57], cospi[7], x[56], x[39], x[56]); + btf_16_ssse3(cospi[59], cospi[5], x[40], x[40], x[55]); + btf_16_ssse3(-cospi[61], cospi[3], x[48], x[47], x[48]); + + // stage 3 + btf_16_ssse3(cospi[62], cospi[2], x[16], x[16], x[31]); + btf_16_ssse3(-cospi[58], cospi[6], x[24], x[23], x[24]); + x[33] = x[32]; + x[38] = x[39]; + x[41] = x[40]; + x[46] = x[47]; + x[49] = x[48]; + x[54] = x[55]; + x[57] = x[56]; + x[62] = x[63]; + + // stage 4 + btf_16_ssse3(cospi[60], cospi[4], x[8], x[8], x[15]); + x[17] = x[16]; + x[22] = x[23]; + x[25] = x[24]; + x[30] = x[31]; + btf_16_sse2(cospi_m04_p60, cospi_p60_p04, x[33], x[62], x[33], x[62]); + btf_16_sse2(cospi_m28_m36, cospi_m36_p28, x[38], x[57], x[38], x[57]); + btf_16_sse2(cospi_m20_p44, cospi_p44_p20, x[41], x[54], x[41], x[54]); + btf_16_sse2(cospi_m12_m52, cospi_m52_p12, x[46], x[49], x[46], x[49]); + + // stage 5 + x[9] = x[8]; + x[14] = x[15]; + btf_16_sse2(cospi_m08_p56, cospi_p56_p08, x[17], x[30], x[17], x[30]); + btf_16_sse2(cospi_m24_m40, cospi_m40_p24, x[22], x[25], x[22], x[25]); + x[35] = x[32]; + x[34] = x[33]; + x[36] = x[39]; + x[37] = x[38]; + x[43] = x[40]; + x[42] = x[41]; + x[44] = x[47]; + x[45] = x[46]; + x[51] = x[48]; + x[50] = x[49]; + x[52] = x[55]; + x[53] = x[54]; + x[59] = x[56]; + x[58] = x[57]; + x[60] = x[63]; + x[61] = x[62]; + + // stage 6 + btf_16_ssse3(cospi[32], cospi[32], x[0], x[0], x[1]); + btf_16_sse2(cospi_m16_p48, cospi_p48_p16, x[9], x[14], x[9], x[14]); + x[19] = x[16]; + x[18] = x[17]; + x[20] = x[23]; + x[21] = x[22]; + x[27] = x[24]; + x[26] = x[25]; + x[28] = x[31]; + x[29] = x[30]; + idct64_stage6_high32_sse2(x, cospi, __rounding, cos_bit); + + // stage 7 + x[3] = x[0]; + x[2] = x[1]; + x[11] = x[8]; + x[10] = x[9]; + x[12] = x[15]; + x[13] = x[14]; + idct64_stage7_high48_sse2(x, cospi, __rounding, cos_bit); + + // stage 8 + x[7] = x[0]; + x[6] = x[1]; + x[5] = x[2]; + x[4] = x[3]; + x[9] = x[9]; + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[10], x[13], x[10], x[13]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[11], x[12], x[11], x[12]); + idct64_stage8_high48_sse2(x, cospi, __rounding, cos_bit); + + idct64_stage9_sse2(x, cospi, __rounding, cos_bit); + idct64_stage10_sse2(x, cospi, __rounding, cos_bit); + idct64_stage11_sse2(output, x); +} + +static void idct64_low16_new_ssse3(const __m128i *input, __m128i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m128i __rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + const __m128i cospi_m16_p48 = pair_set_epi16(-cospi[16], cospi[48]); + const __m128i cospi_p48_p16 = pair_set_epi16(cospi[48], cospi[16]); + const __m128i cospi_m48_m16 = pair_set_epi16(-cospi[48], -cospi[16]); + const __m128i cospi_m32_p32 = pair_set_epi16(-cospi[32], cospi[32]); + + // stage 1 + __m128i x[64]; + x[0] = input[0]; + x[4] = input[8]; + x[8] = input[4]; + x[12] = input[12]; + x[16] = input[2]; + x[20] = input[10]; + x[24] = input[6]; + x[28] = input[14]; + x[32] = input[1]; + x[36] = input[9]; + x[40] = input[5]; + x[44] = input[13]; + x[48] = input[3]; + x[52] = input[11]; + x[56] = input[7]; + x[60] = input[15]; + + // stage 2 + btf_16_ssse3(cospi[63], cospi[1], x[32], x[32], x[63]); + btf_16_ssse3(-cospi[49], cospi[15], x[60], x[35], x[60]); + btf_16_ssse3(cospi[55], cospi[9], x[36], x[36], x[59]); + btf_16_ssse3(-cospi[57], cospi[7], x[56], x[39], x[56]); + btf_16_ssse3(cospi[59], cospi[5], x[40], x[40], x[55]); + btf_16_ssse3(-cospi[53], cospi[11], x[52], x[43], x[52]); + btf_16_ssse3(cospi[51], cospi[13], x[44], x[44], x[51]); + btf_16_ssse3(-cospi[61], cospi[3], x[48], x[47], x[48]); + + // stage 3 + btf_16_ssse3(cospi[62], cospi[2], x[16], x[16], x[31]); + btf_16_ssse3(-cospi[50], cospi[14], x[28], x[19], x[28]); + btf_16_ssse3(cospi[54], cospi[10], x[20], x[20], x[27]); + btf_16_ssse3(-cospi[58], cospi[6], x[24], x[23], x[24]); + x[33] = x[32]; + x[34] = x[35]; + x[37] = x[36]; + x[38] = x[39]; + x[41] = x[40]; + x[42] = x[43]; + x[45] = x[44]; + x[46] = x[47]; + x[49] = x[48]; + x[50] = x[51]; + x[53] = x[52]; + x[54] = x[55]; + x[57] = x[56]; + x[58] = x[59]; + x[61] = x[60]; + x[62] = x[63]; + + // stage 4 + btf_16_ssse3(cospi[60], cospi[4], x[8], x[8], x[15]); + btf_16_ssse3(-cospi[52], cospi[12], x[12], x[11], x[12]); + x[17] = x[16]; + x[18] = x[19]; + x[21] = x[20]; + x[22] = x[23]; + x[25] = x[24]; + x[26] = x[27]; + x[29] = x[28]; + x[30] = x[31]; + idct64_stage4_high32_sse2(x, cospi, __rounding, cos_bit); + + // stage 5 + btf_16_ssse3(cospi[56], cospi[8], x[4], x[4], x[7]); + x[9] = x[8]; + x[10] = x[11]; + x[13] = x[12]; + x[14] = x[15]; + idct64_stage5_high48_sse2(x, cospi, __rounding, cos_bit); + + // stage 6 + btf_16_ssse3(cospi[32], cospi[32], x[0], x[0], x[1]); + x[5] = x[4]; + x[6] = x[7]; + btf_16_sse2(cospi_m16_p48, cospi_p48_p16, x[9], x[14], x[9], x[14]); + btf_16_sse2(cospi_m48_m16, cospi_m16_p48, x[10], x[13], x[10], x[13]); + idct64_stage6_high48_sse2(x, cospi, __rounding, cos_bit); + + // stage 7 + x[3] = x[0]; + x[2] = x[1]; + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[5], x[6], x[5], x[6]); + btf_16_adds_subs_sse2(x[8], x[11]); + btf_16_adds_subs_sse2(x[9], x[10]); + btf_16_subs_adds_sse2(x[15], x[12]); + btf_16_subs_adds_sse2(x[14], x[13]); + idct64_stage7_high48_sse2(x, cospi, __rounding, cos_bit); + + // stage 8 + btf_16_adds_subs_sse2(x[0], x[7]); + btf_16_adds_subs_sse2(x[1], x[6]); + btf_16_adds_subs_sse2(x[2], x[5]); + btf_16_adds_subs_sse2(x[3], x[4]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[10], x[13], x[10], x[13]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[11], x[12], x[11], x[12]); + idct64_stage8_high48_sse2(x, cospi, __rounding, cos_bit); + + idct64_stage9_sse2(x, cospi, __rounding, cos_bit); + idct64_stage10_sse2(x, cospi, __rounding, cos_bit); + idct64_stage11_sse2(output, x); +} + +static void idct64_low32_new_ssse3(const __m128i *input, __m128i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m128i __rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + const __m128i cospi_m16_p48 = pair_set_epi16(-cospi[16], cospi[48]); + const __m128i cospi_p48_p16 = pair_set_epi16(cospi[48], cospi[16]); + const __m128i cospi_m48_m16 = pair_set_epi16(-cospi[48], -cospi[16]); + const __m128i cospi_m32_p32 = pair_set_epi16(-cospi[32], cospi[32]); + + // stage 1 + __m128i x[64]; + x[0] = input[0]; + x[2] = input[16]; + x[4] = input[8]; + x[6] = input[24]; + x[8] = input[4]; + x[10] = input[20]; + x[12] = input[12]; + x[14] = input[28]; + x[16] = input[2]; + x[18] = input[18]; + x[20] = input[10]; + x[22] = input[26]; + x[24] = input[6]; + x[26] = input[22]; + x[28] = input[14]; + x[30] = input[30]; + x[32] = input[1]; + x[34] = input[17]; + x[36] = input[9]; + x[38] = input[25]; + x[40] = input[5]; + x[42] = input[21]; + x[44] = input[13]; + x[46] = input[29]; + x[48] = input[3]; + x[50] = input[19]; + x[52] = input[11]; + x[54] = input[27]; + x[56] = input[7]; + x[58] = input[23]; + x[60] = input[15]; + x[62] = input[31]; + + // stage 2 + btf_16_ssse3(cospi[63], cospi[1], x[32], x[32], x[63]); + btf_16_ssse3(-cospi[33], cospi[31], x[62], x[33], x[62]); + btf_16_ssse3(cospi[47], cospi[17], x[34], x[34], x[61]); + btf_16_ssse3(-cospi[49], cospi[15], x[60], x[35], x[60]); + btf_16_ssse3(cospi[55], cospi[9], x[36], x[36], x[59]); + btf_16_ssse3(-cospi[41], cospi[23], x[58], x[37], x[58]); + btf_16_ssse3(cospi[39], cospi[25], x[38], x[38], x[57]); + btf_16_ssse3(-cospi[57], cospi[7], x[56], x[39], x[56]); + btf_16_ssse3(cospi[59], cospi[5], x[40], x[40], x[55]); + btf_16_ssse3(-cospi[37], cospi[27], x[54], x[41], x[54]); + btf_16_ssse3(cospi[43], cospi[21], x[42], x[42], x[53]); + btf_16_ssse3(-cospi[53], cospi[11], x[52], x[43], x[52]); + btf_16_ssse3(cospi[51], cospi[13], x[44], x[44], x[51]); + btf_16_ssse3(-cospi[45], cospi[19], x[50], x[45], x[50]); + btf_16_ssse3(cospi[35], cospi[29], x[46], x[46], x[49]); + btf_16_ssse3(-cospi[61], cospi[3], x[48], x[47], x[48]); + + // stage 3 + btf_16_ssse3(cospi[62], cospi[2], x[16], x[16], x[31]); + btf_16_ssse3(-cospi[34], cospi[30], x[30], x[17], x[30]); + btf_16_ssse3(cospi[46], cospi[18], x[18], x[18], x[29]); + btf_16_ssse3(-cospi[50], cospi[14], x[28], x[19], x[28]); + btf_16_ssse3(cospi[54], cospi[10], x[20], x[20], x[27]); + btf_16_ssse3(-cospi[42], cospi[22], x[26], x[21], x[26]); + btf_16_ssse3(cospi[38], cospi[26], x[22], x[22], x[25]); + btf_16_ssse3(-cospi[58], cospi[6], x[24], x[23], x[24]); + btf_16_adds_subs_sse2(x[32], x[33]); + btf_16_subs_adds_sse2(x[35], x[34]); + btf_16_adds_subs_sse2(x[36], x[37]); + btf_16_subs_adds_sse2(x[39], x[38]); + btf_16_adds_subs_sse2(x[40], x[41]); + btf_16_subs_adds_sse2(x[43], x[42]); + btf_16_adds_subs_sse2(x[44], x[45]); + btf_16_subs_adds_sse2(x[47], x[46]); + btf_16_adds_subs_sse2(x[48], x[49]); + btf_16_subs_adds_sse2(x[51], x[50]); + btf_16_adds_subs_sse2(x[52], x[53]); + btf_16_subs_adds_sse2(x[55], x[54]); + btf_16_adds_subs_sse2(x[56], x[57]); + btf_16_subs_adds_sse2(x[59], x[58]); + btf_16_adds_subs_sse2(x[60], x[61]); + btf_16_subs_adds_sse2(x[63], x[62]); + + // stage 4 + btf_16_ssse3(cospi[60], cospi[4], x[8], x[8], x[15]); + btf_16_ssse3(-cospi[36], cospi[28], x[14], x[9], x[14]); + btf_16_ssse3(cospi[44], cospi[20], x[10], x[10], x[13]); + btf_16_ssse3(-cospi[52], cospi[12], x[12], x[11], x[12]); + btf_16_adds_subs_sse2(x[16], x[17]); + btf_16_subs_adds_sse2(x[19], x[18]); + btf_16_adds_subs_sse2(x[20], x[21]); + btf_16_subs_adds_sse2(x[23], x[22]); + btf_16_adds_subs_sse2(x[24], x[25]); + btf_16_subs_adds_sse2(x[27], x[26]); + btf_16_adds_subs_sse2(x[28], x[29]); + btf_16_subs_adds_sse2(x[31], x[30]); + idct64_stage4_high32_sse2(x, cospi, __rounding, cos_bit); + + // stage 5 + btf_16_ssse3(cospi[56], cospi[8], x[4], x[4], x[7]); + btf_16_ssse3(-cospi[40], cospi[24], x[6], x[5], x[6]); + btf_16_adds_subs_sse2(x[8], x[9]); + btf_16_subs_adds_sse2(x[11], x[10]); + btf_16_adds_subs_sse2(x[12], x[13]); + btf_16_subs_adds_sse2(x[15], x[14]); + idct64_stage5_high48_sse2(x, cospi, __rounding, cos_bit); + + // stage 6 + btf_16_ssse3(cospi[32], cospi[32], x[0], x[0], x[1]); + btf_16_ssse3(cospi[48], cospi[16], x[2], x[2], x[3]); + btf_16_adds_subs_sse2(x[4], x[5]); + btf_16_subs_adds_sse2(x[7], x[6]); + btf_16_sse2(cospi_m16_p48, cospi_p48_p16, x[9], x[14], x[9], x[14]); + btf_16_sse2(cospi_m48_m16, cospi_m16_p48, x[10], x[13], x[10], x[13]); + idct64_stage6_high48_sse2(x, cospi, __rounding, cos_bit); + + // stage 7 + btf_16_adds_subs_sse2(x[0], x[3]); + btf_16_adds_subs_sse2(x[1], x[2]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[5], x[6], x[5], x[6]); + btf_16_adds_subs_sse2(x[8], x[11]); + btf_16_adds_subs_sse2(x[9], x[10]); + btf_16_subs_adds_sse2(x[15], x[12]); + btf_16_subs_adds_sse2(x[14], x[13]); + idct64_stage7_high48_sse2(x, cospi, __rounding, cos_bit); + + // stage 8 + btf_16_adds_subs_sse2(x[0], x[7]); + btf_16_adds_subs_sse2(x[1], x[6]); + btf_16_adds_subs_sse2(x[2], x[5]); + btf_16_adds_subs_sse2(x[3], x[4]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[10], x[13], x[10], x[13]); + btf_16_sse2(cospi_m32_p32, cospi_p32_p32, x[11], x[12], x[11], x[12]); + idct64_stage8_high48_sse2(x, cospi, __rounding, cos_bit); + + // stage 9~11 + idct64_stage9_sse2(x, cospi, __rounding, cos_bit); + idct64_stage10_sse2(x, cospi, __rounding, cos_bit); + idct64_stage11_sse2(output, x); +} + +void iadst4_new_sse2(const __m128i *input, __m128i *output, int8_t cos_bit) { + (void)cos_bit; + const int32_t *sinpi = sinpi_arr(INV_COS_BIT); + const __m128i sinpi_p01_p04 = pair_set_epi16(sinpi[1], sinpi[4]); + const __m128i sinpi_p02_m01 = pair_set_epi16(sinpi[2], -sinpi[1]); + const __m128i sinpi_p03_p02 = pair_set_epi16(sinpi[3], sinpi[2]); + const __m128i sinpi_p03_m04 = pair_set_epi16(sinpi[3], -sinpi[4]); + const __m128i sinpi_p03_m03 = pair_set_epi16(sinpi[3], -sinpi[3]); + const __m128i sinpi_0_p03 = pair_set_epi16(0, sinpi[3]); + const __m128i sinpi_p04_p02 = pair_set_epi16(sinpi[4], sinpi[2]); + const __m128i sinpi_m03_m01 = pair_set_epi16(-sinpi[3], -sinpi[1]); + __m128i x0[4]; + x0[0] = input[0]; + x0[1] = input[1]; + x0[2] = input[2]; + x0[3] = input[3]; + + __m128i u[4]; + u[0] = _mm_unpacklo_epi16(x0[0], x0[2]); + u[1] = _mm_unpackhi_epi16(x0[0], x0[2]); + u[2] = _mm_unpacklo_epi16(x0[1], x0[3]); + u[3] = _mm_unpackhi_epi16(x0[1], x0[3]); + + __m128i x1[16]; + x1[0] = _mm_madd_epi16(u[0], sinpi_p01_p04); // x0*sin1 + x2*sin4 + x1[1] = _mm_madd_epi16(u[1], sinpi_p01_p04); + x1[2] = _mm_madd_epi16(u[0], sinpi_p02_m01); // x0*sin2 - x2*sin1 + x1[3] = _mm_madd_epi16(u[1], sinpi_p02_m01); + x1[4] = _mm_madd_epi16(u[2], sinpi_p03_p02); // x1*sin3 + x3*sin2 + x1[5] = _mm_madd_epi16(u[3], sinpi_p03_p02); + x1[6] = _mm_madd_epi16(u[2], sinpi_p03_m04); // x1*sin3 - x3*sin4 + x1[7] = _mm_madd_epi16(u[3], sinpi_p03_m04); + x1[8] = _mm_madd_epi16(u[0], sinpi_p03_m03); // x0*sin3 - x2*sin3 + x1[9] = _mm_madd_epi16(u[1], sinpi_p03_m03); + x1[10] = _mm_madd_epi16(u[2], sinpi_0_p03); // x2*sin3 + x1[11] = _mm_madd_epi16(u[3], sinpi_0_p03); + x1[12] = _mm_madd_epi16(u[0], sinpi_p04_p02); // x0*sin4 + x2*sin2 + x1[13] = _mm_madd_epi16(u[1], sinpi_p04_p02); + x1[14] = _mm_madd_epi16(u[2], sinpi_m03_m01); // -x1*sin3 - x3*sin1 + x1[15] = _mm_madd_epi16(u[3], sinpi_m03_m01); + + __m128i x2[8]; + x2[0] = _mm_add_epi32(x1[0], x1[4]); // x0*sin1 +x2*sin4 +x1*sin3 +x3*sin2 + x2[1] = _mm_add_epi32(x1[1], x1[5]); + x2[2] = _mm_add_epi32(x1[2], x1[6]); // x0*sin2 -x2*sin1 +x1*sin3 -x3*sin4 + x2[3] = _mm_add_epi32(x1[3], x1[7]); + x2[4] = _mm_add_epi32(x1[8], x1[10]); // x0*sin3 -x2*sin3 +x3*sin3 + x2[5] = _mm_add_epi32(x1[9], x1[11]); + x2[6] = _mm_add_epi32(x1[12], x1[14]); // x0*sin1 +x2*sin4 +x0*sin2 -x2*sin1 + x2[7] = _mm_add_epi32(x1[13], x1[15]); + + const __m128i rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + for (int i = 0; i < 4; ++i) { + __m128i out0 = _mm_add_epi32(x2[2 * i], rounding); + __m128i out1 = _mm_add_epi32(x2[2 * i + 1], rounding); + out0 = _mm_srai_epi32(out0, INV_COS_BIT); + out1 = _mm_srai_epi32(out1, INV_COS_BIT); + output[i] = _mm_packs_epi32(out0, out1); + } +} + +// TODO(binpengsmail@gmail.com): +// To explore the reuse of VP9 versions of corresponding SSE2 functions and +// evaluate whether there is a possibility for further speedup. +void iadst4_w4_new_sse2(const __m128i *input, __m128i *output, int8_t cos_bit) { + (void)cos_bit; + const int32_t *sinpi = sinpi_arr(INV_COS_BIT); + const __m128i sinpi_p01_p04 = pair_set_epi16(sinpi[1], sinpi[4]); + const __m128i sinpi_p02_m01 = pair_set_epi16(sinpi[2], -sinpi[1]); + const __m128i sinpi_p03_p02 = pair_set_epi16(sinpi[3], sinpi[2]); + const __m128i sinpi_p03_m04 = pair_set_epi16(sinpi[3], -sinpi[4]); + const __m128i sinpi_p03_m03 = pair_set_epi16(sinpi[3], -sinpi[3]); + const __m128i sinpi_0_p03 = pair_set_epi16(0, sinpi[3]); + const __m128i sinpi_p04_p02 = pair_set_epi16(sinpi[4], sinpi[2]); + const __m128i sinpi_m03_m01 = pair_set_epi16(-sinpi[3], -sinpi[1]); + __m128i x0[4]; + x0[0] = input[0]; + x0[1] = input[1]; + x0[2] = input[2]; + x0[3] = input[3]; + + __m128i u[2]; + u[0] = _mm_unpacklo_epi16(x0[0], x0[2]); + u[1] = _mm_unpacklo_epi16(x0[1], x0[3]); + + __m128i x1[8]; + x1[0] = _mm_madd_epi16(u[0], sinpi_p01_p04); // x0*sin1 + x2*sin4 + x1[1] = _mm_madd_epi16(u[0], sinpi_p02_m01); // x0*sin2 - x2*sin1 + x1[2] = _mm_madd_epi16(u[1], sinpi_p03_p02); // x1*sin3 + x3*sin2 + x1[3] = _mm_madd_epi16(u[1], sinpi_p03_m04); // x1*sin3 - x3*sin4 + x1[4] = _mm_madd_epi16(u[0], sinpi_p03_m03); // x0*sin3 - x2*sin3 + x1[5] = _mm_madd_epi16(u[1], sinpi_0_p03); // x2*sin3 + x1[6] = _mm_madd_epi16(u[0], sinpi_p04_p02); // x0*sin4 + x2*sin2 + x1[7] = _mm_madd_epi16(u[1], sinpi_m03_m01); // -x1*sin3 - x3*sin1 + + __m128i x2[4]; + x2[0] = _mm_add_epi32(x1[0], x1[2]); // x0*sin1 + x2*sin4 + x1*sin3 + x3*sin2 + x2[1] = _mm_add_epi32(x1[1], x1[3]); // x0*sin2 - x2*sin1 + x1*sin3 - x3*sin4 + x2[2] = _mm_add_epi32(x1[4], x1[5]); // x0*sin3 - x2*sin3 + x3*sin3 + x2[3] = _mm_add_epi32(x1[6], x1[7]); // x0*sin4 + x2*sin2 - x1*sin3 - x3*sin1 + + const __m128i rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + for (int i = 0; i < 4; ++i) { + __m128i out0 = _mm_add_epi32(x2[i], rounding); + out0 = _mm_srai_epi32(out0, INV_COS_BIT); + output[i] = _mm_packs_epi32(out0, out0); + } +} + +static void iadst8_low1_new_ssse3(const __m128i *input, __m128i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m128i __zero = _mm_setzero_si128(); + const __m128i __rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + + const __m128i cospi_p16_p48 = pair_set_epi16(cospi[16], cospi[48]); + const __m128i cospi_p48_m16 = pair_set_epi16(cospi[48], -cospi[16]); + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + const __m128i cospi_p32_m32 = pair_set_epi16(cospi[32], -cospi[32]); + + // stage 1 + __m128i x[8]; + x[1] = input[0]; + + // stage 2 + btf_16_ssse3(cospi[60], -cospi[4], x[1], x[0], x[1]); + + // stage 3 + x[4] = x[0]; + x[5] = x[1]; + + // stage 4 + btf_16_sse2(cospi_p16_p48, cospi_p48_m16, x[4], x[5], x[4], x[5]); + + // stage 5 + x[2] = x[0]; + x[3] = x[1]; + x[6] = x[4]; + x[7] = x[5]; + + // stage 6 + btf_16_sse2(cospi_p32_p32, cospi_p32_m32, x[2], x[3], x[2], x[3]); + btf_16_sse2(cospi_p32_p32, cospi_p32_m32, x[6], x[7], x[6], x[7]); + + // stage 7 + output[0] = x[0]; + output[1] = _mm_subs_epi16(__zero, x[4]); + output[2] = x[6]; + output[3] = _mm_subs_epi16(__zero, x[2]); + output[4] = x[3]; + output[5] = _mm_subs_epi16(__zero, x[7]); + output[6] = x[5]; + output[7] = _mm_subs_epi16(__zero, x[1]); +} + +void iadst8_new_sse2(const __m128i *input, __m128i *output, int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m128i __zero = _mm_setzero_si128(); + const __m128i __rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + + const __m128i cospi_p04_p60 = pair_set_epi16(cospi[4], cospi[60]); + const __m128i cospi_p60_m04 = pair_set_epi16(cospi[60], -cospi[4]); + const __m128i cospi_p20_p44 = pair_set_epi16(cospi[20], cospi[44]); + const __m128i cospi_p44_m20 = pair_set_epi16(cospi[44], -cospi[20]); + const __m128i cospi_p36_p28 = pair_set_epi16(cospi[36], cospi[28]); + const __m128i cospi_p28_m36 = pair_set_epi16(cospi[28], -cospi[36]); + const __m128i cospi_p52_p12 = pair_set_epi16(cospi[52], cospi[12]); + const __m128i cospi_p12_m52 = pair_set_epi16(cospi[12], -cospi[52]); + const __m128i cospi_p16_p48 = pair_set_epi16(cospi[16], cospi[48]); + const __m128i cospi_p48_m16 = pair_set_epi16(cospi[48], -cospi[16]); + const __m128i cospi_m48_p16 = pair_set_epi16(-cospi[48], cospi[16]); + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + const __m128i cospi_p32_m32 = pair_set_epi16(cospi[32], -cospi[32]); + + // stage 1 + __m128i x[8]; + x[0] = input[7]; + x[1] = input[0]; + x[2] = input[5]; + x[3] = input[2]; + x[4] = input[3]; + x[5] = input[4]; + x[6] = input[1]; + x[7] = input[6]; + + // stage 2 + btf_16_sse2(cospi_p04_p60, cospi_p60_m04, x[0], x[1], x[0], x[1]); + btf_16_sse2(cospi_p20_p44, cospi_p44_m20, x[2], x[3], x[2], x[3]); + btf_16_sse2(cospi_p36_p28, cospi_p28_m36, x[4], x[5], x[4], x[5]); + btf_16_sse2(cospi_p52_p12, cospi_p12_m52, x[6], x[7], x[6], x[7]); + + // stage 3 + btf_16_adds_subs_sse2(x[0], x[4]); + btf_16_adds_subs_sse2(x[1], x[5]); + btf_16_adds_subs_sse2(x[2], x[6]); + btf_16_adds_subs_sse2(x[3], x[7]); + + // stage 4 + btf_16_sse2(cospi_p16_p48, cospi_p48_m16, x[4], x[5], x[4], x[5]); + btf_16_sse2(cospi_m48_p16, cospi_p16_p48, x[6], x[7], x[6], x[7]); + + // stage 5 + btf_16_adds_subs_sse2(x[0], x[2]); + btf_16_adds_subs_sse2(x[1], x[3]); + btf_16_adds_subs_sse2(x[4], x[6]); + btf_16_adds_subs_sse2(x[5], x[7]); + + // stage 6 + btf_16_sse2(cospi_p32_p32, cospi_p32_m32, x[2], x[3], x[2], x[3]); + btf_16_sse2(cospi_p32_p32, cospi_p32_m32, x[6], x[7], x[6], x[7]); + + // stage 7 + output[0] = x[0]; + output[1] = _mm_subs_epi16(__zero, x[4]); + output[2] = x[6]; + output[3] = _mm_subs_epi16(__zero, x[2]); + output[4] = x[3]; + output[5] = _mm_subs_epi16(__zero, x[7]); + output[6] = x[5]; + output[7] = _mm_subs_epi16(__zero, x[1]); +} + +void iadst8_w4_new_sse2(const __m128i *input, __m128i *output, int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m128i __zero = _mm_setzero_si128(); + const __m128i __rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + + const __m128i cospi_p04_p60 = pair_set_epi16(cospi[4], cospi[60]); + const __m128i cospi_p60_m04 = pair_set_epi16(cospi[60], -cospi[4]); + const __m128i cospi_p20_p44 = pair_set_epi16(cospi[20], cospi[44]); + const __m128i cospi_p44_m20 = pair_set_epi16(cospi[44], -cospi[20]); + const __m128i cospi_p36_p28 = pair_set_epi16(cospi[36], cospi[28]); + const __m128i cospi_p28_m36 = pair_set_epi16(cospi[28], -cospi[36]); + const __m128i cospi_p52_p12 = pair_set_epi16(cospi[52], cospi[12]); + const __m128i cospi_p12_m52 = pair_set_epi16(cospi[12], -cospi[52]); + const __m128i cospi_p16_p48 = pair_set_epi16(cospi[16], cospi[48]); + const __m128i cospi_p48_m16 = pair_set_epi16(cospi[48], -cospi[16]); + const __m128i cospi_m48_p16 = pair_set_epi16(-cospi[48], cospi[16]); + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + const __m128i cospi_p32_m32 = pair_set_epi16(cospi[32], -cospi[32]); + + // stage 1 + __m128i x[8]; + x[0] = input[7]; + x[1] = input[0]; + x[2] = input[5]; + x[3] = input[2]; + x[4] = input[3]; + x[5] = input[4]; + x[6] = input[1]; + x[7] = input[6]; + + // stage 2 + btf_16_4p_sse2(cospi_p04_p60, cospi_p60_m04, x[0], x[1], x[0], x[1]); + btf_16_4p_sse2(cospi_p20_p44, cospi_p44_m20, x[2], x[3], x[2], x[3]); + btf_16_4p_sse2(cospi_p36_p28, cospi_p28_m36, x[4], x[5], x[4], x[5]); + btf_16_4p_sse2(cospi_p52_p12, cospi_p12_m52, x[6], x[7], x[6], x[7]); + + // stage 3 + btf_16_adds_subs_sse2(x[0], x[4]); + btf_16_adds_subs_sse2(x[1], x[5]); + btf_16_adds_subs_sse2(x[2], x[6]); + btf_16_adds_subs_sse2(x[3], x[7]); + + // stage 4 + btf_16_4p_sse2(cospi_p16_p48, cospi_p48_m16, x[4], x[5], x[4], x[5]); + btf_16_4p_sse2(cospi_m48_p16, cospi_p16_p48, x[6], x[7], x[6], x[7]); + + // stage 5 + btf_16_adds_subs_sse2(x[0], x[2]); + btf_16_adds_subs_sse2(x[1], x[3]); + btf_16_adds_subs_sse2(x[4], x[6]); + btf_16_adds_subs_sse2(x[5], x[7]); + + // stage 6 + btf_16_4p_sse2(cospi_p32_p32, cospi_p32_m32, x[2], x[3], x[2], x[3]); + btf_16_4p_sse2(cospi_p32_p32, cospi_p32_m32, x[6], x[7], x[6], x[7]); + + // stage 7 + output[0] = x[0]; + output[1] = _mm_subs_epi16(__zero, x[4]); + output[2] = x[6]; + output[3] = _mm_subs_epi16(__zero, x[2]); + output[4] = x[3]; + output[5] = _mm_subs_epi16(__zero, x[7]); + output[6] = x[5]; + output[7] = _mm_subs_epi16(__zero, x[1]); +} + +static INLINE void iadst16_stage3_ssse3(__m128i *x) { + btf_16_adds_subs_sse2(x[0], x[8]); + btf_16_adds_subs_sse2(x[1], x[9]); + btf_16_adds_subs_sse2(x[2], x[10]); + btf_16_adds_subs_sse2(x[3], x[11]); + btf_16_adds_subs_sse2(x[4], x[12]); + btf_16_adds_subs_sse2(x[5], x[13]); + btf_16_adds_subs_sse2(x[6], x[14]); + btf_16_adds_subs_sse2(x[7], x[15]); +} + +static INLINE void iadst16_stage4_ssse3(__m128i *x, const int32_t *cospi, + const __m128i __rounding, + int8_t cos_bit) { + const __m128i cospi_p08_p56 = pair_set_epi16(cospi[8], cospi[56]); + const __m128i cospi_p56_m08 = pair_set_epi16(cospi[56], -cospi[8]); + const __m128i cospi_p40_p24 = pair_set_epi16(cospi[40], cospi[24]); + const __m128i cospi_p24_m40 = pair_set_epi16(cospi[24], -cospi[40]); + const __m128i cospi_m56_p08 = pair_set_epi16(-cospi[56], cospi[8]); + const __m128i cospi_m24_p40 = pair_set_epi16(-cospi[24], cospi[40]); + btf_16_sse2(cospi_p08_p56, cospi_p56_m08, x[8], x[9], x[8], x[9]); + btf_16_sse2(cospi_p40_p24, cospi_p24_m40, x[10], x[11], x[10], x[11]); + btf_16_sse2(cospi_m56_p08, cospi_p08_p56, x[12], x[13], x[12], x[13]); + btf_16_sse2(cospi_m24_p40, cospi_p40_p24, x[14], x[15], x[14], x[15]); +} + +static INLINE void iadst16_stage5_ssse3(__m128i *x) { + btf_16_adds_subs_sse2(x[0], x[4]); + btf_16_adds_subs_sse2(x[1], x[5]); + btf_16_adds_subs_sse2(x[2], x[6]); + btf_16_adds_subs_sse2(x[3], x[7]); + btf_16_adds_subs_sse2(x[8], x[12]); + btf_16_adds_subs_sse2(x[9], x[13]); + btf_16_adds_subs_sse2(x[10], x[14]); + btf_16_adds_subs_sse2(x[11], x[15]); +} + +static INLINE void iadst16_stage6_ssse3(__m128i *x, const int32_t *cospi, + const __m128i __rounding, + int8_t cos_bit) { + const __m128i cospi_p16_p48 = pair_set_epi16(cospi[16], cospi[48]); + const __m128i cospi_p48_m16 = pair_set_epi16(cospi[48], -cospi[16]); + const __m128i cospi_m48_p16 = pair_set_epi16(-cospi[48], cospi[16]); + btf_16_sse2(cospi_p16_p48, cospi_p48_m16, x[4], x[5], x[4], x[5]); + btf_16_sse2(cospi_m48_p16, cospi_p16_p48, x[6], x[7], x[6], x[7]); + btf_16_sse2(cospi_p16_p48, cospi_p48_m16, x[12], x[13], x[12], x[13]); + btf_16_sse2(cospi_m48_p16, cospi_p16_p48, x[14], x[15], x[14], x[15]); +} + +static INLINE void iadst16_stage7_ssse3(__m128i *x) { + btf_16_adds_subs_sse2(x[0], x[2]); + btf_16_adds_subs_sse2(x[1], x[3]); + btf_16_adds_subs_sse2(x[4], x[6]); + btf_16_adds_subs_sse2(x[5], x[7]); + btf_16_adds_subs_sse2(x[8], x[10]); + btf_16_adds_subs_sse2(x[9], x[11]); + btf_16_adds_subs_sse2(x[12], x[14]); + btf_16_adds_subs_sse2(x[13], x[15]); +} + +static INLINE void iadst16_stage8_ssse3(__m128i *x, const int32_t *cospi, + const __m128i __rounding, + int8_t cos_bit) { + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + const __m128i cospi_p32_m32 = pair_set_epi16(cospi[32], -cospi[32]); + btf_16_sse2(cospi_p32_p32, cospi_p32_m32, x[2], x[3], x[2], x[3]); + btf_16_sse2(cospi_p32_p32, cospi_p32_m32, x[6], x[7], x[6], x[7]); + btf_16_sse2(cospi_p32_p32, cospi_p32_m32, x[10], x[11], x[10], x[11]); + btf_16_sse2(cospi_p32_p32, cospi_p32_m32, x[14], x[15], x[14], x[15]); +} + +static INLINE void iadst16_stage9_ssse3(__m128i *output, __m128i *x) { + const __m128i __zero = _mm_setzero_si128(); + output[0] = x[0]; + output[1] = _mm_subs_epi16(__zero, x[8]); + output[2] = x[12]; + output[3] = _mm_subs_epi16(__zero, x[4]); + output[4] = x[6]; + output[5] = _mm_subs_epi16(__zero, x[14]); + output[6] = x[10]; + output[7] = _mm_subs_epi16(__zero, x[2]); + output[8] = x[3]; + output[9] = _mm_subs_epi16(__zero, x[11]); + output[10] = x[15]; + output[11] = _mm_subs_epi16(__zero, x[7]); + output[12] = x[5]; + output[13] = _mm_subs_epi16(__zero, x[13]); + output[14] = x[9]; + output[15] = _mm_subs_epi16(__zero, x[1]); +} + +static void iadst16_low1_new_ssse3(const __m128i *input, __m128i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m128i __rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + + const __m128i cospi_p08_p56 = pair_set_epi16(cospi[8], cospi[56]); + const __m128i cospi_p56_m08 = pair_set_epi16(cospi[56], -cospi[8]); + const __m128i cospi_p16_p48 = pair_set_epi16(cospi[16], cospi[48]); + const __m128i cospi_p48_m16 = pair_set_epi16(cospi[48], -cospi[16]); + + // stage 1 + __m128i x[16]; + x[1] = input[0]; + + // stage 2 + btf_16_ssse3(cospi[62], -cospi[2], x[1], x[0], x[1]); + + // stage 3 + x[8] = x[0]; + x[9] = x[1]; + + // stage 4 + btf_16_sse2(cospi_p08_p56, cospi_p56_m08, x[8], x[9], x[8], x[9]); + + // stage 5 + x[4] = x[0]; + x[5] = x[1]; + x[12] = x[8]; + x[13] = x[9]; + + // stage 6 + btf_16_sse2(cospi_p16_p48, cospi_p48_m16, x[4], x[5], x[4], x[5]); + btf_16_sse2(cospi_p16_p48, cospi_p48_m16, x[12], x[13], x[12], x[13]); + + // stage 7 + x[2] = x[0]; + x[3] = x[1]; + x[6] = x[4]; + x[7] = x[5]; + x[10] = x[8]; + x[11] = x[9]; + x[14] = x[12]; + x[15] = x[13]; + + iadst16_stage8_ssse3(x, cospi, __rounding, cos_bit); + iadst16_stage9_ssse3(output, x); +} + +static void iadst16_low8_new_ssse3(const __m128i *input, __m128i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m128i __rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + + // stage 1 + __m128i x[16]; + x[1] = input[0]; + x[3] = input[2]; + x[5] = input[4]; + x[7] = input[6]; + x[8] = input[7]; + x[10] = input[5]; + x[12] = input[3]; + x[14] = input[1]; + + // stage 2 + btf_16_ssse3(cospi[62], -cospi[2], x[1], x[0], x[1]); + btf_16_ssse3(cospi[54], -cospi[10], x[3], x[2], x[3]); + btf_16_ssse3(cospi[46], -cospi[18], x[5], x[4], x[5]); + btf_16_ssse3(cospi[38], -cospi[26], x[7], x[6], x[7]); + btf_16_ssse3(cospi[34], cospi[30], x[8], x[8], x[9]); + btf_16_ssse3(cospi[42], cospi[22], x[10], x[10], x[11]); + btf_16_ssse3(cospi[50], cospi[14], x[12], x[12], x[13]); + btf_16_ssse3(cospi[58], cospi[6], x[14], x[14], x[15]); + + // stage 3 + iadst16_stage3_ssse3(x); + iadst16_stage4_ssse3(x, cospi, __rounding, cos_bit); + iadst16_stage5_ssse3(x); + iadst16_stage6_ssse3(x, cospi, __rounding, cos_bit); + iadst16_stage7_ssse3(x); + iadst16_stage8_ssse3(x, cospi, __rounding, cos_bit); + iadst16_stage9_ssse3(output, x); +} +void iadst16_new_sse2(const __m128i *input, __m128i *output, int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m128i __rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + const __m128i cospi_p02_p62 = pair_set_epi16(cospi[2], cospi[62]); + const __m128i cospi_p62_m02 = pair_set_epi16(cospi[62], -cospi[2]); + const __m128i cospi_p10_p54 = pair_set_epi16(cospi[10], cospi[54]); + const __m128i cospi_p54_m10 = pair_set_epi16(cospi[54], -cospi[10]); + const __m128i cospi_p18_p46 = pair_set_epi16(cospi[18], cospi[46]); + const __m128i cospi_p46_m18 = pair_set_epi16(cospi[46], -cospi[18]); + const __m128i cospi_p26_p38 = pair_set_epi16(cospi[26], cospi[38]); + const __m128i cospi_p38_m26 = pair_set_epi16(cospi[38], -cospi[26]); + const __m128i cospi_p34_p30 = pair_set_epi16(cospi[34], cospi[30]); + const __m128i cospi_p30_m34 = pair_set_epi16(cospi[30], -cospi[34]); + const __m128i cospi_p42_p22 = pair_set_epi16(cospi[42], cospi[22]); + const __m128i cospi_p22_m42 = pair_set_epi16(cospi[22], -cospi[42]); + const __m128i cospi_p50_p14 = pair_set_epi16(cospi[50], cospi[14]); + const __m128i cospi_p14_m50 = pair_set_epi16(cospi[14], -cospi[50]); + const __m128i cospi_p58_p06 = pair_set_epi16(cospi[58], cospi[6]); + const __m128i cospi_p06_m58 = pair_set_epi16(cospi[6], -cospi[58]); + + // stage 1 + __m128i x[16]; + x[0] = input[15]; + x[1] = input[0]; + x[2] = input[13]; + x[3] = input[2]; + x[4] = input[11]; + x[5] = input[4]; + x[6] = input[9]; + x[7] = input[6]; + x[8] = input[7]; + x[9] = input[8]; + x[10] = input[5]; + x[11] = input[10]; + x[12] = input[3]; + x[13] = input[12]; + x[14] = input[1]; + x[15] = input[14]; + + // stage 2 + btf_16_sse2(cospi_p02_p62, cospi_p62_m02, x[0], x[1], x[0], x[1]); + btf_16_sse2(cospi_p10_p54, cospi_p54_m10, x[2], x[3], x[2], x[3]); + btf_16_sse2(cospi_p18_p46, cospi_p46_m18, x[4], x[5], x[4], x[5]); + btf_16_sse2(cospi_p26_p38, cospi_p38_m26, x[6], x[7], x[6], x[7]); + btf_16_sse2(cospi_p34_p30, cospi_p30_m34, x[8], x[9], x[8], x[9]); + btf_16_sse2(cospi_p42_p22, cospi_p22_m42, x[10], x[11], x[10], x[11]); + btf_16_sse2(cospi_p50_p14, cospi_p14_m50, x[12], x[13], x[12], x[13]); + btf_16_sse2(cospi_p58_p06, cospi_p06_m58, x[14], x[15], x[14], x[15]); + + // stage 3~9 + iadst16_stage3_ssse3(x); + iadst16_stage4_ssse3(x, cospi, __rounding, cos_bit); + iadst16_stage5_ssse3(x); + iadst16_stage6_ssse3(x, cospi, __rounding, cos_bit); + iadst16_stage7_ssse3(x); + iadst16_stage8_ssse3(x, cospi, __rounding, cos_bit); + iadst16_stage9_ssse3(output, x); +} + +void iadst16_w4_new_sse2(const __m128i *input, __m128i *output, + int8_t cos_bit) { + (void)cos_bit; + const int32_t *cospi = cospi_arr(INV_COS_BIT); + const __m128i __rounding = _mm_set1_epi32(1 << (INV_COS_BIT - 1)); + + const __m128i cospi_p02_p62 = pair_set_epi16(cospi[2], cospi[62]); + const __m128i cospi_p62_m02 = pair_set_epi16(cospi[62], -cospi[2]); + const __m128i cospi_p10_p54 = pair_set_epi16(cospi[10], cospi[54]); + const __m128i cospi_p54_m10 = pair_set_epi16(cospi[54], -cospi[10]); + const __m128i cospi_p18_p46 = pair_set_epi16(cospi[18], cospi[46]); + const __m128i cospi_p46_m18 = pair_set_epi16(cospi[46], -cospi[18]); + const __m128i cospi_p26_p38 = pair_set_epi16(cospi[26], cospi[38]); + const __m128i cospi_p38_m26 = pair_set_epi16(cospi[38], -cospi[26]); + const __m128i cospi_p34_p30 = pair_set_epi16(cospi[34], cospi[30]); + const __m128i cospi_p30_m34 = pair_set_epi16(cospi[30], -cospi[34]); + const __m128i cospi_p42_p22 = pair_set_epi16(cospi[42], cospi[22]); + const __m128i cospi_p22_m42 = pair_set_epi16(cospi[22], -cospi[42]); + const __m128i cospi_p50_p14 = pair_set_epi16(cospi[50], cospi[14]); + const __m128i cospi_p14_m50 = pair_set_epi16(cospi[14], -cospi[50]); + const __m128i cospi_p58_p06 = pair_set_epi16(cospi[58], cospi[6]); + const __m128i cospi_p06_m58 = pair_set_epi16(cospi[6], -cospi[58]); + const __m128i cospi_p08_p56 = pair_set_epi16(cospi[8], cospi[56]); + const __m128i cospi_p56_m08 = pair_set_epi16(cospi[56], -cospi[8]); + const __m128i cospi_p40_p24 = pair_set_epi16(cospi[40], cospi[24]); + const __m128i cospi_p24_m40 = pair_set_epi16(cospi[24], -cospi[40]); + const __m128i cospi_m56_p08 = pair_set_epi16(-cospi[56], cospi[8]); + const __m128i cospi_m24_p40 = pair_set_epi16(-cospi[24], cospi[40]); + const __m128i cospi_p16_p48 = pair_set_epi16(cospi[16], cospi[48]); + const __m128i cospi_p48_m16 = pair_set_epi16(cospi[48], -cospi[16]); + const __m128i cospi_m48_p16 = pair_set_epi16(-cospi[48], cospi[16]); + const __m128i cospi_p32_p32 = pair_set_epi16(cospi[32], cospi[32]); + const __m128i cospi_p32_m32 = pair_set_epi16(cospi[32], -cospi[32]); + + // stage 1 + __m128i x[16]; + x[0] = input[15]; + x[1] = input[0]; + x[2] = input[13]; + x[3] = input[2]; + x[4] = input[11]; + x[5] = input[4]; + x[6] = input[9]; + x[7] = input[6]; + x[8] = input[7]; + x[9] = input[8]; + x[10] = input[5]; + x[11] = input[10]; + x[12] = input[3]; + x[13] = input[12]; + x[14] = input[1]; + x[15] = input[14]; + + // stage 2 + btf_16_4p_sse2(cospi_p02_p62, cospi_p62_m02, x[0], x[1], x[0], x[1]); + btf_16_4p_sse2(cospi_p10_p54, cospi_p54_m10, x[2], x[3], x[2], x[3]); + btf_16_4p_sse2(cospi_p18_p46, cospi_p46_m18, x[4], x[5], x[4], x[5]); + btf_16_4p_sse2(cospi_p26_p38, cospi_p38_m26, x[6], x[7], x[6], x[7]); + btf_16_4p_sse2(cospi_p34_p30, cospi_p30_m34, x[8], x[9], x[8], x[9]); + btf_16_4p_sse2(cospi_p42_p22, cospi_p22_m42, x[10], x[11], x[10], x[11]); + btf_16_4p_sse2(cospi_p50_p14, cospi_p14_m50, x[12], x[13], x[12], x[13]); + btf_16_4p_sse2(cospi_p58_p06, cospi_p06_m58, x[14], x[15], x[14], x[15]); + + // stage 3 + iadst16_stage3_ssse3(x); + + // stage 4 + btf_16_4p_sse2(cospi_p08_p56, cospi_p56_m08, x[8], x[9], x[8], x[9]); + btf_16_4p_sse2(cospi_p40_p24, cospi_p24_m40, x[10], x[11], x[10], x[11]); + btf_16_4p_sse2(cospi_m56_p08, cospi_p08_p56, x[12], x[13], x[12], x[13]); + btf_16_4p_sse2(cospi_m24_p40, cospi_p40_p24, x[14], x[15], x[14], x[15]); + + // stage 5 + iadst16_stage5_ssse3(x); + + // stage 6 + btf_16_4p_sse2(cospi_p16_p48, cospi_p48_m16, x[4], x[5], x[4], x[5]); + btf_16_4p_sse2(cospi_m48_p16, cospi_p16_p48, x[6], x[7], x[6], x[7]); + btf_16_4p_sse2(cospi_p16_p48, cospi_p48_m16, x[12], x[13], x[12], x[13]); + btf_16_4p_sse2(cospi_m48_p16, cospi_p16_p48, x[14], x[15], x[14], x[15]); + + // stage 7 + iadst16_stage7_ssse3(x); + + // stage 8 + btf_16_4p_sse2(cospi_p32_p32, cospi_p32_m32, x[2], x[3], x[2], x[3]); + btf_16_4p_sse2(cospi_p32_p32, cospi_p32_m32, x[6], x[7], x[6], x[7]); + btf_16_4p_sse2(cospi_p32_p32, cospi_p32_m32, x[10], x[11], x[10], x[11]); + btf_16_4p_sse2(cospi_p32_p32, cospi_p32_m32, x[14], x[15], x[14], x[15]); + + // stage 9 + iadst16_stage9_ssse3(output, x); +} + +static void iidentity4_new_ssse3(const __m128i *input, __m128i *output, + int8_t cos_bit) { + (void)cos_bit; + const int16_t scale_fractional = (NewSqrt2 - (1 << NewSqrt2Bits)); + const __m128i scale = _mm_set1_epi16(scale_fractional << (15 - NewSqrt2Bits)); + for (int i = 0; i < 4; ++i) { + __m128i x = _mm_mulhrs_epi16(input[i], scale); + output[i] = _mm_adds_epi16(x, input[i]); + } +} + +static void iidentity8_new_sse2(const __m128i *input, __m128i *output, + int8_t cos_bit) { + (void)cos_bit; + for (int i = 0; i < 8; ++i) { + output[i] = _mm_adds_epi16(input[i], input[i]); + } +} + +static void iidentity16_new_ssse3(const __m128i *input, __m128i *output, + int8_t cos_bit) { + (void)cos_bit; + const int16_t scale_fractional = 2 * (NewSqrt2 - (1 << NewSqrt2Bits)); + const __m128i scale = _mm_set1_epi16(scale_fractional << (15 - NewSqrt2Bits)); + for (int i = 0; i < 16; ++i) { + __m128i x = _mm_mulhrs_epi16(input[i], scale); + __m128i srcx2 = _mm_adds_epi16(input[i], input[i]); + output[i] = _mm_adds_epi16(x, srcx2); + } +} + +static INLINE __m128i lowbd_get_recon_8x8_sse2(const __m128i pred, + __m128i res) { + const __m128i zero = _mm_setzero_si128(); + __m128i x0 = _mm_adds_epi16(res, _mm_unpacklo_epi8(pred, zero)); + return _mm_packus_epi16(x0, x0); +} + +static INLINE void lowbd_write_buffer_4xn_sse2(__m128i *in, uint8_t *output, + int stride, int flipud, + const int height) { + int j = flipud ? (height - 1) : 0; + const int step = flipud ? -1 : 1; + const __m128i zero = _mm_setzero_si128(); + for (int i = 0; i < height; ++i, j += step) { + const __m128i v = _mm_cvtsi32_si128(*((uint32_t *)(output + i * stride))); + __m128i u = _mm_adds_epi16(in[j], _mm_unpacklo_epi8(v, zero)); + u = _mm_packus_epi16(u, zero); + *((uint32_t *)(output + i * stride)) = _mm_cvtsi128_si32(u); + } +} + +static INLINE void lowbd_write_buffer_8xn_sse2(__m128i *in, uint8_t *output, + int stride, int flipud, + const int height) { + int j = flipud ? (height - 1) : 0; + const int step = flipud ? -1 : 1; + for (int i = 0; i < height; ++i, j += step) { + const __m128i v = _mm_loadl_epi64((__m128i const *)(output + i * stride)); + const __m128i u = lowbd_get_recon_8x8_sse2(v, in[j]); + _mm_storel_epi64((__m128i *)(output + i * stride), u); + } +} + +// 1D functions process process 8 pixels at one time. +static const transform_1d_ssse3 + lowbd_txfm_all_1d_w8_arr[TX_SIZES][ITX_TYPES_1D] = { + { idct4_new_sse2, iadst4_new_sse2, iidentity4_new_ssse3 }, + { idct8_new_sse2, iadst8_new_sse2, iidentity8_new_sse2 }, + { idct16_new_sse2, iadst16_new_sse2, iidentity16_new_ssse3 }, + { idct32_new_sse2, NULL, NULL }, + { idct64_low32_new_ssse3, NULL, NULL }, + }; + +// functions for blocks with eob at DC and within +// topleft 8x8, 16x16, 32x32 corner +static const transform_1d_ssse3 + lowbd_txfm_all_1d_zeros_w8_arr[TX_SIZES][ITX_TYPES_1D][4] = { + { + { idct4_new_sse2, idct4_new_sse2, NULL, NULL }, + { iadst4_new_sse2, iadst4_new_sse2, NULL, NULL }, + { iidentity4_new_ssse3, iidentity4_new_ssse3, NULL, NULL }, + }, + { { idct8_low1_new_ssse3, idct8_new_sse2, NULL, NULL }, + { iadst8_low1_new_ssse3, iadst8_new_sse2, NULL, NULL }, + { iidentity8_new_sse2, iidentity8_new_sse2, NULL, NULL } }, + { + { idct16_low1_new_ssse3, idct16_low8_new_ssse3, idct16_new_sse2, + NULL }, + { iadst16_low1_new_ssse3, iadst16_low8_new_ssse3, iadst16_new_sse2, + NULL }, + { NULL, NULL, NULL, NULL }, + }, + { { idct32_low1_new_ssse3, idct32_low8_new_ssse3, idct32_low16_new_ssse3, + idct32_new_sse2 }, + { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL } }, + { { idct64_low1_new_ssse3, idct64_low8_new_ssse3, idct64_low16_new_ssse3, + idct64_low32_new_ssse3 }, + { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL } } + }; + +// 1D functions process process 4 pixels at one time. +// used in 4x4, 4x8, 4x16, 8x4, 16x4 +static const transform_1d_ssse3 + lowbd_txfm_all_1d_w4_arr[TX_SIZES][ITX_TYPES_1D] = { + { idct4_w4_new_sse2, iadst4_w4_new_sse2, iidentity4_new_ssse3 }, + { idct8_w4_new_sse2, iadst8_w4_new_sse2, iidentity8_new_sse2 }, + { idct16_w4_new_sse2, iadst16_w4_new_sse2, iidentity16_new_ssse3 }, + { NULL, NULL, NULL }, + { NULL, NULL, NULL }, + }; + +static INLINE void iidentity_row_8xn_ssse3(__m128i *out, const int32_t *input, + int stride, int shift, int height, + int txw_idx, int rect_type) { + const int32_t *input_row = input; + const __m128i scale = _mm_set1_epi16(NewSqrt2list[txw_idx]); + const __m128i rounding = _mm_set1_epi16((1 << (NewSqrt2Bits - 1)) + + (1 << (NewSqrt2Bits - shift - 1))); + const __m128i one = _mm_set1_epi16(1); + const __m128i scale_rounding = _mm_unpacklo_epi16(scale, rounding); + if (rect_type != 1 && rect_type != -1) { + for (int i = 0; i < height; ++i) { + const __m128i src = load_32bit_to_16bit(input_row); + input_row += stride; + __m128i lo = _mm_unpacklo_epi16(src, one); + __m128i hi = _mm_unpackhi_epi16(src, one); + lo = _mm_madd_epi16(lo, scale_rounding); + hi = _mm_madd_epi16(hi, scale_rounding); + lo = _mm_srai_epi32(lo, NewSqrt2Bits - shift); + hi = _mm_srai_epi32(hi, NewSqrt2Bits - shift); + out[i] = _mm_packs_epi32(lo, hi); + } + } else { + const __m128i rect_scale = + _mm_set1_epi16(NewInvSqrt2 << (15 - NewSqrt2Bits)); + for (int i = 0; i < height; ++i) { + __m128i src = load_32bit_to_16bit(input_row); + src = _mm_mulhrs_epi16(src, rect_scale); + input_row += stride; + __m128i lo = _mm_unpacklo_epi16(src, one); + __m128i hi = _mm_unpackhi_epi16(src, one); + lo = _mm_madd_epi16(lo, scale_rounding); + hi = _mm_madd_epi16(hi, scale_rounding); + lo = _mm_srai_epi32(lo, NewSqrt2Bits - shift); + hi = _mm_srai_epi32(hi, NewSqrt2Bits - shift); + out[i] = _mm_packs_epi32(lo, hi); + } + } +} + +static INLINE void iidentity_col_8xn_ssse3(uint8_t *output, int stride, + __m128i *buf, int shift, int height, + int txh_idx) { + const __m128i scale = _mm_set1_epi16(NewSqrt2list[txh_idx]); + const __m128i scale_rounding = _mm_set1_epi16(1 << (NewSqrt2Bits - 1)); + const __m128i shift_rounding = _mm_set1_epi32(1 << (-shift - 1)); + const __m128i one = _mm_set1_epi16(1); + const __m128i scale_coeff = _mm_unpacklo_epi16(scale, scale_rounding); + const __m128i zero = _mm_setzero_si128(); + for (int h = 0; h < height; ++h) { + __m128i lo = _mm_unpacklo_epi16(buf[h], one); + __m128i hi = _mm_unpackhi_epi16(buf[h], one); + lo = _mm_madd_epi16(lo, scale_coeff); + hi = _mm_madd_epi16(hi, scale_coeff); + lo = _mm_srai_epi32(lo, NewSqrt2Bits); + hi = _mm_srai_epi32(hi, NewSqrt2Bits); + lo = _mm_add_epi32(lo, shift_rounding); + hi = _mm_add_epi32(hi, shift_rounding); + lo = _mm_srai_epi32(lo, -shift); + hi = _mm_srai_epi32(hi, -shift); + __m128i x = _mm_packs_epi32(lo, hi); + + const __m128i pred = _mm_loadl_epi64((__m128i const *)(output)); + x = _mm_adds_epi16(x, _mm_unpacklo_epi8(pred, zero)); + const __m128i u = _mm_packus_epi16(x, x); + _mm_storel_epi64((__m128i *)(output), u); + output += stride; + } +} + +static INLINE void lowbd_inv_txfm2d_add_idtx_ssse3(const int32_t *input, + uint8_t *output, int stride, + TX_SIZE tx_size) { + const int8_t *shift = inv_txfm_shift_ls[tx_size]; + const int txw_idx = get_txw_idx(tx_size); + const int txh_idx = get_txh_idx(tx_size); + const int txfm_size_col = tx_size_wide[tx_size]; + const int txfm_size_row = tx_size_high[tx_size]; + const int input_stride = AOMMIN(32, txfm_size_col); + const int row_max = AOMMIN(32, txfm_size_row); + const int rect_type = get_rect_tx_log_ratio(txfm_size_col, txfm_size_row); + __m128i buf[32]; + + for (int i = 0; i < (input_stride >> 3); ++i) { + iidentity_row_8xn_ssse3(buf, input + 8 * i, input_stride, shift[0], row_max, + txw_idx, rect_type); + iidentity_col_8xn_ssse3(output + 8 * i, stride, buf, shift[1], row_max, + txh_idx); + } +} + +void lowbd_inv_txfm2d_add_4x4_ssse3(const int32_t *input, uint8_t *output, + int stride, TX_TYPE tx_type, + TX_SIZE tx_size_, int eob) { + (void)tx_size_; + (void)eob; + __m128i buf[4]; + const TX_SIZE tx_size = TX_4X4; + const int8_t *shift = inv_txfm_shift_ls[tx_size]; + const int txw_idx = get_txw_idx(tx_size); + const int txh_idx = get_txh_idx(tx_size); + const int cos_bit_row = inv_cos_bit_row[txw_idx][txh_idx]; + const int cos_bit_col = inv_cos_bit_col[txw_idx][txh_idx]; + const int txfm_size_col = tx_size_wide[tx_size]; + const int txfm_size_row = tx_size_high[tx_size]; + + const transform_1d_ssse3 row_txfm = + lowbd_txfm_all_1d_w4_arr[txw_idx][hitx_1d_tab[tx_type]]; + const transform_1d_ssse3 col_txfm = + lowbd_txfm_all_1d_w4_arr[txh_idx][vitx_1d_tab[tx_type]]; + + int ud_flip, lr_flip; + get_flip_cfg(tx_type, &ud_flip, &lr_flip); + load_buffer_32bit_to_16bit_w4(input, txfm_size_col, buf, txfm_size_row); + transpose_16bit_4x4(buf, buf); + row_txfm(buf, buf, cos_bit_row); + if (lr_flip) { + __m128i temp[4]; + flip_buf_sse2(buf, temp, txfm_size_col); + transpose_16bit_4x4(temp, buf); + } else { + transpose_16bit_4x4(buf, buf); + } + col_txfm(buf, buf, cos_bit_col); + round_shift_16bit_ssse3(buf, txfm_size_row, shift[1]); + lowbd_write_buffer_4xn_sse2(buf, output, stride, ud_flip, txfm_size_row); +} + +static INLINE __m128i lowbd_get_recon_16x16_sse2(const __m128i pred, + __m128i res0, __m128i res1) { + const __m128i zero = _mm_setzero_si128(); + __m128i x0 = _mm_unpacklo_epi8(pred, zero); + __m128i x1 = _mm_unpackhi_epi8(pred, zero); + x0 = _mm_adds_epi16(res0, x0); + x1 = _mm_adds_epi16(res1, x1); + return _mm_packus_epi16(x0, x1); +} + +static INLINE void lowbd_write_buffer_16xn_sse2(__m128i *in, uint8_t *output, + int stride, int flipud, + int height) { + int j = flipud ? (height - 1) : 0; + const int step = flipud ? -1 : 1; + for (int i = 0; i < height; ++i, j += step) { + __m128i v = _mm_loadu_si128((__m128i const *)(output + i * stride)); + __m128i u = lowbd_get_recon_16x16_sse2(v, in[j], in[j + height]); + _mm_storeu_si128((__m128i *)(output + i * stride), u); + } +} + +static INLINE void round_shift_ssse3(const __m128i *input, __m128i *output, + int size) { + const __m128i scale = _mm_set1_epi16(NewInvSqrt2 * 8); + for (int i = 0; i < size; ++i) { + output[i] = _mm_mulhrs_epi16(input[i], scale); + } +} + +static INLINE void lowbd_inv_txfm2d_add_no_identity_ssse3( + const int32_t *input, uint8_t *output, int stride, TX_TYPE tx_type, + TX_SIZE tx_size, int eob) { + __m128i buf1[64 * 8]; + int eobx, eoby; + get_eobx_eoby_scan_default(&eobx, &eoby, tx_size, eob); + const int8_t *shift = inv_txfm_shift_ls[tx_size]; + const int txw_idx = get_txw_idx(tx_size); + const int txh_idx = get_txh_idx(tx_size); + const int cos_bit_col = inv_cos_bit_col[txw_idx][txh_idx]; + const int cos_bit_row = inv_cos_bit_row[txw_idx][txh_idx]; + const int txfm_size_col = tx_size_wide[tx_size]; + const int txfm_size_row = tx_size_high[tx_size]; + const int buf_size_w_div8 = txfm_size_col >> 3; + const int buf_size_nonzero_w_div8 = (eobx + 8) >> 3; + const int buf_size_nonzero_h_div8 = (eoby + 8) >> 3; + const int input_stride = AOMMIN(32, txfm_size_col); + const int rect_type = get_rect_tx_log_ratio(txfm_size_col, txfm_size_row); + + const int fun_idx_x = lowbd_txfm_all_1d_zeros_idx[eobx]; + const int fun_idx_y = lowbd_txfm_all_1d_zeros_idx[eoby]; + const transform_1d_ssse3 row_txfm = + lowbd_txfm_all_1d_zeros_w8_arr[txw_idx][hitx_1d_tab[tx_type]][fun_idx_x]; + const transform_1d_ssse3 col_txfm = + lowbd_txfm_all_1d_zeros_w8_arr[txh_idx][vitx_1d_tab[tx_type]][fun_idx_y]; + + assert(col_txfm != NULL); + assert(row_txfm != NULL); + int ud_flip, lr_flip; + get_flip_cfg(tx_type, &ud_flip, &lr_flip); + for (int i = 0; i < buf_size_nonzero_h_div8; i++) { + __m128i buf0[64]; + const int32_t *input_row = input + i * input_stride * 8; + for (int j = 0; j < buf_size_nonzero_w_div8; ++j) { + __m128i *buf0_cur = buf0 + j * 8; + load_buffer_32bit_to_16bit(input_row + j * 8, input_stride, buf0_cur, 8); + transpose_16bit_8x8(buf0_cur, buf0_cur); + } + if (rect_type == 1 || rect_type == -1) { + round_shift_ssse3(buf0, buf0, input_stride); // rect special code + } + row_txfm(buf0, buf0, cos_bit_row); + round_shift_16bit_ssse3(buf0, txfm_size_col, shift[0]); + __m128i *_buf1 = buf1 + i * 8; + if (lr_flip) { + for (int j = 0; j < buf_size_w_div8; ++j) { + __m128i temp[8]; + flip_buf_sse2(buf0 + 8 * j, temp, 8); + transpose_16bit_8x8(temp, + _buf1 + txfm_size_row * (buf_size_w_div8 - 1 - j)); + } + } else { + for (int j = 0; j < buf_size_w_div8; ++j) { + transpose_16bit_8x8(buf0 + 8 * j, _buf1 + txfm_size_row * j); + } + } + } + for (int i = 0; i < buf_size_w_div8; i++) { + col_txfm(buf1 + i * txfm_size_row, buf1 + i * txfm_size_row, cos_bit_col); + round_shift_16bit_ssse3(buf1 + i * txfm_size_row, txfm_size_row, shift[1]); + } + + if (txfm_size_col >= 16) { + for (int i = 0; i < (txfm_size_col >> 4); i++) { + lowbd_write_buffer_16xn_sse2(buf1 + i * txfm_size_row * 2, + output + 16 * i, stride, ud_flip, + txfm_size_row); + } + } else if (txfm_size_col == 8) { + lowbd_write_buffer_8xn_sse2(buf1, output, stride, ud_flip, txfm_size_row); + } +} + +static INLINE void lowbd_inv_txfm2d_add_h_identity_ssse3( + const int32_t *input, uint8_t *output, int stride, TX_TYPE tx_type, + TX_SIZE tx_size, int eob) { + const int8_t *shift = inv_txfm_shift_ls[tx_size]; + int eobx, eoby; + get_eobx_eoby_scan_h_identity(&eobx, &eoby, tx_size, eob); + const int txw_idx = get_txw_idx(tx_size); + const int txh_idx = get_txh_idx(tx_size); + const int cos_bit_col = inv_cos_bit_col[txw_idx][txh_idx]; + const int txfm_size_col = tx_size_wide[tx_size]; + const int txfm_size_row = tx_size_high[tx_size]; + const int buf_size_w_div8 = (eobx + 8) >> 3; + const int input_stride = AOMMIN(32, txfm_size_col); + const int rect_type = get_rect_tx_log_ratio(txfm_size_col, txfm_size_row); + + const int fun_idx = lowbd_txfm_all_1d_zeros_idx[eoby]; + assert(fun_idx < 5); + const transform_1d_ssse3 col_txfm = + lowbd_txfm_all_1d_zeros_w8_arr[txh_idx][vitx_1d_tab[tx_type]][fun_idx]; + + assert(col_txfm != NULL); + + int ud_flip, lr_flip; + get_flip_cfg(tx_type, &ud_flip, &lr_flip); + for (int i = 0; i < buf_size_w_div8; i++) { + __m128i buf0[64]; + iidentity_row_8xn_ssse3(buf0, input + 8 * i, input_stride, shift[0], + eoby + 1, txw_idx, rect_type); + col_txfm(buf0, buf0, cos_bit_col); + __m128i mshift = _mm_set1_epi16(1 << (15 + shift[1])); + int k = ud_flip ? (txfm_size_row - 1) : 0; + const int step = ud_flip ? -1 : 1; + uint8_t *out = output + 8 * i; + for (int j = 0; j < txfm_size_row; ++j, k += step) { + const __m128i v = _mm_loadl_epi64((__m128i const *)(out)); + __m128i res = _mm_mulhrs_epi16(buf0[k], mshift); + const __m128i u = lowbd_get_recon_8x8_sse2(v, res); + _mm_storel_epi64((__m128i *)(out), u); + out += stride; + } + } +} + +static INLINE void lowbd_inv_txfm2d_add_v_identity_ssse3( + const int32_t *input, uint8_t *output, int stride, TX_TYPE tx_type, + TX_SIZE tx_size, int eob) { + __m128i buf1[64]; + int eobx, eoby; + get_eobx_eoby_scan_v_identity(&eobx, &eoby, tx_size, eob); + const int8_t *shift = inv_txfm_shift_ls[tx_size]; + const int txw_idx = get_txw_idx(tx_size); + const int txh_idx = get_txh_idx(tx_size); + const int cos_bit_row = inv_cos_bit_row[txw_idx][txh_idx]; + const int txfm_size_col = tx_size_wide[tx_size]; + const int txfm_size_row = tx_size_high[tx_size]; + const int buf_size_w_div8 = txfm_size_col >> 3; + const int buf_size_h_div8 = (eoby + 8) >> 3; + const int input_stride = AOMMIN(32, txfm_size_col); + const int rect_type = get_rect_tx_log_ratio(txfm_size_col, txfm_size_row); + + const int fun_idx = lowbd_txfm_all_1d_zeros_idx[eobx]; + const transform_1d_ssse3 row_txfm = + lowbd_txfm_all_1d_zeros_w8_arr[txw_idx][hitx_1d_tab[tx_type]][fun_idx]; + + assert(row_txfm != NULL); + int ud_flip, lr_flip; + get_flip_cfg(tx_type, &ud_flip, &lr_flip); + for (int i = 0; i < buf_size_h_div8; i++) { + __m128i buf0[64]; + const int32_t *input_row = input + i * input_stride * 8; + for (int j = 0; j < AOMMIN(4, buf_size_w_div8); ++j) { + __m128i *buf0_cur = buf0 + j * 8; + load_buffer_32bit_to_16bit(input_row + j * 8, input_stride, buf0_cur, 8); + transpose_16bit_8x8(buf0_cur, buf0_cur); + } + if (rect_type == 1 || rect_type == -1) { + round_shift_ssse3(buf0, buf0, input_stride); // rect special code + } + row_txfm(buf0, buf0, cos_bit_row); + round_shift_16bit_ssse3(buf0, txfm_size_col, shift[0]); + __m128i *_buf1 = buf1; + if (lr_flip) { + for (int j = 0; j < buf_size_w_div8; ++j) { + __m128i temp[8]; + flip_buf_sse2(buf0 + 8 * j, temp, 8); + transpose_16bit_8x8(temp, _buf1 + 8 * (buf_size_w_div8 - 1 - j)); + } + } else { + for (int j = 0; j < buf_size_w_div8; ++j) { + transpose_16bit_8x8(buf0 + 8 * j, _buf1 + 8 * j); + } + } + + for (int j = 0; j < buf_size_w_div8; ++j) { + iidentity_col_8xn_ssse3(output + i * 8 * stride + j * 8, stride, + buf1 + j * 8, shift[1], 8, txh_idx); + } + } +} + +// for 32x32,32x64,64x32,64x64,32x8,8x32,16x32,32x16,64x16,16x64 +static INLINE void lowbd_inv_txfm2d_add_universe_ssse3( + const int32_t *input, uint8_t *output, int stride, TX_TYPE tx_type, + TX_SIZE tx_size, int eob) { + switch (tx_type) { + case DCT_DCT: + lowbd_inv_txfm2d_add_no_identity_ssse3(input, output, stride, tx_type, + tx_size, eob); + break; + case IDTX: + lowbd_inv_txfm2d_add_idtx_ssse3(input, output, stride, tx_size); + break; + case V_DCT: + case V_ADST: + case V_FLIPADST: + lowbd_inv_txfm2d_add_h_identity_ssse3(input, output, stride, tx_type, + tx_size, eob); + break; + case H_DCT: + case H_ADST: + case H_FLIPADST: + lowbd_inv_txfm2d_add_v_identity_ssse3(input, output, stride, tx_type, + tx_size, eob); + break; + default: + lowbd_inv_txfm2d_add_no_identity_ssse3(input, output, stride, tx_type, + tx_size, eob); + break; + } +} + +void lowbd_inv_txfm2d_add_4x8_ssse3(const int32_t *input, uint8_t *output, + int stride, TX_TYPE tx_type, + TX_SIZE tx_size_, int eob) { + (void)tx_size_; + (void)eob; + __m128i buf[8]; + const TX_SIZE tx_size = TX_4X8; + const int8_t *shift = inv_txfm_shift_ls[tx_size]; + const int txw_idx = get_txw_idx(tx_size); + const int txh_idx = get_txh_idx(tx_size); + const int cos_bit_row = inv_cos_bit_row[txw_idx][txh_idx]; + const int cos_bit_col = inv_cos_bit_col[txw_idx][txh_idx]; + const int txfm_size_col = tx_size_wide[tx_size]; + const int txfm_size_row = tx_size_high[tx_size]; + + const transform_1d_ssse3 row_txfm = + lowbd_txfm_all_1d_w8_arr[txw_idx][hitx_1d_tab[tx_type]]; + const transform_1d_ssse3 col_txfm = + lowbd_txfm_all_1d_w4_arr[txh_idx][vitx_1d_tab[tx_type]]; + + int ud_flip, lr_flip; + get_flip_cfg(tx_type, &ud_flip, &lr_flip); + load_buffer_32bit_to_16bit_w4(input, txfm_size_col, buf, txfm_size_row); + transpose_16bit_4x8(buf, buf); + round_shift_ssse3(buf, buf, txfm_size_col); // rect special code + row_txfm(buf, buf, cos_bit_row); + // round_shift_16bit_ssse3(buf, txfm_size_col, shift[0]);// shift[0] is 0 + if (lr_flip) { + __m128i temp[4]; + flip_buf_sse2(buf, temp, txfm_size_col); + transpose_16bit_8x4(temp, buf); + } else { + transpose_16bit_8x4(buf, buf); + } + col_txfm(buf, buf, cos_bit_col); + round_shift_16bit_ssse3(buf, txfm_size_row, shift[1]); + lowbd_write_buffer_4xn_sse2(buf, output, stride, ud_flip, txfm_size_row); +} + +void lowbd_inv_txfm2d_add_8x4_ssse3(const int32_t *input, uint8_t *output, + int stride, TX_TYPE tx_type, + TX_SIZE tx_size_, int eob) { + (void)tx_size_; + (void)eob; + __m128i buf[8]; + const TX_SIZE tx_size = TX_8X4; + const int8_t *shift = inv_txfm_shift_ls[tx_size]; + const int txw_idx = get_txw_idx(tx_size); + const int txh_idx = get_txh_idx(tx_size); + const int cos_bit_row = inv_cos_bit_row[txw_idx][txh_idx]; + const int cos_bit_col = inv_cos_bit_col[txw_idx][txh_idx]; + const int txfm_size_col = tx_size_wide[tx_size]; + const int txfm_size_row = tx_size_high[tx_size]; + + const transform_1d_ssse3 row_txfm = + lowbd_txfm_all_1d_w4_arr[txw_idx][hitx_1d_tab[tx_type]]; + const transform_1d_ssse3 col_txfm = + lowbd_txfm_all_1d_w8_arr[txh_idx][vitx_1d_tab[tx_type]]; + + int ud_flip, lr_flip; + get_flip_cfg(tx_type, &ud_flip, &lr_flip); + load_buffer_32bit_to_16bit(input, txfm_size_col, buf, txfm_size_row); + transpose_16bit_8x4(buf, buf); + round_shift_ssse3(buf, buf, txfm_size_col); // rect special code + row_txfm(buf, buf, cos_bit_row); + // round_shift_16bit_ssse3(buf, txfm_size_col, shift[0]); // shift[0] is 0 + if (lr_flip) { + __m128i temp[8]; + flip_buf_sse2(buf, temp, txfm_size_col); + transpose_16bit_4x8(temp, buf); + } else { + transpose_16bit_4x8(buf, buf); + } + col_txfm(buf, buf, cos_bit_col); + round_shift_16bit_ssse3(buf, txfm_size_row, shift[1]); + lowbd_write_buffer_8xn_sse2(buf, output, stride, ud_flip, txfm_size_row); +} + +void lowbd_inv_txfm2d_add_4x16_ssse3(const int32_t *input, uint8_t *output, + int stride, TX_TYPE tx_type, + TX_SIZE tx_size_, int eob) { + (void)tx_size_; + (void)eob; + __m128i buf[16]; + const TX_SIZE tx_size = TX_4X16; + const int8_t *shift = inv_txfm_shift_ls[tx_size]; + const int txw_idx = get_txw_idx(tx_size); + const int txh_idx = get_txh_idx(tx_size); + const int cos_bit_row = inv_cos_bit_row[txw_idx][txh_idx]; + const int cos_bit_col = inv_cos_bit_col[txw_idx][txh_idx]; + const int txfm_size_col = tx_size_wide[tx_size]; + const int txfm_size_row = tx_size_high[tx_size]; + + const transform_1d_ssse3 row_txfm = + lowbd_txfm_all_1d_w8_arr[txw_idx][hitx_1d_tab[tx_type]]; + const transform_1d_ssse3 col_txfm = + lowbd_txfm_all_1d_w4_arr[txh_idx][vitx_1d_tab[tx_type]]; + + int ud_flip, lr_flip; + get_flip_cfg(tx_type, &ud_flip, &lr_flip); + + const int row_one_loop = 8; + for (int i = 0; i < 2; ++i) { + const int32_t *input_cur = input + i * txfm_size_col * row_one_loop; + __m128i *buf_cur = buf + i * row_one_loop; + load_buffer_32bit_to_16bit_w4(input_cur, txfm_size_col, buf_cur, + row_one_loop); + transpose_16bit_4x8(buf_cur, buf_cur); + row_txfm(buf_cur, buf_cur, cos_bit_row); + round_shift_16bit_ssse3(buf_cur, row_one_loop, shift[0]); + if (lr_flip) { + __m128i temp[8]; + flip_buf_sse2(buf_cur, temp, txfm_size_col); + transpose_16bit_8x4(temp, buf_cur); + } else { + transpose_16bit_8x4(buf_cur, buf_cur); + } + } + col_txfm(buf, buf, cos_bit_col); + round_shift_16bit_ssse3(buf, txfm_size_row, shift[1]); + lowbd_write_buffer_4xn_sse2(buf, output, stride, ud_flip, txfm_size_row); +} + +void lowbd_inv_txfm2d_add_16x4_ssse3(const int32_t *input, uint8_t *output, + int stride, TX_TYPE tx_type, + TX_SIZE tx_size_, int eob) { + (void)tx_size_; + (void)eob; + __m128i buf[16]; + const TX_SIZE tx_size = TX_16X4; + const int8_t *shift = inv_txfm_shift_ls[tx_size]; + const int txw_idx = get_txw_idx(tx_size); + const int txh_idx = get_txh_idx(tx_size); + const int cos_bit_row = inv_cos_bit_row[txw_idx][txh_idx]; + const int cos_bit_col = inv_cos_bit_col[txw_idx][txh_idx]; + const int txfm_size_col = tx_size_wide[tx_size]; + const int txfm_size_row = tx_size_high[tx_size]; + const int buf_size_w_div8 = txfm_size_col >> 3; + + const transform_1d_ssse3 row_txfm = + lowbd_txfm_all_1d_w4_arr[txw_idx][hitx_1d_tab[tx_type]]; + const transform_1d_ssse3 col_txfm = + lowbd_txfm_all_1d_w8_arr[txh_idx][vitx_1d_tab[tx_type]]; + + int ud_flip, lr_flip; + get_flip_cfg(tx_type, &ud_flip, &lr_flip); + const int row_one_loop = 8; + for (int i = 0; i < buf_size_w_div8; ++i) { + const int32_t *input_cur = input + i * row_one_loop; + __m128i *buf_cur = buf + i * row_one_loop; + load_buffer_32bit_to_16bit(input_cur, txfm_size_col, buf_cur, + txfm_size_row); + transpose_16bit_8x4(buf_cur, buf_cur); + } + row_txfm(buf, buf, cos_bit_row); + round_shift_16bit_ssse3(buf, txfm_size_col, shift[0]); + if (lr_flip) { + __m128i temp[16]; + flip_buf_sse2(buf, temp, 16); + transpose_16bit_4x8(temp, buf); + transpose_16bit_4x8(temp + 8, buf + 8); + } else { + transpose_16bit_4x8(buf, buf); + transpose_16bit_4x8(buf + row_one_loop, buf + row_one_loop); + } + for (int i = 0; i < buf_size_w_div8; i++) { + col_txfm(buf + i * row_one_loop, buf + i * row_one_loop, cos_bit_col); + round_shift_16bit_ssse3(buf + i * row_one_loop, txfm_size_row, shift[1]); + } + lowbd_write_buffer_8xn_sse2(buf, output, stride, ud_flip, 4); + lowbd_write_buffer_8xn_sse2(buf + 8, output + 8, stride, ud_flip, 4); +} + +void av1_lowbd_inv_txfm2d_add_ssse3(const int32_t *input, uint8_t *output, + int stride, TX_TYPE tx_type, + TX_SIZE tx_size, int eob) { + switch (tx_size) { + case TX_4X4: + lowbd_inv_txfm2d_add_4x4_ssse3(input, output, stride, tx_type, tx_size, + eob); + break; + case TX_4X8: + lowbd_inv_txfm2d_add_4x8_ssse3(input, output, stride, tx_type, tx_size, + eob); + break; + case TX_8X4: + lowbd_inv_txfm2d_add_8x4_ssse3(input, output, stride, tx_type, tx_size, + eob); + break; + case TX_4X16: + lowbd_inv_txfm2d_add_4x16_ssse3(input, output, stride, tx_type, tx_size, + eob); + break; + case TX_16X4: + lowbd_inv_txfm2d_add_16x4_ssse3(input, output, stride, tx_type, tx_size, + eob); + break; + default: + lowbd_inv_txfm2d_add_universe_ssse3(input, output, stride, tx_type, + tx_size, eob); + break; + } +} +void av1_inv_txfm_add_ssse3(const tran_low_t *dqcoeff, uint8_t *dst, int stride, + const TxfmParam *txfm_param) { + const TX_TYPE tx_type = txfm_param->tx_type; + if (!txfm_param->lossless) { + av1_lowbd_inv_txfm2d_add_ssse3(dqcoeff, dst, stride, tx_type, + txfm_param->tx_size, txfm_param->eob); + } else { + av1_inv_txfm_add_c(dqcoeff, dst, stride, txfm_param); + } +} diff --git a/third_party/aom/av1/common/x86/av1_inv_txfm_ssse3.h b/third_party/aom/av1/common/x86/av1_inv_txfm_ssse3.h new file mode 100644 index 0000000000..66bd339d11 --- /dev/null +++ b/third_party/aom/av1/common/x86/av1_inv_txfm_ssse3.h @@ -0,0 +1,232 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ +#ifndef AOM_AV1_COMMON_X86_AV1_INV_TXFM_SSSE3_H_ +#define AOM_AV1_COMMON_X86_AV1_INV_TXFM_SSSE3_H_ + +#include <emmintrin.h> // SSE2 +#include <tmmintrin.h> // SSSE3 + +#include "config/aom_config.h" +#include "config/av1_rtcd.h" + +#include "aom/aom_integer.h" +#include "aom_dsp/x86/transpose_sse2.h" +#include "aom_dsp/x86/txfm_common_sse2.h" + +#ifdef __cplusplus +extern "C" { +#endif + +#define btf_16_ssse3(w0, w1, in, out0, out1) \ + do { \ + const __m128i _w0 = _mm_set1_epi16(w0 * 8); \ + const __m128i _w1 = _mm_set1_epi16(w1 * 8); \ + const __m128i _in = in; \ + out0 = _mm_mulhrs_epi16(_in, _w0); \ + out1 = _mm_mulhrs_epi16(_in, _w1); \ + } while (0) + +#define btf_16_adds_subs_sse2(in0, in1) \ + do { \ + const __m128i _in0 = in0; \ + const __m128i _in1 = in1; \ + in0 = _mm_adds_epi16(_in0, _in1); \ + in1 = _mm_subs_epi16(_in0, _in1); \ + } while (0) + +#define btf_16_subs_adds_sse2(in0, in1) \ + do { \ + const __m128i _in0 = in0; \ + const __m128i _in1 = in1; \ + in1 = _mm_subs_epi16(_in0, _in1); \ + in0 = _mm_adds_epi16(_in0, _in1); \ + } while (0) + +#define btf_16_adds_subs_out_sse2(out0, out1, in0, in1) \ + do { \ + const __m128i _in0 = in0; \ + const __m128i _in1 = in1; \ + out0 = _mm_adds_epi16(_in0, _in1); \ + out1 = _mm_subs_epi16(_in0, _in1); \ + } while (0) + +static INLINE void round_shift_16bit_ssse3(__m128i *in, int size, int bit) { + if (bit < 0) { + const __m128i scale = _mm_set1_epi16(1 << (15 + bit)); + for (int i = 0; i < size; ++i) { + in[i] = _mm_mulhrs_epi16(in[i], scale); + } + } else if (bit > 0) { + for (int i = 0; i < size; ++i) { + in[i] = _mm_slli_epi16(in[i], bit); + } + } +} + +// 1D itx types +typedef enum ATTRIBUTE_PACKED { + IDCT_1D, + IADST_1D, + IFLIPADST_1D = IADST_1D, + IIDENTITY_1D, + ITX_TYPES_1D, +} ITX_TYPE_1D; + +static const ITX_TYPE_1D vitx_1d_tab[TX_TYPES] = { + IDCT_1D, IADST_1D, IDCT_1D, IADST_1D, + IFLIPADST_1D, IDCT_1D, IFLIPADST_1D, IADST_1D, + IFLIPADST_1D, IIDENTITY_1D, IDCT_1D, IIDENTITY_1D, + IADST_1D, IIDENTITY_1D, IFLIPADST_1D, IIDENTITY_1D, +}; + +static const ITX_TYPE_1D hitx_1d_tab[TX_TYPES] = { + IDCT_1D, IDCT_1D, IADST_1D, IADST_1D, + IDCT_1D, IFLIPADST_1D, IFLIPADST_1D, IFLIPADST_1D, + IADST_1D, IIDENTITY_1D, IIDENTITY_1D, IDCT_1D, + IIDENTITY_1D, IADST_1D, IIDENTITY_1D, IFLIPADST_1D, +}; + +DECLARE_ALIGNED(16, static const int16_t, av1_eob_to_eobxy_8x8_default[8]) = { + 0x0707, 0x0707, 0x0707, 0x0707, 0x0707, 0x0707, 0x0707, 0x0707, +}; + +DECLARE_ALIGNED(16, static const int16_t, + av1_eob_to_eobxy_16x16_default[16]) = { + 0x0707, 0x0707, 0x0f0f, 0x0f0f, 0x0f0f, 0x0f0f, 0x0f0f, 0x0f0f, + 0x0f0f, 0x0f0f, 0x0f0f, 0x0f0f, 0x0f0f, 0x0f0f, 0x0f0f, 0x0f0f, +}; + +DECLARE_ALIGNED(16, static const int16_t, + av1_eob_to_eobxy_32x32_default[32]) = { + 0x0707, 0x0f0f, 0x0f0f, 0x0f0f, 0x1f1f, 0x1f1f, 0x1f1f, 0x1f1f, + 0x1f1f, 0x1f1f, 0x1f1f, 0x1f1f, 0x1f1f, 0x1f1f, 0x1f1f, 0x1f1f, + 0x1f1f, 0x1f1f, 0x1f1f, 0x1f1f, 0x1f1f, 0x1f1f, 0x1f1f, 0x1f1f, + 0x1f1f, 0x1f1f, 0x1f1f, 0x1f1f, 0x1f1f, 0x1f1f, 0x1f1f, 0x1f1f, +}; + +DECLARE_ALIGNED(16, static const int16_t, av1_eob_to_eobxy_8x16_default[16]) = { + 0x0707, 0x0707, 0x0707, 0x0707, 0x0707, 0x0f07, 0x0f07, 0x0f07, + 0x0f07, 0x0f07, 0x0f07, 0x0f07, 0x0f07, 0x0f07, 0x0f07, 0x0f07, +}; + +DECLARE_ALIGNED(16, static const int16_t, av1_eob_to_eobxy_16x8_default[8]) = { + 0x0707, 0x0707, 0x070f, 0x070f, 0x070f, 0x070f, 0x070f, 0x070f, +}; + +DECLARE_ALIGNED(16, static const int16_t, + av1_eob_to_eobxy_16x32_default[32]) = { + 0x0707, 0x0707, 0x0f0f, 0x0f0f, 0x0f0f, 0x0f0f, 0x0f0f, 0x0f0f, + 0x0f0f, 0x1f0f, 0x1f0f, 0x1f0f, 0x1f0f, 0x1f0f, 0x1f0f, 0x1f0f, + 0x1f0f, 0x1f0f, 0x1f0f, 0x1f0f, 0x1f0f, 0x1f0f, 0x1f0f, 0x1f0f, + 0x1f0f, 0x1f0f, 0x1f0f, 0x1f0f, 0x1f0f, 0x1f0f, 0x1f0f, 0x1f0f, +}; + +DECLARE_ALIGNED(16, static const int16_t, + av1_eob_to_eobxy_32x16_default[16]) = { + 0x0707, 0x0f0f, 0x0f0f, 0x0f0f, 0x0f1f, 0x0f1f, 0x0f1f, 0x0f1f, + 0x0f1f, 0x0f1f, 0x0f1f, 0x0f1f, 0x0f1f, 0x0f1f, 0x0f1f, 0x0f1f, +}; + +DECLARE_ALIGNED(16, static const int16_t, av1_eob_to_eobxy_8x32_default[32]) = { + 0x0707, 0x0707, 0x0707, 0x0707, 0x0707, 0x0f07, 0x0f07, 0x0f07, + 0x0f07, 0x0f07, 0x0f07, 0x0f07, 0x0f07, 0x1f07, 0x1f07, 0x1f07, + 0x1f07, 0x1f07, 0x1f07, 0x1f07, 0x1f07, 0x1f07, 0x1f07, 0x1f07, + 0x1f07, 0x1f07, 0x1f07, 0x1f07, 0x1f07, 0x1f07, 0x1f07, 0x1f07, +}; + +DECLARE_ALIGNED(16, static const int16_t, av1_eob_to_eobxy_32x8_default[8]) = { + 0x0707, 0x070f, 0x070f, 0x071f, 0x071f, 0x071f, 0x071f, 0x071f, +}; + +DECLARE_ALIGNED(16, static const int16_t *, + av1_eob_to_eobxy_default[TX_SIZES_ALL]) = { + NULL, + av1_eob_to_eobxy_8x8_default, + av1_eob_to_eobxy_16x16_default, + av1_eob_to_eobxy_32x32_default, + av1_eob_to_eobxy_32x32_default, + NULL, + NULL, + av1_eob_to_eobxy_8x16_default, + av1_eob_to_eobxy_16x8_default, + av1_eob_to_eobxy_16x32_default, + av1_eob_to_eobxy_32x16_default, + av1_eob_to_eobxy_32x32_default, + av1_eob_to_eobxy_32x32_default, + NULL, + NULL, + av1_eob_to_eobxy_8x32_default, + av1_eob_to_eobxy_32x8_default, + av1_eob_to_eobxy_16x32_default, + av1_eob_to_eobxy_32x16_default, +}; + +static const int lowbd_txfm_all_1d_zeros_idx[32] = { + 0, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, + 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, +}; + +// Transform block width in log2 for eob (size of 64 map to 32) +static const int tx_size_wide_log2_eob[TX_SIZES_ALL] = { + 2, 3, 4, 5, 5, 2, 3, 3, 4, 4, 5, 5, 5, 2, 4, 3, 5, 4, 5, +}; + +static INLINE void get_eobx_eoby_scan_default(int *eobx, int *eoby, + TX_SIZE tx_size, int eob) { + if (eob == 1) { + *eobx = 0; + *eoby = 0; + return; + } + + const int tx_w_log2 = tx_size_wide_log2_eob[tx_size]; + const int eob_row = (eob - 1) >> tx_w_log2; + const int eobxy = av1_eob_to_eobxy_default[tx_size][eob_row]; + *eobx = eobxy & 0xFF; + *eoby = eobxy >> 8; +} + +static int eob_fill[32] = { + 0, 7, 7, 7, 7, 7, 7, 7, 15, 15, 15, 15, 15, 15, 15, 15, + 31, 31, 31, 31, 31, 31, 31, 31, 31, 31, 31, 31, 31, 31, 31, 31, +}; + +static INLINE void get_eobx_eoby_scan_h_identity(int *eobx, int *eoby, + TX_SIZE tx_size, int eob) { + eob -= 1; + const int txfm_size_col = tx_size_wide[tx_size]; + const int eobx_max = AOMMIN(32, txfm_size_col) - 1; + *eobx = (eob >= eobx_max) ? eobx_max : eob_fill[eob]; + const int temp_eoby = eob / (eobx_max + 1); + assert(temp_eoby < 32); + *eoby = eob_fill[temp_eoby]; +} + +static INLINE void get_eobx_eoby_scan_v_identity(int *eobx, int *eoby, + TX_SIZE tx_size, int eob) { + eob -= 1; + const int txfm_size_row = tx_size_high[tx_size]; + const int eoby_max = AOMMIN(32, txfm_size_row) - 1; + *eobx = eob / (eoby_max + 1); + *eoby = (eob >= eoby_max) ? eoby_max : eob_fill[eob]; +} + +typedef void (*transform_1d_ssse3)(const __m128i *input, __m128i *output, + int8_t cos_bit); + +void av1_lowbd_inv_txfm2d_add_ssse3(const int32_t *input, uint8_t *output, + int stride, TX_TYPE tx_type, + TX_SIZE tx_size, int eob); +#ifdef __cplusplus +} // extern "C" +#endif + +#endif // AOM_AV1_COMMON_X86_AV1_INV_TXFM_SSSE3_H_ diff --git a/third_party/aom/av1/common/x86/av1_txfm_sse2.h b/third_party/aom/av1/common/x86/av1_txfm_sse2.h new file mode 100644 index 0000000000..77aeb6eb13 --- /dev/null +++ b/third_party/aom/av1/common/x86/av1_txfm_sse2.h @@ -0,0 +1,317 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ +#ifndef AOM_AV1_COMMON_X86_AV1_TXFM_SSE2_H_ +#define AOM_AV1_COMMON_X86_AV1_TXFM_SSE2_H_ + +#include <emmintrin.h> // SSE2 + +#include "config/aom_config.h" +#include "config/av1_rtcd.h" + +#include "aom/aom_integer.h" +#include "aom_dsp/x86/transpose_sse2.h" +#include "aom_dsp/x86/txfm_common_sse2.h" +#include "av1/common/av1_txfm.h" + +#ifdef __cplusplus +extern "C" { +#endif + +static INLINE void btf_16_w4_sse2( + const __m128i *const w0, const __m128i *const w1, const __m128i __rounding, + const int8_t cos_bit, const __m128i *const in0, const __m128i *const in1, + __m128i *const out0, __m128i *const out1) { + const __m128i t0 = _mm_unpacklo_epi16(*in0, *in1); + const __m128i u0 = _mm_madd_epi16(t0, *w0); + const __m128i v0 = _mm_madd_epi16(t0, *w1); + const __m128i a0 = _mm_add_epi32(u0, __rounding); + const __m128i b0 = _mm_add_epi32(v0, __rounding); + const __m128i c0 = _mm_srai_epi32(a0, cos_bit); + const __m128i d0 = _mm_srai_epi32(b0, cos_bit); + + *out0 = _mm_packs_epi32(c0, c0); + *out1 = _mm_packs_epi32(d0, c0); +} + +#define btf_16_4p_sse2(w0, w1, in0, in1, out0, out1) \ + { \ + __m128i t0 = _mm_unpacklo_epi16(in0, in1); \ + __m128i u0 = _mm_madd_epi16(t0, w0); \ + __m128i v0 = _mm_madd_epi16(t0, w1); \ + \ + __m128i a0 = _mm_add_epi32(u0, __rounding); \ + __m128i b0 = _mm_add_epi32(v0, __rounding); \ + \ + __m128i c0 = _mm_srai_epi32(a0, cos_bit); \ + __m128i d0 = _mm_srai_epi32(b0, cos_bit); \ + \ + out0 = _mm_packs_epi32(c0, c0); \ + out1 = _mm_packs_epi32(d0, d0); \ + } + +#define btf_16_sse2(w0, w1, in0, in1, out0, out1) \ + { \ + __m128i t0 = _mm_unpacklo_epi16(in0, in1); \ + __m128i t1 = _mm_unpackhi_epi16(in0, in1); \ + __m128i u0 = _mm_madd_epi16(t0, w0); \ + __m128i u1 = _mm_madd_epi16(t1, w0); \ + __m128i v0 = _mm_madd_epi16(t0, w1); \ + __m128i v1 = _mm_madd_epi16(t1, w1); \ + \ + __m128i a0 = _mm_add_epi32(u0, __rounding); \ + __m128i a1 = _mm_add_epi32(u1, __rounding); \ + __m128i b0 = _mm_add_epi32(v0, __rounding); \ + __m128i b1 = _mm_add_epi32(v1, __rounding); \ + \ + __m128i c0 = _mm_srai_epi32(a0, cos_bit); \ + __m128i c1 = _mm_srai_epi32(a1, cos_bit); \ + __m128i d0 = _mm_srai_epi32(b0, cos_bit); \ + __m128i d1 = _mm_srai_epi32(b1, cos_bit); \ + \ + out0 = _mm_packs_epi32(c0, c1); \ + out1 = _mm_packs_epi32(d0, d1); \ + } + +static INLINE __m128i load_16bit_to_16bit(const int16_t *a) { + return _mm_load_si128((const __m128i *)a); +} + +static INLINE __m128i load_32bit_to_16bit(const int32_t *a) { + const __m128i a_low = _mm_load_si128((const __m128i *)a); + return _mm_packs_epi32(a_low, *(const __m128i *)(a + 4)); +} + +static INLINE __m128i load_32bit_to_16bit_w4(const int32_t *a) { + const __m128i a_low = _mm_load_si128((const __m128i *)a); + return _mm_packs_epi32(a_low, a_low); +} + +// Store 4 16 bit values. Sign extend the values. +static INLINE void store_16bit_to_32bit_w4(const __m128i a, int32_t *const b) { + const __m128i a_lo = _mm_unpacklo_epi16(a, a); + const __m128i a_1 = _mm_srai_epi32(a_lo, 16); + _mm_store_si128((__m128i *)b, a_1); +} + +// Store 8 16 bit values. Sign extend the values. +static INLINE void store_16bit_to_32bit(__m128i a, int32_t *b) { + const __m128i a_lo = _mm_unpacklo_epi16(a, a); + const __m128i a_hi = _mm_unpackhi_epi16(a, a); + const __m128i a_1 = _mm_srai_epi32(a_lo, 16); + const __m128i a_2 = _mm_srai_epi32(a_hi, 16); + _mm_store_si128((__m128i *)b, a_1); + _mm_store_si128((__m128i *)(b + 4), a_2); +} + +static INLINE __m128i scale_round_sse2(const __m128i a, const int scale) { + const __m128i scale_rounding = pair_set_epi16(scale, 1 << (NewSqrt2Bits - 1)); + const __m128i b = _mm_madd_epi16(a, scale_rounding); + return _mm_srai_epi32(b, NewSqrt2Bits); +} + +static INLINE void store_rect_16bit_to_32bit_w4(const __m128i a, + int32_t *const b) { + const __m128i one = _mm_set1_epi16(1); + const __m128i a_lo = _mm_unpacklo_epi16(a, one); + const __m128i b_lo = scale_round_sse2(a_lo, NewSqrt2); + _mm_store_si128((__m128i *)b, b_lo); +} + +static INLINE void store_rect_16bit_to_32bit(const __m128i a, + int32_t *const b) { + const __m128i one = _mm_set1_epi16(1); + const __m128i a_lo = _mm_unpacklo_epi16(a, one); + const __m128i a_hi = _mm_unpackhi_epi16(a, one); + const __m128i b_lo = scale_round_sse2(a_lo, NewSqrt2); + const __m128i b_hi = scale_round_sse2(a_hi, NewSqrt2); + _mm_store_si128((__m128i *)b, b_lo); + _mm_store_si128((__m128i *)(b + 4), b_hi); +} + +static INLINE void load_buffer_16bit_to_16bit_w4(const int16_t *const in, + const int stride, + __m128i *const out, + const int out_size) { + for (int i = 0; i < out_size; ++i) { + out[i] = _mm_loadl_epi64((const __m128i *)(in + i * stride)); + } +} + +static INLINE void load_buffer_16bit_to_16bit_w4_flip(const int16_t *const in, + const int stride, + __m128i *const out, + const int out_size) { + for (int i = 0; i < out_size; ++i) { + out[out_size - i - 1] = _mm_loadl_epi64((const __m128i *)(in + i * stride)); + } +} + +static INLINE void load_buffer_16bit_to_16bit(const int16_t *in, int stride, + __m128i *out, int out_size) { + for (int i = 0; i < out_size; ++i) { + out[i] = load_16bit_to_16bit(in + i * stride); + } +} + +static INLINE void load_buffer_16bit_to_16bit_flip(const int16_t *in, + int stride, __m128i *out, + int out_size) { + for (int i = 0; i < out_size; ++i) { + out[out_size - i - 1] = load_16bit_to_16bit(in + i * stride); + } +} + +static INLINE void load_buffer_32bit_to_16bit(const int32_t *in, int stride, + __m128i *out, int out_size) { + for (int i = 0; i < out_size; ++i) { + out[i] = load_32bit_to_16bit(in + i * stride); + } +} + +static INLINE void load_buffer_32bit_to_16bit_w4(const int32_t *in, int stride, + __m128i *out, int out_size) { + for (int i = 0; i < out_size; ++i) { + out[i] = load_32bit_to_16bit_w4(in + i * stride); + } +} + +static INLINE void load_buffer_32bit_to_16bit_flip(const int32_t *in, + int stride, __m128i *out, + int out_size) { + for (int i = 0; i < out_size; ++i) { + out[out_size - i - 1] = load_32bit_to_16bit(in + i * stride); + } +} + +static INLINE void store_buffer_16bit_to_32bit_w4(const __m128i *const in, + int32_t *const out, + const int stride, + const int out_size) { + for (int i = 0; i < out_size; ++i) { + store_16bit_to_32bit_w4(in[i], out + i * stride); + } +} + +static INLINE void store_buffer_16bit_to_32bit_w8(const __m128i *const in, + int32_t *const out, + const int stride, + const int out_size) { + for (int i = 0; i < out_size; ++i) { + store_16bit_to_32bit(in[i], out + i * stride); + } +} + +static INLINE void store_rect_buffer_16bit_to_32bit_w4(const __m128i *const in, + int32_t *const out, + const int stride, + const int out_size) { + for (int i = 0; i < out_size; ++i) { + store_rect_16bit_to_32bit_w4(in[i], out + i * stride); + } +} + +static INLINE void store_rect_buffer_16bit_to_32bit_w8(const __m128i *const in, + int32_t *const out, + const int stride, + const int out_size) { + for (int i = 0; i < out_size; ++i) { + store_rect_16bit_to_32bit(in[i], out + i * stride); + } +} + +static INLINE void store_buffer_16bit_to_16bit_8x8(const __m128i *in, + uint16_t *out, + const int stride) { + for (int i = 0; i < 8; ++i) { + _mm_store_si128((__m128i *)(out + i * stride), in[i]); + } +} + +static INLINE void round_shift_16bit(__m128i *in, int size, int bit) { + if (bit < 0) { + bit = -bit; + __m128i rounding = _mm_set1_epi16(1 << (bit - 1)); + for (int i = 0; i < size; ++i) { + in[i] = _mm_adds_epi16(in[i], rounding); + in[i] = _mm_srai_epi16(in[i], bit); + } + } else if (bit > 0) { + for (int i = 0; i < size; ++i) { + in[i] = _mm_slli_epi16(in[i], bit); + } + } +} + +static INLINE void flip_buf_sse2(__m128i *in, __m128i *out, int size) { + for (int i = 0; i < size; ++i) { + out[size - i - 1] = in[i]; + } +} + +void av1_lowbd_fwd_txfm2d_4x4_sse2(const int16_t *input, int32_t *output, + int stride, TX_TYPE tx_type, int bd); + +void av1_lowbd_fwd_txfm2d_4x8_sse2(const int16_t *input, int32_t *output, + int stride, TX_TYPE tx_type, int bd); + +void av1_lowbd_fwd_txfm2d_4x16_sse2(const int16_t *input, int32_t *output, + int stride, TX_TYPE tx_type, int bd); + +void av1_lowbd_fwd_txfm2d_8x4_sse2(const int16_t *input, int32_t *output, + int stride, TX_TYPE tx_type, int bd); + +void av1_lowbd_fwd_txfm2d_8x8_sse2(const int16_t *input, int32_t *output, + int stride, TX_TYPE tx_type, int bd); + +void av1_lowbd_fwd_txfm2d_8x16_sse2(const int16_t *input, int32_t *output, + int stride, TX_TYPE tx_type, int bd); + +void av1_lowbd_fwd_txfm2d_8x32_sse2(const int16_t *input, int32_t *output, + int stride, TX_TYPE tx_type, int bd); + +void av1_lowbd_fwd_txfm2d_16x4_sse2(const int16_t *input, int32_t *output, + int stride, TX_TYPE tx_type, int bd); + +void av1_lowbd_fwd_txfm2d_16x8_sse2(const int16_t *input, int32_t *output, + int stride, TX_TYPE tx_type, int bd); + +void av1_lowbd_fwd_txfm2d_16x16_sse2(const int16_t *input, int32_t *output, + int stride, TX_TYPE tx_type, int bd); + +void av1_lowbd_fwd_txfm2d_16x32_sse2(const int16_t *input, int32_t *output, + int stride, TX_TYPE tx_type, int bd); + +void av1_lowbd_fwd_txfm2d_32x8_sse2(const int16_t *input, int32_t *output, + int stride, TX_TYPE tx_type, int bd); + +void av1_lowbd_fwd_txfm2d_32x16_sse2(const int16_t *input, int32_t *output, + int stride, TX_TYPE tx_type, int bd); + +void av1_lowbd_fwd_txfm2d_32x32_sse2(const int16_t *input, int32_t *output, + int stride, TX_TYPE tx_type, int bd); + +void av1_lowbd_fwd_txfm2d_16x64_sse2(const int16_t *input, int32_t *output, + int stride, TX_TYPE tx_type, int bd); + +void av1_lowbd_fwd_txfm2d_64x16_sse2(const int16_t *input, int32_t *output, + int stride, TX_TYPE tx_type, int bd); + +typedef void (*transform_1d_sse2)(const __m128i *input, __m128i *output, + int8_t cos_bit); + +typedef struct { + transform_1d_sse2 col, row; // vertical and horizontal +} transform_2d_sse2; + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // AOM_AV1_COMMON_X86_AV1_TXFM_SSE2_H_ diff --git a/third_party/aom/av1/common/x86/av1_txfm_sse4.c b/third_party/aom/av1/common/x86/av1_txfm_sse4.c new file mode 100644 index 0000000000..90b9879cc4 --- /dev/null +++ b/third_party/aom/av1/common/x86/av1_txfm_sse4.c @@ -0,0 +1,21 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include "config/aom_dsp_rtcd.h" + +#include "av1/common/av1_txfm.h" +#include "av1/common/x86/av1_txfm_sse4.h" + +void av1_round_shift_array_sse4_1(int32_t *arr, int size, int bit) { + __m128i *const vec = (__m128i *)arr; + const int vec_size = size >> 2; + av1_round_shift_array_32_sse4_1(vec, vec, vec_size, bit); +} diff --git a/third_party/aom/av1/common/x86/av1_txfm_sse4.h b/third_party/aom/av1/common/x86/av1_txfm_sse4.h new file mode 100644 index 0000000000..6cad821b1b --- /dev/null +++ b/third_party/aom/av1/common/x86/av1_txfm_sse4.h @@ -0,0 +1,72 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#ifndef AOM_AV1_COMMON_X86_AV1_TXFM_SSE4_H_ +#define AOM_AV1_COMMON_X86_AV1_TXFM_SSE4_H_ + +#include <smmintrin.h> + +#ifdef __cplusplus +extern "C" { +#endif + +static INLINE __m128i av1_round_shift_32_sse4_1(__m128i vec, int bit) { + __m128i tmp, round; + round = _mm_set1_epi32(1 << (bit - 1)); + tmp = _mm_add_epi32(vec, round); + return _mm_srai_epi32(tmp, bit); +} + +static INLINE void av1_round_shift_array_32_sse4_1(__m128i *input, + __m128i *output, + const int size, + const int bit) { + if (bit > 0) { + int i; + for (i = 0; i < size; i++) { + output[i] = av1_round_shift_32_sse4_1(input[i], bit); + } + } else { + int i; + for (i = 0; i < size; i++) { + output[i] = _mm_slli_epi32(input[i], -bit); + } + } +} + +static INLINE void av1_round_shift_rect_array_32_sse4_1(__m128i *input, + __m128i *output, + const int size, + const int bit, + const int val) { + const __m128i sqrt2 = _mm_set1_epi32(val); + if (bit > 0) { + int i; + for (i = 0; i < size; i++) { + const __m128i r0 = av1_round_shift_32_sse4_1(input[i], bit); + const __m128i r1 = _mm_mullo_epi32(sqrt2, r0); + output[i] = av1_round_shift_32_sse4_1(r1, NewSqrt2Bits); + } + } else { + int i; + for (i = 0; i < size; i++) { + const __m128i r0 = _mm_slli_epi32(input[i], -bit); + const __m128i r1 = _mm_mullo_epi32(sqrt2, r0); + output[i] = av1_round_shift_32_sse4_1(r1, NewSqrt2Bits); + } + } +} + +#ifdef __cplusplus +} +#endif + +#endif // AOM_AV1_COMMON_X86_AV1_TXFM_SSE4_H_ diff --git a/third_party/aom/av1/common/x86/cfl_avx2.c b/third_party/aom/av1/common/x86/cfl_avx2.c new file mode 100644 index 0000000000..a8bfdcce6b --- /dev/null +++ b/third_party/aom/av1/common/x86/cfl_avx2.c @@ -0,0 +1,491 @@ +/* + * Copyright (c) 2017, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ +#include <immintrin.h> + +#include "config/av1_rtcd.h" + +#include "av1/common/cfl.h" + +#include "av1/common/x86/cfl_simd.h" + +#define CFL_GET_SUBSAMPLE_FUNCTION_AVX2(sub, bd) \ + CFL_SUBSAMPLE(avx2, sub, bd, 32, 32) \ + CFL_SUBSAMPLE(avx2, sub, bd, 32, 16) \ + CFL_SUBSAMPLE(avx2, sub, bd, 32, 8) \ + cfl_subsample_##bd##_fn cfl_get_luma_subsampling_##sub##_##bd##_avx2( \ + TX_SIZE tx_size) { \ + static const cfl_subsample_##bd##_fn subfn_##sub[TX_SIZES_ALL] = { \ + subsample_##bd##_##sub##_4x4_ssse3, /* 4x4 */ \ + subsample_##bd##_##sub##_8x8_ssse3, /* 8x8 */ \ + subsample_##bd##_##sub##_16x16_ssse3, /* 16x16 */ \ + subsample_##bd##_##sub##_32x32_avx2, /* 32x32 */ \ + cfl_subsample_##bd##_null, /* 64x64 (invalid CFL size) */ \ + subsample_##bd##_##sub##_4x8_ssse3, /* 4x8 */ \ + subsample_##bd##_##sub##_8x4_ssse3, /* 8x4 */ \ + subsample_##bd##_##sub##_8x16_ssse3, /* 8x16 */ \ + subsample_##bd##_##sub##_16x8_ssse3, /* 16x8 */ \ + subsample_##bd##_##sub##_16x32_ssse3, /* 16x32 */ \ + subsample_##bd##_##sub##_32x16_avx2, /* 32x16 */ \ + cfl_subsample_##bd##_null, /* 32x64 (invalid CFL size) */ \ + cfl_subsample_##bd##_null, /* 64x32 (invalid CFL size) */ \ + subsample_##bd##_##sub##_4x16_ssse3, /* 4x16 */ \ + subsample_##bd##_##sub##_16x4_ssse3, /* 16x4 */ \ + subsample_##bd##_##sub##_8x32_ssse3, /* 8x32 */ \ + subsample_##bd##_##sub##_32x8_avx2, /* 32x8 */ \ + cfl_subsample_##bd##_null, /* 16x64 (invalid CFL size) */ \ + cfl_subsample_##bd##_null, /* 64x16 (invalid CFL size) */ \ + }; \ + return subfn_##sub[tx_size]; \ + } + +/** + * Adds 4 pixels (in a 2x2 grid) and multiplies them by 2. Resulting in a more + * precise version of a box filter 4:2:0 pixel subsampling in Q3. + * + * The CfL prediction buffer is always of size CFL_BUF_SQUARE. However, the + * active area is specified using width and height. + * + * Note: We don't need to worry about going over the active area, as long as we + * stay inside the CfL prediction buffer. + * + * Note: For 4:2:0 luma subsampling, the width will never be greater than 16. + */ +static void cfl_luma_subsampling_420_lbd_avx2(const uint8_t *input, + int input_stride, + uint16_t *pred_buf_q3, int width, + int height) { + (void)width; // Forever 32 + const __m256i twos = _mm256_set1_epi8(2); // Thirty two twos + const int luma_stride = input_stride << 1; + __m256i *row = (__m256i *)pred_buf_q3; + const __m256i *row_end = row + (height >> 1) * CFL_BUF_LINE_I256; + do { + __m256i top = _mm256_loadu_si256((__m256i *)input); + __m256i bot = _mm256_loadu_si256((__m256i *)(input + input_stride)); + + __m256i top_16x16 = _mm256_maddubs_epi16(top, twos); + __m256i bot_16x16 = _mm256_maddubs_epi16(bot, twos); + __m256i sum_16x16 = _mm256_add_epi16(top_16x16, bot_16x16); + + _mm256_storeu_si256(row, sum_16x16); + + input += luma_stride; + } while ((row += CFL_BUF_LINE_I256) < row_end); +} + +CFL_GET_SUBSAMPLE_FUNCTION_AVX2(420, lbd) + +/** + * Adds 2 pixels (in a 2x1 grid) and multiplies them by 4. Resulting in a more + * precise version of a box filter 4:2:2 pixel subsampling in Q3. + * + * The CfL prediction buffer is always of size CFL_BUF_SQUARE. However, the + * active area is specified using width and height. + * + * Note: We don't need to worry about going over the active area, as long as we + * stay inside the CfL prediction buffer. + */ +static void cfl_luma_subsampling_422_lbd_avx2(const uint8_t *input, + int input_stride, + uint16_t *pred_buf_q3, int width, + int height) { + (void)width; // Forever 32 + const __m256i fours = _mm256_set1_epi8(4); // Thirty two fours + __m256i *row = (__m256i *)pred_buf_q3; + const __m256i *row_end = row + height * CFL_BUF_LINE_I256; + do { + __m256i top = _mm256_loadu_si256((__m256i *)input); + __m256i top_16x16 = _mm256_maddubs_epi16(top, fours); + _mm256_storeu_si256(row, top_16x16); + input += input_stride; + } while ((row += CFL_BUF_LINE_I256) < row_end); +} + +CFL_GET_SUBSAMPLE_FUNCTION_AVX2(422, lbd) + +/** + * Multiplies the pixels by 8 (scaling in Q3). The AVX2 subsampling is only + * performed on block of width 32. + * + * The CfL prediction buffer is always of size CFL_BUF_SQUARE. However, the + * active area is specified using width and height. + * + * Note: We don't need to worry about going over the active area, as long as we + * stay inside the CfL prediction buffer. + */ +static void cfl_luma_subsampling_444_lbd_avx2(const uint8_t *input, + int input_stride, + uint16_t *pred_buf_q3, int width, + int height) { + (void)width; // Forever 32 + __m256i *row = (__m256i *)pred_buf_q3; + const __m256i *row_end = row + height * CFL_BUF_LINE_I256; + const __m256i zeros = _mm256_setzero_si256(); + do { + __m256i top = _mm256_loadu_si256((__m256i *)input); + top = _mm256_permute4x64_epi64(top, _MM_SHUFFLE(3, 1, 2, 0)); + + __m256i row_lo = _mm256_unpacklo_epi8(top, zeros); + row_lo = _mm256_slli_epi16(row_lo, 3); + __m256i row_hi = _mm256_unpackhi_epi8(top, zeros); + row_hi = _mm256_slli_epi16(row_hi, 3); + + _mm256_storeu_si256(row, row_lo); + _mm256_storeu_si256(row + 1, row_hi); + + input += input_stride; + } while ((row += CFL_BUF_LINE_I256) < row_end); +} + +CFL_GET_SUBSAMPLE_FUNCTION_AVX2(444, lbd) + +/** + * Adds 4 pixels (in a 2x2 grid) and multiplies them by 2. Resulting in a more + * precise version of a box filter 4:2:0 pixel subsampling in Q3. + * + * The CfL prediction buffer is always of size CFL_BUF_SQUARE. However, the + * active area is specified using width and height. + * + * Note: We don't need to worry about going over the active area, as long as we + * stay inside the CfL prediction buffer. + * + * Note: For 4:2:0 luma subsampling, the width will never be greater than 16. + */ +static void cfl_luma_subsampling_420_hbd_avx2(const uint16_t *input, + int input_stride, + uint16_t *pred_buf_q3, int width, + int height) { + (void)width; // Forever 32 + const int luma_stride = input_stride << 1; + __m256i *row = (__m256i *)pred_buf_q3; + const __m256i *row_end = row + (height >> 1) * CFL_BUF_LINE_I256; + do { + __m256i top = _mm256_loadu_si256((__m256i *)input); + __m256i bot = _mm256_loadu_si256((__m256i *)(input + input_stride)); + __m256i sum = _mm256_add_epi16(top, bot); + + __m256i top_1 = _mm256_loadu_si256((__m256i *)(input + 16)); + __m256i bot_1 = _mm256_loadu_si256((__m256i *)(input + 16 + input_stride)); + __m256i sum_1 = _mm256_add_epi16(top_1, bot_1); + + __m256i hsum = _mm256_hadd_epi16(sum, sum_1); + hsum = _mm256_permute4x64_epi64(hsum, _MM_SHUFFLE(3, 1, 2, 0)); + hsum = _mm256_add_epi16(hsum, hsum); + + _mm256_storeu_si256(row, hsum); + + input += luma_stride; + } while ((row += CFL_BUF_LINE_I256) < row_end); +} + +CFL_GET_SUBSAMPLE_FUNCTION_AVX2(420, hbd) + +/** + * Adds 2 pixels (in a 2x1 grid) and multiplies them by 4. Resulting in a more + * precise version of a box filter 4:2:2 pixel subsampling in Q3. + * + * The CfL prediction buffer is always of size CFL_BUF_SQUARE. However, the + * active area is specified using width and height. + * + * Note: We don't need to worry about going over the active area, as long as we + * stay inside the CfL prediction buffer. + * + */ +static void cfl_luma_subsampling_422_hbd_avx2(const uint16_t *input, + int input_stride, + uint16_t *pred_buf_q3, int width, + int height) { + (void)width; // Forever 32 + __m256i *row = (__m256i *)pred_buf_q3; + const __m256i *row_end = row + height * CFL_BUF_LINE_I256; + do { + __m256i top = _mm256_loadu_si256((__m256i *)input); + __m256i top_1 = _mm256_loadu_si256((__m256i *)(input + 16)); + __m256i hsum = _mm256_hadd_epi16(top, top_1); + hsum = _mm256_permute4x64_epi64(hsum, _MM_SHUFFLE(3, 1, 2, 0)); + hsum = _mm256_slli_epi16(hsum, 2); + + _mm256_storeu_si256(row, hsum); + + input += input_stride; + } while ((row += CFL_BUF_LINE_I256) < row_end); +} + +CFL_GET_SUBSAMPLE_FUNCTION_AVX2(422, hbd) + +static void cfl_luma_subsampling_444_hbd_avx2(const uint16_t *input, + int input_stride, + uint16_t *pred_buf_q3, int width, + int height) { + (void)width; // Forever 32 + __m256i *row = (__m256i *)pred_buf_q3; + const __m256i *row_end = row + height * CFL_BUF_LINE_I256; + do { + __m256i top = _mm256_loadu_si256((__m256i *)input); + __m256i top_1 = _mm256_loadu_si256((__m256i *)(input + 16)); + _mm256_storeu_si256(row, _mm256_slli_epi16(top, 3)); + _mm256_storeu_si256(row + 1, _mm256_slli_epi16(top_1, 3)); + input += input_stride; + } while ((row += CFL_BUF_LINE_I256) < row_end); +} + +CFL_GET_SUBSAMPLE_FUNCTION_AVX2(444, hbd) + +static INLINE __m256i predict_unclipped(const __m256i *input, __m256i alpha_q12, + __m256i alpha_sign, __m256i dc_q0) { + __m256i ac_q3 = _mm256_loadu_si256(input); + __m256i ac_sign = _mm256_sign_epi16(alpha_sign, ac_q3); + __m256i scaled_luma_q0 = + _mm256_mulhrs_epi16(_mm256_abs_epi16(ac_q3), alpha_q12); + scaled_luma_q0 = _mm256_sign_epi16(scaled_luma_q0, ac_sign); + return _mm256_add_epi16(scaled_luma_q0, dc_q0); +} + +static INLINE void cfl_predict_lbd_avx2(const int16_t *pred_buf_q3, + uint8_t *dst, int dst_stride, + int alpha_q3, int width, int height) { + (void)width; + const __m256i alpha_sign = _mm256_set1_epi16(alpha_q3); + const __m256i alpha_q12 = _mm256_slli_epi16(_mm256_abs_epi16(alpha_sign), 9); + const __m256i dc_q0 = _mm256_set1_epi16(*dst); + __m256i *row = (__m256i *)pred_buf_q3; + const __m256i *row_end = row + height * CFL_BUF_LINE_I256; + + do { + __m256i res = predict_unclipped(row, alpha_q12, alpha_sign, dc_q0); + __m256i next = predict_unclipped(row + 1, alpha_q12, alpha_sign, dc_q0); + res = _mm256_packus_epi16(res, next); + res = _mm256_permute4x64_epi64(res, _MM_SHUFFLE(3, 1, 2, 0)); + _mm256_storeu_si256((__m256i *)dst, res); + dst += dst_stride; + } while ((row += CFL_BUF_LINE_I256) < row_end); +} + +CFL_PREDICT_X(avx2, 32, 8, lbd); +CFL_PREDICT_X(avx2, 32, 16, lbd); +CFL_PREDICT_X(avx2, 32, 32, lbd); + +cfl_predict_lbd_fn get_predict_lbd_fn_avx2(TX_SIZE tx_size) { + static const cfl_predict_lbd_fn pred[TX_SIZES_ALL] = { + predict_lbd_4x4_ssse3, /* 4x4 */ + predict_lbd_8x8_ssse3, /* 8x8 */ + predict_lbd_16x16_ssse3, /* 16x16 */ + predict_lbd_32x32_avx2, /* 32x32 */ + cfl_predict_lbd_null, /* 64x64 (invalid CFL size) */ + predict_lbd_4x8_ssse3, /* 4x8 */ + predict_lbd_8x4_ssse3, /* 8x4 */ + predict_lbd_8x16_ssse3, /* 8x16 */ + predict_lbd_16x8_ssse3, /* 16x8 */ + predict_lbd_16x32_ssse3, /* 16x32 */ + predict_lbd_32x16_avx2, /* 32x16 */ + cfl_predict_lbd_null, /* 32x64 (invalid CFL size) */ + cfl_predict_lbd_null, /* 64x32 (invalid CFL size) */ + predict_lbd_4x16_ssse3, /* 4x16 */ + predict_lbd_16x4_ssse3, /* 16x4 */ + predict_lbd_8x32_ssse3, /* 8x32 */ + predict_lbd_32x8_avx2, /* 32x8 */ + cfl_predict_lbd_null, /* 16x64 (invalid CFL size) */ + cfl_predict_lbd_null, /* 64x16 (invalid CFL size) */ + }; + // Modulo TX_SIZES_ALL to ensure that an attacker won't be able to index the + // function pointer array out of bounds. + return pred[tx_size % TX_SIZES_ALL]; +} + +static __m256i highbd_max_epi16(int bd) { + const __m256i neg_one = _mm256_set1_epi16(-1); + // (1 << bd) - 1 => -(-1 << bd) -1 => -1 - (-1 << bd) => -1 ^ (-1 << bd) + return _mm256_xor_si256(_mm256_slli_epi16(neg_one, bd), neg_one); +} + +static __m256i highbd_clamp_epi16(__m256i u, __m256i zero, __m256i max) { + return _mm256_max_epi16(_mm256_min_epi16(u, max), zero); +} + +static INLINE void cfl_predict_hbd_avx2(const int16_t *pred_buf_q3, + uint16_t *dst, int dst_stride, + int alpha_q3, int bd, int width, + int height) { + // Use SSSE3 version for smaller widths + assert(width == 16 || width == 32); + const __m256i alpha_sign = _mm256_set1_epi16(alpha_q3); + const __m256i alpha_q12 = _mm256_slli_epi16(_mm256_abs_epi16(alpha_sign), 9); + const __m256i dc_q0 = _mm256_loadu_si256((__m256i *)dst); + const __m256i max = highbd_max_epi16(bd); + + __m256i *row = (__m256i *)pred_buf_q3; + const __m256i *row_end = row + height * CFL_BUF_LINE_I256; + do { + const __m256i res = predict_unclipped(row, alpha_q12, alpha_sign, dc_q0); + _mm256_storeu_si256((__m256i *)dst, + highbd_clamp_epi16(res, _mm256_setzero_si256(), max)); + if (width == 32) { + const __m256i res_1 = + predict_unclipped(row + 1, alpha_q12, alpha_sign, dc_q0); + _mm256_storeu_si256( + (__m256i *)(dst + 16), + highbd_clamp_epi16(res_1, _mm256_setzero_si256(), max)); + } + dst += dst_stride; + } while ((row += CFL_BUF_LINE_I256) < row_end); +} + +CFL_PREDICT_X(avx2, 16, 4, hbd) +CFL_PREDICT_X(avx2, 16, 8, hbd) +CFL_PREDICT_X(avx2, 16, 16, hbd) +CFL_PREDICT_X(avx2, 16, 32, hbd) +CFL_PREDICT_X(avx2, 32, 8, hbd) +CFL_PREDICT_X(avx2, 32, 16, hbd) +CFL_PREDICT_X(avx2, 32, 32, hbd) + +cfl_predict_hbd_fn get_predict_hbd_fn_avx2(TX_SIZE tx_size) { + static const cfl_predict_hbd_fn pred[TX_SIZES_ALL] = { + predict_hbd_4x4_ssse3, /* 4x4 */ + predict_hbd_8x8_ssse3, /* 8x8 */ + predict_hbd_16x16_avx2, /* 16x16 */ + predict_hbd_32x32_avx2, /* 32x32 */ + cfl_predict_hbd_null, /* 64x64 (invalid CFL size) */ + predict_hbd_4x8_ssse3, /* 4x8 */ + predict_hbd_8x4_ssse3, /* 8x4 */ + predict_hbd_8x16_ssse3, /* 8x16 */ + predict_hbd_16x8_avx2, /* 16x8 */ + predict_hbd_16x32_avx2, /* 16x32 */ + predict_hbd_32x16_avx2, /* 32x16 */ + cfl_predict_hbd_null, /* 32x64 (invalid CFL size) */ + cfl_predict_hbd_null, /* 64x32 (invalid CFL size) */ + predict_hbd_4x16_ssse3, /* 4x16 */ + predict_hbd_16x4_avx2, /* 16x4 */ + predict_hbd_8x32_ssse3, /* 8x32 */ + predict_hbd_32x8_avx2, /* 32x8 */ + cfl_predict_hbd_null, /* 16x64 (invalid CFL size) */ + cfl_predict_hbd_null, /* 64x16 (invalid CFL size) */ + }; + // Modulo TX_SIZES_ALL to ensure that an attacker won't be able to index the + // function pointer array out of bounds. + return pred[tx_size % TX_SIZES_ALL]; +} + +// Returns a vector where all the (32-bits) elements are the sum of all the +// lanes in a. +static INLINE __m256i fill_sum_epi32(__m256i a) { + // Given that a == [A, B, C, D, E, F, G, H] + a = _mm256_hadd_epi32(a, a); + // Given that A' == A + B, C' == C + D, E' == E + F, G' == G + H + // a == [A', C', A', C', E', G', E', G'] + a = _mm256_permute4x64_epi64(a, _MM_SHUFFLE(3, 1, 2, 0)); + // a == [A', C', E', G', A', C', E', G'] + a = _mm256_hadd_epi32(a, a); + // Given that A'' == A' + C' and E'' == E' + G' + // a == [A'', E'', A'', E'', A'', E'', A'', E''] + return _mm256_hadd_epi32(a, a); + // Given that A''' == A'' + E'' + // a == [A''', A''', A''', A''', A''', A''', A''', A'''] +} + +static INLINE __m256i _mm256_addl_epi16(__m256i a) { + return _mm256_add_epi32(_mm256_unpacklo_epi16(a, _mm256_setzero_si256()), + _mm256_unpackhi_epi16(a, _mm256_setzero_si256())); +} + +static INLINE void subtract_average_avx2(const uint16_t *src_ptr, + int16_t *dst_ptr, int width, + int height, int round_offset, + int num_pel_log2) { + // Use SSE2 version for smaller widths + assert(width == 16 || width == 32); + + const __m256i *src = (__m256i *)src_ptr; + const __m256i *const end = src + height * CFL_BUF_LINE_I256; + // To maximize usage of the AVX2 registers, we sum two rows per loop + // iteration + const int step = 2 * CFL_BUF_LINE_I256; + + __m256i sum = _mm256_setzero_si256(); + // For width 32, we use a second sum accumulator to reduce accumulator + // dependencies in the loop. + __m256i sum2; + if (width == 32) sum2 = _mm256_setzero_si256(); + + do { + // Add top row to the bottom row + __m256i l0 = _mm256_add_epi16(_mm256_loadu_si256(src), + _mm256_loadu_si256(src + CFL_BUF_LINE_I256)); + sum = _mm256_add_epi32(sum, _mm256_addl_epi16(l0)); + if (width == 32) { /* Don't worry, this if it gets optimized out. */ + // Add the second part of the top row to the second part of the bottom row + __m256i l1 = + _mm256_add_epi16(_mm256_loadu_si256(src + 1), + _mm256_loadu_si256(src + 1 + CFL_BUF_LINE_I256)); + sum2 = _mm256_add_epi32(sum2, _mm256_addl_epi16(l1)); + } + src += step; + } while (src < end); + // Combine both sum accumulators + if (width == 32) sum = _mm256_add_epi32(sum, sum2); + + __m256i fill = fill_sum_epi32(sum); + + __m256i avg_epi16 = _mm256_srli_epi32( + _mm256_add_epi32(fill, _mm256_set1_epi32(round_offset)), num_pel_log2); + avg_epi16 = _mm256_packs_epi32(avg_epi16, avg_epi16); + + // Store and subtract loop + src = (__m256i *)src_ptr; + __m256i *dst = (__m256i *)dst_ptr; + do { + _mm256_storeu_si256(dst, + _mm256_sub_epi16(_mm256_loadu_si256(src), avg_epi16)); + if (width == 32) { + _mm256_storeu_si256( + dst + 1, _mm256_sub_epi16(_mm256_loadu_si256(src + 1), avg_epi16)); + } + src += CFL_BUF_LINE_I256; + dst += CFL_BUF_LINE_I256; + } while (src < end); +} + +// Declare wrappers for AVX2 sizes +CFL_SUB_AVG_X(avx2, 16, 4, 32, 6) +CFL_SUB_AVG_X(avx2, 16, 8, 64, 7) +CFL_SUB_AVG_X(avx2, 16, 16, 128, 8) +CFL_SUB_AVG_X(avx2, 16, 32, 256, 9) +CFL_SUB_AVG_X(avx2, 32, 8, 128, 8) +CFL_SUB_AVG_X(avx2, 32, 16, 256, 9) +CFL_SUB_AVG_X(avx2, 32, 32, 512, 10) + +// Based on the observation that for small blocks AVX2 does not outperform +// SSE2, we call the SSE2 code for block widths 4 and 8. +cfl_subtract_average_fn get_subtract_average_fn_avx2(TX_SIZE tx_size) { + static const cfl_subtract_average_fn sub_avg[TX_SIZES_ALL] = { + subtract_average_4x4_sse2, /* 4x4 */ + subtract_average_8x8_sse2, /* 8x8 */ + subtract_average_16x16_avx2, /* 16x16 */ + subtract_average_32x32_avx2, /* 32x32 */ + cfl_subtract_average_null, /* 64x64 (invalid CFL size) */ + subtract_average_4x8_sse2, /* 4x8 */ + subtract_average_8x4_sse2, /* 8x4 */ + subtract_average_8x16_sse2, /* 8x16 */ + subtract_average_16x8_avx2, /* 16x8 */ + subtract_average_16x32_avx2, /* 16x32 */ + subtract_average_32x16_avx2, /* 32x16 */ + cfl_subtract_average_null, /* 32x64 (invalid CFL size) */ + cfl_subtract_average_null, /* 64x32 (invalid CFL size) */ + subtract_average_4x16_sse2, /* 4x16 */ + subtract_average_16x4_avx2, /* 16x4 */ + subtract_average_8x32_sse2, /* 8x32 */ + subtract_average_32x8_avx2, /* 32x8 */ + cfl_subtract_average_null, /* 16x64 (invalid CFL size) */ + cfl_subtract_average_null, /* 64x16 (invalid CFL size) */ + }; + // Modulo TX_SIZES_ALL to ensure that an attacker won't be able to + // index the function pointer array out of bounds. + return sub_avg[tx_size % TX_SIZES_ALL]; +} diff --git a/third_party/aom/av1/common/x86/cfl_simd.h b/third_party/aom/av1/common/x86/cfl_simd.h new file mode 100644 index 0000000000..3b342cd4e4 --- /dev/null +++ b/third_party/aom/av1/common/x86/cfl_simd.h @@ -0,0 +1,243 @@ +/* + * Copyright (c) 2017, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#ifndef AOM_AV1_COMMON_X86_CFL_SIMD_H_ +#define AOM_AV1_COMMON_X86_CFL_SIMD_H_ + +#include "av1/common/blockd.h" + +// SSSE3 version is optimal for with == 4, we reuse them in AVX2 +void subsample_lbd_420_4x4_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_420_4x8_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_420_4x16_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); + +// SSSE3 version is optimal for with == 8, we reuse it in AVX2 +void subsample_lbd_420_8x4_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_420_8x8_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_420_8x16_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_420_8x32_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); + +// SSSE3 version is optimal for with == 16, we reuse it in AVX2 +void subsample_lbd_420_16x4_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_420_16x8_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_420_16x16_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_420_16x32_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); + +// SSSE3 version is optimal for with == 4, we reuse them in AVX2 +void subsample_lbd_422_4x4_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_422_4x8_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_422_4x16_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); + +// SSSE3 version is optimal for with == 8, we reuse it in AVX2 +void subsample_lbd_422_8x4_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_422_8x8_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_422_8x16_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_422_8x32_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); + +// SSSE3 version is optimal for with == 16, we reuse it in AVX2 +void subsample_lbd_422_16x4_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_422_16x8_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_422_16x16_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_422_16x32_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); + +// SSSE3 version is optimal for with == 4, we reuse them in AVX2 +void subsample_lbd_444_4x4_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_444_4x8_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_444_4x16_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); + +// SSSE3 version is optimal for with == 8, we reuse it in AVX2 +void subsample_lbd_444_8x4_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_444_8x8_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_444_8x16_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_444_8x32_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); + +// SSSE3 version is optimal for with == 16, we reuse it in AVX2 +void subsample_lbd_444_16x4_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_444_16x8_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_444_16x16_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); +void subsample_lbd_444_16x32_ssse3(const uint8_t *input, int input_stride, + uint16_t *output_q3); + +void subsample_hbd_420_4x4_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_420_4x8_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_420_4x16_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); + +// SSSE3 version is optimal for with == 8, we reuse it in AVX2 +void subsample_hbd_420_8x4_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_420_8x8_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_420_8x16_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_420_8x32_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); + +// SSSE3 version is faster for with == 16, we reuse it in AVX2 +void subsample_hbd_420_16x4_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_420_16x8_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_420_16x16_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_420_16x32_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); + +void subsample_hbd_422_4x4_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_422_4x8_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_422_4x16_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); + +// SSSE3 version is optimal for with == 8, we reuse it in AVX2 +void subsample_hbd_422_8x4_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_422_8x8_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_422_8x16_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_422_8x32_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); + +// SSSE3 version is faster for with == 16, we reuse it in AVX2 +void subsample_hbd_422_16x4_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_422_16x8_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_422_16x16_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_422_16x32_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); + +void subsample_hbd_444_4x4_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_444_4x8_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_444_4x16_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); + +// SSSE3 version is optimal for with == 8, we reuse it in AVX2 +void subsample_hbd_444_8x4_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_444_8x8_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_444_8x16_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_444_8x32_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); + +// SSSE3 version is faster for with == 16, we reuse it in AVX2 +void subsample_hbd_444_16x4_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_444_16x8_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_444_16x16_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); +void subsample_hbd_444_16x32_ssse3(const uint16_t *input, int input_stride, + uint16_t *output_q3); + +// SSE2 version is optimal for with == 4, we reuse them in AVX2 +void subtract_average_4x4_sse2(const uint16_t *src, int16_t *dst); +void subtract_average_4x8_sse2(const uint16_t *src, int16_t *dst); +void subtract_average_4x16_sse2(const uint16_t *src, int16_t *dst); + +// SSE2 version is optimal for with == 8, we reuse them in AVX2 +void subtract_average_8x4_sse2(const uint16_t *src, int16_t *dst); +void subtract_average_8x8_sse2(const uint16_t *src, int16_t *dst); +void subtract_average_8x16_sse2(const uint16_t *src, int16_t *dst); +void subtract_average_8x32_sse2(const uint16_t *src, int16_t *dst); + +void predict_lbd_4x4_ssse3(const int16_t *pred_buf_q3, uint8_t *dst, + int dst_stride, int alpha_q3); +void predict_lbd_4x8_ssse3(const int16_t *pred_buf_q3, uint8_t *dst, + int dst_stride, int alpha_q3); +void predict_lbd_4x16_ssse3(const int16_t *pred_buf_q3, uint8_t *dst, + int dst_stride, int alpha_q3); + +void predict_lbd_8x4_ssse3(const int16_t *pred_buf_q3, uint8_t *dst, + int dst_stride, int alpha_q3); +void predict_lbd_8x8_ssse3(const int16_t *pred_buf_q3, uint8_t *dst, + int dst_stride, int alpha_q3); +void predict_lbd_8x16_ssse3(const int16_t *pred_buf_q3, uint8_t *dst, + int dst_stride, int alpha_q3); +void predict_lbd_8x32_ssse3(const int16_t *pred_buf_q3, uint8_t *dst, + int dst_stride, int alpha_q3); + +void predict_lbd_16x4_ssse3(const int16_t *pred_buf_q3, uint8_t *dst, + int dst_stride, int alpha_q3); +void predict_lbd_16x8_ssse3(const int16_t *pred_buf_q3, uint8_t *dst, + int dst_stride, int alpha_q3); +void predict_lbd_16x16_ssse3(const int16_t *pred_buf_q3, uint8_t *dst, + int dst_stride, int alpha_q3); +void predict_lbd_16x32_ssse3(const int16_t *pred_buf_q3, uint8_t *dst, + int dst_stride, int alpha_q3); + +void predict_hbd_4x4_ssse3(const int16_t *pred_buf_q3, uint16_t *dst, + int dst_stride, int alpha_q3, int bd); +void predict_hbd_4x8_ssse3(const int16_t *pred_buf_q3, uint16_t *dst, + int dst_stride, int alpha_q3, int bd); +void predict_hbd_4x16_ssse3(const int16_t *pred_buf_q3, uint16_t *dst, + int dst_stride, int alpha_q3, int bd); + +void predict_hbd_8x4_ssse3(const int16_t *pred_buf_q3, uint16_t *dst, + int dst_stride, int alpha_q3, int bd); +void predict_hbd_8x8_ssse3(const int16_t *pred_buf_q3, uint16_t *dst, + int dst_stride, int alpha_q3, int bd); +void predict_hbd_8x16_ssse3(const int16_t *pred_buf_q3, uint16_t *dst, + int dst_stride, int alpha_q3, int bd); +void predict_hbd_8x32_ssse3(const int16_t *pred_buf_q3, uint16_t *dst, + int dst_stride, int alpha_q3, int bd); + +void predict_hbd_16x4_ssse3(const int16_t *pred_buf_q3, uint16_t *dst, + int dst_stride, int alpha_q3, int bd); +void predict_hbd_16x8_ssse3(const int16_t *pred_buf_q3, uint16_t *dst, + int dst_stride, int alpha_q3, int bd); +void predict_hbd_16x16_ssse3(const int16_t *pred_buf_q3, uint16_t *dst, + int dst_stride, int alpha_q3, int bd); +void predict_hbd_16x32_ssse3(const int16_t *pred_buf_q3, uint16_t *dst, + int dst_stride, int alpha_q3, int bd); + +#endif // AOM_AV1_COMMON_X86_CFL_SIMD_H_ diff --git a/third_party/aom/av1/common/x86/cfl_sse2.c b/third_party/aom/av1/common/x86/cfl_sse2.c new file mode 100644 index 0000000000..4783fe098c --- /dev/null +++ b/third_party/aom/av1/common/x86/cfl_sse2.c @@ -0,0 +1,89 @@ +/* + * Copyright (c) 2017, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <emmintrin.h> + +#include "av1/common/cfl.h" +#include "config/av1_rtcd.h" + +static INLINE __m128i fill_sum_epi32(__m128i l0) { + l0 = _mm_add_epi32(l0, _mm_shuffle_epi32(l0, _MM_SHUFFLE(1, 0, 3, 2))); + return _mm_add_epi32(l0, _mm_shuffle_epi32(l0, _MM_SHUFFLE(2, 3, 0, 1))); +} + +static INLINE void subtract_average_sse2(const uint16_t *src_ptr, + int16_t *dst_ptr, int width, + int height, int round_offset, + int num_pel_log2) { + const __m128i zeros = _mm_setzero_si128(); + const __m128i round_offset_epi32 = _mm_set1_epi32(round_offset); + const __m128i *src = (__m128i *)src_ptr; + const __m128i *const end = src + height * CFL_BUF_LINE_I128; + const int step = CFL_BUF_LINE_I128 * (1 + (width == 8) + 3 * (width == 4)); + + __m128i sum = zeros; + do { + __m128i l0; + if (width == 4) { + l0 = _mm_add_epi16(_mm_loadl_epi64(src), + _mm_loadl_epi64(src + CFL_BUF_LINE_I128)); + __m128i l1 = _mm_add_epi16(_mm_loadl_epi64(src + 2 * CFL_BUF_LINE_I128), + _mm_loadl_epi64(src + 3 * CFL_BUF_LINE_I128)); + sum = _mm_add_epi32(sum, _mm_add_epi32(_mm_unpacklo_epi16(l0, zeros), + _mm_unpacklo_epi16(l1, zeros))); + } else { + if (width == 8) { + l0 = _mm_add_epi16(_mm_loadu_si128(src), + _mm_loadu_si128(src + CFL_BUF_LINE_I128)); + } else { + l0 = _mm_add_epi16(_mm_loadu_si128(src), _mm_loadu_si128(src + 1)); + } + sum = _mm_add_epi32(sum, _mm_add_epi32(_mm_unpacklo_epi16(l0, zeros), + _mm_unpackhi_epi16(l0, zeros))); + if (width == 32) { + l0 = _mm_add_epi16(_mm_loadu_si128(src + 2), _mm_loadu_si128(src + 3)); + sum = _mm_add_epi32(sum, _mm_add_epi32(_mm_unpacklo_epi16(l0, zeros), + _mm_unpackhi_epi16(l0, zeros))); + } + } + src += step; + } while (src < end); + + sum = fill_sum_epi32(sum); + + __m128i avg_epi16 = + _mm_srli_epi32(_mm_add_epi32(sum, round_offset_epi32), num_pel_log2); + avg_epi16 = _mm_packs_epi32(avg_epi16, avg_epi16); + + src = (__m128i *)src_ptr; + __m128i *dst = (__m128i *)dst_ptr; + do { + if (width == 4) { + _mm_storel_epi64(dst, _mm_sub_epi16(_mm_loadl_epi64(src), avg_epi16)); + } else { + _mm_storeu_si128(dst, _mm_sub_epi16(_mm_loadu_si128(src), avg_epi16)); + if (width > 8) { + _mm_storeu_si128(dst + 1, + _mm_sub_epi16(_mm_loadu_si128(src + 1), avg_epi16)); + if (width == 32) { + _mm_storeu_si128(dst + 2, + _mm_sub_epi16(_mm_loadu_si128(src + 2), avg_epi16)); + _mm_storeu_si128(dst + 3, + _mm_sub_epi16(_mm_loadu_si128(src + 3), avg_epi16)); + } + } + } + src += CFL_BUF_LINE_I128; + dst += CFL_BUF_LINE_I128; + } while (src < end); +} + +CFL_SUB_AVG_FN(sse2) diff --git a/third_party/aom/av1/common/x86/cfl_ssse3.c b/third_party/aom/av1/common/x86/cfl_ssse3.c new file mode 100644 index 0000000000..bbf0072955 --- /dev/null +++ b/third_party/aom/av1/common/x86/cfl_ssse3.c @@ -0,0 +1,393 @@ +/* + * Copyright (c) 2017, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <tmmintrin.h> + +#include "config/av1_rtcd.h" + +#include "av1/common/cfl.h" + +#include "av1/common/x86/cfl_simd.h" + +// Load 32-bit integer from memory into the first element of dst. +static INLINE __m128i _mm_loadh_epi32(__m128i const *mem_addr) { + return _mm_cvtsi32_si128(*((int *)mem_addr)); +} + +// Store 32-bit integer from the first element of a into memory. +static INLINE void _mm_storeh_epi32(__m128i const *mem_addr, __m128i a) { + *((int *)mem_addr) = _mm_cvtsi128_si32(a); +} + +/** + * Adds 4 pixels (in a 2x2 grid) and multiplies them by 2. Resulting in a more + * precise version of a box filter 4:2:0 pixel subsampling in Q3. + * + * The CfL prediction buffer is always of size CFL_BUF_SQUARE. However, the + * active area is specified using width and height. + * + * Note: We don't need to worry about going over the active area, as long as we + * stay inside the CfL prediction buffer. + */ +static INLINE void cfl_luma_subsampling_420_lbd_ssse3(const uint8_t *input, + int input_stride, + uint16_t *pred_buf_q3, + int width, int height) { + const __m128i twos = _mm_set1_epi8(2); + __m128i *pred_buf_m128i = (__m128i *)pred_buf_q3; + const __m128i *end = pred_buf_m128i + (height >> 1) * CFL_BUF_LINE_I128; + const int luma_stride = input_stride << 1; + do { + if (width == 4) { + __m128i top = _mm_loadh_epi32((__m128i *)input); + top = _mm_maddubs_epi16(top, twos); + __m128i bot = _mm_loadh_epi32((__m128i *)(input + input_stride)); + bot = _mm_maddubs_epi16(bot, twos); + const __m128i sum = _mm_add_epi16(top, bot); + _mm_storeh_epi32(pred_buf_m128i, sum); + } else if (width == 8) { + __m128i top = _mm_loadl_epi64((__m128i *)input); + top = _mm_maddubs_epi16(top, twos); + __m128i bot = _mm_loadl_epi64((__m128i *)(input + input_stride)); + bot = _mm_maddubs_epi16(bot, twos); + const __m128i sum = _mm_add_epi16(top, bot); + _mm_storel_epi64(pred_buf_m128i, sum); + } else { + __m128i top = _mm_loadu_si128((__m128i *)input); + top = _mm_maddubs_epi16(top, twos); + __m128i bot = _mm_loadu_si128((__m128i *)(input + input_stride)); + bot = _mm_maddubs_epi16(bot, twos); + const __m128i sum = _mm_add_epi16(top, bot); + _mm_storeu_si128(pred_buf_m128i, sum); + if (width == 32) { + __m128i top_1 = _mm_loadu_si128(((__m128i *)input) + 1); + __m128i bot_1 = + _mm_loadu_si128(((__m128i *)(input + input_stride)) + 1); + top_1 = _mm_maddubs_epi16(top_1, twos); + bot_1 = _mm_maddubs_epi16(bot_1, twos); + __m128i sum_1 = _mm_add_epi16(top_1, bot_1); + _mm_storeu_si128(pred_buf_m128i + 1, sum_1); + } + } + input += luma_stride; + pred_buf_m128i += CFL_BUF_LINE_I128; + } while (pred_buf_m128i < end); +} + +/** + * Adds 2 pixels (in a 2x1 grid) and multiplies them by 4. Resulting in a more + * precise version of a box filter 4:2:2 pixel subsampling in Q3. + * + * The CfL prediction buffer is always of size CFL_BUF_SQUARE. However, the + * active area is specified using width and height. + * + * Note: We don't need to worry about going over the active area, as long as we + * stay inside the CfL prediction buffer. + */ +static INLINE void cfl_luma_subsampling_422_lbd_ssse3(const uint8_t *input, + int input_stride, + uint16_t *pred_buf_q3, + int width, int height) { + const __m128i fours = _mm_set1_epi8(4); + __m128i *pred_buf_m128i = (__m128i *)pred_buf_q3; + const __m128i *end = pred_buf_m128i + height * CFL_BUF_LINE_I128; + do { + if (width == 4) { + __m128i top = _mm_loadh_epi32((__m128i *)input); + top = _mm_maddubs_epi16(top, fours); + _mm_storeh_epi32(pred_buf_m128i, top); + } else if (width == 8) { + __m128i top = _mm_loadl_epi64((__m128i *)input); + top = _mm_maddubs_epi16(top, fours); + _mm_storel_epi64(pred_buf_m128i, top); + } else { + __m128i top = _mm_loadu_si128((__m128i *)input); + top = _mm_maddubs_epi16(top, fours); + _mm_storeu_si128(pred_buf_m128i, top); + if (width == 32) { + __m128i top_1 = _mm_loadu_si128(((__m128i *)input) + 1); + top_1 = _mm_maddubs_epi16(top_1, fours); + _mm_storeu_si128(pred_buf_m128i + 1, top_1); + } + } + input += input_stride; + pred_buf_m128i += CFL_BUF_LINE_I128; + } while (pred_buf_m128i < end); +} + +/** + * Multiplies the pixels by 8 (scaling in Q3). + * + * The CfL prediction buffer is always of size CFL_BUF_SQUARE. However, the + * active area is specified using width and height. + * + * Note: We don't need to worry about going over the active area, as long as we + * stay inside the CfL prediction buffer. + */ +static INLINE void cfl_luma_subsampling_444_lbd_ssse3(const uint8_t *input, + int input_stride, + uint16_t *pred_buf_q3, + int width, int height) { + const __m128i zeros = _mm_setzero_si128(); + const int luma_stride = input_stride; + __m128i *pred_buf_m128i = (__m128i *)pred_buf_q3; + const __m128i *end = pred_buf_m128i + height * CFL_BUF_LINE_I128; + do { + if (width == 4) { + __m128i row = _mm_loadh_epi32((__m128i *)input); + row = _mm_unpacklo_epi8(row, zeros); + _mm_storel_epi64(pred_buf_m128i, _mm_slli_epi16(row, 3)); + } else if (width == 8) { + __m128i row = _mm_loadl_epi64((__m128i *)input); + row = _mm_unpacklo_epi8(row, zeros); + _mm_storeu_si128(pred_buf_m128i, _mm_slli_epi16(row, 3)); + } else { + __m128i row = _mm_loadu_si128((__m128i *)input); + const __m128i row_lo = _mm_unpacklo_epi8(row, zeros); + const __m128i row_hi = _mm_unpackhi_epi8(row, zeros); + _mm_storeu_si128(pred_buf_m128i, _mm_slli_epi16(row_lo, 3)); + _mm_storeu_si128(pred_buf_m128i + 1, _mm_slli_epi16(row_hi, 3)); + if (width == 32) { + __m128i row_1 = _mm_loadu_si128(((__m128i *)input) + 1); + const __m128i row_1_lo = _mm_unpacklo_epi8(row_1, zeros); + const __m128i row_1_hi = _mm_unpackhi_epi8(row_1, zeros); + _mm_storeu_si128(pred_buf_m128i + 2, _mm_slli_epi16(row_1_lo, 3)); + _mm_storeu_si128(pred_buf_m128i + 3, _mm_slli_epi16(row_1_hi, 3)); + } + } + input += luma_stride; + pred_buf_m128i += CFL_BUF_LINE_I128; + } while (pred_buf_m128i < end); +} + +/** + * Adds 4 pixels (in a 2x2 grid) and multiplies them by 2. Resulting in a more + * precise version of a box filter 4:2:0 pixel subsampling in Q3. + * + * The CfL prediction buffer is always of size CFL_BUF_SQUARE. However, the + * active area is specified using width and height. + * + * Note: We don't need to worry about going over the active area, as long as we + * stay inside the CfL prediction buffer. + */ +static INLINE void cfl_luma_subsampling_420_hbd_ssse3(const uint16_t *input, + int input_stride, + uint16_t *pred_buf_q3, + int width, int height) { + const uint16_t *end = pred_buf_q3 + (height >> 1) * CFL_BUF_LINE; + const int luma_stride = input_stride << 1; + do { + if (width == 4) { + const __m128i top = _mm_loadl_epi64((__m128i *)input); + const __m128i bot = _mm_loadl_epi64((__m128i *)(input + input_stride)); + __m128i sum = _mm_add_epi16(top, bot); + sum = _mm_hadd_epi16(sum, sum); + *((int *)pred_buf_q3) = _mm_cvtsi128_si32(_mm_add_epi16(sum, sum)); + } else { + const __m128i top = _mm_loadu_si128((__m128i *)input); + const __m128i bot = _mm_loadu_si128((__m128i *)(input + input_stride)); + __m128i sum = _mm_add_epi16(top, bot); + if (width == 8) { + sum = _mm_hadd_epi16(sum, sum); + _mm_storel_epi64((__m128i *)pred_buf_q3, _mm_add_epi16(sum, sum)); + } else { + const __m128i top_1 = _mm_loadu_si128(((__m128i *)input) + 1); + const __m128i bot_1 = + _mm_loadu_si128(((__m128i *)(input + input_stride)) + 1); + sum = _mm_hadd_epi16(sum, _mm_add_epi16(top_1, bot_1)); + _mm_storeu_si128((__m128i *)pred_buf_q3, _mm_add_epi16(sum, sum)); + if (width == 32) { + const __m128i top_2 = _mm_loadu_si128(((__m128i *)input) + 2); + const __m128i bot_2 = + _mm_loadu_si128(((__m128i *)(input + input_stride)) + 2); + const __m128i top_3 = _mm_loadu_si128(((__m128i *)input) + 3); + const __m128i bot_3 = + _mm_loadu_si128(((__m128i *)(input + input_stride)) + 3); + const __m128i sum_2 = _mm_add_epi16(top_2, bot_2); + const __m128i sum_3 = _mm_add_epi16(top_3, bot_3); + __m128i next_sum = _mm_hadd_epi16(sum_2, sum_3); + _mm_storeu_si128(((__m128i *)pred_buf_q3) + 1, + _mm_add_epi16(next_sum, next_sum)); + } + } + } + input += luma_stride; + } while ((pred_buf_q3 += CFL_BUF_LINE) < end); +} + +/** + * Adds 2 pixels (in a 2x1 grid) and multiplies them by 4. Resulting in a more + * precise version of a box filter 4:2:2 pixel subsampling in Q3. + * + * The CfL prediction buffer is always of size CFL_BUF_SQUARE. However, the + * active area is specified using width and height. + * + * Note: We don't need to worry about going over the active area, as long as we + * stay inside the CfL prediction buffer. + */ +static INLINE void cfl_luma_subsampling_422_hbd_ssse3(const uint16_t *input, + int input_stride, + uint16_t *pred_buf_q3, + int width, int height) { + __m128i *pred_buf_m128i = (__m128i *)pred_buf_q3; + const __m128i *end = pred_buf_m128i + height * CFL_BUF_LINE_I128; + do { + if (width == 4) { + const __m128i top = _mm_loadl_epi64((__m128i *)input); + const __m128i sum = _mm_slli_epi16(_mm_hadd_epi16(top, top), 2); + _mm_storeh_epi32(pred_buf_m128i, sum); + } else { + const __m128i top = _mm_loadu_si128((__m128i *)input); + if (width == 8) { + const __m128i sum = _mm_slli_epi16(_mm_hadd_epi16(top, top), 2); + _mm_storel_epi64(pred_buf_m128i, sum); + } else { + const __m128i top_1 = _mm_loadu_si128(((__m128i *)input) + 1); + const __m128i sum = _mm_slli_epi16(_mm_hadd_epi16(top, top_1), 2); + _mm_storeu_si128(pred_buf_m128i, sum); + if (width == 32) { + const __m128i top_2 = _mm_loadu_si128(((__m128i *)input) + 2); + const __m128i top_3 = _mm_loadu_si128(((__m128i *)input) + 3); + const __m128i sum_1 = _mm_slli_epi16(_mm_hadd_epi16(top_2, top_3), 2); + _mm_storeu_si128(pred_buf_m128i + 1, sum_1); + } + } + } + pred_buf_m128i += CFL_BUF_LINE_I128; + input += input_stride; + } while (pred_buf_m128i < end); +} + +static INLINE void cfl_luma_subsampling_444_hbd_ssse3(const uint16_t *input, + int input_stride, + uint16_t *pred_buf_q3, + int width, int height) { + const uint16_t *end = pred_buf_q3 + height * CFL_BUF_LINE; + do { + if (width == 4) { + const __m128i row = _mm_slli_epi16(_mm_loadl_epi64((__m128i *)input), 3); + _mm_storel_epi64((__m128i *)pred_buf_q3, row); + } else { + const __m128i row = _mm_slli_epi16(_mm_loadu_si128((__m128i *)input), 3); + _mm_storeu_si128((__m128i *)pred_buf_q3, row); + if (width >= 16) { + __m128i row_1 = _mm_loadu_si128(((__m128i *)input) + 1); + row_1 = _mm_slli_epi16(row_1, 3); + _mm_storeu_si128(((__m128i *)pred_buf_q3) + 1, row_1); + if (width == 32) { + __m128i row_2 = _mm_loadu_si128(((__m128i *)input) + 2); + row_2 = _mm_slli_epi16(row_2, 3); + _mm_storeu_si128(((__m128i *)pred_buf_q3) + 2, row_2); + __m128i row_3 = _mm_loadu_si128(((__m128i *)input) + 3); + row_3 = _mm_slli_epi16(row_3, 3); + _mm_storeu_si128(((__m128i *)pred_buf_q3) + 3, row_3); + } + } + } + input += input_stride; + pred_buf_q3 += CFL_BUF_LINE; + } while (pred_buf_q3 < end); +} + +CFL_GET_SUBSAMPLE_FUNCTION(ssse3) + +static INLINE __m128i predict_unclipped(const __m128i *input, __m128i alpha_q12, + __m128i alpha_sign, __m128i dc_q0) { + __m128i ac_q3 = _mm_loadu_si128(input); + __m128i ac_sign = _mm_sign_epi16(alpha_sign, ac_q3); + __m128i scaled_luma_q0 = _mm_mulhrs_epi16(_mm_abs_epi16(ac_q3), alpha_q12); + scaled_luma_q0 = _mm_sign_epi16(scaled_luma_q0, ac_sign); + return _mm_add_epi16(scaled_luma_q0, dc_q0); +} + +static INLINE void cfl_predict_lbd_ssse3(const int16_t *pred_buf_q3, + uint8_t *dst, int dst_stride, + int alpha_q3, int width, int height) { + const __m128i alpha_sign = _mm_set1_epi16(alpha_q3); + const __m128i alpha_q12 = _mm_slli_epi16(_mm_abs_epi16(alpha_sign), 9); + const __m128i dc_q0 = _mm_set1_epi16(*dst); + __m128i *row = (__m128i *)pred_buf_q3; + const __m128i *row_end = row + height * CFL_BUF_LINE_I128; + do { + __m128i res = predict_unclipped(row, alpha_q12, alpha_sign, dc_q0); + if (width < 16) { + res = _mm_packus_epi16(res, res); + if (width == 4) + _mm_storeh_epi32((__m128i *)dst, res); + else + _mm_storel_epi64((__m128i *)dst, res); + } else { + __m128i next = predict_unclipped(row + 1, alpha_q12, alpha_sign, dc_q0); + res = _mm_packus_epi16(res, next); + _mm_storeu_si128((__m128i *)dst, res); + if (width == 32) { + res = predict_unclipped(row + 2, alpha_q12, alpha_sign, dc_q0); + next = predict_unclipped(row + 3, alpha_q12, alpha_sign, dc_q0); + res = _mm_packus_epi16(res, next); + _mm_storeu_si128((__m128i *)(dst + 16), res); + } + } + dst += dst_stride; + } while ((row += CFL_BUF_LINE_I128) < row_end); +} + +CFL_PREDICT_FN(ssse3, lbd) + +static INLINE __m128i highbd_max_epi16(int bd) { + const __m128i neg_one = _mm_set1_epi16(-1); + // (1 << bd) - 1 => -(-1 << bd) -1 => -1 - (-1 << bd) => -1 ^ (-1 << bd) + return _mm_xor_si128(_mm_slli_epi16(neg_one, bd), neg_one); +} + +static INLINE __m128i highbd_clamp_epi16(__m128i u, __m128i zero, __m128i max) { + return _mm_max_epi16(_mm_min_epi16(u, max), zero); +} + +static INLINE void cfl_predict_hbd_ssse3(const int16_t *pred_buf_q3, + uint16_t *dst, int dst_stride, + int alpha_q3, int bd, int width, + int height) { + const __m128i alpha_sign = _mm_set1_epi16(alpha_q3); + const __m128i alpha_q12 = _mm_slli_epi16(_mm_abs_epi16(alpha_sign), 9); + const __m128i dc_q0 = _mm_set1_epi16(*dst); + const __m128i max = highbd_max_epi16(bd); + const __m128i zeros = _mm_setzero_si128(); + __m128i *row = (__m128i *)pred_buf_q3; + const __m128i *row_end = row + height * CFL_BUF_LINE_I128; + do { + __m128i res = predict_unclipped(row, alpha_q12, alpha_sign, dc_q0); + res = highbd_clamp_epi16(res, zeros, max); + if (width == 4) { + _mm_storel_epi64((__m128i *)dst, res); + } else { + _mm_storeu_si128((__m128i *)dst, res); + } + if (width >= 16) { + const __m128i res_1 = + predict_unclipped(row + 1, alpha_q12, alpha_sign, dc_q0); + _mm_storeu_si128(((__m128i *)dst) + 1, + highbd_clamp_epi16(res_1, zeros, max)); + } + if (width == 32) { + const __m128i res_2 = + predict_unclipped(row + 2, alpha_q12, alpha_sign, dc_q0); + _mm_storeu_si128((__m128i *)(dst + 16), + highbd_clamp_epi16(res_2, zeros, max)); + const __m128i res_3 = + predict_unclipped(row + 3, alpha_q12, alpha_sign, dc_q0); + _mm_storeu_si128((__m128i *)(dst + 24), + highbd_clamp_epi16(res_3, zeros, max)); + } + dst += dst_stride; + } while ((row += CFL_BUF_LINE_I128) < row_end); +} + +CFL_PREDICT_FN(ssse3, hbd) diff --git a/third_party/aom/av1/common/x86/convolve_2d_avx2.c b/third_party/aom/av1/common/x86/convolve_2d_avx2.c new file mode 100644 index 0000000000..0acafd0446 --- /dev/null +++ b/third_party/aom/av1/common/x86/convolve_2d_avx2.c @@ -0,0 +1,283 @@ +/* + * Copyright (c) 2017, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <immintrin.h> + +#include "config/av1_rtcd.h" + +#include "aom_dsp/x86/convolve_avx2.h" +#include "aom_dsp/x86/convolve_common_intrin.h" +#include "aom_dsp/aom_dsp_common.h" +#include "aom_dsp/aom_filter.h" +#include "aom_dsp/x86/synonyms.h" +#include "av1/common/convolve.h" + +void av1_convolve_2d_sr_avx2(const uint8_t *src, int src_stride, uint8_t *dst, + int dst_stride, int w, int h, + const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, + const int subpel_x_q4, const int subpel_y_q4, + ConvolveParams *conv_params) { + const int bd = 8; + + DECLARE_ALIGNED(32, int16_t, im_block[(MAX_SB_SIZE + MAX_FILTER_TAP) * 8]); + int im_h = h + filter_params_y->taps - 1; + int im_stride = 8; + int i, j; + const int fo_vert = filter_params_y->taps / 2 - 1; + const int fo_horiz = filter_params_x->taps / 2 - 1; + const uint8_t *const src_ptr = src - fo_vert * src_stride - fo_horiz; + + const int bits = + FILTER_BITS * 2 - conv_params->round_0 - conv_params->round_1; + const int offset_bits = bd + 2 * FILTER_BITS - conv_params->round_0; + + __m256i filt[4], coeffs_h[4], coeffs_v[4]; + + assert(conv_params->round_0 > 0); + + filt[0] = _mm256_load_si256((__m256i const *)filt_global_avx2); + filt[1] = _mm256_load_si256((__m256i const *)(filt_global_avx2 + 32)); + filt[2] = _mm256_load_si256((__m256i const *)(filt_global_avx2 + 32 * 2)); + filt[3] = _mm256_load_si256((__m256i const *)(filt_global_avx2 + 32 * 3)); + + prepare_coeffs_lowbd(filter_params_x, subpel_x_q4, coeffs_h); + prepare_coeffs(filter_params_y, subpel_y_q4, coeffs_v); + + const __m256i round_const_h = _mm256_set1_epi16( + ((1 << (conv_params->round_0 - 1)) >> 1) + (1 << (bd + FILTER_BITS - 2))); + const __m128i round_shift_h = _mm_cvtsi32_si128(conv_params->round_0 - 1); + + const __m256i sum_round_v = _mm256_set1_epi32( + (1 << offset_bits) + ((1 << conv_params->round_1) >> 1)); + const __m128i sum_shift_v = _mm_cvtsi32_si128(conv_params->round_1); + + const __m256i round_const_v = _mm256_set1_epi32( + ((1 << bits) >> 1) - (1 << (offset_bits - conv_params->round_1)) - + ((1 << (offset_bits - conv_params->round_1)) >> 1)); + const __m128i round_shift_v = _mm_cvtsi32_si128(bits); + + for (j = 0; j < w; j += 8) { + for (i = 0; i < im_h; i += 2) { + __m256i data = _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)&src_ptr[(i * src_stride) + j])); + + // Load the next line + if (i + 1 < im_h) + data = _mm256_inserti128_si256( + data, + _mm_loadu_si128( + (__m128i *)&src_ptr[(i * src_stride) + j + src_stride]), + 1); + + __m256i res = convolve_lowbd_x(data, coeffs_h, filt); + + res = + _mm256_sra_epi16(_mm256_add_epi16(res, round_const_h), round_shift_h); + + _mm256_store_si256((__m256i *)&im_block[i * im_stride], res); + } + + /* Vertical filter */ + { + __m256i src_0 = _mm256_loadu_si256((__m256i *)(im_block + 0 * im_stride)); + __m256i src_1 = _mm256_loadu_si256((__m256i *)(im_block + 1 * im_stride)); + __m256i src_2 = _mm256_loadu_si256((__m256i *)(im_block + 2 * im_stride)); + __m256i src_3 = _mm256_loadu_si256((__m256i *)(im_block + 3 * im_stride)); + __m256i src_4 = _mm256_loadu_si256((__m256i *)(im_block + 4 * im_stride)); + __m256i src_5 = _mm256_loadu_si256((__m256i *)(im_block + 5 * im_stride)); + + __m256i s[8]; + s[0] = _mm256_unpacklo_epi16(src_0, src_1); + s[1] = _mm256_unpacklo_epi16(src_2, src_3); + s[2] = _mm256_unpacklo_epi16(src_4, src_5); + + s[4] = _mm256_unpackhi_epi16(src_0, src_1); + s[5] = _mm256_unpackhi_epi16(src_2, src_3); + s[6] = _mm256_unpackhi_epi16(src_4, src_5); + + for (i = 0; i < h; i += 2) { + const int16_t *data = &im_block[i * im_stride]; + + const __m256i s6 = + _mm256_loadu_si256((__m256i *)(data + 6 * im_stride)); + const __m256i s7 = + _mm256_loadu_si256((__m256i *)(data + 7 * im_stride)); + + s[3] = _mm256_unpacklo_epi16(s6, s7); + s[7] = _mm256_unpackhi_epi16(s6, s7); + + __m256i res_a = convolve(s, coeffs_v); + __m256i res_b = convolve(s + 4, coeffs_v); + + // Combine V round and 2F-H-V round into a single rounding + res_a = + _mm256_sra_epi32(_mm256_add_epi32(res_a, sum_round_v), sum_shift_v); + res_b = + _mm256_sra_epi32(_mm256_add_epi32(res_b, sum_round_v), sum_shift_v); + + const __m256i res_a_round = _mm256_sra_epi32( + _mm256_add_epi32(res_a, round_const_v), round_shift_v); + const __m256i res_b_round = _mm256_sra_epi32( + _mm256_add_epi32(res_b, round_const_v), round_shift_v); + + /* rounding code */ + // 16 bit conversion + const __m256i res_16bit = _mm256_packs_epi32(res_a_round, res_b_round); + // 8 bit conversion and saturation to uint8 + const __m256i res_8b = _mm256_packus_epi16(res_16bit, res_16bit); + + const __m128i res_0 = _mm256_castsi256_si128(res_8b); + const __m128i res_1 = _mm256_extracti128_si256(res_8b, 1); + + // Store values into the destination buffer + __m128i *const p_0 = (__m128i *)&dst[i * dst_stride + j]; + __m128i *const p_1 = (__m128i *)&dst[i * dst_stride + j + dst_stride]; + if (w - j > 4) { + _mm_storel_epi64(p_0, res_0); + _mm_storel_epi64(p_1, res_1); + } else if (w == 4) { + xx_storel_32(p_0, res_0); + xx_storel_32(p_1, res_1); + } else { + *(uint16_t *)p_0 = _mm_cvtsi128_si32(res_0); + *(uint16_t *)p_1 = _mm_cvtsi128_si32(res_1); + } + + s[0] = s[1]; + s[1] = s[2]; + s[2] = s[3]; + + s[4] = s[5]; + s[5] = s[6]; + s[6] = s[7]; + } + } + } +} + +static INLINE void copy_128(const uint8_t *src, uint8_t *dst) { + __m256i s[4]; + s[0] = _mm256_loadu_si256((__m256i *)(src + 0 * 32)); + s[1] = _mm256_loadu_si256((__m256i *)(src + 1 * 32)); + s[2] = _mm256_loadu_si256((__m256i *)(src + 2 * 32)); + s[3] = _mm256_loadu_si256((__m256i *)(src + 3 * 32)); + _mm256_storeu_si256((__m256i *)(dst + 0 * 32), s[0]); + _mm256_storeu_si256((__m256i *)(dst + 1 * 32), s[1]); + _mm256_storeu_si256((__m256i *)(dst + 2 * 32), s[2]); + _mm256_storeu_si256((__m256i *)(dst + 3 * 32), s[3]); +} + +void av1_convolve_2d_copy_sr_avx2(const uint8_t *src, int src_stride, + uint8_t *dst, int dst_stride, int w, int h, + const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, + const int subpel_x_q4, const int subpel_y_q4, + ConvolveParams *conv_params) { + (void)filter_params_x; + (void)filter_params_y; + (void)subpel_x_q4; + (void)subpel_y_q4; + (void)conv_params; + + if (w >= 16) { + assert(!((intptr_t)dst % 16)); + assert(!(dst_stride % 16)); + } + + if (w == 2) { + do { + memcpy(dst, src, 2 * sizeof(*src)); + src += src_stride; + dst += dst_stride; + memcpy(dst, src, 2 * sizeof(*src)); + src += src_stride; + dst += dst_stride; + h -= 2; + } while (h); + } else if (w == 4) { + do { + memcpy(dst, src, 4 * sizeof(*src)); + src += src_stride; + dst += dst_stride; + memcpy(dst, src, 4 * sizeof(*src)); + src += src_stride; + dst += dst_stride; + h -= 2; + } while (h); + } else if (w == 8) { + do { + __m128i s[2]; + s[0] = _mm_loadl_epi64((__m128i *)src); + src += src_stride; + s[1] = _mm_loadl_epi64((__m128i *)src); + src += src_stride; + _mm_storel_epi64((__m128i *)dst, s[0]); + dst += dst_stride; + _mm_storel_epi64((__m128i *)dst, s[1]); + dst += dst_stride; + h -= 2; + } while (h); + } else if (w == 16) { + do { + __m128i s[2]; + s[0] = _mm_loadu_si128((__m128i *)src); + src += src_stride; + s[1] = _mm_loadu_si128((__m128i *)src); + src += src_stride; + _mm_store_si128((__m128i *)dst, s[0]); + dst += dst_stride; + _mm_store_si128((__m128i *)dst, s[1]); + dst += dst_stride; + h -= 2; + } while (h); + } else if (w == 32) { + do { + __m256i s[2]; + s[0] = _mm256_loadu_si256((__m256i *)src); + src += src_stride; + s[1] = _mm256_loadu_si256((__m256i *)src); + src += src_stride; + _mm256_storeu_si256((__m256i *)dst, s[0]); + dst += dst_stride; + _mm256_storeu_si256((__m256i *)dst, s[1]); + dst += dst_stride; + h -= 2; + } while (h); + } else if (w == 64) { + do { + __m256i s[4]; + s[0] = _mm256_loadu_si256((__m256i *)(src + 0 * 32)); + s[1] = _mm256_loadu_si256((__m256i *)(src + 1 * 32)); + src += src_stride; + s[2] = _mm256_loadu_si256((__m256i *)(src + 0 * 32)); + s[3] = _mm256_loadu_si256((__m256i *)(src + 1 * 32)); + src += src_stride; + _mm256_storeu_si256((__m256i *)(dst + 0 * 32), s[0]); + _mm256_storeu_si256((__m256i *)(dst + 1 * 32), s[1]); + dst += dst_stride; + _mm256_storeu_si256((__m256i *)(dst + 0 * 32), s[2]); + _mm256_storeu_si256((__m256i *)(dst + 1 * 32), s[3]); + dst += dst_stride; + h -= 2; + } while (h); + } else { + do { + copy_128(src, dst); + src += src_stride; + dst += dst_stride; + copy_128(src, dst); + src += src_stride; + dst += dst_stride; + h -= 2; + } while (h); + } +} diff --git a/third_party/aom/av1/common/x86/convolve_2d_sse2.c b/third_party/aom/av1/common/x86/convolve_2d_sse2.c new file mode 100644 index 0000000000..b1a62a4f69 --- /dev/null +++ b/third_party/aom/av1/common/x86/convolve_2d_sse2.c @@ -0,0 +1,472 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <emmintrin.h> + +#include "config/av1_rtcd.h" + +#include "aom_dsp/aom_dsp_common.h" +#include "aom_dsp/aom_filter.h" +#include "aom_dsp/x86/convolve_sse2.h" +#include "av1/common/convolve.h" + +void av1_convolve_2d_sr_sse2(const uint8_t *src, int src_stride, uint8_t *dst, + int dst_stride, int w, int h, + const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, + const int subpel_x_q4, const int subpel_y_q4, + ConvolveParams *conv_params) { + const int bd = 8; + + DECLARE_ALIGNED(16, int16_t, + im_block[(MAX_SB_SIZE + MAX_FILTER_TAP - 1) * MAX_SB_SIZE]); + int im_h = h + filter_params_y->taps - 1; + int im_stride = MAX_SB_SIZE; + int i, j; + const int fo_vert = filter_params_y->taps / 2 - 1; + const int fo_horiz = filter_params_x->taps / 2 - 1; + const uint8_t *const src_ptr = src - fo_vert * src_stride - fo_horiz; + + const __m128i zero = _mm_setzero_si128(); + const int bits = + FILTER_BITS * 2 - conv_params->round_0 - conv_params->round_1; + const int offset_bits = bd + 2 * FILTER_BITS - conv_params->round_0; + + assert(conv_params->round_0 > 0); + + /* Horizontal filter */ + { + const int16_t *x_filter = av1_get_interp_filter_subpel_kernel( + filter_params_x, subpel_x_q4 & SUBPEL_MASK); + const __m128i coeffs_x = _mm_loadu_si128((__m128i *)x_filter); + + // coeffs 0 1 0 1 2 3 2 3 + const __m128i tmp_0 = _mm_unpacklo_epi32(coeffs_x, coeffs_x); + // coeffs 4 5 4 5 6 7 6 7 + const __m128i tmp_1 = _mm_unpackhi_epi32(coeffs_x, coeffs_x); + + // coeffs 0 1 0 1 0 1 0 1 + const __m128i coeff_01 = _mm_unpacklo_epi64(tmp_0, tmp_0); + // coeffs 2 3 2 3 2 3 2 3 + const __m128i coeff_23 = _mm_unpackhi_epi64(tmp_0, tmp_0); + // coeffs 4 5 4 5 4 5 4 5 + const __m128i coeff_45 = _mm_unpacklo_epi64(tmp_1, tmp_1); + // coeffs 6 7 6 7 6 7 6 7 + const __m128i coeff_67 = _mm_unpackhi_epi64(tmp_1, tmp_1); + + const __m128i round_const = _mm_set1_epi32( + (1 << (bd + FILTER_BITS - 1)) + ((1 << conv_params->round_0) >> 1)); + const __m128i round_shift = _mm_cvtsi32_si128(conv_params->round_0); + + for (i = 0; i < im_h; ++i) { + for (j = 0; j < w; j += 8) { + const __m128i data = + _mm_loadu_si128((__m128i *)&src_ptr[i * src_stride + j]); + + // Filter even-index pixels + const __m128i src_0 = _mm_unpacklo_epi8(data, zero); + const __m128i res_0 = _mm_madd_epi16(src_0, coeff_01); + const __m128i src_2 = _mm_unpacklo_epi8(_mm_srli_si128(data, 2), zero); + const __m128i res_2 = _mm_madd_epi16(src_2, coeff_23); + const __m128i src_4 = _mm_unpacklo_epi8(_mm_srli_si128(data, 4), zero); + const __m128i res_4 = _mm_madd_epi16(src_4, coeff_45); + const __m128i src_6 = _mm_unpacklo_epi8(_mm_srli_si128(data, 6), zero); + const __m128i res_6 = _mm_madd_epi16(src_6, coeff_67); + + __m128i res_even = _mm_add_epi32(_mm_add_epi32(res_0, res_4), + _mm_add_epi32(res_2, res_6)); + res_even = + _mm_sra_epi32(_mm_add_epi32(res_even, round_const), round_shift); + + // Filter odd-index pixels + const __m128i src_1 = _mm_unpacklo_epi8(_mm_srli_si128(data, 1), zero); + const __m128i res_1 = _mm_madd_epi16(src_1, coeff_01); + const __m128i src_3 = _mm_unpacklo_epi8(_mm_srli_si128(data, 3), zero); + const __m128i res_3 = _mm_madd_epi16(src_3, coeff_23); + const __m128i src_5 = _mm_unpacklo_epi8(_mm_srli_si128(data, 5), zero); + const __m128i res_5 = _mm_madd_epi16(src_5, coeff_45); + const __m128i src_7 = _mm_unpacklo_epi8(_mm_srli_si128(data, 7), zero); + const __m128i res_7 = _mm_madd_epi16(src_7, coeff_67); + + __m128i res_odd = _mm_add_epi32(_mm_add_epi32(res_1, res_5), + _mm_add_epi32(res_3, res_7)); + res_odd = + _mm_sra_epi32(_mm_add_epi32(res_odd, round_const), round_shift); + + // Pack in the column order 0, 2, 4, 6, 1, 3, 5, 7 + __m128i res = _mm_packs_epi32(res_even, res_odd); + _mm_storeu_si128((__m128i *)&im_block[i * im_stride + j], res); + } + } + } + + /* Vertical filter */ + { + const int16_t *y_filter = av1_get_interp_filter_subpel_kernel( + filter_params_y, subpel_y_q4 & SUBPEL_MASK); + const __m128i coeffs_y = _mm_loadu_si128((__m128i *)y_filter); + + // coeffs 0 1 0 1 2 3 2 3 + const __m128i tmp_0 = _mm_unpacklo_epi32(coeffs_y, coeffs_y); + // coeffs 4 5 4 5 6 7 6 7 + const __m128i tmp_1 = _mm_unpackhi_epi32(coeffs_y, coeffs_y); + + // coeffs 0 1 0 1 0 1 0 1 + const __m128i coeff_01 = _mm_unpacklo_epi64(tmp_0, tmp_0); + // coeffs 2 3 2 3 2 3 2 3 + const __m128i coeff_23 = _mm_unpackhi_epi64(tmp_0, tmp_0); + // coeffs 4 5 4 5 4 5 4 5 + const __m128i coeff_45 = _mm_unpacklo_epi64(tmp_1, tmp_1); + // coeffs 6 7 6 7 6 7 6 7 + const __m128i coeff_67 = _mm_unpackhi_epi64(tmp_1, tmp_1); + + const __m128i sum_round = + _mm_set1_epi32((1 << offset_bits) + ((1 << conv_params->round_1) >> 1)); + const __m128i sum_shift = _mm_cvtsi32_si128(conv_params->round_1); + + const __m128i round_const = _mm_set1_epi32( + ((1 << bits) >> 1) - (1 << (offset_bits - conv_params->round_1)) - + ((1 << (offset_bits - conv_params->round_1)) >> 1)); + const __m128i round_shift = _mm_cvtsi32_si128(bits); + + for (i = 0; i < h; ++i) { + for (j = 0; j < w; j += 8) { + // Filter even-index pixels + const int16_t *data = &im_block[i * im_stride + j]; + const __m128i src_0 = + _mm_unpacklo_epi16(*(__m128i *)(data + 0 * im_stride), + *(__m128i *)(data + 1 * im_stride)); + const __m128i src_2 = + _mm_unpacklo_epi16(*(__m128i *)(data + 2 * im_stride), + *(__m128i *)(data + 3 * im_stride)); + const __m128i src_4 = + _mm_unpacklo_epi16(*(__m128i *)(data + 4 * im_stride), + *(__m128i *)(data + 5 * im_stride)); + const __m128i src_6 = + _mm_unpacklo_epi16(*(__m128i *)(data + 6 * im_stride), + *(__m128i *)(data + 7 * im_stride)); + + const __m128i res_0 = _mm_madd_epi16(src_0, coeff_01); + const __m128i res_2 = _mm_madd_epi16(src_2, coeff_23); + const __m128i res_4 = _mm_madd_epi16(src_4, coeff_45); + const __m128i res_6 = _mm_madd_epi16(src_6, coeff_67); + + const __m128i res_even = _mm_add_epi32(_mm_add_epi32(res_0, res_2), + _mm_add_epi32(res_4, res_6)); + + // Filter odd-index pixels + const __m128i src_1 = + _mm_unpackhi_epi16(*(__m128i *)(data + 0 * im_stride), + *(__m128i *)(data + 1 * im_stride)); + const __m128i src_3 = + _mm_unpackhi_epi16(*(__m128i *)(data + 2 * im_stride), + *(__m128i *)(data + 3 * im_stride)); + const __m128i src_5 = + _mm_unpackhi_epi16(*(__m128i *)(data + 4 * im_stride), + *(__m128i *)(data + 5 * im_stride)); + const __m128i src_7 = + _mm_unpackhi_epi16(*(__m128i *)(data + 6 * im_stride), + *(__m128i *)(data + 7 * im_stride)); + + const __m128i res_1 = _mm_madd_epi16(src_1, coeff_01); + const __m128i res_3 = _mm_madd_epi16(src_3, coeff_23); + const __m128i res_5 = _mm_madd_epi16(src_5, coeff_45); + const __m128i res_7 = _mm_madd_epi16(src_7, coeff_67); + + const __m128i res_odd = _mm_add_epi32(_mm_add_epi32(res_1, res_3), + _mm_add_epi32(res_5, res_7)); + + // Rearrange pixels back into the order 0 ... 7 + const __m128i res_lo = _mm_unpacklo_epi32(res_even, res_odd); + const __m128i res_hi = _mm_unpackhi_epi32(res_even, res_odd); + + __m128i res_lo_round = + _mm_sra_epi32(_mm_add_epi32(res_lo, sum_round), sum_shift); + __m128i res_hi_round = + _mm_sra_epi32(_mm_add_epi32(res_hi, sum_round), sum_shift); + + res_lo_round = _mm_sra_epi32(_mm_add_epi32(res_lo_round, round_const), + round_shift); + res_hi_round = _mm_sra_epi32(_mm_add_epi32(res_hi_round, round_const), + round_shift); + + const __m128i res16 = _mm_packs_epi32(res_lo_round, res_hi_round); + const __m128i res = _mm_packus_epi16(res16, res16); + + // Accumulate values into the destination buffer + __m128i *const p = (__m128i *)&dst[i * dst_stride + j]; + + if (w == 2) { + *(uint16_t *)p = _mm_cvtsi128_si32(res); + } else if (w == 4) { + *(uint32_t *)p = _mm_cvtsi128_si32(res); + } else { + _mm_storel_epi64(p, res); + } + } + } + } +} + +static INLINE void copy_128(const uint8_t *src, uint8_t *dst) { + __m128i s[8]; + s[0] = _mm_loadu_si128((__m128i *)(src + 0 * 16)); + s[1] = _mm_loadu_si128((__m128i *)(src + 1 * 16)); + s[2] = _mm_loadu_si128((__m128i *)(src + 2 * 16)); + s[3] = _mm_loadu_si128((__m128i *)(src + 3 * 16)); + s[4] = _mm_loadu_si128((__m128i *)(src + 4 * 16)); + s[5] = _mm_loadu_si128((__m128i *)(src + 5 * 16)); + s[6] = _mm_loadu_si128((__m128i *)(src + 6 * 16)); + s[7] = _mm_loadu_si128((__m128i *)(src + 7 * 16)); + _mm_store_si128((__m128i *)(dst + 0 * 16), s[0]); + _mm_store_si128((__m128i *)(dst + 1 * 16), s[1]); + _mm_store_si128((__m128i *)(dst + 2 * 16), s[2]); + _mm_store_si128((__m128i *)(dst + 3 * 16), s[3]); + _mm_store_si128((__m128i *)(dst + 4 * 16), s[4]); + _mm_store_si128((__m128i *)(dst + 5 * 16), s[5]); + _mm_store_si128((__m128i *)(dst + 6 * 16), s[6]); + _mm_store_si128((__m128i *)(dst + 7 * 16), s[7]); +} + +void av1_convolve_2d_copy_sr_sse2(const uint8_t *src, int src_stride, + uint8_t *dst, int dst_stride, int w, int h, + const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, + const int subpel_x_q4, const int subpel_y_q4, + ConvolveParams *conv_params) { + (void)filter_params_x; + (void)filter_params_y; + (void)subpel_x_q4; + (void)subpel_y_q4; + (void)conv_params; + + if (w >= 16) { + assert(!((intptr_t)dst % 16)); + assert(!(dst_stride % 16)); + } + + if (w == 2) { + do { + memcpy(dst, src, 2 * sizeof(*src)); + src += src_stride; + dst += dst_stride; + memcpy(dst, src, 2 * sizeof(*src)); + src += src_stride; + dst += dst_stride; + h -= 2; + } while (h); + } else if (w == 4) { + do { + memcpy(dst, src, 4 * sizeof(*src)); + src += src_stride; + dst += dst_stride; + memcpy(dst, src, 4 * sizeof(*src)); + src += src_stride; + dst += dst_stride; + h -= 2; + } while (h); + } else if (w == 8) { + do { + __m128i s[2]; + s[0] = _mm_loadl_epi64((__m128i *)src); + src += src_stride; + s[1] = _mm_loadl_epi64((__m128i *)src); + src += src_stride; + _mm_storel_epi64((__m128i *)dst, s[0]); + dst += dst_stride; + _mm_storel_epi64((__m128i *)dst, s[1]); + dst += dst_stride; + h -= 2; + } while (h); + } else if (w == 16) { + do { + __m128i s[2]; + s[0] = _mm_loadu_si128((__m128i *)src); + src += src_stride; + s[1] = _mm_loadu_si128((__m128i *)src); + src += src_stride; + _mm_store_si128((__m128i *)dst, s[0]); + dst += dst_stride; + _mm_store_si128((__m128i *)dst, s[1]); + dst += dst_stride; + h -= 2; + } while (h); + } else if (w == 32) { + do { + __m128i s[4]; + s[0] = _mm_loadu_si128((__m128i *)(src + 0 * 16)); + s[1] = _mm_loadu_si128((__m128i *)(src + 1 * 16)); + src += src_stride; + s[2] = _mm_loadu_si128((__m128i *)(src + 0 * 16)); + s[3] = _mm_loadu_si128((__m128i *)(src + 1 * 16)); + src += src_stride; + _mm_store_si128((__m128i *)(dst + 0 * 16), s[0]); + _mm_store_si128((__m128i *)(dst + 1 * 16), s[1]); + dst += dst_stride; + _mm_store_si128((__m128i *)(dst + 0 * 16), s[2]); + _mm_store_si128((__m128i *)(dst + 1 * 16), s[3]); + dst += dst_stride; + h -= 2; + } while (h); + } else if (w == 64) { + do { + __m128i s[8]; + s[0] = _mm_loadu_si128((__m128i *)(src + 0 * 16)); + s[1] = _mm_loadu_si128((__m128i *)(src + 1 * 16)); + s[2] = _mm_loadu_si128((__m128i *)(src + 2 * 16)); + s[3] = _mm_loadu_si128((__m128i *)(src + 3 * 16)); + src += src_stride; + s[4] = _mm_loadu_si128((__m128i *)(src + 0 * 16)); + s[5] = _mm_loadu_si128((__m128i *)(src + 1 * 16)); + s[6] = _mm_loadu_si128((__m128i *)(src + 2 * 16)); + s[7] = _mm_loadu_si128((__m128i *)(src + 3 * 16)); + src += src_stride; + _mm_store_si128((__m128i *)(dst + 0 * 16), s[0]); + _mm_store_si128((__m128i *)(dst + 1 * 16), s[1]); + _mm_store_si128((__m128i *)(dst + 2 * 16), s[2]); + _mm_store_si128((__m128i *)(dst + 3 * 16), s[3]); + dst += dst_stride; + _mm_store_si128((__m128i *)(dst + 0 * 16), s[4]); + _mm_store_si128((__m128i *)(dst + 1 * 16), s[5]); + _mm_store_si128((__m128i *)(dst + 2 * 16), s[6]); + _mm_store_si128((__m128i *)(dst + 3 * 16), s[7]); + dst += dst_stride; + h -= 2; + } while (h); + } else { + do { + copy_128(src, dst); + src += src_stride; + dst += dst_stride; + copy_128(src, dst); + src += src_stride; + dst += dst_stride; + h -= 2; + } while (h); + } +} + +void av1_jnt_convolve_2d_copy_sse2(const uint8_t *src, int src_stride, + uint8_t *dst0, int dst_stride0, int w, int h, + const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, + const int subpel_x_q4, const int subpel_y_q4, + ConvolveParams *conv_params) { + const int bd = 8; + CONV_BUF_TYPE *dst = conv_params->dst; + int dst_stride = conv_params->dst_stride; + (void)filter_params_x; + (void)filter_params_y; + (void)subpel_x_q4; + (void)subpel_y_q4; + + const int bits = + FILTER_BITS * 2 - conv_params->round_1 - conv_params->round_0; + const int do_average = conv_params->do_average; + const int use_jnt_comp_avg = conv_params->use_jnt_comp_avg; + const __m128i zero = _mm_setzero_si128(); + const __m128i left_shift = _mm_cvtsi32_si128(bits); + int i, j; + + const int w0 = conv_params->fwd_offset; + const int w1 = conv_params->bck_offset; + const __m128i wt0 = _mm_set1_epi16(w0); + const __m128i wt1 = _mm_set1_epi16(w1); + const __m128i wt = _mm_unpacklo_epi16(wt0, wt1); + + const int offset_0 = + bd + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const int offset = (1 << offset_0) + (1 << (offset_0 - 1)); + const __m128i offset_const = _mm_set1_epi16(offset); + const int rounding_shift = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const __m128i rounding_const = _mm_set1_epi16((1 << rounding_shift) >> 1); + + assert((w % 4) == 0); + + if (!(w % 16)) { + for (i = 0; i < h; ++i) { + for (j = 0; j < w; j += 16) { + const __m128i d8 = _mm_loadu_si128((__m128i *)&src[j]); + + const __m128i d16_lo = _mm_unpacklo_epi8(d8, zero); + const __m128i d16_hi = _mm_unpackhi_epi8(d8, zero); + + const __m128i res_lo = _mm_sll_epi16(d16_lo, left_shift); + const __m128i res_unsigned_lo = _mm_add_epi16(res_lo, offset_const); + + const __m128i res_hi = _mm_sll_epi16(d16_hi, left_shift); + const __m128i res_unsigned_hi = _mm_add_epi16(res_hi, offset_const); + + if (do_average) { + const __m128i data_ref_0_lo = _mm_loadu_si128((__m128i *)(&dst[j])); + const __m128i data_ref_0_hi = + _mm_loadu_si128((__m128i *)(&dst[j + 8])); + + const __m128i comp_avg_res_lo = + comp_avg(&data_ref_0_lo, &res_unsigned_lo, &wt, use_jnt_comp_avg); + + const __m128i round_result_lo = convolve_rounding( + &comp_avg_res_lo, &offset_const, &rounding_const, rounding_shift); + + const __m128i comp_avg_res_hi = + comp_avg(&data_ref_0_hi, &res_unsigned_hi, &wt, use_jnt_comp_avg); + + const __m128i round_result_hi = convolve_rounding( + &comp_avg_res_hi, &offset_const, &rounding_const, rounding_shift); + + const __m128i res_8 = + _mm_packus_epi16(round_result_lo, round_result_hi); + + _mm_store_si128((__m128i *)(&dst0[j]), res_8); + } else { + _mm_store_si128((__m128i *)(&dst[j]), res_unsigned_lo); + _mm_store_si128((__m128i *)(&dst[j + 8]), res_unsigned_hi); + } + } + src += src_stride; + dst += dst_stride; + dst0 += dst_stride0; + } + } else { + for (i = 0; i < h; ++i) { + for (j = 0; j < w; j += 8) { + const __m128i d8 = _mm_loadl_epi64((__m128i *)&src[j]); + const __m128i d16_0 = _mm_unpacklo_epi8(d8, zero); + + const __m128i res = _mm_sll_epi16(d16_0, left_shift); + const __m128i res_unsigned = _mm_add_epi16(res, offset_const); + + if (do_average) { + const __m128i data_ref_0 = _mm_loadu_si128((__m128i *)(&dst[j])); + + const __m128i comp_avg_res = + comp_avg(&data_ref_0, &res_unsigned, &wt, use_jnt_comp_avg); + + const __m128i round_result = convolve_rounding( + &comp_avg_res, &offset_const, &rounding_const, rounding_shift); + + const __m128i res_8 = _mm_packus_epi16(round_result, round_result); + + if (w > 4) + _mm_storel_epi64((__m128i *)(&dst0[j]), res_8); + else + *(uint32_t *)(&dst0[j]) = _mm_cvtsi128_si32(res_8); + } else { + _mm_store_si128((__m128i *)(&dst[j]), res_unsigned); + } + } + src += src_stride; + dst += dst_stride; + dst0 += dst_stride0; + } + } +} diff --git a/third_party/aom/av1/common/x86/convolve_avx2.c b/third_party/aom/av1/common/x86/convolve_avx2.c new file mode 100644 index 0000000000..0e91ea9475 --- /dev/null +++ b/third_party/aom/av1/common/x86/convolve_avx2.c @@ -0,0 +1,277 @@ +/* + * Copyright (c) 2017, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <immintrin.h> + +#include "config/av1_rtcd.h" + +#include "aom_dsp/aom_dsp_common.h" +#include "aom_dsp/x86/convolve_avx2.h" +#include "aom_dsp/x86/synonyms.h" + +void av1_convolve_y_sr_avx2(const uint8_t *src, int src_stride, uint8_t *dst, + int dst_stride, int w, int h, + const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, + const int subpel_x_q4, const int subpel_y_q4, + ConvolveParams *conv_params) { + int i, j; + const int fo_vert = filter_params_y->taps / 2 - 1; + const uint8_t *const src_ptr = src - fo_vert * src_stride; + + // right shift is F-1 because we are already dividing + // filter co-efficients by 2 + const int right_shift_bits = (FILTER_BITS - 1); + const __m128i right_shift = _mm_cvtsi32_si128(right_shift_bits); + const __m256i right_shift_const = + _mm256_set1_epi16((1 << right_shift_bits) >> 1); + __m256i coeffs[4], s[8]; + + assert(conv_params->round_0 <= FILTER_BITS); + assert(((conv_params->round_0 + conv_params->round_1) <= (FILTER_BITS + 1)) || + ((conv_params->round_0 + conv_params->round_1) == (2 * FILTER_BITS))); + + prepare_coeffs_lowbd(filter_params_y, subpel_y_q4, coeffs); + + (void)filter_params_x; + (void)subpel_x_q4; + (void)conv_params; + + for (j = 0; j < w; j += 16) { + const uint8_t *data = &src_ptr[j]; + __m256i src6; + + // Load lines a and b. Line a to lower 128, line b to upper 128 + const __m256i src_01a = _mm256_permute2x128_si256( + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 0 * src_stride))), + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 1 * src_stride))), + 0x20); + + const __m256i src_12a = _mm256_permute2x128_si256( + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 1 * src_stride))), + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 2 * src_stride))), + 0x20); + + const __m256i src_23a = _mm256_permute2x128_si256( + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 2 * src_stride))), + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 3 * src_stride))), + 0x20); + + const __m256i src_34a = _mm256_permute2x128_si256( + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 3 * src_stride))), + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 4 * src_stride))), + 0x20); + + const __m256i src_45a = _mm256_permute2x128_si256( + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 4 * src_stride))), + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 5 * src_stride))), + 0x20); + + src6 = _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 6 * src_stride))); + const __m256i src_56a = _mm256_permute2x128_si256( + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 5 * src_stride))), + src6, 0x20); + + s[0] = _mm256_unpacklo_epi8(src_01a, src_12a); + s[1] = _mm256_unpacklo_epi8(src_23a, src_34a); + s[2] = _mm256_unpacklo_epi8(src_45a, src_56a); + + s[4] = _mm256_unpackhi_epi8(src_01a, src_12a); + s[5] = _mm256_unpackhi_epi8(src_23a, src_34a); + s[6] = _mm256_unpackhi_epi8(src_45a, src_56a); + + for (i = 0; i < h; i += 2) { + data = &src_ptr[i * src_stride + j]; + const __m256i src_67a = _mm256_permute2x128_si256( + src6, + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 7 * src_stride))), + 0x20); + + src6 = _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 8 * src_stride))); + const __m256i src_78a = _mm256_permute2x128_si256( + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 7 * src_stride))), + src6, 0x20); + + s[3] = _mm256_unpacklo_epi8(src_67a, src_78a); + s[7] = _mm256_unpackhi_epi8(src_67a, src_78a); + + const __m256i res_lo = convolve_lowbd(s, coeffs); + + /* rounding code */ + // shift by F - 1 + const __m256i res_16b_lo = _mm256_sra_epi16( + _mm256_add_epi16(res_lo, right_shift_const), right_shift); + // 8 bit conversion and saturation to uint8 + __m256i res_8b_lo = _mm256_packus_epi16(res_16b_lo, res_16b_lo); + + if (w - j > 8) { + const __m256i res_hi = convolve_lowbd(s + 4, coeffs); + + /* rounding code */ + // shift by F - 1 + const __m256i res_16b_hi = _mm256_sra_epi16( + _mm256_add_epi16(res_hi, right_shift_const), right_shift); + // 8 bit conversion and saturation to uint8 + __m256i res_8b_hi = _mm256_packus_epi16(res_16b_hi, res_16b_hi); + + __m256i res_a = _mm256_unpacklo_epi64(res_8b_lo, res_8b_hi); + + const __m128i res_0 = _mm256_castsi256_si128(res_a); + const __m128i res_1 = _mm256_extracti128_si256(res_a, 1); + + _mm_storeu_si128((__m128i *)&dst[i * dst_stride + j], res_0); + _mm_storeu_si128((__m128i *)&dst[i * dst_stride + j + dst_stride], + res_1); + } else { + const __m128i res_0 = _mm256_castsi256_si128(res_8b_lo); + const __m128i res_1 = _mm256_extracti128_si256(res_8b_lo, 1); + if (w - j > 4) { + _mm_storel_epi64((__m128i *)&dst[i * dst_stride + j], res_0); + _mm_storel_epi64((__m128i *)&dst[i * dst_stride + j + dst_stride], + res_1); + } else if (w - j > 2) { + xx_storel_32(&dst[i * dst_stride + j], res_0); + xx_storel_32(&dst[i * dst_stride + j + dst_stride], res_1); + } else { + __m128i *const p_0 = (__m128i *)&dst[i * dst_stride + j]; + __m128i *const p_1 = (__m128i *)&dst[i * dst_stride + j + dst_stride]; + *(uint16_t *)p_0 = _mm_cvtsi128_si32(res_0); + *(uint16_t *)p_1 = _mm_cvtsi128_si32(res_1); + } + } + + s[0] = s[1]; + s[1] = s[2]; + s[2] = s[3]; + + s[4] = s[5]; + s[5] = s[6]; + s[6] = s[7]; + } + } +} + +void av1_convolve_x_sr_avx2(const uint8_t *src, int src_stride, uint8_t *dst, + int dst_stride, int w, int h, + const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, + const int subpel_x_q4, const int subpel_y_q4, + ConvolveParams *conv_params) { + int i, j; + const int fo_horiz = filter_params_x->taps / 2 - 1; + const uint8_t *const src_ptr = src - fo_horiz; + const int bits = FILTER_BITS - conv_params->round_0; + + __m256i filt[4], coeffs[4]; + + filt[0] = _mm256_load_si256((__m256i const *)filt_global_avx2); + filt[1] = _mm256_load_si256((__m256i const *)(filt_global_avx2 + 32)); + filt[2] = _mm256_load_si256((__m256i const *)(filt_global_avx2 + 32 * 2)); + filt[3] = _mm256_load_si256((__m256i const *)(filt_global_avx2 + 32 * 3)); + + prepare_coeffs_lowbd(filter_params_x, subpel_x_q4, coeffs); + + const __m256i round_0_const = + _mm256_set1_epi16((1 << (conv_params->round_0 - 1)) >> 1); + const __m128i round_0_shift = _mm_cvtsi32_si128(conv_params->round_0 - 1); + const __m256i round_const = _mm256_set1_epi16((1 << bits) >> 1); + const __m128i round_shift = _mm_cvtsi32_si128(bits); + + (void)filter_params_y; + (void)subpel_y_q4; + + assert(bits >= 0); + assert((FILTER_BITS - conv_params->round_1) >= 0 || + ((conv_params->round_0 + conv_params->round_1) == 2 * FILTER_BITS)); + assert(conv_params->round_0 > 0); + + if (w <= 8) { + for (i = 0; i < h; i += 2) { + const __m256i data = _mm256_permute2x128_si256( + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(&src_ptr[i * src_stride]))), + _mm256_castsi128_si256(_mm_loadu_si128( + (__m128i *)(&src_ptr[i * src_stride + src_stride]))), + 0x20); + + __m256i res_16b = convolve_lowbd_x(data, coeffs, filt); + + res_16b = _mm256_sra_epi16(_mm256_add_epi16(res_16b, round_0_const), + round_0_shift); + + res_16b = + _mm256_sra_epi16(_mm256_add_epi16(res_16b, round_const), round_shift); + + /* rounding code */ + // 8 bit conversion and saturation to uint8 + __m256i res_8b = _mm256_packus_epi16(res_16b, res_16b); + + const __m128i res_0 = _mm256_castsi256_si128(res_8b); + const __m128i res_1 = _mm256_extracti128_si256(res_8b, 1); + if (w > 4) { + _mm_storel_epi64((__m128i *)&dst[i * dst_stride], res_0); + _mm_storel_epi64((__m128i *)&dst[i * dst_stride + dst_stride], res_1); + } else if (w > 2) { + xx_storel_32(&dst[i * dst_stride], res_0); + xx_storel_32(&dst[i * dst_stride + dst_stride], res_1); + } else { + __m128i *const p_0 = (__m128i *)&dst[i * dst_stride]; + __m128i *const p_1 = (__m128i *)&dst[i * dst_stride + dst_stride]; + *(uint16_t *)p_0 = _mm_cvtsi128_si32(res_0); + *(uint16_t *)p_1 = _mm_cvtsi128_si32(res_1); + } + } + } else { + for (i = 0; i < h; ++i) { + for (j = 0; j < w; j += 16) { + // 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 8 9 10 11 12 13 14 15 16 17 18 + // 19 20 21 22 23 + const __m256i data = _mm256_inserti128_si256( + _mm256_loadu_si256((__m256i *)&src_ptr[(i * src_stride) + j]), + _mm_loadu_si128((__m128i *)&src_ptr[(i * src_stride) + (j + 8)]), + 1); + + __m256i res_16b = convolve_lowbd_x(data, coeffs, filt); + + res_16b = _mm256_sra_epi16(_mm256_add_epi16(res_16b, round_0_const), + round_0_shift); + + res_16b = _mm256_sra_epi16(_mm256_add_epi16(res_16b, round_const), + round_shift); + + /* rounding code */ + // 8 bit conversion and saturation to uint8 + __m256i res_8b = _mm256_packus_epi16(res_16b, res_16b); + + // Store values into the destination buffer + // 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 + res_8b = _mm256_permute4x64_epi64(res_8b, 216); + __m128i res = _mm256_castsi256_si128(res_8b); + _mm_storeu_si128((__m128i *)&dst[i * dst_stride + j], res); + } + } + } +} diff --git a/third_party/aom/av1/common/x86/convolve_sse2.c b/third_party/aom/av1/common/x86/convolve_sse2.c new file mode 100644 index 0000000000..5016642dee --- /dev/null +++ b/third_party/aom/av1/common/x86/convolve_sse2.c @@ -0,0 +1,338 @@ +/* + * Copyright (c) 2017, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <emmintrin.h> + +#include "config/av1_rtcd.h" + +#include "aom_dsp/aom_dsp_common.h" +#include "aom_dsp/aom_filter.h" +#include "aom_dsp/x86/convolve_common_intrin.h" +#include "av1/common/convolve.h" + +static INLINE void prepare_coeffs(const InterpFilterParams *const filter_params, + const int subpel_q4, + __m128i *const coeffs /* [4] */) { + const int16_t *const y_filter = av1_get_interp_filter_subpel_kernel( + filter_params, subpel_q4 & SUBPEL_MASK); + const __m128i coeffs_y = _mm_loadu_si128((__m128i *)y_filter); + // coeffs 0 1 0 1 2 3 2 3 + const __m128i tmp_0 = _mm_unpacklo_epi32(coeffs_y, coeffs_y); + // coeffs 4 5 4 5 6 7 6 7 + const __m128i tmp_1 = _mm_unpackhi_epi32(coeffs_y, coeffs_y); + + coeffs[0] = _mm_unpacklo_epi64(tmp_0, tmp_0); // coeffs 0 1 0 1 0 1 0 1 + coeffs[1] = _mm_unpackhi_epi64(tmp_0, tmp_0); // coeffs 2 3 2 3 2 3 2 3 + coeffs[2] = _mm_unpacklo_epi64(tmp_1, tmp_1); // coeffs 4 5 4 5 4 5 4 5 + coeffs[3] = _mm_unpackhi_epi64(tmp_1, tmp_1); // coeffs 6 7 6 7 6 7 6 7 +} + +static INLINE __m128i convolve(const __m128i *const s, + const __m128i *const coeffs) { + const __m128i d0 = _mm_madd_epi16(s[0], coeffs[0]); + const __m128i d1 = _mm_madd_epi16(s[1], coeffs[1]); + const __m128i d2 = _mm_madd_epi16(s[2], coeffs[2]); + const __m128i d3 = _mm_madd_epi16(s[3], coeffs[3]); + const __m128i d = _mm_add_epi32(_mm_add_epi32(d0, d1), _mm_add_epi32(d2, d3)); + return d; +} + +static INLINE __m128i convolve_lo_x(const __m128i *const s, + const __m128i *const coeffs) { + __m128i ss[4]; + ss[0] = _mm_unpacklo_epi8(s[0], _mm_setzero_si128()); + ss[1] = _mm_unpacklo_epi8(s[1], _mm_setzero_si128()); + ss[2] = _mm_unpacklo_epi8(s[2], _mm_setzero_si128()); + ss[3] = _mm_unpacklo_epi8(s[3], _mm_setzero_si128()); + return convolve(ss, coeffs); +} + +static INLINE __m128i convolve_lo_y(const __m128i *const s, + const __m128i *const coeffs) { + __m128i ss[4]; + ss[0] = _mm_unpacklo_epi8(s[0], _mm_setzero_si128()); + ss[1] = _mm_unpacklo_epi8(s[2], _mm_setzero_si128()); + ss[2] = _mm_unpacklo_epi8(s[4], _mm_setzero_si128()); + ss[3] = _mm_unpacklo_epi8(s[6], _mm_setzero_si128()); + return convolve(ss, coeffs); +} + +static INLINE __m128i convolve_hi_y(const __m128i *const s, + const __m128i *const coeffs) { + __m128i ss[4]; + ss[0] = _mm_unpackhi_epi8(s[0], _mm_setzero_si128()); + ss[1] = _mm_unpackhi_epi8(s[2], _mm_setzero_si128()); + ss[2] = _mm_unpackhi_epi8(s[4], _mm_setzero_si128()); + ss[3] = _mm_unpackhi_epi8(s[6], _mm_setzero_si128()); + return convolve(ss, coeffs); +} + +void av1_convolve_y_sr_sse2(const uint8_t *src, int src_stride, uint8_t *dst, + int dst_stride, int w, int h, + const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, + const int subpel_x_q4, const int subpel_y_q4, + ConvolveParams *conv_params) { + const int fo_vert = filter_params_y->taps / 2 - 1; + const uint8_t *src_ptr = src - fo_vert * src_stride; + const __m128i round_const = _mm_set1_epi32((1 << FILTER_BITS) >> 1); + const __m128i round_shift = _mm_cvtsi32_si128(FILTER_BITS); + __m128i coeffs[4]; + + (void)filter_params_x; + (void)subpel_x_q4; + (void)conv_params; + + assert(conv_params->round_0 <= FILTER_BITS); + assert(((conv_params->round_0 + conv_params->round_1) <= (FILTER_BITS + 1)) || + ((conv_params->round_0 + conv_params->round_1) == (2 * FILTER_BITS))); + + prepare_coeffs(filter_params_y, subpel_y_q4, coeffs); + + if (w <= 4) { + __m128i s[8], src6, res, res_round, res16; + uint32_t res_int; + src6 = _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 6 * src_stride)); + s[0] = _mm_unpacklo_epi8( + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 0 * src_stride)), + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 1 * src_stride))); + s[1] = _mm_unpacklo_epi8( + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 1 * src_stride)), + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 2 * src_stride))); + s[2] = _mm_unpacklo_epi8( + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 2 * src_stride)), + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 3 * src_stride))); + s[3] = _mm_unpacklo_epi8( + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 3 * src_stride)), + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 4 * src_stride))); + s[4] = _mm_unpacklo_epi8( + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 4 * src_stride)), + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 5 * src_stride))); + s[5] = _mm_unpacklo_epi8( + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 5 * src_stride)), src6); + + do { + s[6] = _mm_unpacklo_epi8( + src6, _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 7 * src_stride))); + src6 = _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 8 * src_stride)); + s[7] = _mm_unpacklo_epi8( + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 7 * src_stride)), src6); + + res = convolve_lo_y(s + 0, coeffs); + res_round = _mm_sra_epi32(_mm_add_epi32(res, round_const), round_shift); + res16 = _mm_packs_epi32(res_round, res_round); + res_int = _mm_cvtsi128_si32(_mm_packus_epi16(res16, res16)); + + if (w == 2) + *(uint16_t *)dst = res_int; + else + *(uint32_t *)dst = res_int; + + src_ptr += src_stride; + dst += dst_stride; + + res = convolve_lo_y(s + 1, coeffs); + res_round = _mm_sra_epi32(_mm_add_epi32(res, round_const), round_shift); + res16 = _mm_packs_epi32(res_round, res_round); + res_int = _mm_cvtsi128_si32(_mm_packus_epi16(res16, res16)); + + if (w == 2) + *(uint16_t *)dst = res_int; + else + *(uint32_t *)dst = res_int; + + src_ptr += src_stride; + dst += dst_stride; + + s[0] = s[2]; + s[1] = s[3]; + s[2] = s[4]; + s[3] = s[5]; + s[4] = s[6]; + s[5] = s[7]; + h -= 2; + } while (h); + } else { + assert(!(w % 8)); + int j = 0; + do { + __m128i s[8], src6, res_lo, res_hi; + __m128i res_lo_round, res_hi_round, res16, res; + const uint8_t *data = &src_ptr[j]; + + src6 = _mm_loadl_epi64((__m128i *)(data + 6 * src_stride)); + s[0] = _mm_unpacklo_epi8( + _mm_loadl_epi64((__m128i *)(data + 0 * src_stride)), + _mm_loadl_epi64((__m128i *)(data + 1 * src_stride))); + s[1] = _mm_unpacklo_epi8( + _mm_loadl_epi64((__m128i *)(data + 1 * src_stride)), + _mm_loadl_epi64((__m128i *)(data + 2 * src_stride))); + s[2] = _mm_unpacklo_epi8( + _mm_loadl_epi64((__m128i *)(data + 2 * src_stride)), + _mm_loadl_epi64((__m128i *)(data + 3 * src_stride))); + s[3] = _mm_unpacklo_epi8( + _mm_loadl_epi64((__m128i *)(data + 3 * src_stride)), + _mm_loadl_epi64((__m128i *)(data + 4 * src_stride))); + s[4] = _mm_unpacklo_epi8( + _mm_loadl_epi64((__m128i *)(data + 4 * src_stride)), + _mm_loadl_epi64((__m128i *)(data + 5 * src_stride))); + s[5] = _mm_unpacklo_epi8( + _mm_loadl_epi64((__m128i *)(data + 5 * src_stride)), src6); + + int i = 0; + do { + data = &src_ptr[i * src_stride + j]; + s[6] = _mm_unpacklo_epi8( + src6, _mm_loadl_epi64((__m128i *)(data + 7 * src_stride))); + src6 = _mm_loadl_epi64((__m128i *)(data + 8 * src_stride)); + s[7] = _mm_unpacklo_epi8( + _mm_loadl_epi64((__m128i *)(data + 7 * src_stride)), src6); + + res_lo = convolve_lo_y(s, coeffs); // Filter low index pixels + res_hi = convolve_hi_y(s, coeffs); // Filter high index pixels + + res_lo_round = + _mm_sra_epi32(_mm_add_epi32(res_lo, round_const), round_shift); + res_hi_round = + _mm_sra_epi32(_mm_add_epi32(res_hi, round_const), round_shift); + + res16 = _mm_packs_epi32(res_lo_round, res_hi_round); + res = _mm_packus_epi16(res16, res16); + + _mm_storel_epi64((__m128i *)(dst + i * dst_stride + j), res); + i++; + + res_lo = convolve_lo_y(s + 1, coeffs); // Filter low index pixels + res_hi = convolve_hi_y(s + 1, coeffs); // Filter high index pixels + + res_lo_round = + _mm_sra_epi32(_mm_add_epi32(res_lo, round_const), round_shift); + res_hi_round = + _mm_sra_epi32(_mm_add_epi32(res_hi, round_const), round_shift); + + res16 = _mm_packs_epi32(res_lo_round, res_hi_round); + res = _mm_packus_epi16(res16, res16); + + _mm_storel_epi64((__m128i *)(dst + i * dst_stride + j), res); + i++; + + s[0] = s[2]; + s[1] = s[3]; + s[2] = s[4]; + s[3] = s[5]; + s[4] = s[6]; + s[5] = s[7]; + } while (i < h); + j += 8; + } while (j < w); + } +} + +void av1_convolve_x_sr_sse2(const uint8_t *src, int src_stride, uint8_t *dst, + int dst_stride, int w, int h, + const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, + const int subpel_x_q4, const int subpel_y_q4, + ConvolveParams *conv_params) { + const int fo_horiz = filter_params_x->taps / 2 - 1; + const uint8_t *src_ptr = src - fo_horiz; + const int bits = FILTER_BITS - conv_params->round_0; + const __m128i round_0_const = + _mm_set1_epi32((1 << conv_params->round_0) >> 1); + const __m128i round_const = _mm_set1_epi32((1 << bits) >> 1); + const __m128i round_0_shift = _mm_cvtsi32_si128(conv_params->round_0); + const __m128i round_shift = _mm_cvtsi32_si128(bits); + __m128i coeffs[4]; + + (void)filter_params_y; + (void)subpel_y_q4; + + assert(bits >= 0); + assert((FILTER_BITS - conv_params->round_1) >= 0 || + ((conv_params->round_0 + conv_params->round_1) == 2 * FILTER_BITS)); + + prepare_coeffs(filter_params_x, subpel_x_q4, coeffs); + + if (w <= 4) { + do { + const __m128i data = _mm_loadu_si128((__m128i *)src_ptr); + __m128i s[4]; + + s[0] = _mm_unpacklo_epi8(data, _mm_srli_si128(data, 1)); + s[1] = + _mm_unpacklo_epi8(_mm_srli_si128(data, 2), _mm_srli_si128(data, 3)); + s[2] = + _mm_unpacklo_epi8(_mm_srli_si128(data, 4), _mm_srli_si128(data, 5)); + s[3] = + _mm_unpacklo_epi8(_mm_srli_si128(data, 6), _mm_srli_si128(data, 7)); + const __m128i res_lo = convolve_lo_x(s, coeffs); + __m128i res_lo_round = + _mm_sra_epi32(_mm_add_epi32(res_lo, round_0_const), round_0_shift); + res_lo_round = + _mm_sra_epi32(_mm_add_epi32(res_lo_round, round_const), round_shift); + + const __m128i res16 = _mm_packs_epi32(res_lo_round, res_lo_round); + const __m128i res = _mm_packus_epi16(res16, res16); + + uint32_t r = _mm_cvtsi128_si32(res); + if (w == 2) + *(uint16_t *)dst = r; + else + *(uint32_t *)dst = r; + + src_ptr += src_stride; + dst += dst_stride; + } while (--h); + } else { + assert(!(w % 8)); + int i = 0; + do { + int j = 0; + do { + const __m128i data = + _mm_loadu_si128((__m128i *)&src_ptr[i * src_stride + j]); + __m128i s[4]; + + // Filter even-index pixels + s[0] = data; + s[1] = _mm_srli_si128(data, 2); + s[2] = _mm_srli_si128(data, 4); + s[3] = _mm_srli_si128(data, 6); + const __m128i res_even = convolve_lo_x(s, coeffs); + + // Filter odd-index pixels + s[0] = _mm_srli_si128(data, 1); + s[1] = _mm_srli_si128(data, 3); + s[2] = _mm_srli_si128(data, 5); + s[3] = _mm_srli_si128(data, 7); + const __m128i res_odd = convolve_lo_x(s, coeffs); + + // Rearrange pixels back into the order 0 ... 7 + const __m128i res_lo = _mm_unpacklo_epi32(res_even, res_odd); + const __m128i res_hi = _mm_unpackhi_epi32(res_even, res_odd); + __m128i res_lo_round = + _mm_sra_epi32(_mm_add_epi32(res_lo, round_0_const), round_0_shift); + res_lo_round = _mm_sra_epi32(_mm_add_epi32(res_lo_round, round_const), + round_shift); + __m128i res_hi_round = + _mm_sra_epi32(_mm_add_epi32(res_hi, round_0_const), round_0_shift); + res_hi_round = _mm_sra_epi32(_mm_add_epi32(res_hi_round, round_const), + round_shift); + + const __m128i res16 = _mm_packs_epi32(res_lo_round, res_hi_round); + const __m128i res = _mm_packus_epi16(res16, res16); + + _mm_storel_epi64((__m128i *)(dst + i * dst_stride + j), res); + j += 8; + } while (j < w); + } while (++i < h); + } +} diff --git a/third_party/aom/av1/common/x86/filterintra_sse4.c b/third_party/aom/av1/common/x86/filterintra_sse4.c new file mode 100644 index 0000000000..c11edc1d45 --- /dev/null +++ b/third_party/aom/av1/common/x86/filterintra_sse4.c @@ -0,0 +1,75 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <smmintrin.h> + +#include "config/av1_rtcd.h" + +#include "aom_dsp/x86/synonyms.h" +#include "av1/common/enums.h" +#include "av1/common/reconintra.h" + +void av1_filter_intra_predictor_sse4_1(uint8_t *dst, ptrdiff_t stride, + TX_SIZE tx_size, const uint8_t *above, + const uint8_t *left, int mode) { + int r, c; + uint8_t buffer[33][33]; + const int bw = tx_size_wide[tx_size]; + const int bh = tx_size_high[tx_size]; + + assert(bw <= 32 && bh <= 32); + + // The initialization is just for silencing Jenkins static analysis warnings + for (r = 0; r < bh + 1; ++r) + memset(buffer[r], 0, (bw + 1) * sizeof(buffer[0][0])); + + for (r = 0; r < bh; ++r) buffer[r + 1][0] = left[r]; + memcpy(buffer[0], &above[-1], (bw + 1) * sizeof(uint8_t)); + + const __m128i f1f0 = xx_load_128(av1_filter_intra_taps[mode][0]); + const __m128i f3f2 = xx_load_128(av1_filter_intra_taps[mode][2]); + const __m128i f5f4 = xx_load_128(av1_filter_intra_taps[mode][4]); + const __m128i f7f6 = xx_load_128(av1_filter_intra_taps[mode][6]); + const __m128i filter_intra_scale_bits = + _mm_set1_epi16(1 << (15 - FILTER_INTRA_SCALE_BITS)); + + for (r = 1; r < bh + 1; r += 2) { + for (c = 1; c < bw + 1; c += 4) { + DECLARE_ALIGNED(16, uint8_t, p[8]); + memcpy(p, &buffer[r - 1][c - 1], 5 * sizeof(uint8_t)); + p[5] = buffer[r][c - 1]; + p[6] = buffer[r + 1][c - 1]; + p[7] = 0; + const __m128i p_b = xx_loadl_64(p); + const __m128i in = _mm_unpacklo_epi64(p_b, p_b); + const __m128i out_01 = _mm_maddubs_epi16(in, f1f0); + const __m128i out_23 = _mm_maddubs_epi16(in, f3f2); + const __m128i out_45 = _mm_maddubs_epi16(in, f5f4); + const __m128i out_67 = _mm_maddubs_epi16(in, f7f6); + const __m128i out_0123 = _mm_hadd_epi16(out_01, out_23); + const __m128i out_4567 = _mm_hadd_epi16(out_45, out_67); + const __m128i out_01234567 = _mm_hadd_epi16(out_0123, out_4567); + // Rounding + const __m128i round_w = + _mm_mulhrs_epi16(out_01234567, filter_intra_scale_bits); + const __m128i out_r = _mm_packus_epi16(round_w, round_w); + const __m128i out_r1 = _mm_srli_si128(out_r, 4); + // Storing + xx_storel_32(&buffer[r][c], out_r); + xx_storel_32(&buffer[r + 1][c], out_r1); + } + } + + for (r = 0; r < bh; ++r) { + memcpy(dst, &buffer[r + 1][1], bw * sizeof(uint8_t)); + dst += stride; + } +} diff --git a/third_party/aom/av1/common/x86/highbd_convolve_2d_avx2.c b/third_party/aom/av1/common/x86/highbd_convolve_2d_avx2.c new file mode 100644 index 0000000000..ae68f0bbb3 --- /dev/null +++ b/third_party/aom/av1/common/x86/highbd_convolve_2d_avx2.c @@ -0,0 +1,326 @@ +/* + * Copyright (c) 2017, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <immintrin.h> +#include <assert.h> + +#include "config/aom_dsp_rtcd.h" + +#include "aom_dsp/x86/convolve_avx2.h" +#include "aom_dsp/x86/synonyms.h" +#include "aom_dsp/aom_dsp_common.h" +#include "aom_dsp/aom_filter.h" +#include "av1/common/convolve.h" + +void av1_highbd_convolve_2d_sr_avx2(const uint16_t *src, int src_stride, + uint16_t *dst, int dst_stride, int w, int h, + const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, + const int subpel_x_q4, + const int subpel_y_q4, + ConvolveParams *conv_params, int bd) { + DECLARE_ALIGNED(32, int16_t, im_block[(MAX_SB_SIZE + MAX_FILTER_TAP) * 8]); + int im_h = h + filter_params_y->taps - 1; + int im_stride = 8; + int i, j; + const int fo_vert = filter_params_y->taps / 2 - 1; + const int fo_horiz = filter_params_x->taps / 2 - 1; + const uint16_t *const src_ptr = src - fo_vert * src_stride - fo_horiz; + + // Check that, even with 12-bit input, the intermediate values will fit + // into an unsigned 16-bit intermediate array. + assert(bd + FILTER_BITS + 2 - conv_params->round_0 <= 16); + + __m256i s[8], coeffs_y[4], coeffs_x[4]; + + const __m256i round_const_x = _mm256_set1_epi32( + ((1 << conv_params->round_0) >> 1) + (1 << (bd + FILTER_BITS - 1))); + const __m128i round_shift_x = _mm_cvtsi32_si128(conv_params->round_0); + + const __m256i round_const_y = _mm256_set1_epi32( + ((1 << conv_params->round_1) >> 1) - + (1 << (bd + 2 * FILTER_BITS - conv_params->round_0 - 1))); + const __m128i round_shift_y = _mm_cvtsi32_si128(conv_params->round_1); + + const int bits = + FILTER_BITS * 2 - conv_params->round_0 - conv_params->round_1; + const __m128i round_shift_bits = _mm_cvtsi32_si128(bits); + const __m256i round_const_bits = _mm256_set1_epi32((1 << bits) >> 1); + const __m256i clip_pixel = + _mm256_set1_epi16(bd == 10 ? 1023 : (bd == 12 ? 4095 : 255)); + const __m256i zero = _mm256_setzero_si256(); + + prepare_coeffs(filter_params_x, subpel_x_q4, coeffs_x); + prepare_coeffs(filter_params_y, subpel_y_q4, coeffs_y); + + for (j = 0; j < w; j += 8) { + /* Horizontal filter */ + { + for (i = 0; i < im_h; i += 2) { + const __m256i row0 = + _mm256_loadu_si256((__m256i *)&src_ptr[i * src_stride + j]); + __m256i row1 = _mm256_set1_epi16(0); + if (i + 1 < im_h) + row1 = + _mm256_loadu_si256((__m256i *)&src_ptr[(i + 1) * src_stride + j]); + + const __m256i r0 = _mm256_permute2x128_si256(row0, row1, 0x20); + const __m256i r1 = _mm256_permute2x128_si256(row0, row1, 0x31); + + // even pixels + s[0] = _mm256_alignr_epi8(r1, r0, 0); + s[1] = _mm256_alignr_epi8(r1, r0, 4); + s[2] = _mm256_alignr_epi8(r1, r0, 8); + s[3] = _mm256_alignr_epi8(r1, r0, 12); + + __m256i res_even = convolve(s, coeffs_x); + res_even = _mm256_sra_epi32(_mm256_add_epi32(res_even, round_const_x), + round_shift_x); + + // odd pixels + s[0] = _mm256_alignr_epi8(r1, r0, 2); + s[1] = _mm256_alignr_epi8(r1, r0, 6); + s[2] = _mm256_alignr_epi8(r1, r0, 10); + s[3] = _mm256_alignr_epi8(r1, r0, 14); + + __m256i res_odd = convolve(s, coeffs_x); + res_odd = _mm256_sra_epi32(_mm256_add_epi32(res_odd, round_const_x), + round_shift_x); + + __m256i res_even1 = _mm256_packs_epi32(res_even, res_even); + __m256i res_odd1 = _mm256_packs_epi32(res_odd, res_odd); + __m256i res = _mm256_unpacklo_epi16(res_even1, res_odd1); + + _mm256_store_si256((__m256i *)&im_block[i * im_stride], res); + } + } + + /* Vertical filter */ + { + __m256i s0 = _mm256_loadu_si256((__m256i *)(im_block + 0 * im_stride)); + __m256i s1 = _mm256_loadu_si256((__m256i *)(im_block + 1 * im_stride)); + __m256i s2 = _mm256_loadu_si256((__m256i *)(im_block + 2 * im_stride)); + __m256i s3 = _mm256_loadu_si256((__m256i *)(im_block + 3 * im_stride)); + __m256i s4 = _mm256_loadu_si256((__m256i *)(im_block + 4 * im_stride)); + __m256i s5 = _mm256_loadu_si256((__m256i *)(im_block + 5 * im_stride)); + + s[0] = _mm256_unpacklo_epi16(s0, s1); + s[1] = _mm256_unpacklo_epi16(s2, s3); + s[2] = _mm256_unpacklo_epi16(s4, s5); + + s[4] = _mm256_unpackhi_epi16(s0, s1); + s[5] = _mm256_unpackhi_epi16(s2, s3); + s[6] = _mm256_unpackhi_epi16(s4, s5); + + for (i = 0; i < h; i += 2) { + const int16_t *data = &im_block[i * im_stride]; + + const __m256i s6 = + _mm256_loadu_si256((__m256i *)(data + 6 * im_stride)); + const __m256i s7 = + _mm256_loadu_si256((__m256i *)(data + 7 * im_stride)); + + s[3] = _mm256_unpacklo_epi16(s6, s7); + s[7] = _mm256_unpackhi_epi16(s6, s7); + + const __m256i res_a = convolve(s, coeffs_y); + __m256i res_a_round = _mm256_sra_epi32( + _mm256_add_epi32(res_a, round_const_y), round_shift_y); + + res_a_round = _mm256_sra_epi32( + _mm256_add_epi32(res_a_round, round_const_bits), round_shift_bits); + + if (w - j > 4) { + const __m256i res_b = convolve(s + 4, coeffs_y); + __m256i res_b_round = _mm256_sra_epi32( + _mm256_add_epi32(res_b, round_const_y), round_shift_y); + res_b_round = + _mm256_sra_epi32(_mm256_add_epi32(res_b_round, round_const_bits), + round_shift_bits); + + __m256i res_16bit = _mm256_packs_epi32(res_a_round, res_b_round); + res_16bit = _mm256_min_epi16(res_16bit, clip_pixel); + res_16bit = _mm256_max_epi16(res_16bit, zero); + + _mm_storeu_si128((__m128i *)&dst[i * dst_stride + j], + _mm256_castsi256_si128(res_16bit)); + _mm_storeu_si128((__m128i *)&dst[i * dst_stride + j + dst_stride], + _mm256_extracti128_si256(res_16bit, 1)); + } else if (w == 4) { + res_a_round = _mm256_packs_epi32(res_a_round, res_a_round); + res_a_round = _mm256_min_epi16(res_a_round, clip_pixel); + res_a_round = _mm256_max_epi16(res_a_round, zero); + + _mm_storel_epi64((__m128i *)&dst[i * dst_stride + j], + _mm256_castsi256_si128(res_a_round)); + _mm_storel_epi64((__m128i *)&dst[i * dst_stride + j + dst_stride], + _mm256_extracti128_si256(res_a_round, 1)); + } else { + res_a_round = _mm256_packs_epi32(res_a_round, res_a_round); + res_a_round = _mm256_min_epi16(res_a_round, clip_pixel); + res_a_round = _mm256_max_epi16(res_a_round, zero); + + xx_storel_32((__m128i *)&dst[i * dst_stride + j], + _mm256_castsi256_si128(res_a_round)); + xx_storel_32((__m128i *)&dst[i * dst_stride + j + dst_stride], + _mm256_extracti128_si256(res_a_round, 1)); + } + + s[0] = s[1]; + s[1] = s[2]; + s[2] = s[3]; + + s[4] = s[5]; + s[5] = s[6]; + s[6] = s[7]; + } + } + } +} + +static INLINE void copy_64(const uint16_t *src, uint16_t *dst) { + __m256i s[4]; + s[0] = _mm256_loadu_si256((__m256i *)(src + 0 * 16)); + s[1] = _mm256_loadu_si256((__m256i *)(src + 1 * 16)); + s[2] = _mm256_loadu_si256((__m256i *)(src + 2 * 16)); + s[3] = _mm256_loadu_si256((__m256i *)(src + 3 * 16)); + _mm256_storeu_si256((__m256i *)(dst + 0 * 16), s[0]); + _mm256_storeu_si256((__m256i *)(dst + 1 * 16), s[1]); + _mm256_storeu_si256((__m256i *)(dst + 2 * 16), s[2]); + _mm256_storeu_si256((__m256i *)(dst + 3 * 16), s[3]); +} + +static INLINE void copy_128(const uint16_t *src, uint16_t *dst) { + __m256i s[8]; + s[0] = _mm256_loadu_si256((__m256i *)(src + 0 * 16)); + s[1] = _mm256_loadu_si256((__m256i *)(src + 1 * 16)); + s[2] = _mm256_loadu_si256((__m256i *)(src + 2 * 16)); + s[3] = _mm256_loadu_si256((__m256i *)(src + 3 * 16)); + s[4] = _mm256_loadu_si256((__m256i *)(src + 4 * 16)); + s[5] = _mm256_loadu_si256((__m256i *)(src + 5 * 16)); + s[6] = _mm256_loadu_si256((__m256i *)(src + 6 * 16)); + s[7] = _mm256_loadu_si256((__m256i *)(src + 7 * 16)); + + _mm256_storeu_si256((__m256i *)(dst + 0 * 16), s[0]); + _mm256_storeu_si256((__m256i *)(dst + 1 * 16), s[1]); + _mm256_storeu_si256((__m256i *)(dst + 2 * 16), s[2]); + _mm256_storeu_si256((__m256i *)(dst + 3 * 16), s[3]); + _mm256_storeu_si256((__m256i *)(dst + 4 * 16), s[4]); + _mm256_storeu_si256((__m256i *)(dst + 5 * 16), s[5]); + _mm256_storeu_si256((__m256i *)(dst + 6 * 16), s[6]); + _mm256_storeu_si256((__m256i *)(dst + 7 * 16), s[7]); +} + +void av1_highbd_convolve_2d_copy_sr_avx2( + const uint16_t *src, int src_stride, uint16_t *dst, int dst_stride, int w, + int h, const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, const int subpel_x_q4, + const int subpel_y_q4, ConvolveParams *conv_params, int bd) { + (void)filter_params_x; + (void)filter_params_y; + (void)subpel_x_q4; + (void)subpel_y_q4; + (void)conv_params; + (void)bd; + + if (w >= 16) { + assert(!((intptr_t)dst % 16)); + assert(!(dst_stride % 16)); + } + + if (w == 2) { + do { + memcpy(dst, src, 2 * sizeof(*src)); + src += src_stride; + dst += dst_stride; + memcpy(dst, src, 2 * sizeof(*src)); + src += src_stride; + dst += dst_stride; + h -= 2; + } while (h); + } else if (w == 4) { + do { + __m128i s[2]; + s[0] = _mm_loadl_epi64((__m128i *)src); + src += src_stride; + s[1] = _mm_loadl_epi64((__m128i *)src); + src += src_stride; + _mm_storel_epi64((__m128i *)dst, s[0]); + dst += dst_stride; + _mm_storel_epi64((__m128i *)dst, s[1]); + dst += dst_stride; + h -= 2; + } while (h); + } else if (w == 8) { + do { + __m128i s[2]; + s[0] = _mm_loadu_si128((__m128i *)src); + src += src_stride; + s[1] = _mm_loadu_si128((__m128i *)src); + src += src_stride; + _mm_store_si128((__m128i *)dst, s[0]); + dst += dst_stride; + _mm_store_si128((__m128i *)dst, s[1]); + dst += dst_stride; + h -= 2; + } while (h); + } else if (w == 16) { + do { + __m256i s[2]; + s[0] = _mm256_loadu_si256((__m256i *)src); + src += src_stride; + s[1] = _mm256_loadu_si256((__m256i *)src); + src += src_stride; + _mm256_storeu_si256((__m256i *)dst, s[0]); + dst += dst_stride; + _mm256_storeu_si256((__m256i *)dst, s[1]); + dst += dst_stride; + h -= 2; + } while (h); + } else if (w == 32) { + do { + __m256i s[4]; + s[0] = _mm256_loadu_si256((__m256i *)(src + 0 * 16)); + s[1] = _mm256_loadu_si256((__m256i *)(src + 1 * 16)); + src += src_stride; + s[2] = _mm256_loadu_si256((__m256i *)(src + 0 * 16)); + s[3] = _mm256_loadu_si256((__m256i *)(src + 1 * 16)); + src += src_stride; + _mm256_storeu_si256((__m256i *)(dst + 0 * 16), s[0]); + _mm256_storeu_si256((__m256i *)(dst + 1 * 16), s[1]); + dst += dst_stride; + _mm256_storeu_si256((__m256i *)(dst + 0 * 16), s[2]); + _mm256_storeu_si256((__m256i *)(dst + 1 * 16), s[3]); + dst += dst_stride; + h -= 2; + } while (h); + } else if (w == 64) { + do { + copy_64(src, dst); + src += src_stride; + dst += dst_stride; + copy_64(src, dst); + src += src_stride; + dst += dst_stride; + h -= 2; + } while (h); + } else { + do { + copy_128(src, dst); + src += src_stride; + dst += dst_stride; + copy_128(src, dst); + src += src_stride; + dst += dst_stride; + h -= 2; + } while (h); + } +} diff --git a/third_party/aom/av1/common/x86/highbd_convolve_2d_sse2.c b/third_party/aom/av1/common/x86/highbd_convolve_2d_sse2.c new file mode 100644 index 0000000000..15f8872c18 --- /dev/null +++ b/third_party/aom/av1/common/x86/highbd_convolve_2d_sse2.c @@ -0,0 +1,191 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ +#include <emmintrin.h> +#include <assert.h> + +#include "config/aom_dsp_rtcd.h" + +#include "aom_dsp/aom_filter.h" + +static INLINE void copy_64(const uint16_t *src, uint16_t *dst) { + __m128i s[8]; + s[0] = _mm_loadu_si128((__m128i *)(src + 0 * 8)); + s[1] = _mm_loadu_si128((__m128i *)(src + 1 * 8)); + s[2] = _mm_loadu_si128((__m128i *)(src + 2 * 8)); + s[3] = _mm_loadu_si128((__m128i *)(src + 3 * 8)); + s[4] = _mm_loadu_si128((__m128i *)(src + 4 * 8)); + s[5] = _mm_loadu_si128((__m128i *)(src + 5 * 8)); + s[6] = _mm_loadu_si128((__m128i *)(src + 6 * 8)); + s[7] = _mm_loadu_si128((__m128i *)(src + 7 * 8)); + _mm_store_si128((__m128i *)(dst + 0 * 8), s[0]); + _mm_store_si128((__m128i *)(dst + 1 * 8), s[1]); + _mm_store_si128((__m128i *)(dst + 2 * 8), s[2]); + _mm_store_si128((__m128i *)(dst + 3 * 8), s[3]); + _mm_store_si128((__m128i *)(dst + 4 * 8), s[4]); + _mm_store_si128((__m128i *)(dst + 5 * 8), s[5]); + _mm_store_si128((__m128i *)(dst + 6 * 8), s[6]); + _mm_store_si128((__m128i *)(dst + 7 * 8), s[7]); +} + +static INLINE void copy_128(const uint16_t *src, uint16_t *dst) { + __m128i s[16]; + s[0] = _mm_loadu_si128((__m128i *)(src + 0 * 8)); + s[1] = _mm_loadu_si128((__m128i *)(src + 1 * 8)); + s[2] = _mm_loadu_si128((__m128i *)(src + 2 * 8)); + s[3] = _mm_loadu_si128((__m128i *)(src + 3 * 8)); + s[4] = _mm_loadu_si128((__m128i *)(src + 4 * 8)); + s[5] = _mm_loadu_si128((__m128i *)(src + 5 * 8)); + s[6] = _mm_loadu_si128((__m128i *)(src + 6 * 8)); + s[7] = _mm_loadu_si128((__m128i *)(src + 7 * 8)); + s[8] = _mm_loadu_si128((__m128i *)(src + 8 * 8)); + s[9] = _mm_loadu_si128((__m128i *)(src + 9 * 8)); + s[10] = _mm_loadu_si128((__m128i *)(src + 10 * 8)); + s[11] = _mm_loadu_si128((__m128i *)(src + 11 * 8)); + s[12] = _mm_loadu_si128((__m128i *)(src + 12 * 8)); + s[13] = _mm_loadu_si128((__m128i *)(src + 13 * 8)); + s[14] = _mm_loadu_si128((__m128i *)(src + 14 * 8)); + s[15] = _mm_loadu_si128((__m128i *)(src + 15 * 8)); + _mm_store_si128((__m128i *)(dst + 0 * 8), s[0]); + _mm_store_si128((__m128i *)(dst + 1 * 8), s[1]); + _mm_store_si128((__m128i *)(dst + 2 * 8), s[2]); + _mm_store_si128((__m128i *)(dst + 3 * 8), s[3]); + _mm_store_si128((__m128i *)(dst + 4 * 8), s[4]); + _mm_store_si128((__m128i *)(dst + 5 * 8), s[5]); + _mm_store_si128((__m128i *)(dst + 6 * 8), s[6]); + _mm_store_si128((__m128i *)(dst + 7 * 8), s[7]); + _mm_store_si128((__m128i *)(dst + 8 * 8), s[8]); + _mm_store_si128((__m128i *)(dst + 9 * 8), s[9]); + _mm_store_si128((__m128i *)(dst + 10 * 8), s[10]); + _mm_store_si128((__m128i *)(dst + 11 * 8), s[11]); + _mm_store_si128((__m128i *)(dst + 12 * 8), s[12]); + _mm_store_si128((__m128i *)(dst + 13 * 8), s[13]); + _mm_store_si128((__m128i *)(dst + 14 * 8), s[14]); + _mm_store_si128((__m128i *)(dst + 15 * 8), s[15]); +} + +void av1_highbd_convolve_2d_copy_sr_sse2( + const uint16_t *src, int src_stride, uint16_t *dst, int dst_stride, int w, + int h, const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, const int subpel_x_q4, + const int subpel_y_q4, ConvolveParams *conv_params, int bd) { + (void)filter_params_x; + (void)filter_params_y; + (void)subpel_x_q4; + (void)subpel_y_q4; + (void)conv_params; + (void)bd; + if (w >= 16) { + assert(!((intptr_t)dst % 16)); + assert(!(dst_stride % 16)); + } + + if (w == 2) { + do { + __m128i s = _mm_loadl_epi64((__m128i *)src); + *(uint32_t *)dst = _mm_cvtsi128_si32(s); + src += src_stride; + dst += dst_stride; + s = _mm_loadl_epi64((__m128i *)src); + *(uint32_t *)dst = _mm_cvtsi128_si32(s); + src += src_stride; + dst += dst_stride; + h -= 2; + } while (h); + } else if (w == 4) { + do { + __m128i s[2]; + s[0] = _mm_loadl_epi64((__m128i *)src); + src += src_stride; + s[1] = _mm_loadl_epi64((__m128i *)src); + src += src_stride; + _mm_storel_epi64((__m128i *)dst, s[0]); + dst += dst_stride; + _mm_storel_epi64((__m128i *)dst, s[1]); + dst += dst_stride; + h -= 2; + } while (h); + } else if (w == 8) { + do { + __m128i s[2]; + s[0] = _mm_loadu_si128((__m128i *)src); + src += src_stride; + s[1] = _mm_loadu_si128((__m128i *)src); + src += src_stride; + _mm_store_si128((__m128i *)dst, s[0]); + dst += dst_stride; + _mm_store_si128((__m128i *)dst, s[1]); + dst += dst_stride; + h -= 2; + } while (h); + } else if (w == 16) { + do { + __m128i s[4]; + s[0] = _mm_loadu_si128((__m128i *)(src + 0 * 8)); + s[1] = _mm_loadu_si128((__m128i *)(src + 1 * 8)); + src += src_stride; + s[2] = _mm_loadu_si128((__m128i *)(src + 0 * 8)); + s[3] = _mm_loadu_si128((__m128i *)(src + 1 * 8)); + src += src_stride; + _mm_store_si128((__m128i *)(dst + 0 * 8), s[0]); + _mm_store_si128((__m128i *)(dst + 1 * 8), s[1]); + dst += dst_stride; + _mm_store_si128((__m128i *)(dst + 0 * 8), s[2]); + _mm_store_si128((__m128i *)(dst + 1 * 8), s[3]); + dst += dst_stride; + h -= 2; + } while (h); + } else if (w == 32) { + do { + __m128i s[8]; + s[0] = _mm_loadu_si128((__m128i *)(src + 0 * 8)); + s[1] = _mm_loadu_si128((__m128i *)(src + 1 * 8)); + s[2] = _mm_loadu_si128((__m128i *)(src + 2 * 8)); + s[3] = _mm_loadu_si128((__m128i *)(src + 3 * 8)); + src += src_stride; + s[4] = _mm_loadu_si128((__m128i *)(src + 0 * 8)); + s[5] = _mm_loadu_si128((__m128i *)(src + 1 * 8)); + s[6] = _mm_loadu_si128((__m128i *)(src + 2 * 8)); + s[7] = _mm_loadu_si128((__m128i *)(src + 3 * 8)); + src += src_stride; + _mm_store_si128((__m128i *)(dst + 0 * 8), s[0]); + _mm_store_si128((__m128i *)(dst + 1 * 8), s[1]); + _mm_store_si128((__m128i *)(dst + 2 * 8), s[2]); + _mm_store_si128((__m128i *)(dst + 3 * 8), s[3]); + dst += dst_stride; + _mm_store_si128((__m128i *)(dst + 0 * 8), s[4]); + _mm_store_si128((__m128i *)(dst + 1 * 8), s[5]); + _mm_store_si128((__m128i *)(dst + 2 * 8), s[6]); + _mm_store_si128((__m128i *)(dst + 3 * 8), s[7]); + dst += dst_stride; + h -= 2; + } while (h); + } else if (w == 64) { + do { + copy_64(src, dst); + src += src_stride; + dst += dst_stride; + copy_64(src, dst); + src += src_stride; + dst += dst_stride; + h -= 2; + } while (h); + } else { + do { + copy_128(src, dst); + src += src_stride; + dst += dst_stride; + copy_128(src, dst); + src += src_stride; + dst += dst_stride; + h -= 2; + } while (h); + } +} diff --git a/third_party/aom/av1/common/x86/highbd_convolve_2d_sse4.c b/third_party/aom/av1/common/x86/highbd_convolve_2d_sse4.c new file mode 100644 index 0000000000..3f8dafb4b3 --- /dev/null +++ b/third_party/aom/av1/common/x86/highbd_convolve_2d_sse4.c @@ -0,0 +1,420 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <tmmintrin.h> +#include <smmintrin.h> +#include <assert.h> + +#include "config/aom_dsp_rtcd.h" + +#include "aom_dsp/aom_dsp_common.h" +#include "aom_dsp/aom_filter.h" +#include "aom_dsp/x86/convolve_sse2.h" +#include "aom_dsp/x86/convolve_sse4_1.h" +#include "av1/common/convolve.h" + +void av1_highbd_jnt_convolve_2d_copy_sse4_1( + const uint16_t *src, int src_stride, uint16_t *dst0, int dst_stride0, int w, + int h, const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, const int subpel_x_q4, + const int subpel_y_q4, ConvolveParams *conv_params, int bd) { + CONV_BUF_TYPE *dst = conv_params->dst; + int dst_stride = conv_params->dst_stride; + (void)filter_params_x; + (void)filter_params_y; + (void)subpel_x_q4; + (void)subpel_y_q4; + + const int bits = + FILTER_BITS * 2 - conv_params->round_1 - conv_params->round_0; + const __m128i left_shift = _mm_cvtsi32_si128(bits); + const int do_average = conv_params->do_average; + const int use_jnt_comp_avg = conv_params->use_jnt_comp_avg; + const int w0 = conv_params->fwd_offset; + const int w1 = conv_params->bck_offset; + const __m128i wt0 = _mm_set1_epi32(w0); + const __m128i wt1 = _mm_set1_epi32(w1); + const __m128i zero = _mm_setzero_si128(); + int i, j; + + const int offset_0 = + bd + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const int offset = (1 << offset_0) + (1 << (offset_0 - 1)); + const __m128i offset_const = _mm_set1_epi32(offset); + const __m128i offset_const_16b = _mm_set1_epi16(offset); + const int rounding_shift = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const __m128i rounding_const = _mm_set1_epi32((1 << rounding_shift) >> 1); + const __m128i clip_pixel_to_bd = + _mm_set1_epi16(bd == 10 ? 1023 : (bd == 12 ? 4095 : 255)); + + assert(bits <= 4); + + if (!(w % 8)) { + for (i = 0; i < h; i += 1) { + for (j = 0; j < w; j += 8) { + const __m128i src_16bit = + _mm_loadu_si128((__m128i *)(&src[i * src_stride + j])); + const __m128i res = _mm_sll_epi16(src_16bit, left_shift); + if (do_average) { + const __m128i data_0 = + _mm_loadu_si128((__m128i *)(&dst[i * dst_stride + j])); + + const __m128i data_ref_0_lo = _mm_unpacklo_epi16(data_0, zero); + const __m128i data_ref_0_hi = _mm_unpackhi_epi16(data_0, zero); + + const __m128i res_32b_lo = _mm_unpacklo_epi16(res, zero); + const __m128i res_unsigned_lo = + _mm_add_epi32(res_32b_lo, offset_const); + + const __m128i comp_avg_res_lo = highbd_comp_avg_sse4_1( + &data_ref_0_lo, &res_unsigned_lo, &wt0, &wt1, use_jnt_comp_avg); + + const __m128i res_32b_hi = _mm_unpackhi_epi16(res, zero); + const __m128i res_unsigned_hi = + _mm_add_epi32(res_32b_hi, offset_const); + + const __m128i comp_avg_res_hi = highbd_comp_avg_sse4_1( + &data_ref_0_hi, &res_unsigned_hi, &wt0, &wt1, use_jnt_comp_avg); + + const __m128i round_result_lo = highbd_convolve_rounding_sse2( + &comp_avg_res_lo, &offset_const, &rounding_const, rounding_shift); + const __m128i round_result_hi = highbd_convolve_rounding_sse2( + &comp_avg_res_hi, &offset_const, &rounding_const, rounding_shift); + + const __m128i res_16b = + _mm_packus_epi32(round_result_lo, round_result_hi); + const __m128i res_clip = _mm_min_epi16(res_16b, clip_pixel_to_bd); + + _mm_store_si128((__m128i *)(&dst0[i * dst_stride0 + j]), res_clip); + } else { + const __m128i res_unsigned_16b = + _mm_adds_epu16(res, offset_const_16b); + + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j]), + res_unsigned_16b); + } + } + } + } else if (!(w % 4)) { + for (i = 0; i < h; i += 2) { + for (j = 0; j < w; j += 4) { + const __m128i src_row_0 = + _mm_loadl_epi64((__m128i *)(&src[i * src_stride + j])); + const __m128i src_row_1 = + _mm_loadl_epi64((__m128i *)(&src[i * src_stride + j + src_stride])); + const __m128i src_10 = _mm_unpacklo_epi64(src_row_0, src_row_1); + + const __m128i res = _mm_sll_epi16(src_10, left_shift); + + if (do_average) { + const __m128i data_0 = + _mm_loadl_epi64((__m128i *)(&dst[i * dst_stride + j])); + const __m128i data_1 = _mm_loadl_epi64( + (__m128i *)(&dst[i * dst_stride + j + dst_stride])); + + const __m128i data_ref_0 = _mm_unpacklo_epi16(data_0, zero); + const __m128i data_ref_1 = _mm_unpacklo_epi16(data_1, zero); + + const __m128i res_32b = _mm_unpacklo_epi16(res, zero); + const __m128i res_unsigned_lo = _mm_add_epi32(res_32b, offset_const); + + const __m128i res_32b_hi = _mm_unpackhi_epi16(res, zero); + const __m128i res_unsigned_hi = + _mm_add_epi32(res_32b_hi, offset_const); + + const __m128i comp_avg_res_lo = highbd_comp_avg_sse4_1( + &data_ref_0, &res_unsigned_lo, &wt0, &wt1, use_jnt_comp_avg); + const __m128i comp_avg_res_hi = highbd_comp_avg_sse4_1( + &data_ref_1, &res_unsigned_hi, &wt0, &wt1, use_jnt_comp_avg); + + const __m128i round_result_lo = highbd_convolve_rounding_sse2( + &comp_avg_res_lo, &offset_const, &rounding_const, rounding_shift); + const __m128i round_result_hi = highbd_convolve_rounding_sse2( + &comp_avg_res_hi, &offset_const, &rounding_const, rounding_shift); + + const __m128i res_16b = + _mm_packus_epi32(round_result_lo, round_result_hi); + const __m128i res_clip = _mm_min_epi16(res_16b, clip_pixel_to_bd); + + const __m128i res_1 = _mm_srli_si128(res_clip, 8); + + _mm_storel_epi64((__m128i *)(&dst0[i * dst_stride0 + j]), res_clip); + _mm_storel_epi64( + (__m128i *)(&dst0[i * dst_stride0 + j + dst_stride0]), res_1); + } else { + const __m128i res_unsigned_16b = + _mm_adds_epu16(res, offset_const_16b); + + const __m128i res_1 = _mm_srli_si128(res_unsigned_16b, 8); + + _mm_storel_epi64((__m128i *)(&dst[i * dst_stride + j]), + res_unsigned_16b); + _mm_storel_epi64((__m128i *)(&dst[i * dst_stride + j + dst_stride]), + res_1); + } + } + } + } +} + +void av1_highbd_jnt_convolve_2d_sse4_1( + const uint16_t *src, int src_stride, uint16_t *dst0, int dst_stride0, int w, + int h, const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, const int subpel_x_q4, + const int subpel_y_q4, ConvolveParams *conv_params, int bd) { + DECLARE_ALIGNED(16, int16_t, + im_block[(MAX_SB_SIZE + MAX_FILTER_TAP - 1) * MAX_SB_SIZE]); + CONV_BUF_TYPE *dst = conv_params->dst; + int dst_stride = conv_params->dst_stride; + int im_h = h + filter_params_y->taps - 1; + int im_stride = MAX_SB_SIZE; + int i, j; + const int do_average = conv_params->do_average; + const int use_jnt_comp_avg = conv_params->use_jnt_comp_avg; + const int fo_vert = filter_params_y->taps / 2 - 1; + const int fo_horiz = filter_params_x->taps / 2 - 1; + const uint16_t *const src_ptr = src - fo_vert * src_stride - fo_horiz; + + const int w0 = conv_params->fwd_offset; + const int w1 = conv_params->bck_offset; + const __m128i wt0 = _mm_set1_epi32(w0); + const __m128i wt1 = _mm_set1_epi32(w1); + + const int offset_0 = + bd + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const int offset = (1 << offset_0) + (1 << (offset_0 - 1)); + const __m128i offset_const = _mm_set1_epi32(offset); + const int rounding_shift = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const __m128i rounding_const = _mm_set1_epi32((1 << rounding_shift) >> 1); + const __m128i clip_pixel_to_bd = + _mm_set1_epi16(bd == 10 ? 1023 : (bd == 12 ? 4095 : 255)); + + // Check that, even with 12-bit input, the intermediate values will fit + // into an unsigned 16-bit intermediate array. + assert(bd + FILTER_BITS + 2 - conv_params->round_0 <= 16); + + /* Horizontal filter */ + { + const int16_t *x_filter = av1_get_interp_filter_subpel_kernel( + filter_params_x, subpel_x_q4 & SUBPEL_MASK); + const __m128i coeffs_x = _mm_loadu_si128((__m128i *)x_filter); + + // coeffs 0 1 0 1 2 3 2 3 + const __m128i tmp_0 = _mm_unpacklo_epi32(coeffs_x, coeffs_x); + // coeffs 4 5 4 5 6 7 6 7 + const __m128i tmp_1 = _mm_unpackhi_epi32(coeffs_x, coeffs_x); + + // coeffs 0 1 0 1 0 1 0 1 + const __m128i coeff_01 = _mm_unpacklo_epi64(tmp_0, tmp_0); + // coeffs 2 3 2 3 2 3 2 3 + const __m128i coeff_23 = _mm_unpackhi_epi64(tmp_0, tmp_0); + // coeffs 4 5 4 5 4 5 4 5 + const __m128i coeff_45 = _mm_unpacklo_epi64(tmp_1, tmp_1); + // coeffs 6 7 6 7 6 7 6 7 + const __m128i coeff_67 = _mm_unpackhi_epi64(tmp_1, tmp_1); + + const __m128i round_const = _mm_set1_epi32( + ((1 << conv_params->round_0) >> 1) + (1 << (bd + FILTER_BITS - 1))); + const __m128i round_shift = _mm_cvtsi32_si128(conv_params->round_0); + + for (i = 0; i < im_h; ++i) { + for (j = 0; j < w; j += 8) { + const __m128i data = + _mm_loadu_si128((__m128i *)&src_ptr[i * src_stride + j]); + const __m128i data2 = + _mm_loadu_si128((__m128i *)&src_ptr[i * src_stride + j + 8]); + + // Filter even-index pixels + const __m128i res_0 = _mm_madd_epi16(data, coeff_01); + const __m128i res_2 = + _mm_madd_epi16(_mm_alignr_epi8(data2, data, 4), coeff_23); + const __m128i res_4 = + _mm_madd_epi16(_mm_alignr_epi8(data2, data, 8), coeff_45); + const __m128i res_6 = + _mm_madd_epi16(_mm_alignr_epi8(data2, data, 12), coeff_67); + + __m128i res_even = _mm_add_epi32(_mm_add_epi32(res_0, res_4), + _mm_add_epi32(res_2, res_6)); + res_even = + _mm_sra_epi32(_mm_add_epi32(res_even, round_const), round_shift); + + // Filter odd-index pixels + const __m128i res_1 = + _mm_madd_epi16(_mm_alignr_epi8(data2, data, 2), coeff_01); + const __m128i res_3 = + _mm_madd_epi16(_mm_alignr_epi8(data2, data, 6), coeff_23); + const __m128i res_5 = + _mm_madd_epi16(_mm_alignr_epi8(data2, data, 10), coeff_45); + const __m128i res_7 = + _mm_madd_epi16(_mm_alignr_epi8(data2, data, 14), coeff_67); + + __m128i res_odd = _mm_add_epi32(_mm_add_epi32(res_1, res_5), + _mm_add_epi32(res_3, res_7)); + res_odd = + _mm_sra_epi32(_mm_add_epi32(res_odd, round_const), round_shift); + + // Pack in the column order 0, 2, 4, 6, 1, 3, 5, 7 + __m128i res = _mm_packs_epi32(res_even, res_odd); + _mm_storeu_si128((__m128i *)&im_block[i * im_stride + j], res); + } + } + } + + /* Vertical filter */ + { + const int16_t *y_filter = av1_get_interp_filter_subpel_kernel( + filter_params_y, subpel_y_q4 & SUBPEL_MASK); + const __m128i coeffs_y = _mm_loadu_si128((__m128i *)y_filter); + + // coeffs 0 1 0 1 2 3 2 3 + const __m128i tmp_0 = _mm_unpacklo_epi32(coeffs_y, coeffs_y); + // coeffs 4 5 4 5 6 7 6 7 + const __m128i tmp_1 = _mm_unpackhi_epi32(coeffs_y, coeffs_y); + + // coeffs 0 1 0 1 0 1 0 1 + const __m128i coeff_01 = _mm_unpacklo_epi64(tmp_0, tmp_0); + // coeffs 2 3 2 3 2 3 2 3 + const __m128i coeff_23 = _mm_unpackhi_epi64(tmp_0, tmp_0); + // coeffs 4 5 4 5 4 5 4 5 + const __m128i coeff_45 = _mm_unpacklo_epi64(tmp_1, tmp_1); + // coeffs 6 7 6 7 6 7 6 7 + const __m128i coeff_67 = _mm_unpackhi_epi64(tmp_1, tmp_1); + + const __m128i round_const = _mm_set1_epi32( + ((1 << conv_params->round_1) >> 1) - + (1 << (bd + 2 * FILTER_BITS - conv_params->round_0 - 1))); + const __m128i round_shift = _mm_cvtsi32_si128(conv_params->round_1); + + for (i = 0; i < h; ++i) { + for (j = 0; j < w; j += 8) { + // Filter even-index pixels + const int16_t *data = &im_block[i * im_stride + j]; + const __m128i src_0 = + _mm_unpacklo_epi16(*(__m128i *)(data + 0 * im_stride), + *(__m128i *)(data + 1 * im_stride)); + const __m128i src_2 = + _mm_unpacklo_epi16(*(__m128i *)(data + 2 * im_stride), + *(__m128i *)(data + 3 * im_stride)); + const __m128i src_4 = + _mm_unpacklo_epi16(*(__m128i *)(data + 4 * im_stride), + *(__m128i *)(data + 5 * im_stride)); + const __m128i src_6 = + _mm_unpacklo_epi16(*(__m128i *)(data + 6 * im_stride), + *(__m128i *)(data + 7 * im_stride)); + + const __m128i res_0 = _mm_madd_epi16(src_0, coeff_01); + const __m128i res_2 = _mm_madd_epi16(src_2, coeff_23); + const __m128i res_4 = _mm_madd_epi16(src_4, coeff_45); + const __m128i res_6 = _mm_madd_epi16(src_6, coeff_67); + + const __m128i res_even = _mm_add_epi32(_mm_add_epi32(res_0, res_2), + _mm_add_epi32(res_4, res_6)); + + // Filter odd-index pixels + const __m128i src_1 = + _mm_unpackhi_epi16(*(__m128i *)(data + 0 * im_stride), + *(__m128i *)(data + 1 * im_stride)); + const __m128i src_3 = + _mm_unpackhi_epi16(*(__m128i *)(data + 2 * im_stride), + *(__m128i *)(data + 3 * im_stride)); + const __m128i src_5 = + _mm_unpackhi_epi16(*(__m128i *)(data + 4 * im_stride), + *(__m128i *)(data + 5 * im_stride)); + const __m128i src_7 = + _mm_unpackhi_epi16(*(__m128i *)(data + 6 * im_stride), + *(__m128i *)(data + 7 * im_stride)); + + const __m128i res_1 = _mm_madd_epi16(src_1, coeff_01); + const __m128i res_3 = _mm_madd_epi16(src_3, coeff_23); + const __m128i res_5 = _mm_madd_epi16(src_5, coeff_45); + const __m128i res_7 = _mm_madd_epi16(src_7, coeff_67); + + const __m128i res_odd = _mm_add_epi32(_mm_add_epi32(res_1, res_3), + _mm_add_epi32(res_5, res_7)); + + // Rearrange pixels back into the order 0 ... 7 + const __m128i res_lo = _mm_unpacklo_epi32(res_even, res_odd); + const __m128i res_hi = _mm_unpackhi_epi32(res_even, res_odd); + + const __m128i res_lo_round = + _mm_sra_epi32(_mm_add_epi32(res_lo, round_const), round_shift); + + const __m128i res_unsigned_lo = + _mm_add_epi32(res_lo_round, offset_const); + + if (w < 8) { + if (do_average) { + const __m128i data_0 = + _mm_loadl_epi64((__m128i *)(&dst[i * dst_stride + j])); + + const __m128i data_ref_0 = _mm_cvtepu16_epi32(data_0); + + const __m128i comp_avg_res = highbd_comp_avg_sse4_1( + &data_ref_0, &res_unsigned_lo, &wt0, &wt1, use_jnt_comp_avg); + + const __m128i round_result = highbd_convolve_rounding_sse2( + &comp_avg_res, &offset_const, &rounding_const, rounding_shift); + + const __m128i res_16b = + _mm_packus_epi32(round_result, round_result); + const __m128i res_clip = _mm_min_epi16(res_16b, clip_pixel_to_bd); + + _mm_storel_epi64((__m128i *)(&dst0[i * dst_stride0 + j]), res_clip); + } else { + const __m128i res_16b = + _mm_packus_epi32(res_unsigned_lo, res_unsigned_lo); + _mm_storel_epi64((__m128i *)(&dst[i * dst_stride + j]), res_16b); + } + } else { + const __m128i res_hi_round = + _mm_sra_epi32(_mm_add_epi32(res_hi, round_const), round_shift); + + const __m128i res_unsigned_hi = + _mm_add_epi32(res_hi_round, offset_const); + + if (do_average) { + const __m128i data_lo = + _mm_loadl_epi64((__m128i *)(&dst[i * dst_stride + j])); + const __m128i data_hi = + _mm_loadl_epi64((__m128i *)(&dst[i * dst_stride + j + 4])); + + const __m128i data_ref_0_lo = _mm_cvtepu16_epi32(data_lo); + const __m128i data_ref_0_hi = _mm_cvtepu16_epi32(data_hi); + + const __m128i comp_avg_res_lo = highbd_comp_avg_sse4_1( + &data_ref_0_lo, &res_unsigned_lo, &wt0, &wt1, use_jnt_comp_avg); + const __m128i comp_avg_res_hi = highbd_comp_avg_sse4_1( + &data_ref_0_hi, &res_unsigned_hi, &wt0, &wt1, use_jnt_comp_avg); + + const __m128i round_result_lo = + highbd_convolve_rounding_sse2(&comp_avg_res_lo, &offset_const, + &rounding_const, rounding_shift); + const __m128i round_result_hi = + highbd_convolve_rounding_sse2(&comp_avg_res_hi, &offset_const, + &rounding_const, rounding_shift); + + const __m128i res_16b = + _mm_packus_epi32(round_result_lo, round_result_hi); + const __m128i res_clip = _mm_min_epi16(res_16b, clip_pixel_to_bd); + + _mm_store_si128((__m128i *)(&dst0[i * dst_stride0 + j]), res_clip); + } else { + const __m128i res_16b = + _mm_packus_epi32(res_unsigned_lo, res_unsigned_hi); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j]), res_16b); + } + } + } + } + } +} diff --git a/third_party/aom/av1/common/x86/highbd_convolve_2d_ssse3.c b/third_party/aom/av1/common/x86/highbd_convolve_2d_ssse3.c new file mode 100644 index 0000000000..1d029db39a --- /dev/null +++ b/third_party/aom/av1/common/x86/highbd_convolve_2d_ssse3.c @@ -0,0 +1,217 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <tmmintrin.h> +#include <assert.h> + +#include "config/aom_dsp_rtcd.h" + +#include "aom_dsp/aom_dsp_common.h" +#include "aom_dsp/aom_filter.h" +#include "aom_dsp/x86/convolve_sse2.h" +#include "av1/common/convolve.h" + +void av1_highbd_convolve_2d_sr_ssse3( + const uint16_t *src, int src_stride, uint16_t *dst, int dst_stride, int w, + int h, const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, const int subpel_x_q4, + const int subpel_y_q4, ConvolveParams *conv_params, int bd) { + DECLARE_ALIGNED(32, int16_t, im_block[(MAX_SB_SIZE + MAX_FILTER_TAP) * 8]); + int im_h = h + filter_params_y->taps - 1; + int im_stride = 8; + int i, j; + const int fo_vert = filter_params_y->taps / 2 - 1; + const int fo_horiz = filter_params_x->taps / 2 - 1; + const uint16_t *const src_ptr = src - fo_vert * src_stride - fo_horiz; + + // Check that, even with 12-bit input, the intermediate values will fit + // into an unsigned 16-bit intermediate array. + assert(bd + FILTER_BITS + 2 - conv_params->round_0 <= 16); + __m128i coeffs_x[4], coeffs_y[4], s[16]; + + const __m128i round_const_x = _mm_set1_epi32( + ((1 << conv_params->round_0) >> 1) + (1 << (bd + FILTER_BITS - 1))); + const __m128i round_shift_x = _mm_cvtsi32_si128(conv_params->round_0); + + const __m128i round_const_y = + _mm_set1_epi32(((1 << conv_params->round_1) >> 1) - + (1 << (bd + 2 * FILTER_BITS - conv_params->round_0 - 1))); + const __m128i round_shift_y = _mm_cvtsi32_si128(conv_params->round_1); + + const int bits = + FILTER_BITS * 2 - conv_params->round_0 - conv_params->round_1; + const __m128i round_shift_bits = _mm_cvtsi32_si128(bits); + const __m128i round_const_bits = _mm_set1_epi32((1 << bits) >> 1); + const __m128i clip_pixel = + _mm_set1_epi16(bd == 10 ? 1023 : (bd == 12 ? 4095 : 255)); + const __m128i zero = _mm_setzero_si128(); + + prepare_coeffs(filter_params_x, subpel_x_q4, coeffs_x); + prepare_coeffs(filter_params_y, subpel_y_q4, coeffs_y); + + for (j = 0; j < w; j += 8) { + /* Horizontal filter */ + { + for (i = 0; i < im_h; i += 1) { + const __m128i row00 = + _mm_loadu_si128((__m128i *)&src_ptr[i * src_stride + j]); + const __m128i row01 = + _mm_loadu_si128((__m128i *)&src_ptr[i * src_stride + (j + 8)]); + + // even pixels + s[0] = _mm_alignr_epi8(row01, row00, 0); + s[1] = _mm_alignr_epi8(row01, row00, 4); + s[2] = _mm_alignr_epi8(row01, row00, 8); + s[3] = _mm_alignr_epi8(row01, row00, 12); + + __m128i res_even = convolve(s, coeffs_x); + res_even = _mm_sra_epi32(_mm_add_epi32(res_even, round_const_x), + round_shift_x); + + // odd pixels + s[0] = _mm_alignr_epi8(row01, row00, 2); + s[1] = _mm_alignr_epi8(row01, row00, 6); + s[2] = _mm_alignr_epi8(row01, row00, 10); + s[3] = _mm_alignr_epi8(row01, row00, 14); + + __m128i res_odd = convolve(s, coeffs_x); + res_odd = + _mm_sra_epi32(_mm_add_epi32(res_odd, round_const_x), round_shift_x); + + __m128i res_even1 = _mm_packs_epi32(res_even, res_even); + __m128i res_odd1 = _mm_packs_epi32(res_odd, res_odd); + __m128i res = _mm_unpacklo_epi16(res_even1, res_odd1); + + _mm_store_si128((__m128i *)&im_block[i * im_stride], res); + } + } + /* Vertical filter */ + { + __m128i s0 = _mm_loadu_si128((__m128i *)(im_block + 0 * im_stride)); + __m128i s1 = _mm_loadu_si128((__m128i *)(im_block + 1 * im_stride)); + __m128i s2 = _mm_loadu_si128((__m128i *)(im_block + 2 * im_stride)); + __m128i s3 = _mm_loadu_si128((__m128i *)(im_block + 3 * im_stride)); + __m128i s4 = _mm_loadu_si128((__m128i *)(im_block + 4 * im_stride)); + __m128i s5 = _mm_loadu_si128((__m128i *)(im_block + 5 * im_stride)); + __m128i s6 = _mm_loadu_si128((__m128i *)(im_block + 6 * im_stride)); + + s[0] = _mm_unpacklo_epi16(s0, s1); + s[1] = _mm_unpacklo_epi16(s2, s3); + s[2] = _mm_unpacklo_epi16(s4, s5); + + s[4] = _mm_unpackhi_epi16(s0, s1); + s[5] = _mm_unpackhi_epi16(s2, s3); + s[6] = _mm_unpackhi_epi16(s4, s5); + + s[0 + 8] = _mm_unpacklo_epi16(s1, s2); + s[1 + 8] = _mm_unpacklo_epi16(s3, s4); + s[2 + 8] = _mm_unpacklo_epi16(s5, s6); + + s[4 + 8] = _mm_unpackhi_epi16(s1, s2); + s[5 + 8] = _mm_unpackhi_epi16(s3, s4); + s[6 + 8] = _mm_unpackhi_epi16(s5, s6); + + for (i = 0; i < h; i += 2) { + const int16_t *data = &im_block[i * im_stride]; + + __m128i s7 = _mm_loadu_si128((__m128i *)(data + 7 * im_stride)); + __m128i s8 = _mm_loadu_si128((__m128i *)(data + 8 * im_stride)); + + s[3] = _mm_unpacklo_epi16(s6, s7); + s[7] = _mm_unpackhi_epi16(s6, s7); + + s[3 + 8] = _mm_unpacklo_epi16(s7, s8); + s[7 + 8] = _mm_unpackhi_epi16(s7, s8); + + const __m128i res_a0 = convolve(s, coeffs_y); + __m128i res_a_round0 = + _mm_sra_epi32(_mm_add_epi32(res_a0, round_const_y), round_shift_y); + res_a_round0 = _mm_sra_epi32( + _mm_add_epi32(res_a_round0, round_const_bits), round_shift_bits); + + const __m128i res_a1 = convolve(s + 8, coeffs_y); + __m128i res_a_round1 = + _mm_sra_epi32(_mm_add_epi32(res_a1, round_const_y), round_shift_y); + res_a_round1 = _mm_sra_epi32( + _mm_add_epi32(res_a_round1, round_const_bits), round_shift_bits); + + if (w - j > 4) { + const __m128i res_b0 = convolve(s + 4, coeffs_y); + __m128i res_b_round0 = _mm_sra_epi32( + _mm_add_epi32(res_b0, round_const_y), round_shift_y); + res_b_round0 = _mm_sra_epi32( + _mm_add_epi32(res_b_round0, round_const_bits), round_shift_bits); + + const __m128i res_b1 = convolve(s + 4 + 8, coeffs_y); + __m128i res_b_round1 = _mm_sra_epi32( + _mm_add_epi32(res_b1, round_const_y), round_shift_y); + res_b_round1 = _mm_sra_epi32( + _mm_add_epi32(res_b_round1, round_const_bits), round_shift_bits); + + __m128i res_16bit0 = _mm_packs_epi32(res_a_round0, res_b_round0); + res_16bit0 = _mm_min_epi16(res_16bit0, clip_pixel); + res_16bit0 = _mm_max_epi16(res_16bit0, zero); + + __m128i res_16bit1 = _mm_packs_epi32(res_a_round1, res_b_round1); + res_16bit1 = _mm_min_epi16(res_16bit1, clip_pixel); + res_16bit1 = _mm_max_epi16(res_16bit1, zero); + + _mm_storeu_si128((__m128i *)&dst[i * dst_stride + j], res_16bit0); + _mm_storeu_si128((__m128i *)&dst[i * dst_stride + j + dst_stride], + res_16bit1); + } else if (w == 4) { + res_a_round0 = _mm_packs_epi32(res_a_round0, res_a_round0); + res_a_round0 = _mm_min_epi16(res_a_round0, clip_pixel); + res_a_round0 = _mm_max_epi16(res_a_round0, zero); + + res_a_round1 = _mm_packs_epi32(res_a_round1, res_a_round1); + res_a_round1 = _mm_min_epi16(res_a_round1, clip_pixel); + res_a_round1 = _mm_max_epi16(res_a_round1, zero); + + _mm_storel_epi64((__m128i *)&dst[i * dst_stride + j], res_a_round0); + _mm_storel_epi64((__m128i *)&dst[i * dst_stride + j + dst_stride], + res_a_round1); + } else { + res_a_round0 = _mm_packs_epi32(res_a_round0, res_a_round0); + res_a_round0 = _mm_min_epi16(res_a_round0, clip_pixel); + res_a_round0 = _mm_max_epi16(res_a_round0, zero); + + res_a_round1 = _mm_packs_epi32(res_a_round1, res_a_round1); + res_a_round1 = _mm_min_epi16(res_a_round1, clip_pixel); + res_a_round1 = _mm_max_epi16(res_a_round1, zero); + + *((uint32_t *)(&dst[i * dst_stride + j])) = + _mm_cvtsi128_si32(res_a_round0); + + *((uint32_t *)(&dst[i * dst_stride + j + dst_stride])) = + _mm_cvtsi128_si32(res_a_round1); + } + s[0] = s[1]; + s[1] = s[2]; + s[2] = s[3]; + + s[4] = s[5]; + s[5] = s[6]; + s[6] = s[7]; + + s[0 + 8] = s[1 + 8]; + s[1 + 8] = s[2 + 8]; + s[2 + 8] = s[3 + 8]; + + s[4 + 8] = s[5 + 8]; + s[5 + 8] = s[6 + 8]; + s[6 + 8] = s[7 + 8]; + + s6 = s8; + } + } + } +} diff --git a/third_party/aom/av1/common/x86/highbd_inv_txfm_avx2.c b/third_party/aom/av1/common/x86/highbd_inv_txfm_avx2.c new file mode 100644 index 0000000000..ade2af03e4 --- /dev/null +++ b/third_party/aom/av1/common/x86/highbd_inv_txfm_avx2.c @@ -0,0 +1,1349 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ +#include <assert.h> +#include <immintrin.h> + +#include "config/aom_config.h" +#include "config/av1_rtcd.h" + +#include "av1/common/av1_inv_txfm1d_cfg.h" +#include "av1/common/idct.h" +#include "av1/common/x86/av1_inv_txfm_ssse3.h" +#include "av1/common/x86/highbd_txfm_utility_sse4.h" + +// Note: +// Total 32x4 registers to represent 32x32 block coefficients. +// For high bit depth, each coefficient is 4-byte. +// Each __m256i register holds 8 coefficients. +// So each "row" we needs 4 register. Totally 32 rows +// Register layout: +// v0, v1, v2, v3, +// v4, v5, v6, v7, +// ... ... +// v124, v125, v126, v127 + +static INLINE __m256i highbd_clamp_epi16_avx2(__m256i u, int bd) { + const __m256i zero = _mm256_setzero_si256(); + const __m256i one = _mm256_set1_epi16(1); + const __m256i max = _mm256_sub_epi16(_mm256_slli_epi16(one, bd), one); + __m256i clamped, mask; + + mask = _mm256_cmpgt_epi16(u, max); + clamped = _mm256_andnot_si256(mask, u); + mask = _mm256_and_si256(mask, max); + clamped = _mm256_or_si256(mask, clamped); + mask = _mm256_cmpgt_epi16(clamped, zero); + clamped = _mm256_and_si256(clamped, mask); + + return clamped; +} + +static INLINE __m256i highbd_get_recon_16x8_avx2(const __m256i pred, + __m256i res0, __m256i res1, + const int bd) { + __m256i x0 = _mm256_cvtepi16_epi32(_mm256_castsi256_si128(pred)); + __m256i x1 = _mm256_cvtepi16_epi32(_mm256_extractf128_si256(pred, 1)); + + x0 = _mm256_add_epi32(res0, x0); + x1 = _mm256_add_epi32(res1, x1); + x0 = _mm256_packus_epi32(x0, x1); + x0 = _mm256_permute4x64_epi64(x0, 0xd8); + x0 = highbd_clamp_epi16_avx2(x0, bd); + return x0; +} + +static INLINE void highbd_write_buffer_16xn_avx2(__m256i *in, uint16_t *output, + int stride, int flipud, + int height, const int bd) { + int j = flipud ? (height - 1) : 0; + const int step = flipud ? -1 : 1; + for (int i = 0; i < height; ++i, j += step) { + __m256i v = _mm256_loadu_si256((__m256i const *)(output + i * stride)); + __m256i u = highbd_get_recon_16x8_avx2(v, in[j], in[j + height], bd); + + _mm256_storeu_si256((__m256i *)(output + i * stride), u); + } +} + +static INLINE __m256i av1_round_shift_32_avx2(__m256i vec, int bit) { + __m256i tmp, round; + round = _mm256_set1_epi32(1 << (bit - 1)); + tmp = _mm256_add_epi32(vec, round); + return _mm256_srai_epi32(tmp, bit); +} + +static INLINE void av1_round_shift_array_32_avx2(__m256i *input, + __m256i *output, + const int size, + const int bit) { + if (bit > 0) { + int i; + for (i = 0; i < size; i++) { + output[i] = av1_round_shift_32_avx2(input[i], bit); + } + } else { + int i; + for (i = 0; i < size; i++) { + output[i] = _mm256_slli_epi32(input[i], -bit); + } + } +} + +static void transpose_8x8_avx2(const __m256i *in, __m256i *out) { + __m256i u0, u1, u2, u3, u4, u5, u6, u7; + __m256i x0, x1; + + u0 = _mm256_unpacklo_epi32(in[0], in[1]); + u1 = _mm256_unpackhi_epi32(in[0], in[1]); + + u2 = _mm256_unpacklo_epi32(in[2], in[3]); + u3 = _mm256_unpackhi_epi32(in[2], in[3]); + + u4 = _mm256_unpacklo_epi32(in[4], in[5]); + u5 = _mm256_unpackhi_epi32(in[4], in[5]); + + u6 = _mm256_unpacklo_epi32(in[6], in[7]); + u7 = _mm256_unpackhi_epi32(in[6], in[7]); + + x0 = _mm256_unpacklo_epi64(u0, u2); + x1 = _mm256_unpacklo_epi64(u4, u6); + out[0] = _mm256_permute2f128_si256(x0, x1, 0x20); + out[4] = _mm256_permute2f128_si256(x0, x1, 0x31); + + x0 = _mm256_unpackhi_epi64(u0, u2); + x1 = _mm256_unpackhi_epi64(u4, u6); + out[1] = _mm256_permute2f128_si256(x0, x1, 0x20); + out[5] = _mm256_permute2f128_si256(x0, x1, 0x31); + + x0 = _mm256_unpacklo_epi64(u1, u3); + x1 = _mm256_unpacklo_epi64(u5, u7); + out[2] = _mm256_permute2f128_si256(x0, x1, 0x20); + out[6] = _mm256_permute2f128_si256(x0, x1, 0x31); + + x0 = _mm256_unpackhi_epi64(u1, u3); + x1 = _mm256_unpackhi_epi64(u5, u7); + out[3] = _mm256_permute2f128_si256(x0, x1, 0x20); + out[7] = _mm256_permute2f128_si256(x0, x1, 0x31); +} + +static void load_buffer_32x32(const int32_t *coeff, __m256i *in, + int input_stiride, int size) { + int i; + for (i = 0; i < size; ++i) { + in[i] = _mm256_loadu_si256((const __m256i *)(coeff + i * input_stiride)); + } +} + +static INLINE __m256i half_btf_0_avx2(const __m256i *w0, const __m256i *n0, + const __m256i *rounding, int bit) { + __m256i x; + x = _mm256_mullo_epi32(*w0, *n0); + x = _mm256_add_epi32(x, *rounding); + x = _mm256_srai_epi32(x, bit); + return x; +} + +static INLINE __m256i half_btf_avx2(const __m256i *w0, const __m256i *n0, + const __m256i *w1, const __m256i *n1, + const __m256i *rounding, int bit) { + __m256i x, y; + + x = _mm256_mullo_epi32(*w0, *n0); + y = _mm256_mullo_epi32(*w1, *n1); + x = _mm256_add_epi32(x, y); + x = _mm256_add_epi32(x, *rounding); + x = _mm256_srai_epi32(x, bit); + return x; +} + +static void addsub_avx2(const __m256i in0, const __m256i in1, __m256i *out0, + __m256i *out1, const __m256i *clamp_lo, + const __m256i *clamp_hi) { + __m256i a0 = _mm256_add_epi32(in0, in1); + __m256i a1 = _mm256_sub_epi32(in0, in1); + + a0 = _mm256_max_epi32(a0, *clamp_lo); + a0 = _mm256_min_epi32(a0, *clamp_hi); + a1 = _mm256_max_epi32(a1, *clamp_lo); + a1 = _mm256_min_epi32(a1, *clamp_hi); + + *out0 = a0; + *out1 = a1; +} + +static void addsub_no_clamp_avx2(const __m256i in0, const __m256i in1, + __m256i *out0, __m256i *out1) { + __m256i a0 = _mm256_add_epi32(in0, in1); + __m256i a1 = _mm256_sub_epi32(in0, in1); + + *out0 = a0; + *out1 = a1; +} + +static void addsub_shift_avx2(const __m256i in0, const __m256i in1, + __m256i *out0, __m256i *out1, + const __m256i *clamp_lo, const __m256i *clamp_hi, + int shift) { + __m256i offset = _mm256_set1_epi32((1 << shift) >> 1); + __m256i in0_w_offset = _mm256_add_epi32(in0, offset); + __m256i a0 = _mm256_add_epi32(in0_w_offset, in1); + __m256i a1 = _mm256_sub_epi32(in0_w_offset, in1); + + a0 = _mm256_sra_epi32(a0, _mm_cvtsi32_si128(shift)); + a1 = _mm256_sra_epi32(a1, _mm_cvtsi32_si128(shift)); + + a0 = _mm256_max_epi32(a0, *clamp_lo); + a0 = _mm256_min_epi32(a0, *clamp_hi); + a1 = _mm256_max_epi32(a1, *clamp_lo); + a1 = _mm256_min_epi32(a1, *clamp_hi); + + *out0 = a0; + *out1 = a1; +} + +static INLINE void idct32_stage4_avx2( + __m256i *bf1, const __m256i *cospim8, const __m256i *cospi56, + const __m256i *cospi8, const __m256i *cospim56, const __m256i *cospim40, + const __m256i *cospi24, const __m256i *cospi40, const __m256i *cospim24, + const __m256i *rounding, int bit) { + __m256i temp1, temp2; + temp1 = half_btf_avx2(cospim8, &bf1[17], cospi56, &bf1[30], rounding, bit); + bf1[30] = half_btf_avx2(cospi56, &bf1[17], cospi8, &bf1[30], rounding, bit); + bf1[17] = temp1; + + temp2 = half_btf_avx2(cospim56, &bf1[18], cospim8, &bf1[29], rounding, bit); + bf1[29] = half_btf_avx2(cospim8, &bf1[18], cospi56, &bf1[29], rounding, bit); + bf1[18] = temp2; + + temp1 = half_btf_avx2(cospim40, &bf1[21], cospi24, &bf1[26], rounding, bit); + bf1[26] = half_btf_avx2(cospi24, &bf1[21], cospi40, &bf1[26], rounding, bit); + bf1[21] = temp1; + + temp2 = half_btf_avx2(cospim24, &bf1[22], cospim40, &bf1[25], rounding, bit); + bf1[25] = half_btf_avx2(cospim40, &bf1[22], cospi24, &bf1[25], rounding, bit); + bf1[22] = temp2; +} + +static INLINE void idct32_stage5_avx2( + __m256i *bf1, const __m256i *cospim16, const __m256i *cospi48, + const __m256i *cospi16, const __m256i *cospim48, const __m256i *clamp_lo, + const __m256i *clamp_hi, const __m256i *rounding, int bit) { + __m256i temp1, temp2; + temp1 = half_btf_avx2(cospim16, &bf1[9], cospi48, &bf1[14], rounding, bit); + bf1[14] = half_btf_avx2(cospi48, &bf1[9], cospi16, &bf1[14], rounding, bit); + bf1[9] = temp1; + + temp2 = half_btf_avx2(cospim48, &bf1[10], cospim16, &bf1[13], rounding, bit); + bf1[13] = half_btf_avx2(cospim16, &bf1[10], cospi48, &bf1[13], rounding, bit); + bf1[10] = temp2; + + addsub_avx2(bf1[16], bf1[19], bf1 + 16, bf1 + 19, clamp_lo, clamp_hi); + addsub_avx2(bf1[17], bf1[18], bf1 + 17, bf1 + 18, clamp_lo, clamp_hi); + addsub_avx2(bf1[23], bf1[20], bf1 + 23, bf1 + 20, clamp_lo, clamp_hi); + addsub_avx2(bf1[22], bf1[21], bf1 + 22, bf1 + 21, clamp_lo, clamp_hi); + addsub_avx2(bf1[24], bf1[27], bf1 + 24, bf1 + 27, clamp_lo, clamp_hi); + addsub_avx2(bf1[25], bf1[26], bf1 + 25, bf1 + 26, clamp_lo, clamp_hi); + addsub_avx2(bf1[31], bf1[28], bf1 + 31, bf1 + 28, clamp_lo, clamp_hi); + addsub_avx2(bf1[30], bf1[29], bf1 + 30, bf1 + 29, clamp_lo, clamp_hi); +} + +static INLINE void idct32_stage6_avx2( + __m256i *bf1, const __m256i *cospim32, const __m256i *cospi32, + const __m256i *cospim16, const __m256i *cospi48, const __m256i *cospi16, + const __m256i *cospim48, const __m256i *clamp_lo, const __m256i *clamp_hi, + const __m256i *rounding, int bit) { + __m256i temp1, temp2; + temp1 = half_btf_avx2(cospim32, &bf1[5], cospi32, &bf1[6], rounding, bit); + bf1[6] = half_btf_avx2(cospi32, &bf1[5], cospi32, &bf1[6], rounding, bit); + bf1[5] = temp1; + + addsub_avx2(bf1[8], bf1[11], bf1 + 8, bf1 + 11, clamp_lo, clamp_hi); + addsub_avx2(bf1[9], bf1[10], bf1 + 9, bf1 + 10, clamp_lo, clamp_hi); + addsub_avx2(bf1[15], bf1[12], bf1 + 15, bf1 + 12, clamp_lo, clamp_hi); + addsub_avx2(bf1[14], bf1[13], bf1 + 14, bf1 + 13, clamp_lo, clamp_hi); + + temp1 = half_btf_avx2(cospim16, &bf1[18], cospi48, &bf1[29], rounding, bit); + bf1[29] = half_btf_avx2(cospi48, &bf1[18], cospi16, &bf1[29], rounding, bit); + bf1[18] = temp1; + temp2 = half_btf_avx2(cospim16, &bf1[19], cospi48, &bf1[28], rounding, bit); + bf1[28] = half_btf_avx2(cospi48, &bf1[19], cospi16, &bf1[28], rounding, bit); + bf1[19] = temp2; + temp1 = half_btf_avx2(cospim48, &bf1[20], cospim16, &bf1[27], rounding, bit); + bf1[27] = half_btf_avx2(cospim16, &bf1[20], cospi48, &bf1[27], rounding, bit); + bf1[20] = temp1; + temp2 = half_btf_avx2(cospim48, &bf1[21], cospim16, &bf1[26], rounding, bit); + bf1[26] = half_btf_avx2(cospim16, &bf1[21], cospi48, &bf1[26], rounding, bit); + bf1[21] = temp2; +} + +static INLINE void idct32_stage7_avx2(__m256i *bf1, const __m256i *cospim32, + const __m256i *cospi32, + const __m256i *clamp_lo, + const __m256i *clamp_hi, + const __m256i *rounding, int bit) { + __m256i temp1, temp2; + addsub_avx2(bf1[0], bf1[7], bf1 + 0, bf1 + 7, clamp_lo, clamp_hi); + addsub_avx2(bf1[1], bf1[6], bf1 + 1, bf1 + 6, clamp_lo, clamp_hi); + addsub_avx2(bf1[2], bf1[5], bf1 + 2, bf1 + 5, clamp_lo, clamp_hi); + addsub_avx2(bf1[3], bf1[4], bf1 + 3, bf1 + 4, clamp_lo, clamp_hi); + + temp1 = half_btf_avx2(cospim32, &bf1[10], cospi32, &bf1[13], rounding, bit); + bf1[13] = half_btf_avx2(cospi32, &bf1[10], cospi32, &bf1[13], rounding, bit); + bf1[10] = temp1; + temp2 = half_btf_avx2(cospim32, &bf1[11], cospi32, &bf1[12], rounding, bit); + bf1[12] = half_btf_avx2(cospi32, &bf1[11], cospi32, &bf1[12], rounding, bit); + bf1[11] = temp2; + + addsub_avx2(bf1[16], bf1[23], bf1 + 16, bf1 + 23, clamp_lo, clamp_hi); + addsub_avx2(bf1[17], bf1[22], bf1 + 17, bf1 + 22, clamp_lo, clamp_hi); + addsub_avx2(bf1[18], bf1[21], bf1 + 18, bf1 + 21, clamp_lo, clamp_hi); + addsub_avx2(bf1[19], bf1[20], bf1 + 19, bf1 + 20, clamp_lo, clamp_hi); + addsub_avx2(bf1[31], bf1[24], bf1 + 31, bf1 + 24, clamp_lo, clamp_hi); + addsub_avx2(bf1[30], bf1[25], bf1 + 30, bf1 + 25, clamp_lo, clamp_hi); + addsub_avx2(bf1[29], bf1[26], bf1 + 29, bf1 + 26, clamp_lo, clamp_hi); + addsub_avx2(bf1[28], bf1[27], bf1 + 28, bf1 + 27, clamp_lo, clamp_hi); +} + +static INLINE void idct32_stage8_avx2(__m256i *bf1, const __m256i *cospim32, + const __m256i *cospi32, + const __m256i *clamp_lo, + const __m256i *clamp_hi, + const __m256i *rounding, int bit) { + __m256i temp1, temp2; + addsub_avx2(bf1[0], bf1[15], bf1 + 0, bf1 + 15, clamp_lo, clamp_hi); + addsub_avx2(bf1[1], bf1[14], bf1 + 1, bf1 + 14, clamp_lo, clamp_hi); + addsub_avx2(bf1[2], bf1[13], bf1 + 2, bf1 + 13, clamp_lo, clamp_hi); + addsub_avx2(bf1[3], bf1[12], bf1 + 3, bf1 + 12, clamp_lo, clamp_hi); + addsub_avx2(bf1[4], bf1[11], bf1 + 4, bf1 + 11, clamp_lo, clamp_hi); + addsub_avx2(bf1[5], bf1[10], bf1 + 5, bf1 + 10, clamp_lo, clamp_hi); + addsub_avx2(bf1[6], bf1[9], bf1 + 6, bf1 + 9, clamp_lo, clamp_hi); + addsub_avx2(bf1[7], bf1[8], bf1 + 7, bf1 + 8, clamp_lo, clamp_hi); + + temp1 = half_btf_avx2(cospim32, &bf1[20], cospi32, &bf1[27], rounding, bit); + bf1[27] = half_btf_avx2(cospi32, &bf1[20], cospi32, &bf1[27], rounding, bit); + bf1[20] = temp1; + temp2 = half_btf_avx2(cospim32, &bf1[21], cospi32, &bf1[26], rounding, bit); + bf1[26] = half_btf_avx2(cospi32, &bf1[21], cospi32, &bf1[26], rounding, bit); + bf1[21] = temp2; + temp1 = half_btf_avx2(cospim32, &bf1[22], cospi32, &bf1[25], rounding, bit); + bf1[25] = half_btf_avx2(cospi32, &bf1[22], cospi32, &bf1[25], rounding, bit); + bf1[22] = temp1; + temp2 = half_btf_avx2(cospim32, &bf1[23], cospi32, &bf1[24], rounding, bit); + bf1[24] = half_btf_avx2(cospi32, &bf1[23], cospi32, &bf1[24], rounding, bit); + bf1[23] = temp2; +} + +static INLINE void idct32_stage9_avx2(__m256i *bf1, __m256i *out, + const int do_cols, const int bd, + const int out_shift, + const int log_range) { + if (do_cols) { + addsub_no_clamp_avx2(bf1[0], bf1[31], out + 0, out + 31); + addsub_no_clamp_avx2(bf1[1], bf1[30], out + 1, out + 30); + addsub_no_clamp_avx2(bf1[2], bf1[29], out + 2, out + 29); + addsub_no_clamp_avx2(bf1[3], bf1[28], out + 3, out + 28); + addsub_no_clamp_avx2(bf1[4], bf1[27], out + 4, out + 27); + addsub_no_clamp_avx2(bf1[5], bf1[26], out + 5, out + 26); + addsub_no_clamp_avx2(bf1[6], bf1[25], out + 6, out + 25); + addsub_no_clamp_avx2(bf1[7], bf1[24], out + 7, out + 24); + addsub_no_clamp_avx2(bf1[8], bf1[23], out + 8, out + 23); + addsub_no_clamp_avx2(bf1[9], bf1[22], out + 9, out + 22); + addsub_no_clamp_avx2(bf1[10], bf1[21], out + 10, out + 21); + addsub_no_clamp_avx2(bf1[11], bf1[20], out + 11, out + 20); + addsub_no_clamp_avx2(bf1[12], bf1[19], out + 12, out + 19); + addsub_no_clamp_avx2(bf1[13], bf1[18], out + 13, out + 18); + addsub_no_clamp_avx2(bf1[14], bf1[17], out + 14, out + 17); + addsub_no_clamp_avx2(bf1[15], bf1[16], out + 15, out + 16); + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m256i clamp_lo_out = _mm256_set1_epi32(AOMMAX( + -(1 << (log_range_out - 1)), -(1 << (log_range - 1 - out_shift)))); + const __m256i clamp_hi_out = _mm256_set1_epi32(AOMMIN( + (1 << (log_range_out - 1)) - 1, (1 << (log_range - 1 - out_shift)))); + + addsub_shift_avx2(bf1[0], bf1[31], out + 0, out + 31, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf1[1], bf1[30], out + 1, out + 30, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf1[2], bf1[29], out + 2, out + 29, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf1[3], bf1[28], out + 3, out + 28, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf1[4], bf1[27], out + 4, out + 27, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf1[5], bf1[26], out + 5, out + 26, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf1[6], bf1[25], out + 6, out + 25, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf1[7], bf1[24], out + 7, out + 24, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf1[8], bf1[23], out + 8, out + 23, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf1[9], bf1[22], out + 9, out + 22, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf1[10], bf1[21], out + 10, out + 21, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf1[11], bf1[20], out + 11, out + 20, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf1[12], bf1[19], out + 12, out + 19, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf1[13], bf1[18], out + 13, out + 18, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf1[14], bf1[17], out + 14, out + 17, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf1[15], bf1[16], out + 15, out + 16, &clamp_lo_out, + &clamp_hi_out, out_shift); + } +} + +static void idct32_low1_avx2(__m256i *in, __m256i *out, int bit, int do_cols, + int bd, int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m256i cospi32 = _mm256_set1_epi32(cospi[32]); + const __m256i rounding = _mm256_set1_epi32(1 << (bit - 1)); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m256i clamp_lo = _mm256_set1_epi32(-(1 << (log_range - 1))); + const __m256i clamp_hi = _mm256_set1_epi32((1 << (log_range - 1)) - 1); + __m256i x; + // stage 0 + // stage 1 + // stage 2 + // stage 3 + // stage 4 + // stage 5 + x = _mm256_mullo_epi32(in[0], cospi32); + x = _mm256_add_epi32(x, rounding); + x = _mm256_srai_epi32(x, bit); + + // stage 6 + // stage 7 + // stage 8 + // stage 9 + if (do_cols) { + x = _mm256_max_epi32(x, clamp_lo); + x = _mm256_min_epi32(x, clamp_hi); + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m256i clamp_lo_out = _mm256_set1_epi32(AOMMAX( + -(1 << (log_range_out - 1)), -(1 << (log_range - 1 - out_shift)))); + const __m256i clamp_hi_out = _mm256_set1_epi32(AOMMIN( + (1 << (log_range_out - 1)) - 1, (1 << (log_range - 1 - out_shift)))); + __m256i offset = _mm256_set1_epi32((1 << out_shift) >> 1); + x = _mm256_add_epi32(offset, x); + x = _mm256_sra_epi32(x, _mm_cvtsi32_si128(out_shift)); + x = _mm256_max_epi32(x, clamp_lo_out); + x = _mm256_min_epi32(x, clamp_hi_out); + } + + out[0] = x; + out[1] = x; + out[2] = x; + out[3] = x; + out[4] = x; + out[5] = x; + out[6] = x; + out[7] = x; + out[8] = x; + out[9] = x; + out[10] = x; + out[11] = x; + out[12] = x; + out[13] = x; + out[14] = x; + out[15] = x; + out[16] = x; + out[17] = x; + out[18] = x; + out[19] = x; + out[20] = x; + out[21] = x; + out[22] = x; + out[23] = x; + out[24] = x; + out[25] = x; + out[26] = x; + out[27] = x; + out[28] = x; + out[29] = x; + out[30] = x; + out[31] = x; +} + +static void idct32_low8_avx2(__m256i *in, __m256i *out, int bit, int do_cols, + int bd, int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m256i cospi62 = _mm256_set1_epi32(cospi[62]); + const __m256i cospi14 = _mm256_set1_epi32(cospi[14]); + const __m256i cospi54 = _mm256_set1_epi32(cospi[54]); + const __m256i cospi6 = _mm256_set1_epi32(cospi[6]); + const __m256i cospi10 = _mm256_set1_epi32(cospi[10]); + const __m256i cospi2 = _mm256_set1_epi32(cospi[2]); + const __m256i cospim58 = _mm256_set1_epi32(-cospi[58]); + const __m256i cospim50 = _mm256_set1_epi32(-cospi[50]); + const __m256i cospi60 = _mm256_set1_epi32(cospi[60]); + const __m256i cospi12 = _mm256_set1_epi32(cospi[12]); + const __m256i cospi4 = _mm256_set1_epi32(cospi[4]); + const __m256i cospim52 = _mm256_set1_epi32(-cospi[52]); + const __m256i cospi56 = _mm256_set1_epi32(cospi[56]); + const __m256i cospi24 = _mm256_set1_epi32(cospi[24]); + const __m256i cospi40 = _mm256_set1_epi32(cospi[40]); + const __m256i cospi8 = _mm256_set1_epi32(cospi[8]); + const __m256i cospim40 = _mm256_set1_epi32(-cospi[40]); + const __m256i cospim8 = _mm256_set1_epi32(-cospi[8]); + const __m256i cospim56 = _mm256_set1_epi32(-cospi[56]); + const __m256i cospim24 = _mm256_set1_epi32(-cospi[24]); + const __m256i cospi32 = _mm256_set1_epi32(cospi[32]); + const __m256i cospim32 = _mm256_set1_epi32(-cospi[32]); + const __m256i cospi48 = _mm256_set1_epi32(cospi[48]); + const __m256i cospim48 = _mm256_set1_epi32(-cospi[48]); + const __m256i cospi16 = _mm256_set1_epi32(cospi[16]); + const __m256i cospim16 = _mm256_set1_epi32(-cospi[16]); + const __m256i rounding = _mm256_set1_epi32(1 << (bit - 1)); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m256i clamp_lo = _mm256_set1_epi32(-(1 << (log_range - 1))); + const __m256i clamp_hi = _mm256_set1_epi32((1 << (log_range - 1)) - 1); + __m256i bf1[32]; + + { + // stage 0 + // stage 1 + bf1[0] = in[0]; + bf1[4] = in[4]; + bf1[8] = in[2]; + bf1[12] = in[6]; + bf1[16] = in[1]; + bf1[20] = in[5]; + bf1[24] = in[3]; + bf1[28] = in[7]; + + // stage 2 + bf1[31] = half_btf_0_avx2(&cospi2, &bf1[16], &rounding, bit); + bf1[16] = half_btf_0_avx2(&cospi62, &bf1[16], &rounding, bit); + bf1[19] = half_btf_0_avx2(&cospim50, &bf1[28], &rounding, bit); + bf1[28] = half_btf_0_avx2(&cospi14, &bf1[28], &rounding, bit); + bf1[27] = half_btf_0_avx2(&cospi10, &bf1[20], &rounding, bit); + bf1[20] = half_btf_0_avx2(&cospi54, &bf1[20], &rounding, bit); + bf1[23] = half_btf_0_avx2(&cospim58, &bf1[24], &rounding, bit); + bf1[24] = half_btf_0_avx2(&cospi6, &bf1[24], &rounding, bit); + + // stage 3 + bf1[15] = half_btf_0_avx2(&cospi4, &bf1[8], &rounding, bit); + bf1[8] = half_btf_0_avx2(&cospi60, &bf1[8], &rounding, bit); + + bf1[11] = half_btf_0_avx2(&cospim52, &bf1[12], &rounding, bit); + bf1[12] = half_btf_0_avx2(&cospi12, &bf1[12], &rounding, bit); + bf1[17] = bf1[16]; + bf1[18] = bf1[19]; + bf1[21] = bf1[20]; + bf1[22] = bf1[23]; + bf1[25] = bf1[24]; + bf1[26] = bf1[27]; + bf1[29] = bf1[28]; + bf1[30] = bf1[31]; + + // stage 4 + bf1[7] = half_btf_0_avx2(&cospi8, &bf1[4], &rounding, bit); + bf1[4] = half_btf_0_avx2(&cospi56, &bf1[4], &rounding, bit); + + bf1[9] = bf1[8]; + bf1[10] = bf1[11]; + bf1[13] = bf1[12]; + bf1[14] = bf1[15]; + + idct32_stage4_avx2(bf1, &cospim8, &cospi56, &cospi8, &cospim56, &cospim40, + &cospi24, &cospi40, &cospim24, &rounding, bit); + + // stage 5 + bf1[0] = half_btf_0_avx2(&cospi32, &bf1[0], &rounding, bit); + bf1[1] = bf1[0]; + bf1[5] = bf1[4]; + bf1[6] = bf1[7]; + + idct32_stage5_avx2(bf1, &cospim16, &cospi48, &cospi16, &cospim48, &clamp_lo, + &clamp_hi, &rounding, bit); + + // stage 6 + bf1[3] = bf1[0]; + bf1[2] = bf1[1]; + + idct32_stage6_avx2(bf1, &cospim32, &cospi32, &cospim16, &cospi48, &cospi16, + &cospim48, &clamp_lo, &clamp_hi, &rounding, bit); + + // stage 7 + idct32_stage7_avx2(bf1, &cospim32, &cospi32, &clamp_lo, &clamp_hi, + &rounding, bit); + + // stage 8 + idct32_stage8_avx2(bf1, &cospim32, &cospi32, &clamp_lo, &clamp_hi, + &rounding, bit); + + // stage 9 + idct32_stage9_avx2(bf1, out, do_cols, bd, out_shift, log_range); + } +} + +static void idct32_low16_avx2(__m256i *in, __m256i *out, int bit, int do_cols, + int bd, int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m256i cospi62 = _mm256_set1_epi32(cospi[62]); + const __m256i cospi30 = _mm256_set1_epi32(cospi[30]); + const __m256i cospi46 = _mm256_set1_epi32(cospi[46]); + const __m256i cospi14 = _mm256_set1_epi32(cospi[14]); + const __m256i cospi54 = _mm256_set1_epi32(cospi[54]); + const __m256i cospi22 = _mm256_set1_epi32(cospi[22]); + const __m256i cospi38 = _mm256_set1_epi32(cospi[38]); + const __m256i cospi6 = _mm256_set1_epi32(cospi[6]); + const __m256i cospi26 = _mm256_set1_epi32(cospi[26]); + const __m256i cospi10 = _mm256_set1_epi32(cospi[10]); + const __m256i cospi18 = _mm256_set1_epi32(cospi[18]); + const __m256i cospi2 = _mm256_set1_epi32(cospi[2]); + const __m256i cospim58 = _mm256_set1_epi32(-cospi[58]); + const __m256i cospim42 = _mm256_set1_epi32(-cospi[42]); + const __m256i cospim50 = _mm256_set1_epi32(-cospi[50]); + const __m256i cospim34 = _mm256_set1_epi32(-cospi[34]); + const __m256i cospi60 = _mm256_set1_epi32(cospi[60]); + const __m256i cospi28 = _mm256_set1_epi32(cospi[28]); + const __m256i cospi44 = _mm256_set1_epi32(cospi[44]); + const __m256i cospi12 = _mm256_set1_epi32(cospi[12]); + const __m256i cospi20 = _mm256_set1_epi32(cospi[20]); + const __m256i cospi4 = _mm256_set1_epi32(cospi[4]); + const __m256i cospim52 = _mm256_set1_epi32(-cospi[52]); + const __m256i cospim36 = _mm256_set1_epi32(-cospi[36]); + const __m256i cospi56 = _mm256_set1_epi32(cospi[56]); + const __m256i cospi24 = _mm256_set1_epi32(cospi[24]); + const __m256i cospi40 = _mm256_set1_epi32(cospi[40]); + const __m256i cospi8 = _mm256_set1_epi32(cospi[8]); + const __m256i cospim40 = _mm256_set1_epi32(-cospi[40]); + const __m256i cospim8 = _mm256_set1_epi32(-cospi[8]); + const __m256i cospim56 = _mm256_set1_epi32(-cospi[56]); + const __m256i cospim24 = _mm256_set1_epi32(-cospi[24]); + const __m256i cospi32 = _mm256_set1_epi32(cospi[32]); + const __m256i cospim32 = _mm256_set1_epi32(-cospi[32]); + const __m256i cospi48 = _mm256_set1_epi32(cospi[48]); + const __m256i cospim48 = _mm256_set1_epi32(-cospi[48]); + const __m256i cospi16 = _mm256_set1_epi32(cospi[16]); + const __m256i cospim16 = _mm256_set1_epi32(-cospi[16]); + const __m256i rounding = _mm256_set1_epi32(1 << (bit - 1)); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m256i clamp_lo = _mm256_set1_epi32(-(1 << (log_range - 1))); + const __m256i clamp_hi = _mm256_set1_epi32((1 << (log_range - 1)) - 1); + __m256i bf1[32]; + + { + // stage 0 + // stage 1 + bf1[0] = in[0]; + bf1[2] = in[8]; + bf1[4] = in[4]; + bf1[6] = in[12]; + bf1[8] = in[2]; + bf1[10] = in[10]; + bf1[12] = in[6]; + bf1[14] = in[14]; + bf1[16] = in[1]; + bf1[18] = in[9]; + bf1[20] = in[5]; + bf1[22] = in[13]; + bf1[24] = in[3]; + bf1[26] = in[11]; + bf1[28] = in[7]; + bf1[30] = in[15]; + + // stage 2 + bf1[31] = half_btf_0_avx2(&cospi2, &bf1[16], &rounding, bit); + bf1[16] = half_btf_0_avx2(&cospi62, &bf1[16], &rounding, bit); + bf1[17] = half_btf_0_avx2(&cospim34, &bf1[30], &rounding, bit); + bf1[30] = half_btf_0_avx2(&cospi30, &bf1[30], &rounding, bit); + bf1[29] = half_btf_0_avx2(&cospi18, &bf1[18], &rounding, bit); + bf1[18] = half_btf_0_avx2(&cospi46, &bf1[18], &rounding, bit); + bf1[19] = half_btf_0_avx2(&cospim50, &bf1[28], &rounding, bit); + bf1[28] = half_btf_0_avx2(&cospi14, &bf1[28], &rounding, bit); + bf1[27] = half_btf_0_avx2(&cospi10, &bf1[20], &rounding, bit); + bf1[20] = half_btf_0_avx2(&cospi54, &bf1[20], &rounding, bit); + bf1[21] = half_btf_0_avx2(&cospim42, &bf1[26], &rounding, bit); + bf1[26] = half_btf_0_avx2(&cospi22, &bf1[26], &rounding, bit); + bf1[25] = half_btf_0_avx2(&cospi26, &bf1[22], &rounding, bit); + bf1[22] = half_btf_0_avx2(&cospi38, &bf1[22], &rounding, bit); + bf1[23] = half_btf_0_avx2(&cospim58, &bf1[24], &rounding, bit); + bf1[24] = half_btf_0_avx2(&cospi6, &bf1[24], &rounding, bit); + + // stage 3 + bf1[15] = half_btf_0_avx2(&cospi4, &bf1[8], &rounding, bit); + bf1[8] = half_btf_0_avx2(&cospi60, &bf1[8], &rounding, bit); + bf1[9] = half_btf_0_avx2(&cospim36, &bf1[14], &rounding, bit); + bf1[14] = half_btf_0_avx2(&cospi28, &bf1[14], &rounding, bit); + bf1[13] = half_btf_0_avx2(&cospi20, &bf1[10], &rounding, bit); + bf1[10] = half_btf_0_avx2(&cospi44, &bf1[10], &rounding, bit); + bf1[11] = half_btf_0_avx2(&cospim52, &bf1[12], &rounding, bit); + bf1[12] = half_btf_0_avx2(&cospi12, &bf1[12], &rounding, bit); + + addsub_avx2(bf1[16], bf1[17], bf1 + 16, bf1 + 17, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[19], bf1[18], bf1 + 19, bf1 + 18, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[20], bf1[21], bf1 + 20, bf1 + 21, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[23], bf1[22], bf1 + 23, bf1 + 22, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[24], bf1[25], bf1 + 24, bf1 + 25, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[27], bf1[26], bf1 + 27, bf1 + 26, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[28], bf1[29], bf1 + 28, bf1 + 29, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[31], bf1[30], bf1 + 31, bf1 + 30, &clamp_lo, &clamp_hi); + + // stage 4 + bf1[7] = half_btf_0_avx2(&cospi8, &bf1[4], &rounding, bit); + bf1[4] = half_btf_0_avx2(&cospi56, &bf1[4], &rounding, bit); + bf1[5] = half_btf_0_avx2(&cospim40, &bf1[6], &rounding, bit); + bf1[6] = half_btf_0_avx2(&cospi24, &bf1[6], &rounding, bit); + + addsub_avx2(bf1[8], bf1[9], bf1 + 8, bf1 + 9, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[11], bf1[10], bf1 + 11, bf1 + 10, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[12], bf1[13], bf1 + 12, bf1 + 13, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[15], bf1[14], bf1 + 15, bf1 + 14, &clamp_lo, &clamp_hi); + + idct32_stage4_avx2(bf1, &cospim8, &cospi56, &cospi8, &cospim56, &cospim40, + &cospi24, &cospi40, &cospim24, &rounding, bit); + + // stage 5 + bf1[0] = half_btf_0_avx2(&cospi32, &bf1[0], &rounding, bit); + bf1[1] = bf1[0]; + bf1[3] = half_btf_0_avx2(&cospi16, &bf1[2], &rounding, bit); + bf1[2] = half_btf_0_avx2(&cospi48, &bf1[2], &rounding, bit); + + addsub_avx2(bf1[4], bf1[5], bf1 + 4, bf1 + 5, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[7], bf1[6], bf1 + 7, bf1 + 6, &clamp_lo, &clamp_hi); + + idct32_stage5_avx2(bf1, &cospim16, &cospi48, &cospi16, &cospim48, &clamp_lo, + &clamp_hi, &rounding, bit); + + // stage 6 + addsub_avx2(bf1[0], bf1[3], bf1 + 0, bf1 + 3, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[1], bf1[2], bf1 + 1, bf1 + 2, &clamp_lo, &clamp_hi); + + idct32_stage6_avx2(bf1, &cospim32, &cospi32, &cospim16, &cospi48, &cospi16, + &cospim48, &clamp_lo, &clamp_hi, &rounding, bit); + + // stage 7 + idct32_stage7_avx2(bf1, &cospim32, &cospi32, &clamp_lo, &clamp_hi, + &rounding, bit); + + // stage 8 + idct32_stage8_avx2(bf1, &cospim32, &cospi32, &clamp_lo, &clamp_hi, + &rounding, bit); + + // stage 9 + idct32_stage9_avx2(bf1, out, do_cols, bd, out_shift, log_range); + } +} + +static void idct32_avx2(__m256i *in, __m256i *out, int bit, int do_cols, int bd, + int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m256i cospi62 = _mm256_set1_epi32(cospi[62]); + const __m256i cospi30 = _mm256_set1_epi32(cospi[30]); + const __m256i cospi46 = _mm256_set1_epi32(cospi[46]); + const __m256i cospi14 = _mm256_set1_epi32(cospi[14]); + const __m256i cospi54 = _mm256_set1_epi32(cospi[54]); + const __m256i cospi22 = _mm256_set1_epi32(cospi[22]); + const __m256i cospi38 = _mm256_set1_epi32(cospi[38]); + const __m256i cospi6 = _mm256_set1_epi32(cospi[6]); + const __m256i cospi58 = _mm256_set1_epi32(cospi[58]); + const __m256i cospi26 = _mm256_set1_epi32(cospi[26]); + const __m256i cospi42 = _mm256_set1_epi32(cospi[42]); + const __m256i cospi10 = _mm256_set1_epi32(cospi[10]); + const __m256i cospi50 = _mm256_set1_epi32(cospi[50]); + const __m256i cospi18 = _mm256_set1_epi32(cospi[18]); + const __m256i cospi34 = _mm256_set1_epi32(cospi[34]); + const __m256i cospi2 = _mm256_set1_epi32(cospi[2]); + const __m256i cospim58 = _mm256_set1_epi32(-cospi[58]); + const __m256i cospim26 = _mm256_set1_epi32(-cospi[26]); + const __m256i cospim42 = _mm256_set1_epi32(-cospi[42]); + const __m256i cospim10 = _mm256_set1_epi32(-cospi[10]); + const __m256i cospim50 = _mm256_set1_epi32(-cospi[50]); + const __m256i cospim18 = _mm256_set1_epi32(-cospi[18]); + const __m256i cospim34 = _mm256_set1_epi32(-cospi[34]); + const __m256i cospim2 = _mm256_set1_epi32(-cospi[2]); + const __m256i cospi60 = _mm256_set1_epi32(cospi[60]); + const __m256i cospi28 = _mm256_set1_epi32(cospi[28]); + const __m256i cospi44 = _mm256_set1_epi32(cospi[44]); + const __m256i cospi12 = _mm256_set1_epi32(cospi[12]); + const __m256i cospi52 = _mm256_set1_epi32(cospi[52]); + const __m256i cospi20 = _mm256_set1_epi32(cospi[20]); + const __m256i cospi36 = _mm256_set1_epi32(cospi[36]); + const __m256i cospi4 = _mm256_set1_epi32(cospi[4]); + const __m256i cospim52 = _mm256_set1_epi32(-cospi[52]); + const __m256i cospim20 = _mm256_set1_epi32(-cospi[20]); + const __m256i cospim36 = _mm256_set1_epi32(-cospi[36]); + const __m256i cospim4 = _mm256_set1_epi32(-cospi[4]); + const __m256i cospi56 = _mm256_set1_epi32(cospi[56]); + const __m256i cospi24 = _mm256_set1_epi32(cospi[24]); + const __m256i cospi40 = _mm256_set1_epi32(cospi[40]); + const __m256i cospi8 = _mm256_set1_epi32(cospi[8]); + const __m256i cospim40 = _mm256_set1_epi32(-cospi[40]); + const __m256i cospim8 = _mm256_set1_epi32(-cospi[8]); + const __m256i cospim56 = _mm256_set1_epi32(-cospi[56]); + const __m256i cospim24 = _mm256_set1_epi32(-cospi[24]); + const __m256i cospi32 = _mm256_set1_epi32(cospi[32]); + const __m256i cospim32 = _mm256_set1_epi32(-cospi[32]); + const __m256i cospi48 = _mm256_set1_epi32(cospi[48]); + const __m256i cospim48 = _mm256_set1_epi32(-cospi[48]); + const __m256i cospi16 = _mm256_set1_epi32(cospi[16]); + const __m256i cospim16 = _mm256_set1_epi32(-cospi[16]); + const __m256i rounding = _mm256_set1_epi32(1 << (bit - 1)); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m256i clamp_lo = _mm256_set1_epi32(-(1 << (log_range - 1))); + const __m256i clamp_hi = _mm256_set1_epi32((1 << (log_range - 1)) - 1); + __m256i bf1[32], bf0[32]; + + { + // stage 0 + // stage 1 + bf1[0] = in[0]; + bf1[1] = in[16]; + bf1[2] = in[8]; + bf1[3] = in[24]; + bf1[4] = in[4]; + bf1[5] = in[20]; + bf1[6] = in[12]; + bf1[7] = in[28]; + bf1[8] = in[2]; + bf1[9] = in[18]; + bf1[10] = in[10]; + bf1[11] = in[26]; + bf1[12] = in[6]; + bf1[13] = in[22]; + bf1[14] = in[14]; + bf1[15] = in[30]; + bf1[16] = in[1]; + bf1[17] = in[17]; + bf1[18] = in[9]; + bf1[19] = in[25]; + bf1[20] = in[5]; + bf1[21] = in[21]; + bf1[22] = in[13]; + bf1[23] = in[29]; + bf1[24] = in[3]; + bf1[25] = in[19]; + bf1[26] = in[11]; + bf1[27] = in[27]; + bf1[28] = in[7]; + bf1[29] = in[23]; + bf1[30] = in[15]; + bf1[31] = in[31]; + + // stage 2 + bf0[0] = bf1[0]; + bf0[1] = bf1[1]; + bf0[2] = bf1[2]; + bf0[3] = bf1[3]; + bf0[4] = bf1[4]; + bf0[5] = bf1[5]; + bf0[6] = bf1[6]; + bf0[7] = bf1[7]; + bf0[8] = bf1[8]; + bf0[9] = bf1[9]; + bf0[10] = bf1[10]; + bf0[11] = bf1[11]; + bf0[12] = bf1[12]; + bf0[13] = bf1[13]; + bf0[14] = bf1[14]; + bf0[15] = bf1[15]; + bf0[16] = + half_btf_avx2(&cospi62, &bf1[16], &cospim2, &bf1[31], &rounding, bit); + bf0[17] = + half_btf_avx2(&cospi30, &bf1[17], &cospim34, &bf1[30], &rounding, bit); + bf0[18] = + half_btf_avx2(&cospi46, &bf1[18], &cospim18, &bf1[29], &rounding, bit); + bf0[19] = + half_btf_avx2(&cospi14, &bf1[19], &cospim50, &bf1[28], &rounding, bit); + bf0[20] = + half_btf_avx2(&cospi54, &bf1[20], &cospim10, &bf1[27], &rounding, bit); + bf0[21] = + half_btf_avx2(&cospi22, &bf1[21], &cospim42, &bf1[26], &rounding, bit); + bf0[22] = + half_btf_avx2(&cospi38, &bf1[22], &cospim26, &bf1[25], &rounding, bit); + bf0[23] = + half_btf_avx2(&cospi6, &bf1[23], &cospim58, &bf1[24], &rounding, bit); + bf0[24] = + half_btf_avx2(&cospi58, &bf1[23], &cospi6, &bf1[24], &rounding, bit); + bf0[25] = + half_btf_avx2(&cospi26, &bf1[22], &cospi38, &bf1[25], &rounding, bit); + bf0[26] = + half_btf_avx2(&cospi42, &bf1[21], &cospi22, &bf1[26], &rounding, bit); + bf0[27] = + half_btf_avx2(&cospi10, &bf1[20], &cospi54, &bf1[27], &rounding, bit); + bf0[28] = + half_btf_avx2(&cospi50, &bf1[19], &cospi14, &bf1[28], &rounding, bit); + bf0[29] = + half_btf_avx2(&cospi18, &bf1[18], &cospi46, &bf1[29], &rounding, bit); + bf0[30] = + half_btf_avx2(&cospi34, &bf1[17], &cospi30, &bf1[30], &rounding, bit); + bf0[31] = + half_btf_avx2(&cospi2, &bf1[16], &cospi62, &bf1[31], &rounding, bit); + + // stage 3 + bf1[0] = bf0[0]; + bf1[1] = bf0[1]; + bf1[2] = bf0[2]; + bf1[3] = bf0[3]; + bf1[4] = bf0[4]; + bf1[5] = bf0[5]; + bf1[6] = bf0[6]; + bf1[7] = bf0[7]; + bf1[8] = + half_btf_avx2(&cospi60, &bf0[8], &cospim4, &bf0[15], &rounding, bit); + bf1[9] = + half_btf_avx2(&cospi28, &bf0[9], &cospim36, &bf0[14], &rounding, bit); + bf1[10] = + half_btf_avx2(&cospi44, &bf0[10], &cospim20, &bf0[13], &rounding, bit); + bf1[11] = + half_btf_avx2(&cospi12, &bf0[11], &cospim52, &bf0[12], &rounding, bit); + bf1[12] = + half_btf_avx2(&cospi52, &bf0[11], &cospi12, &bf0[12], &rounding, bit); + bf1[13] = + half_btf_avx2(&cospi20, &bf0[10], &cospi44, &bf0[13], &rounding, bit); + bf1[14] = + half_btf_avx2(&cospi36, &bf0[9], &cospi28, &bf0[14], &rounding, bit); + bf1[15] = + half_btf_avx2(&cospi4, &bf0[8], &cospi60, &bf0[15], &rounding, bit); + + addsub_avx2(bf0[16], bf0[17], bf1 + 16, bf1 + 17, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[19], bf0[18], bf1 + 19, bf1 + 18, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[20], bf0[21], bf1 + 20, bf1 + 21, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[23], bf0[22], bf1 + 23, bf1 + 22, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[24], bf0[25], bf1 + 24, bf1 + 25, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[27], bf0[26], bf1 + 27, bf1 + 26, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[28], bf0[29], bf1 + 28, bf1 + 29, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[31], bf0[30], bf1 + 31, bf1 + 30, &clamp_lo, &clamp_hi); + + // stage 4 + bf0[0] = bf1[0]; + bf0[1] = bf1[1]; + bf0[2] = bf1[2]; + bf0[3] = bf1[3]; + bf0[4] = + half_btf_avx2(&cospi56, &bf1[4], &cospim8, &bf1[7], &rounding, bit); + bf0[5] = + half_btf_avx2(&cospi24, &bf1[5], &cospim40, &bf1[6], &rounding, bit); + bf0[6] = + half_btf_avx2(&cospi40, &bf1[5], &cospi24, &bf1[6], &rounding, bit); + bf0[7] = half_btf_avx2(&cospi8, &bf1[4], &cospi56, &bf1[7], &rounding, bit); + + addsub_avx2(bf1[8], bf1[9], bf0 + 8, bf0 + 9, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[11], bf1[10], bf0 + 11, bf0 + 10, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[12], bf1[13], bf0 + 12, bf0 + 13, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[15], bf1[14], bf0 + 15, bf0 + 14, &clamp_lo, &clamp_hi); + + bf0[16] = bf1[16]; + bf0[17] = + half_btf_avx2(&cospim8, &bf1[17], &cospi56, &bf1[30], &rounding, bit); + bf0[18] = + half_btf_avx2(&cospim56, &bf1[18], &cospim8, &bf1[29], &rounding, bit); + bf0[19] = bf1[19]; + bf0[20] = bf1[20]; + bf0[21] = + half_btf_avx2(&cospim40, &bf1[21], &cospi24, &bf1[26], &rounding, bit); + bf0[22] = + half_btf_avx2(&cospim24, &bf1[22], &cospim40, &bf1[25], &rounding, bit); + bf0[23] = bf1[23]; + bf0[24] = bf1[24]; + bf0[25] = + half_btf_avx2(&cospim40, &bf1[22], &cospi24, &bf1[25], &rounding, bit); + bf0[26] = + half_btf_avx2(&cospi24, &bf1[21], &cospi40, &bf1[26], &rounding, bit); + bf0[27] = bf1[27]; + bf0[28] = bf1[28]; + bf0[29] = + half_btf_avx2(&cospim8, &bf1[18], &cospi56, &bf1[29], &rounding, bit); + bf0[30] = + half_btf_avx2(&cospi56, &bf1[17], &cospi8, &bf1[30], &rounding, bit); + bf0[31] = bf1[31]; + + // stage 5 + bf1[0] = + half_btf_avx2(&cospi32, &bf0[0], &cospi32, &bf0[1], &rounding, bit); + bf1[1] = + half_btf_avx2(&cospi32, &bf0[0], &cospim32, &bf0[1], &rounding, bit); + bf1[2] = + half_btf_avx2(&cospi48, &bf0[2], &cospim16, &bf0[3], &rounding, bit); + bf1[3] = + half_btf_avx2(&cospi16, &bf0[2], &cospi48, &bf0[3], &rounding, bit); + addsub_avx2(bf0[4], bf0[5], bf1 + 4, bf1 + 5, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[7], bf0[6], bf1 + 7, bf1 + 6, &clamp_lo, &clamp_hi); + bf1[8] = bf0[8]; + bf1[9] = + half_btf_avx2(&cospim16, &bf0[9], &cospi48, &bf0[14], &rounding, bit); + bf1[10] = + half_btf_avx2(&cospim48, &bf0[10], &cospim16, &bf0[13], &rounding, bit); + bf1[11] = bf0[11]; + bf1[12] = bf0[12]; + bf1[13] = + half_btf_avx2(&cospim16, &bf0[10], &cospi48, &bf0[13], &rounding, bit); + bf1[14] = + half_btf_avx2(&cospi48, &bf0[9], &cospi16, &bf0[14], &rounding, bit); + bf1[15] = bf0[15]; + addsub_avx2(bf0[16], bf0[19], bf1 + 16, bf1 + 19, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[17], bf0[18], bf1 + 17, bf1 + 18, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[23], bf0[20], bf1 + 23, bf1 + 20, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[22], bf0[21], bf1 + 22, bf1 + 21, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[24], bf0[27], bf1 + 24, bf1 + 27, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[25], bf0[26], bf1 + 25, bf1 + 26, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[31], bf0[28], bf1 + 31, bf1 + 28, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[30], bf0[29], bf1 + 30, bf1 + 29, &clamp_lo, &clamp_hi); + + // stage 6 + addsub_avx2(bf1[0], bf1[3], bf0 + 0, bf0 + 3, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[1], bf1[2], bf0 + 1, bf0 + 2, &clamp_lo, &clamp_hi); + bf0[4] = bf1[4]; + bf0[5] = + half_btf_avx2(&cospim32, &bf1[5], &cospi32, &bf1[6], &rounding, bit); + bf0[6] = + half_btf_avx2(&cospi32, &bf1[5], &cospi32, &bf1[6], &rounding, bit); + bf0[7] = bf1[7]; + addsub_avx2(bf1[8], bf1[11], bf0 + 8, bf0 + 11, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[9], bf1[10], bf0 + 9, bf0 + 10, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[15], bf1[12], bf0 + 15, bf0 + 12, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[14], bf1[13], bf0 + 14, bf0 + 13, &clamp_lo, &clamp_hi); + bf0[16] = bf1[16]; + bf0[17] = bf1[17]; + bf0[18] = + half_btf_avx2(&cospim16, &bf1[18], &cospi48, &bf1[29], &rounding, bit); + bf0[19] = + half_btf_avx2(&cospim16, &bf1[19], &cospi48, &bf1[28], &rounding, bit); + bf0[20] = + half_btf_avx2(&cospim48, &bf1[20], &cospim16, &bf1[27], &rounding, bit); + bf0[21] = + half_btf_avx2(&cospim48, &bf1[21], &cospim16, &bf1[26], &rounding, bit); + bf0[22] = bf1[22]; + bf0[23] = bf1[23]; + bf0[24] = bf1[24]; + bf0[25] = bf1[25]; + bf0[26] = + half_btf_avx2(&cospim16, &bf1[21], &cospi48, &bf1[26], &rounding, bit); + bf0[27] = + half_btf_avx2(&cospim16, &bf1[20], &cospi48, &bf1[27], &rounding, bit); + bf0[28] = + half_btf_avx2(&cospi48, &bf1[19], &cospi16, &bf1[28], &rounding, bit); + bf0[29] = + half_btf_avx2(&cospi48, &bf1[18], &cospi16, &bf1[29], &rounding, bit); + bf0[30] = bf1[30]; + bf0[31] = bf1[31]; + + // stage 7 + addsub_avx2(bf0[0], bf0[7], bf1 + 0, bf1 + 7, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[1], bf0[6], bf1 + 1, bf1 + 6, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[2], bf0[5], bf1 + 2, bf1 + 5, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[3], bf0[4], bf1 + 3, bf1 + 4, &clamp_lo, &clamp_hi); + bf1[8] = bf0[8]; + bf1[9] = bf0[9]; + bf1[10] = + half_btf_avx2(&cospim32, &bf0[10], &cospi32, &bf0[13], &rounding, bit); + bf1[11] = + half_btf_avx2(&cospim32, &bf0[11], &cospi32, &bf0[12], &rounding, bit); + bf1[12] = + half_btf_avx2(&cospi32, &bf0[11], &cospi32, &bf0[12], &rounding, bit); + bf1[13] = + half_btf_avx2(&cospi32, &bf0[10], &cospi32, &bf0[13], &rounding, bit); + bf1[14] = bf0[14]; + bf1[15] = bf0[15]; + addsub_avx2(bf0[16], bf0[23], bf1 + 16, bf1 + 23, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[17], bf0[22], bf1 + 17, bf1 + 22, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[18], bf0[21], bf1 + 18, bf1 + 21, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[19], bf0[20], bf1 + 19, bf1 + 20, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[31], bf0[24], bf1 + 31, bf1 + 24, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[30], bf0[25], bf1 + 30, bf1 + 25, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[29], bf0[26], bf1 + 29, bf1 + 26, &clamp_lo, &clamp_hi); + addsub_avx2(bf0[28], bf0[27], bf1 + 28, bf1 + 27, &clamp_lo, &clamp_hi); + + // stage 8 + addsub_avx2(bf1[0], bf1[15], bf0 + 0, bf0 + 15, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[1], bf1[14], bf0 + 1, bf0 + 14, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[2], bf1[13], bf0 + 2, bf0 + 13, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[3], bf1[12], bf0 + 3, bf0 + 12, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[4], bf1[11], bf0 + 4, bf0 + 11, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[5], bf1[10], bf0 + 5, bf0 + 10, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[6], bf1[9], bf0 + 6, bf0 + 9, &clamp_lo, &clamp_hi); + addsub_avx2(bf1[7], bf1[8], bf0 + 7, bf0 + 8, &clamp_lo, &clamp_hi); + bf0[16] = bf1[16]; + bf0[17] = bf1[17]; + bf0[18] = bf1[18]; + bf0[19] = bf1[19]; + bf0[20] = + half_btf_avx2(&cospim32, &bf1[20], &cospi32, &bf1[27], &rounding, bit); + bf0[21] = + half_btf_avx2(&cospim32, &bf1[21], &cospi32, &bf1[26], &rounding, bit); + bf0[22] = + half_btf_avx2(&cospim32, &bf1[22], &cospi32, &bf1[25], &rounding, bit); + bf0[23] = + half_btf_avx2(&cospim32, &bf1[23], &cospi32, &bf1[24], &rounding, bit); + bf0[24] = + half_btf_avx2(&cospi32, &bf1[23], &cospi32, &bf1[24], &rounding, bit); + bf0[25] = + half_btf_avx2(&cospi32, &bf1[22], &cospi32, &bf1[25], &rounding, bit); + bf0[26] = + half_btf_avx2(&cospi32, &bf1[21], &cospi32, &bf1[26], &rounding, bit); + bf0[27] = + half_btf_avx2(&cospi32, &bf1[20], &cospi32, &bf1[27], &rounding, bit); + bf0[28] = bf1[28]; + bf0[29] = bf1[29]; + bf0[30] = bf1[30]; + bf0[31] = bf1[31]; + + // stage 9 + if (do_cols) { + addsub_no_clamp_avx2(bf0[0], bf0[31], out + 0, out + 31); + addsub_no_clamp_avx2(bf0[1], bf0[30], out + 1, out + 30); + addsub_no_clamp_avx2(bf0[2], bf0[29], out + 2, out + 29); + addsub_no_clamp_avx2(bf0[3], bf0[28], out + 3, out + 28); + addsub_no_clamp_avx2(bf0[4], bf0[27], out + 4, out + 27); + addsub_no_clamp_avx2(bf0[5], bf0[26], out + 5, out + 26); + addsub_no_clamp_avx2(bf0[6], bf0[25], out + 6, out + 25); + addsub_no_clamp_avx2(bf0[7], bf0[24], out + 7, out + 24); + addsub_no_clamp_avx2(bf0[8], bf0[23], out + 8, out + 23); + addsub_no_clamp_avx2(bf0[9], bf0[22], out + 9, out + 22); + addsub_no_clamp_avx2(bf0[10], bf0[21], out + 10, out + 21); + addsub_no_clamp_avx2(bf0[11], bf0[20], out + 11, out + 20); + addsub_no_clamp_avx2(bf0[12], bf0[19], out + 12, out + 19); + addsub_no_clamp_avx2(bf0[13], bf0[18], out + 13, out + 18); + addsub_no_clamp_avx2(bf0[14], bf0[17], out + 14, out + 17); + addsub_no_clamp_avx2(bf0[15], bf0[16], out + 15, out + 16); + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m256i clamp_lo_out = _mm256_set1_epi32(AOMMAX( + -(1 << (log_range_out - 1)), -(1 << (log_range - 1 - out_shift)))); + const __m256i clamp_hi_out = _mm256_set1_epi32(AOMMIN( + (1 << (log_range_out - 1)) - 1, (1 << (log_range - 1 - out_shift)))); + + addsub_shift_avx2(bf0[0], bf0[31], out + 0, out + 31, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf0[1], bf0[30], out + 1, out + 30, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf0[2], bf0[29], out + 2, out + 29, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf0[3], bf0[28], out + 3, out + 28, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf0[4], bf0[27], out + 4, out + 27, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf0[5], bf0[26], out + 5, out + 26, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf0[6], bf0[25], out + 6, out + 25, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf0[7], bf0[24], out + 7, out + 24, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf0[8], bf0[23], out + 8, out + 23, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf0[9], bf0[22], out + 9, out + 22, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf0[10], bf0[21], out + 10, out + 21, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf0[11], bf0[20], out + 11, out + 20, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf0[12], bf0[19], out + 12, out + 19, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf0[13], bf0[18], out + 13, out + 18, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf0[14], bf0[17], out + 14, out + 17, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_avx2(bf0[15], bf0[16], out + 15, out + 16, &clamp_lo_out, + &clamp_hi_out, out_shift); + } + } +} + +typedef void (*transform_1d_avx2)(__m256i *in, __m256i *out, int bit, + int do_cols, int bd, int out_shift); + +static const transform_1d_avx2 + highbd_txfm_all_1d_zeros_w8_arr[TX_SIZES][ITX_TYPES_1D][4] = { + { + { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL }, + }, + { { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL } }, + { + { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL }, + }, + { { idct32_low1_avx2, idct32_low8_avx2, idct32_low16_avx2, idct32_avx2 }, + { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL } }, + + { { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL } } + }; + +static void highbd_inv_txfm2d_add_no_identity_avx2(const int32_t *input, + uint16_t *output, int stride, + TX_TYPE tx_type, + TX_SIZE tx_size, int eob, + const int bd) { + __m256i buf1[64 * 2]; + int eobx, eoby; + get_eobx_eoby_scan_default(&eobx, &eoby, tx_size, eob); + const int8_t *shift = inv_txfm_shift_ls[tx_size]; + const int txw_idx = get_txw_idx(tx_size); + const int txh_idx = get_txh_idx(tx_size); + const int txfm_size_col = tx_size_wide[tx_size]; + const int txfm_size_row = tx_size_high[tx_size]; + const int buf_size_w_div8 = txfm_size_col >> 3; + const int buf_size_nonzero_w_div8 = (eobx + 8) >> 3; + const int buf_size_nonzero_h_div8 = (eoby + 8) >> 3; + const int input_stride = AOMMIN(32, txfm_size_col); + + const int fun_idx_x = lowbd_txfm_all_1d_zeros_idx[eobx]; + const int fun_idx_y = lowbd_txfm_all_1d_zeros_idx[eoby]; + const transform_1d_avx2 row_txfm = + highbd_txfm_all_1d_zeros_w8_arr[txw_idx][hitx_1d_tab[tx_type]][fun_idx_x]; + const transform_1d_avx2 col_txfm = + highbd_txfm_all_1d_zeros_w8_arr[txh_idx][vitx_1d_tab[tx_type]][fun_idx_y]; + + assert(col_txfm != NULL); + assert(row_txfm != NULL); + int ud_flip, lr_flip; + get_flip_cfg(tx_type, &ud_flip, &lr_flip); + + // 1st stage: column transform + for (int i = 0; i < buf_size_nonzero_h_div8; i++) { + __m256i buf0[32]; + const int32_t *input_row = input + i * input_stride * 8; + for (int j = 0; j < buf_size_nonzero_w_div8; ++j) { + __m256i *buf0_cur = buf0 + j * 8; + load_buffer_32x32(input_row + j * 8, buf0_cur, input_stride, 8); + + transpose_8x8_avx2(&buf0_cur[0], &buf0_cur[0]); + } + + row_txfm(buf0, buf0, inv_cos_bit_row[txw_idx][txh_idx], 0, bd, -shift[0]); + + __m256i *_buf1 = buf1 + i * 8; + for (int j = 0; j < buf_size_w_div8; ++j) { + transpose_8x8_avx2(&buf0[j * 8], &_buf1[j * txfm_size_row]); + } + } + // 2nd stage: column transform + for (int i = 0; i < buf_size_w_div8; i++) { + col_txfm(buf1 + i * txfm_size_row, buf1 + i * txfm_size_row, + inv_cos_bit_col[txw_idx][txh_idx], 1, bd, 0); + + av1_round_shift_array_32_avx2(buf1 + i * txfm_size_row, + buf1 + i * txfm_size_row, txfm_size_row, + -shift[1]); + } + + // write to buffer + { + for (int i = 0; i < (txfm_size_col >> 4); i++) { + highbd_write_buffer_16xn_avx2(buf1 + i * txfm_size_row * 2, + output + 16 * i, stride, ud_flip, + txfm_size_row, bd); + } + } +} + +void av1_highbd_inv_txfm2d_add_universe_avx2(const int32_t *input, + uint8_t *output, int stride, + TX_TYPE tx_type, TX_SIZE tx_size, + int eob, const int bd) { + switch (tx_type) { + case DCT_DCT: + highbd_inv_txfm2d_add_no_identity_avx2(input, CONVERT_TO_SHORTPTR(output), + stride, tx_type, tx_size, eob, bd); + break; + default: assert(0); break; + } +} + +void av1_highbd_inv_txfm_add_32x32_avx2(const tran_low_t *input, uint8_t *dest, + int stride, + const TxfmParam *txfm_param) { + const int bd = txfm_param->bd; + const TX_TYPE tx_type = txfm_param->tx_type; + const int32_t *src = cast_to_int32(input); + switch (tx_type) { + case DCT_DCT: + av1_highbd_inv_txfm2d_add_universe_avx2(input, dest, stride, tx_type, + txfm_param->tx_size, + txfm_param->eob, bd); + break; + // Assembly version doesn't support IDTX, so use C version for it. + case IDTX: + av1_inv_txfm2d_add_32x32_c(src, CONVERT_TO_SHORTPTR(dest), stride, + tx_type, bd); + break; + + default: assert(0); + } +} + +void av1_highbd_inv_txfm_add_avx2(const tran_low_t *input, uint8_t *dest, + int stride, const TxfmParam *txfm_param) { + assert(av1_ext_tx_used[txfm_param->tx_set_type][txfm_param->tx_type]); + const TX_SIZE tx_size = txfm_param->tx_size; + switch (tx_size) { + case TX_32X32: + av1_highbd_inv_txfm_add_32x32_avx2(input, dest, stride, txfm_param); + break; + case TX_16X16: + av1_highbd_inv_txfm_add_16x16_sse4_1(input, dest, stride, txfm_param); + break; + case TX_8X8: + av1_highbd_inv_txfm_add_8x8_sse4_1(input, dest, stride, txfm_param); + break; + case TX_4X8: + av1_highbd_inv_txfm_add_4x8(input, dest, stride, txfm_param); + break; + case TX_8X4: + av1_highbd_inv_txfm_add_8x4(input, dest, stride, txfm_param); + break; + case TX_8X16: + av1_highbd_inv_txfm_add_8x16_sse4_1(input, dest, stride, txfm_param); + break; + case TX_16X8: + av1_highbd_inv_txfm_add_16x8_sse4_1(input, dest, stride, txfm_param); + break; + case TX_16X32: + av1_highbd_inv_txfm_add_16x32(input, dest, stride, txfm_param); + break; + case TX_32X16: + av1_highbd_inv_txfm_add_32x16(input, dest, stride, txfm_param); + break; + case TX_32X64: + av1_highbd_inv_txfm_add_32x64(input, dest, stride, txfm_param); + break; + case TX_64X32: + av1_highbd_inv_txfm_add_64x32(input, dest, stride, txfm_param); + break; + case TX_4X4: + av1_highbd_inv_txfm_add_4x4_sse4_1(input, dest, stride, txfm_param); + break; + case TX_16X4: + av1_highbd_inv_txfm_add_16x4(input, dest, stride, txfm_param); + break; + case TX_4X16: + av1_highbd_inv_txfm_add_4x16(input, dest, stride, txfm_param); + break; + case TX_8X32: + av1_highbd_inv_txfm_add_8x32(input, dest, stride, txfm_param); + break; + case TX_32X8: + av1_highbd_inv_txfm_add_32x8(input, dest, stride, txfm_param); + break; + case TX_64X64: + case TX_16X64: + case TX_64X16: + av1_highbd_inv_txfm2d_add_universe_sse4_1( + input, dest, stride, txfm_param->tx_type, txfm_param->tx_size, + txfm_param->eob, txfm_param->bd); + break; + default: assert(0 && "Invalid transform size"); break; + } +} diff --git a/third_party/aom/av1/common/x86/highbd_inv_txfm_sse4.c b/third_party/aom/av1/common/x86/highbd_inv_txfm_sse4.c new file mode 100644 index 0000000000..e29e0baf50 --- /dev/null +++ b/third_party/aom/av1/common/x86/highbd_inv_txfm_sse4.c @@ -0,0 +1,5348 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ +#include <assert.h> +#include <smmintrin.h> /* SSE4.1 */ + +#include "config/aom_config.h" +#include "config/av1_rtcd.h" + +#include "av1/common/av1_inv_txfm1d_cfg.h" +#include "av1/common/idct.h" +#include "av1/common/x86/av1_inv_txfm_ssse3.h" +#include "av1/common/x86/av1_txfm_sse4.h" +#include "av1/common/x86/highbd_txfm_utility_sse4.h" + +static INLINE __m128i highbd_clamp_epi16(__m128i u, int bd) { + const __m128i zero = _mm_setzero_si128(); + const __m128i one = _mm_set1_epi16(1); + const __m128i max = _mm_sub_epi16(_mm_slli_epi16(one, bd), one); + __m128i clamped, mask; + + mask = _mm_cmpgt_epi16(u, max); + clamped = _mm_andnot_si128(mask, u); + mask = _mm_and_si128(mask, max); + clamped = _mm_or_si128(mask, clamped); + mask = _mm_cmpgt_epi16(clamped, zero); + clamped = _mm_and_si128(clamped, mask); + + return clamped; +} + +static INLINE __m128i highbd_get_recon_8x8_sse4_1(const __m128i pred, + __m128i res0, __m128i res1, + const int bd) { + __m128i x0 = _mm_cvtepi16_epi32(pred); + __m128i x1 = _mm_cvtepi16_epi32(_mm_srli_si128(pred, 8)); + + x0 = _mm_add_epi32(res0, x0); + x1 = _mm_add_epi32(res1, x1); + x0 = _mm_packus_epi32(x0, x1); + x0 = highbd_clamp_epi16(x0, bd); + return x0; +} + +static INLINE void highbd_write_buffer_8xn_sse4_1(__m128i *in, uint16_t *output, + int stride, int flipud, + int height, const int bd) { + int j = flipud ? (height - 1) : 0; + const int step = flipud ? -1 : 1; + for (int i = 0; i < height; ++i, j += step) { + __m128i v = _mm_loadu_si128((__m128i const *)(output + i * stride)); + __m128i u = highbd_get_recon_8x8_sse4_1(v, in[j], in[j + height], bd); + + _mm_storeu_si128((__m128i *)(output + i * stride), u); + } +} + +static INLINE void load_buffer_32bit_input(const int32_t *in, int stride, + __m128i *out, int out_size) { + for (int i = 0; i < out_size; ++i) { + out[i] = _mm_loadu_si128((const __m128i *)(in + i * stride)); + } +} + +static INLINE void load_buffer_4x4(const int32_t *coeff, __m128i *in) { + in[0] = _mm_load_si128((const __m128i *)(coeff + 0)); + in[1] = _mm_load_si128((const __m128i *)(coeff + 4)); + in[2] = _mm_load_si128((const __m128i *)(coeff + 8)); + in[3] = _mm_load_si128((const __m128i *)(coeff + 12)); +} + +static void addsub_sse4_1(const __m128i in0, const __m128i in1, __m128i *out0, + __m128i *out1, const __m128i *clamp_lo, + const __m128i *clamp_hi) { + __m128i a0 = _mm_add_epi32(in0, in1); + __m128i a1 = _mm_sub_epi32(in0, in1); + + a0 = _mm_max_epi32(a0, *clamp_lo); + a0 = _mm_min_epi32(a0, *clamp_hi); + a1 = _mm_max_epi32(a1, *clamp_lo); + a1 = _mm_min_epi32(a1, *clamp_hi); + + *out0 = a0; + *out1 = a1; +} + +static void addsub_no_clamp_sse4_1(const __m128i in0, const __m128i in1, + __m128i *out0, __m128i *out1) { + __m128i a0 = _mm_add_epi32(in0, in1); + __m128i a1 = _mm_sub_epi32(in0, in1); + + *out0 = a0; + *out1 = a1; +} + +static void addsub_shift_sse4_1(const __m128i in0, const __m128i in1, + __m128i *out0, __m128i *out1, + const __m128i *clamp_lo, + const __m128i *clamp_hi, int shift) { + __m128i offset = _mm_set1_epi32((1 << shift) >> 1); + __m128i in0_w_offset = _mm_add_epi32(in0, offset); + __m128i a0 = _mm_add_epi32(in0_w_offset, in1); + __m128i a1 = _mm_sub_epi32(in0_w_offset, in1); + + a0 = _mm_sra_epi32(a0, _mm_cvtsi32_si128(shift)); + a1 = _mm_sra_epi32(a1, _mm_cvtsi32_si128(shift)); + + a0 = _mm_max_epi32(a0, *clamp_lo); + a0 = _mm_min_epi32(a0, *clamp_hi); + a1 = _mm_max_epi32(a1, *clamp_lo); + a1 = _mm_min_epi32(a1, *clamp_hi); + + *out0 = a0; + *out1 = a1; +} + +static INLINE void idct32_stage4_sse4_1( + __m128i *bf1, const __m128i *cospim8, const __m128i *cospi56, + const __m128i *cospi8, const __m128i *cospim56, const __m128i *cospim40, + const __m128i *cospi24, const __m128i *cospi40, const __m128i *cospim24, + const __m128i *rounding, int bit) { + __m128i temp1, temp2; + temp1 = half_btf_sse4_1(cospim8, &bf1[17], cospi56, &bf1[30], rounding, bit); + bf1[30] = half_btf_sse4_1(cospi56, &bf1[17], cospi8, &bf1[30], rounding, bit); + bf1[17] = temp1; + + temp2 = half_btf_sse4_1(cospim56, &bf1[18], cospim8, &bf1[29], rounding, bit); + bf1[29] = + half_btf_sse4_1(cospim8, &bf1[18], cospi56, &bf1[29], rounding, bit); + bf1[18] = temp2; + + temp1 = half_btf_sse4_1(cospim40, &bf1[21], cospi24, &bf1[26], rounding, bit); + bf1[26] = + half_btf_sse4_1(cospi24, &bf1[21], cospi40, &bf1[26], rounding, bit); + bf1[21] = temp1; + + temp2 = + half_btf_sse4_1(cospim24, &bf1[22], cospim40, &bf1[25], rounding, bit); + bf1[25] = + half_btf_sse4_1(cospim40, &bf1[22], cospi24, &bf1[25], rounding, bit); + bf1[22] = temp2; +} + +static INLINE void idct32_stage5_sse4_1( + __m128i *bf1, const __m128i *cospim16, const __m128i *cospi48, + const __m128i *cospi16, const __m128i *cospim48, const __m128i *clamp_lo, + const __m128i *clamp_hi, const __m128i *rounding, int bit) { + __m128i temp1, temp2; + temp1 = half_btf_sse4_1(cospim16, &bf1[9], cospi48, &bf1[14], rounding, bit); + bf1[14] = half_btf_sse4_1(cospi48, &bf1[9], cospi16, &bf1[14], rounding, bit); + bf1[9] = temp1; + + temp2 = + half_btf_sse4_1(cospim48, &bf1[10], cospim16, &bf1[13], rounding, bit); + bf1[13] = + half_btf_sse4_1(cospim16, &bf1[10], cospi48, &bf1[13], rounding, bit); + bf1[10] = temp2; + + addsub_sse4_1(bf1[16], bf1[19], bf1 + 16, bf1 + 19, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[17], bf1[18], bf1 + 17, bf1 + 18, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[23], bf1[20], bf1 + 23, bf1 + 20, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[22], bf1[21], bf1 + 22, bf1 + 21, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[24], bf1[27], bf1 + 24, bf1 + 27, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[25], bf1[26], bf1 + 25, bf1 + 26, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[31], bf1[28], bf1 + 31, bf1 + 28, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[30], bf1[29], bf1 + 30, bf1 + 29, clamp_lo, clamp_hi); +} + +static INLINE void idct32_stage6_sse4_1( + __m128i *bf1, const __m128i *cospim32, const __m128i *cospi32, + const __m128i *cospim16, const __m128i *cospi48, const __m128i *cospi16, + const __m128i *cospim48, const __m128i *clamp_lo, const __m128i *clamp_hi, + const __m128i *rounding, int bit) { + __m128i temp1, temp2; + temp1 = half_btf_sse4_1(cospim32, &bf1[5], cospi32, &bf1[6], rounding, bit); + bf1[6] = half_btf_sse4_1(cospi32, &bf1[5], cospi32, &bf1[6], rounding, bit); + bf1[5] = temp1; + + addsub_sse4_1(bf1[8], bf1[11], bf1 + 8, bf1 + 11, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[9], bf1[10], bf1 + 9, bf1 + 10, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[15], bf1[12], bf1 + 15, bf1 + 12, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[14], bf1[13], bf1 + 14, bf1 + 13, clamp_lo, clamp_hi); + + temp1 = half_btf_sse4_1(cospim16, &bf1[18], cospi48, &bf1[29], rounding, bit); + bf1[29] = + half_btf_sse4_1(cospi48, &bf1[18], cospi16, &bf1[29], rounding, bit); + bf1[18] = temp1; + temp2 = half_btf_sse4_1(cospim16, &bf1[19], cospi48, &bf1[28], rounding, bit); + bf1[28] = + half_btf_sse4_1(cospi48, &bf1[19], cospi16, &bf1[28], rounding, bit); + bf1[19] = temp2; + temp1 = + half_btf_sse4_1(cospim48, &bf1[20], cospim16, &bf1[27], rounding, bit); + bf1[27] = + half_btf_sse4_1(cospim16, &bf1[20], cospi48, &bf1[27], rounding, bit); + bf1[20] = temp1; + temp2 = + half_btf_sse4_1(cospim48, &bf1[21], cospim16, &bf1[26], rounding, bit); + bf1[26] = + half_btf_sse4_1(cospim16, &bf1[21], cospi48, &bf1[26], rounding, bit); + bf1[21] = temp2; +} + +static INLINE void idct32_stage7_sse4_1(__m128i *bf1, const __m128i *cospim32, + const __m128i *cospi32, + const __m128i *clamp_lo, + const __m128i *clamp_hi, + const __m128i *rounding, int bit) { + __m128i temp1, temp2; + addsub_sse4_1(bf1[0], bf1[7], bf1 + 0, bf1 + 7, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[1], bf1[6], bf1 + 1, bf1 + 6, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[2], bf1[5], bf1 + 2, bf1 + 5, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[3], bf1[4], bf1 + 3, bf1 + 4, clamp_lo, clamp_hi); + + temp1 = half_btf_sse4_1(cospim32, &bf1[10], cospi32, &bf1[13], rounding, bit); + bf1[13] = + half_btf_sse4_1(cospi32, &bf1[10], cospi32, &bf1[13], rounding, bit); + bf1[10] = temp1; + temp2 = half_btf_sse4_1(cospim32, &bf1[11], cospi32, &bf1[12], rounding, bit); + bf1[12] = + half_btf_sse4_1(cospi32, &bf1[11], cospi32, &bf1[12], rounding, bit); + bf1[11] = temp2; + + addsub_sse4_1(bf1[16], bf1[23], bf1 + 16, bf1 + 23, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[17], bf1[22], bf1 + 17, bf1 + 22, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[18], bf1[21], bf1 + 18, bf1 + 21, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[19], bf1[20], bf1 + 19, bf1 + 20, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[31], bf1[24], bf1 + 31, bf1 + 24, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[30], bf1[25], bf1 + 30, bf1 + 25, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[29], bf1[26], bf1 + 29, bf1 + 26, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[28], bf1[27], bf1 + 28, bf1 + 27, clamp_lo, clamp_hi); +} + +static INLINE void idct32_stage8_sse4_1(__m128i *bf1, const __m128i *cospim32, + const __m128i *cospi32, + const __m128i *clamp_lo, + const __m128i *clamp_hi, + const __m128i *rounding, int bit) { + __m128i temp1, temp2; + addsub_sse4_1(bf1[0], bf1[15], bf1 + 0, bf1 + 15, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[1], bf1[14], bf1 + 1, bf1 + 14, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[2], bf1[13], bf1 + 2, bf1 + 13, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[3], bf1[12], bf1 + 3, bf1 + 12, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[4], bf1[11], bf1 + 4, bf1 + 11, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[5], bf1[10], bf1 + 5, bf1 + 10, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[6], bf1[9], bf1 + 6, bf1 + 9, clamp_lo, clamp_hi); + addsub_sse4_1(bf1[7], bf1[8], bf1 + 7, bf1 + 8, clamp_lo, clamp_hi); + + temp1 = half_btf_sse4_1(cospim32, &bf1[20], cospi32, &bf1[27], rounding, bit); + bf1[27] = + half_btf_sse4_1(cospi32, &bf1[20], cospi32, &bf1[27], rounding, bit); + bf1[20] = temp1; + temp2 = half_btf_sse4_1(cospim32, &bf1[21], cospi32, &bf1[26], rounding, bit); + bf1[26] = + half_btf_sse4_1(cospi32, &bf1[21], cospi32, &bf1[26], rounding, bit); + bf1[21] = temp2; + temp1 = half_btf_sse4_1(cospim32, &bf1[22], cospi32, &bf1[25], rounding, bit); + bf1[25] = + half_btf_sse4_1(cospi32, &bf1[22], cospi32, &bf1[25], rounding, bit); + bf1[22] = temp1; + temp2 = half_btf_sse4_1(cospim32, &bf1[23], cospi32, &bf1[24], rounding, bit); + bf1[24] = + half_btf_sse4_1(cospi32, &bf1[23], cospi32, &bf1[24], rounding, bit); + bf1[23] = temp2; +} + +static INLINE void idct32_stage9_sse4_1(__m128i *bf1, __m128i *out, + const int do_cols, const int bd, + const int out_shift, + const int log_range) { + if (do_cols) { + addsub_no_clamp_sse4_1(bf1[0], bf1[31], out + 0, out + 31); + addsub_no_clamp_sse4_1(bf1[1], bf1[30], out + 1, out + 30); + addsub_no_clamp_sse4_1(bf1[2], bf1[29], out + 2, out + 29); + addsub_no_clamp_sse4_1(bf1[3], bf1[28], out + 3, out + 28); + addsub_no_clamp_sse4_1(bf1[4], bf1[27], out + 4, out + 27); + addsub_no_clamp_sse4_1(bf1[5], bf1[26], out + 5, out + 26); + addsub_no_clamp_sse4_1(bf1[6], bf1[25], out + 6, out + 25); + addsub_no_clamp_sse4_1(bf1[7], bf1[24], out + 7, out + 24); + addsub_no_clamp_sse4_1(bf1[8], bf1[23], out + 8, out + 23); + addsub_no_clamp_sse4_1(bf1[9], bf1[22], out + 9, out + 22); + addsub_no_clamp_sse4_1(bf1[10], bf1[21], out + 10, out + 21); + addsub_no_clamp_sse4_1(bf1[11], bf1[20], out + 11, out + 20); + addsub_no_clamp_sse4_1(bf1[12], bf1[19], out + 12, out + 19); + addsub_no_clamp_sse4_1(bf1[13], bf1[18], out + 13, out + 18); + addsub_no_clamp_sse4_1(bf1[14], bf1[17], out + 14, out + 17); + addsub_no_clamp_sse4_1(bf1[15], bf1[16], out + 15, out + 16); + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m128i clamp_lo_out = _mm_set1_epi32(AOMMAX( + -(1 << (log_range_out - 1)), -(1 << (log_range - 1 - out_shift)))); + const __m128i clamp_hi_out = _mm_set1_epi32(AOMMIN( + (1 << (log_range_out - 1)) - 1, (1 << (log_range - 1 - out_shift)))); + + addsub_shift_sse4_1(bf1[0], bf1[31], out + 0, out + 31, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf1[1], bf1[30], out + 1, out + 30, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf1[2], bf1[29], out + 2, out + 29, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf1[3], bf1[28], out + 3, out + 28, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf1[4], bf1[27], out + 4, out + 27, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf1[5], bf1[26], out + 5, out + 26, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf1[6], bf1[25], out + 6, out + 25, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf1[7], bf1[24], out + 7, out + 24, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf1[8], bf1[23], out + 8, out + 23, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf1[9], bf1[22], out + 9, out + 22, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf1[10], bf1[21], out + 10, out + 21, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf1[11], bf1[20], out + 11, out + 20, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf1[12], bf1[19], out + 12, out + 19, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf1[13], bf1[18], out + 13, out + 18, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf1[14], bf1[17], out + 14, out + 17, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf1[15], bf1[16], out + 15, out + 16, &clamp_lo_out, + &clamp_hi_out, out_shift); + } +} + +static void neg_shift_sse4_1(const __m128i in0, const __m128i in1, + __m128i *out0, __m128i *out1, + const __m128i *clamp_lo, const __m128i *clamp_hi, + int shift) { + __m128i offset = _mm_set1_epi32((1 << shift) >> 1); + __m128i a0 = _mm_add_epi32(offset, in0); + __m128i a1 = _mm_sub_epi32(offset, in1); + + a0 = _mm_sra_epi32(a0, _mm_cvtsi32_si128(shift)); + a1 = _mm_sra_epi32(a1, _mm_cvtsi32_si128(shift)); + + a0 = _mm_max_epi32(a0, *clamp_lo); + a0 = _mm_min_epi32(a0, *clamp_hi); + a1 = _mm_max_epi32(a1, *clamp_lo); + a1 = _mm_min_epi32(a1, *clamp_hi); + + *out0 = a0; + *out1 = a1; +} + +static void idct4x4_sse4_1(__m128i *in, int bit, int do_cols, int bd) { + const int32_t *cospi = cospi_arr(bit); + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + const __m128i cospi48 = _mm_set1_epi32(cospi[48]); + const __m128i cospi16 = _mm_set1_epi32(cospi[16]); + const __m128i cospim16 = _mm_set1_epi32(-cospi[16]); + const __m128i rnding = _mm_set1_epi32(1 << (bit - 1)); + + __m128i u0, u1, u2, u3; + __m128i v0, v1, v2, v3, x, y; + + v0 = _mm_unpacklo_epi32(in[0], in[1]); + v1 = _mm_unpackhi_epi32(in[0], in[1]); + v2 = _mm_unpacklo_epi32(in[2], in[3]); + v3 = _mm_unpackhi_epi32(in[2], in[3]); + + u0 = _mm_unpacklo_epi64(v0, v2); + u1 = _mm_unpackhi_epi64(v0, v2); + u2 = _mm_unpacklo_epi64(v1, v3); + u3 = _mm_unpackhi_epi64(v1, v3); + + x = _mm_mullo_epi32(u0, cospi32); + y = _mm_mullo_epi32(u2, cospi32); + v0 = _mm_add_epi32(x, y); + v0 = _mm_add_epi32(v0, rnding); + v0 = _mm_srai_epi32(v0, bit); + + v1 = _mm_sub_epi32(x, y); + v1 = _mm_add_epi32(v1, rnding); + v1 = _mm_srai_epi32(v1, bit); + + x = _mm_mullo_epi32(u1, cospi48); + y = _mm_mullo_epi32(u3, cospim16); + v2 = _mm_add_epi32(x, y); + v2 = _mm_add_epi32(v2, rnding); + v2 = _mm_srai_epi32(v2, bit); + + x = _mm_mullo_epi32(u1, cospi16); + y = _mm_mullo_epi32(u3, cospi48); + v3 = _mm_add_epi32(x, y); + v3 = _mm_add_epi32(v3, rnding); + v3 = _mm_srai_epi32(v3, bit); + + if (do_cols) { + addsub_no_clamp_sse4_1(v0, v3, in + 0, in + 3); + addsub_no_clamp_sse4_1(v1, v2, in + 1, in + 2); + } else { + const int log_range = AOMMAX(16, bd + 6); + const __m128i clamp_lo = _mm_set1_epi32(-(1 << (log_range - 1))); + const __m128i clamp_hi = _mm_set1_epi32((1 << (log_range - 1)) - 1); + addsub_sse4_1(v0, v3, in + 0, in + 3, &clamp_lo, &clamp_hi); + addsub_sse4_1(v1, v2, in + 1, in + 2, &clamp_lo, &clamp_hi); + } +} + +static void iadst4x4_sse4_1(__m128i *in, int bit, int do_cols, int bd) { + const int32_t *sinpi = sinpi_arr(bit); + const __m128i rnding = _mm_set1_epi32(1 << (bit - 1)); + const __m128i sinpi1 = _mm_set1_epi32((int)sinpi[1]); + const __m128i sinpi2 = _mm_set1_epi32((int)sinpi[2]); + const __m128i sinpi3 = _mm_set1_epi32((int)sinpi[3]); + const __m128i sinpi4 = _mm_set1_epi32((int)sinpi[4]); + __m128i t; + __m128i s0, s1, s2, s3, s4, s5, s6, s7; + __m128i x0, x1, x2, x3; + __m128i u0, u1, u2, u3; + __m128i v0, v1, v2, v3; + + v0 = _mm_unpacklo_epi32(in[0], in[1]); + v1 = _mm_unpackhi_epi32(in[0], in[1]); + v2 = _mm_unpacklo_epi32(in[2], in[3]); + v3 = _mm_unpackhi_epi32(in[2], in[3]); + + x0 = _mm_unpacklo_epi64(v0, v2); + x1 = _mm_unpackhi_epi64(v0, v2); + x2 = _mm_unpacklo_epi64(v1, v3); + x3 = _mm_unpackhi_epi64(v1, v3); + + s0 = _mm_mullo_epi32(x0, sinpi1); + s1 = _mm_mullo_epi32(x0, sinpi2); + s2 = _mm_mullo_epi32(x1, sinpi3); + s3 = _mm_mullo_epi32(x2, sinpi4); + s4 = _mm_mullo_epi32(x2, sinpi1); + s5 = _mm_mullo_epi32(x3, sinpi2); + s6 = _mm_mullo_epi32(x3, sinpi4); + t = _mm_sub_epi32(x0, x2); + s7 = _mm_add_epi32(t, x3); + + t = _mm_add_epi32(s0, s3); + s0 = _mm_add_epi32(t, s5); + t = _mm_sub_epi32(s1, s4); + s1 = _mm_sub_epi32(t, s6); + s3 = s2; + s2 = _mm_mullo_epi32(s7, sinpi3); + + u0 = _mm_add_epi32(s0, s3); + u1 = _mm_add_epi32(s1, s3); + u2 = s2; + t = _mm_add_epi32(s0, s1); + u3 = _mm_sub_epi32(t, s3); + + u0 = _mm_add_epi32(u0, rnding); + u0 = _mm_srai_epi32(u0, bit); + + u1 = _mm_add_epi32(u1, rnding); + u1 = _mm_srai_epi32(u1, bit); + + u2 = _mm_add_epi32(u2, rnding); + u2 = _mm_srai_epi32(u2, bit); + + u3 = _mm_add_epi32(u3, rnding); + u3 = _mm_srai_epi32(u3, bit); + + if (!do_cols) { + const int log_range = AOMMAX(16, bd + 6); + const __m128i clamp_lo = _mm_set1_epi32(-(1 << (log_range - 1))); + const __m128i clamp_hi = _mm_set1_epi32((1 << (log_range - 1)) - 1); + + u0 = _mm_max_epi32(u0, clamp_lo); + u0 = _mm_min_epi32(u0, clamp_hi); + u1 = _mm_max_epi32(u1, clamp_lo); + u1 = _mm_min_epi32(u1, clamp_hi); + u2 = _mm_max_epi32(u2, clamp_lo); + u2 = _mm_min_epi32(u2, clamp_hi); + u3 = _mm_max_epi32(u3, clamp_lo); + u3 = _mm_min_epi32(u3, clamp_hi); + } + + in[0] = u0; + in[1] = u1; + in[2] = u2; + in[3] = u3; +} + +static INLINE void round_shift_4x4(__m128i *in, int shift) { + __m128i rnding = _mm_set1_epi32(1 << (shift - 1)); + + in[0] = _mm_add_epi32(in[0], rnding); + in[1] = _mm_add_epi32(in[1], rnding); + in[2] = _mm_add_epi32(in[2], rnding); + in[3] = _mm_add_epi32(in[3], rnding); + + in[0] = _mm_srai_epi32(in[0], shift); + in[1] = _mm_srai_epi32(in[1], shift); + in[2] = _mm_srai_epi32(in[2], shift); + in[3] = _mm_srai_epi32(in[3], shift); +} + +static void write_buffer_4x4(__m128i *in, uint16_t *output, int stride, + int fliplr, int flipud, int shift, int bd) { + const __m128i zero = _mm_setzero_si128(); + __m128i u0, u1, u2, u3; + __m128i v0, v1, v2, v3; + + round_shift_4x4(in, shift); + + v0 = _mm_loadl_epi64((__m128i const *)(output + 0 * stride)); + v1 = _mm_loadl_epi64((__m128i const *)(output + 1 * stride)); + v2 = _mm_loadl_epi64((__m128i const *)(output + 2 * stride)); + v3 = _mm_loadl_epi64((__m128i const *)(output + 3 * stride)); + + v0 = _mm_unpacklo_epi16(v0, zero); + v1 = _mm_unpacklo_epi16(v1, zero); + v2 = _mm_unpacklo_epi16(v2, zero); + v3 = _mm_unpacklo_epi16(v3, zero); + + if (fliplr) { + in[0] = _mm_shuffle_epi32(in[0], 0x1B); + in[1] = _mm_shuffle_epi32(in[1], 0x1B); + in[2] = _mm_shuffle_epi32(in[2], 0x1B); + in[3] = _mm_shuffle_epi32(in[3], 0x1B); + } + + if (flipud) { + u0 = _mm_add_epi32(in[3], v0); + u1 = _mm_add_epi32(in[2], v1); + u2 = _mm_add_epi32(in[1], v2); + u3 = _mm_add_epi32(in[0], v3); + } else { + u0 = _mm_add_epi32(in[0], v0); + u1 = _mm_add_epi32(in[1], v1); + u2 = _mm_add_epi32(in[2], v2); + u3 = _mm_add_epi32(in[3], v3); + } + + v0 = _mm_packus_epi32(u0, u1); + v2 = _mm_packus_epi32(u2, u3); + + u0 = highbd_clamp_epi16(v0, bd); + u2 = highbd_clamp_epi16(v2, bd); + + v0 = _mm_unpacklo_epi64(u0, u0); + v1 = _mm_unpackhi_epi64(u0, u0); + v2 = _mm_unpacklo_epi64(u2, u2); + v3 = _mm_unpackhi_epi64(u2, u2); + + _mm_storel_epi64((__m128i *)(output + 0 * stride), v0); + _mm_storel_epi64((__m128i *)(output + 1 * stride), v1); + _mm_storel_epi64((__m128i *)(output + 2 * stride), v2); + _mm_storel_epi64((__m128i *)(output + 3 * stride), v3); +} + +void av1_inv_txfm2d_add_4x4_sse4_1(const int32_t *coeff, uint16_t *output, + int stride, TX_TYPE tx_type, int bd) { + __m128i in[4]; + const int8_t *shift = inv_txfm_shift_ls[TX_4X4]; + const int txw_idx = get_txw_idx(TX_4X4); + const int txh_idx = get_txh_idx(TX_4X4); + + switch (tx_type) { + case DCT_DCT: + load_buffer_4x4(coeff, in); + idct4x4_sse4_1(in, inv_cos_bit_row[txw_idx][txh_idx], 0, bd); + idct4x4_sse4_1(in, inv_cos_bit_col[txw_idx][txh_idx], 1, bd); + write_buffer_4x4(in, output, stride, 0, 0, -shift[1], bd); + break; + case ADST_DCT: + load_buffer_4x4(coeff, in); + idct4x4_sse4_1(in, inv_cos_bit_row[txw_idx][txh_idx], 0, bd); + iadst4x4_sse4_1(in, inv_cos_bit_col[txw_idx][txh_idx], 1, bd); + write_buffer_4x4(in, output, stride, 0, 0, -shift[1], bd); + break; + case DCT_ADST: + load_buffer_4x4(coeff, in); + iadst4x4_sse4_1(in, inv_cos_bit_row[txw_idx][txh_idx], 0, bd); + idct4x4_sse4_1(in, inv_cos_bit_col[txw_idx][txh_idx], 1, bd); + write_buffer_4x4(in, output, stride, 0, 0, -shift[1], bd); + break; + case ADST_ADST: + load_buffer_4x4(coeff, in); + iadst4x4_sse4_1(in, inv_cos_bit_row[txw_idx][txh_idx], 0, bd); + iadst4x4_sse4_1(in, inv_cos_bit_col[txw_idx][txh_idx], 1, bd); + write_buffer_4x4(in, output, stride, 0, 0, -shift[1], bd); + break; + case FLIPADST_DCT: + load_buffer_4x4(coeff, in); + idct4x4_sse4_1(in, inv_cos_bit_row[txw_idx][txh_idx], 0, bd); + iadst4x4_sse4_1(in, inv_cos_bit_col[txw_idx][txh_idx], 1, bd); + write_buffer_4x4(in, output, stride, 0, 1, -shift[1], bd); + break; + case DCT_FLIPADST: + load_buffer_4x4(coeff, in); + iadst4x4_sse4_1(in, inv_cos_bit_row[txw_idx][txh_idx], 0, bd); + idct4x4_sse4_1(in, inv_cos_bit_col[txw_idx][txh_idx], 1, bd); + write_buffer_4x4(in, output, stride, 1, 0, -shift[1], bd); + break; + case FLIPADST_FLIPADST: + load_buffer_4x4(coeff, in); + iadst4x4_sse4_1(in, inv_cos_bit_row[txw_idx][txh_idx], 0, bd); + iadst4x4_sse4_1(in, inv_cos_bit_col[txw_idx][txh_idx], 1, bd); + write_buffer_4x4(in, output, stride, 1, 1, -shift[1], bd); + break; + case ADST_FLIPADST: + load_buffer_4x4(coeff, in); + iadst4x4_sse4_1(in, inv_cos_bit_row[txw_idx][txh_idx], 0, bd); + iadst4x4_sse4_1(in, inv_cos_bit_col[txw_idx][txh_idx], 1, bd); + write_buffer_4x4(in, output, stride, 1, 0, -shift[1], bd); + break; + case FLIPADST_ADST: + load_buffer_4x4(coeff, in); + iadst4x4_sse4_1(in, inv_cos_bit_row[txw_idx][txh_idx], 0, bd); + iadst4x4_sse4_1(in, inv_cos_bit_col[txw_idx][txh_idx], 1, bd); + write_buffer_4x4(in, output, stride, 0, 1, -shift[1], bd); + break; + default: assert(0); + } +} + +// 8x8 +static void load_buffer_8x8(const int32_t *coeff, __m128i *in) { + in[0] = _mm_load_si128((const __m128i *)(coeff + 0)); + in[1] = _mm_load_si128((const __m128i *)(coeff + 4)); + in[2] = _mm_load_si128((const __m128i *)(coeff + 8)); + in[3] = _mm_load_si128((const __m128i *)(coeff + 12)); + in[4] = _mm_load_si128((const __m128i *)(coeff + 16)); + in[5] = _mm_load_si128((const __m128i *)(coeff + 20)); + in[6] = _mm_load_si128((const __m128i *)(coeff + 24)); + in[7] = _mm_load_si128((const __m128i *)(coeff + 28)); + in[8] = _mm_load_si128((const __m128i *)(coeff + 32)); + in[9] = _mm_load_si128((const __m128i *)(coeff + 36)); + in[10] = _mm_load_si128((const __m128i *)(coeff + 40)); + in[11] = _mm_load_si128((const __m128i *)(coeff + 44)); + in[12] = _mm_load_si128((const __m128i *)(coeff + 48)); + in[13] = _mm_load_si128((const __m128i *)(coeff + 52)); + in[14] = _mm_load_si128((const __m128i *)(coeff + 56)); + in[15] = _mm_load_si128((const __m128i *)(coeff + 60)); +} + +static void idct8x8_sse4_1(__m128i *in, __m128i *out, int bit, int do_cols, + int bd, int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m128i cospi56 = _mm_set1_epi32(cospi[56]); + const __m128i cospim8 = _mm_set1_epi32(-cospi[8]); + const __m128i cospi24 = _mm_set1_epi32(cospi[24]); + const __m128i cospim40 = _mm_set1_epi32(-cospi[40]); + const __m128i cospi40 = _mm_set1_epi32(cospi[40]); + const __m128i cospi8 = _mm_set1_epi32(cospi[8]); + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + const __m128i cospi48 = _mm_set1_epi32(cospi[48]); + const __m128i cospim16 = _mm_set1_epi32(-cospi[16]); + const __m128i cospi16 = _mm_set1_epi32(cospi[16]); + const __m128i rnding = _mm_set1_epi32(1 << (bit - 1)); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m128i clamp_lo = _mm_set1_epi32(-(1 << (log_range - 1))); + const __m128i clamp_hi = _mm_set1_epi32((1 << (log_range - 1)) - 1); + __m128i u0, u1, u2, u3, u4, u5, u6, u7; + __m128i v0, v1, v2, v3, v4, v5, v6, v7; + __m128i x, y; + int col; + + // Note: + // Even column: 0, 2, ..., 14 + // Odd column: 1, 3, ..., 15 + // one even column plus one odd column constructs one row (8 coeffs) + // total we have 8 rows (8x8). + for (col = 0; col < 2; ++col) { + // stage 0 + // stage 1 + // stage 2 + u0 = in[0 * 2 + col]; + u1 = in[4 * 2 + col]; + u2 = in[2 * 2 + col]; + u3 = in[6 * 2 + col]; + + x = _mm_mullo_epi32(in[1 * 2 + col], cospi56); + y = _mm_mullo_epi32(in[7 * 2 + col], cospim8); + u4 = _mm_add_epi32(x, y); + u4 = _mm_add_epi32(u4, rnding); + u4 = _mm_srai_epi32(u4, bit); + + x = _mm_mullo_epi32(in[1 * 2 + col], cospi8); + y = _mm_mullo_epi32(in[7 * 2 + col], cospi56); + u7 = _mm_add_epi32(x, y); + u7 = _mm_add_epi32(u7, rnding); + u7 = _mm_srai_epi32(u7, bit); + + x = _mm_mullo_epi32(in[5 * 2 + col], cospi24); + y = _mm_mullo_epi32(in[3 * 2 + col], cospim40); + u5 = _mm_add_epi32(x, y); + u5 = _mm_add_epi32(u5, rnding); + u5 = _mm_srai_epi32(u5, bit); + + x = _mm_mullo_epi32(in[5 * 2 + col], cospi40); + y = _mm_mullo_epi32(in[3 * 2 + col], cospi24); + u6 = _mm_add_epi32(x, y); + u6 = _mm_add_epi32(u6, rnding); + u6 = _mm_srai_epi32(u6, bit); + + // stage 3 + x = _mm_mullo_epi32(u0, cospi32); + y = _mm_mullo_epi32(u1, cospi32); + v0 = _mm_add_epi32(x, y); + v0 = _mm_add_epi32(v0, rnding); + v0 = _mm_srai_epi32(v0, bit); + + v1 = _mm_sub_epi32(x, y); + v1 = _mm_add_epi32(v1, rnding); + v1 = _mm_srai_epi32(v1, bit); + + x = _mm_mullo_epi32(u2, cospi48); + y = _mm_mullo_epi32(u3, cospim16); + v2 = _mm_add_epi32(x, y); + v2 = _mm_add_epi32(v2, rnding); + v2 = _mm_srai_epi32(v2, bit); + + x = _mm_mullo_epi32(u2, cospi16); + y = _mm_mullo_epi32(u3, cospi48); + v3 = _mm_add_epi32(x, y); + v3 = _mm_add_epi32(v3, rnding); + v3 = _mm_srai_epi32(v3, bit); + + addsub_sse4_1(u4, u5, &v4, &v5, &clamp_lo, &clamp_hi); + addsub_sse4_1(u7, u6, &v7, &v6, &clamp_lo, &clamp_hi); + + // stage 4 + addsub_sse4_1(v0, v3, &u0, &u3, &clamp_lo, &clamp_hi); + addsub_sse4_1(v1, v2, &u1, &u2, &clamp_lo, &clamp_hi); + u4 = v4; + u7 = v7; + + x = _mm_mullo_epi32(v5, cospi32); + y = _mm_mullo_epi32(v6, cospi32); + u6 = _mm_add_epi32(y, x); + u6 = _mm_add_epi32(u6, rnding); + u6 = _mm_srai_epi32(u6, bit); + + u5 = _mm_sub_epi32(y, x); + u5 = _mm_add_epi32(u5, rnding); + u5 = _mm_srai_epi32(u5, bit); + + // stage 5 + if (do_cols) { + addsub_no_clamp_sse4_1(u0, u7, out + 0 * 2 + col, out + 7 * 2 + col); + addsub_no_clamp_sse4_1(u1, u6, out + 1 * 2 + col, out + 6 * 2 + col); + addsub_no_clamp_sse4_1(u2, u5, out + 2 * 2 + col, out + 5 * 2 + col); + addsub_no_clamp_sse4_1(u3, u4, out + 3 * 2 + col, out + 4 * 2 + col); + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m128i clamp_lo_out = _mm_set1_epi32(AOMMAX( + -(1 << (log_range_out - 1)), -(1 << (log_range - 1 - out_shift)))); + const __m128i clamp_hi_out = _mm_set1_epi32(AOMMIN( + (1 << (log_range_out - 1)) - 1, (1 << (log_range - 1 - out_shift)))); + addsub_shift_sse4_1(u0, u7, out + 0 * 2 + col, out + 7 * 2 + col, + &clamp_lo_out, &clamp_hi_out, out_shift); + addsub_shift_sse4_1(u1, u6, out + 1 * 2 + col, out + 6 * 2 + col, + &clamp_lo_out, &clamp_hi_out, out_shift); + addsub_shift_sse4_1(u2, u5, out + 2 * 2 + col, out + 5 * 2 + col, + &clamp_lo_out, &clamp_hi_out, out_shift); + addsub_shift_sse4_1(u3, u4, out + 3 * 2 + col, out + 4 * 2 + col, + &clamp_lo_out, &clamp_hi_out, out_shift); + } + } +} + +static void iadst8x8_sse4_1(__m128i *in, __m128i *out, int bit, int do_cols, + int bd, int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m128i cospi4 = _mm_set1_epi32(cospi[4]); + const __m128i cospi60 = _mm_set1_epi32(cospi[60]); + const __m128i cospi20 = _mm_set1_epi32(cospi[20]); + const __m128i cospi44 = _mm_set1_epi32(cospi[44]); + const __m128i cospi36 = _mm_set1_epi32(cospi[36]); + const __m128i cospi28 = _mm_set1_epi32(cospi[28]); + const __m128i cospi52 = _mm_set1_epi32(cospi[52]); + const __m128i cospi12 = _mm_set1_epi32(cospi[12]); + const __m128i cospi16 = _mm_set1_epi32(cospi[16]); + const __m128i cospi48 = _mm_set1_epi32(cospi[48]); + const __m128i cospim48 = _mm_set1_epi32(-cospi[48]); + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + const __m128i rnding = _mm_set1_epi32(1 << (bit - 1)); + const __m128i kZero = _mm_setzero_si128(); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m128i clamp_lo = _mm_set1_epi32(-(1 << (log_range - 1))); + const __m128i clamp_hi = _mm_set1_epi32((1 << (log_range - 1)) - 1); + __m128i u[8], v[8], x; + + // Even 8 points: 0, 2, ..., 14 + // stage 0 + // stage 1 + // stage 2 + // (1) + u[0] = _mm_mullo_epi32(in[14], cospi4); + x = _mm_mullo_epi32(in[0], cospi60); + u[0] = _mm_add_epi32(u[0], x); + u[0] = _mm_add_epi32(u[0], rnding); + u[0] = _mm_srai_epi32(u[0], bit); + + u[1] = _mm_mullo_epi32(in[14], cospi60); + x = _mm_mullo_epi32(in[0], cospi4); + u[1] = _mm_sub_epi32(u[1], x); + u[1] = _mm_add_epi32(u[1], rnding); + u[1] = _mm_srai_epi32(u[1], bit); + + // (2) + u[2] = _mm_mullo_epi32(in[10], cospi20); + x = _mm_mullo_epi32(in[4], cospi44); + u[2] = _mm_add_epi32(u[2], x); + u[2] = _mm_add_epi32(u[2], rnding); + u[2] = _mm_srai_epi32(u[2], bit); + + u[3] = _mm_mullo_epi32(in[10], cospi44); + x = _mm_mullo_epi32(in[4], cospi20); + u[3] = _mm_sub_epi32(u[3], x); + u[3] = _mm_add_epi32(u[3], rnding); + u[3] = _mm_srai_epi32(u[3], bit); + + // (3) + u[4] = _mm_mullo_epi32(in[6], cospi36); + x = _mm_mullo_epi32(in[8], cospi28); + u[4] = _mm_add_epi32(u[4], x); + u[4] = _mm_add_epi32(u[4], rnding); + u[4] = _mm_srai_epi32(u[4], bit); + + u[5] = _mm_mullo_epi32(in[6], cospi28); + x = _mm_mullo_epi32(in[8], cospi36); + u[5] = _mm_sub_epi32(u[5], x); + u[5] = _mm_add_epi32(u[5], rnding); + u[5] = _mm_srai_epi32(u[5], bit); + + // (4) + u[6] = _mm_mullo_epi32(in[2], cospi52); + x = _mm_mullo_epi32(in[12], cospi12); + u[6] = _mm_add_epi32(u[6], x); + u[6] = _mm_add_epi32(u[6], rnding); + u[6] = _mm_srai_epi32(u[6], bit); + + u[7] = _mm_mullo_epi32(in[2], cospi12); + x = _mm_mullo_epi32(in[12], cospi52); + u[7] = _mm_sub_epi32(u[7], x); + u[7] = _mm_add_epi32(u[7], rnding); + u[7] = _mm_srai_epi32(u[7], bit); + + // stage 3 + addsub_sse4_1(u[0], u[4], &v[0], &v[4], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[1], u[5], &v[1], &v[5], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[2], u[6], &v[2], &v[6], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[3], u[7], &v[3], &v[7], &clamp_lo, &clamp_hi); + + // stage 4 + u[0] = v[0]; + u[1] = v[1]; + u[2] = v[2]; + u[3] = v[3]; + + u[4] = _mm_mullo_epi32(v[4], cospi16); + x = _mm_mullo_epi32(v[5], cospi48); + u[4] = _mm_add_epi32(u[4], x); + u[4] = _mm_add_epi32(u[4], rnding); + u[4] = _mm_srai_epi32(u[4], bit); + + u[5] = _mm_mullo_epi32(v[4], cospi48); + x = _mm_mullo_epi32(v[5], cospi16); + u[5] = _mm_sub_epi32(u[5], x); + u[5] = _mm_add_epi32(u[5], rnding); + u[5] = _mm_srai_epi32(u[5], bit); + + u[6] = _mm_mullo_epi32(v[6], cospim48); + x = _mm_mullo_epi32(v[7], cospi16); + u[6] = _mm_add_epi32(u[6], x); + u[6] = _mm_add_epi32(u[6], rnding); + u[6] = _mm_srai_epi32(u[6], bit); + + u[7] = _mm_mullo_epi32(v[6], cospi16); + x = _mm_mullo_epi32(v[7], cospim48); + u[7] = _mm_sub_epi32(u[7], x); + u[7] = _mm_add_epi32(u[7], rnding); + u[7] = _mm_srai_epi32(u[7], bit); + + // stage 5 + addsub_sse4_1(u[0], u[2], &v[0], &v[2], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[1], u[3], &v[1], &v[3], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[4], u[6], &v[4], &v[6], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[5], u[7], &v[5], &v[7], &clamp_lo, &clamp_hi); + + // stage 6 + u[0] = v[0]; + u[1] = v[1]; + u[4] = v[4]; + u[5] = v[5]; + + v[0] = _mm_mullo_epi32(v[2], cospi32); + x = _mm_mullo_epi32(v[3], cospi32); + u[2] = _mm_add_epi32(v[0], x); + u[2] = _mm_add_epi32(u[2], rnding); + u[2] = _mm_srai_epi32(u[2], bit); + + u[3] = _mm_sub_epi32(v[0], x); + u[3] = _mm_add_epi32(u[3], rnding); + u[3] = _mm_srai_epi32(u[3], bit); + + v[0] = _mm_mullo_epi32(v[6], cospi32); + x = _mm_mullo_epi32(v[7], cospi32); + u[6] = _mm_add_epi32(v[0], x); + u[6] = _mm_add_epi32(u[6], rnding); + u[6] = _mm_srai_epi32(u[6], bit); + + u[7] = _mm_sub_epi32(v[0], x); + u[7] = _mm_add_epi32(u[7], rnding); + u[7] = _mm_srai_epi32(u[7], bit); + + // stage 7 + if (do_cols) { + out[0] = u[0]; + out[2] = _mm_sub_epi32(kZero, u[4]); + out[4] = u[6]; + out[6] = _mm_sub_epi32(kZero, u[2]); + out[8] = u[3]; + out[10] = _mm_sub_epi32(kZero, u[7]); + out[12] = u[5]; + out[14] = _mm_sub_epi32(kZero, u[1]); + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m128i clamp_lo_out = _mm_set1_epi32(-(1 << (log_range_out - 1))); + const __m128i clamp_hi_out = _mm_set1_epi32((1 << (log_range_out - 1)) - 1); + + neg_shift_sse4_1(u[0], u[4], out + 0, out + 2, &clamp_lo_out, &clamp_hi_out, + out_shift); + neg_shift_sse4_1(u[6], u[2], out + 4, out + 6, &clamp_lo_out, &clamp_hi_out, + out_shift); + neg_shift_sse4_1(u[3], u[7], out + 8, out + 10, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(u[5], u[1], out + 12, out + 14, &clamp_lo_out, + &clamp_hi_out, out_shift); + } + + // Odd 8 points: 1, 3, ..., 15 + // stage 0 + // stage 1 + // stage 2 + // (1) + u[0] = _mm_mullo_epi32(in[15], cospi4); + x = _mm_mullo_epi32(in[1], cospi60); + u[0] = _mm_add_epi32(u[0], x); + u[0] = _mm_add_epi32(u[0], rnding); + u[0] = _mm_srai_epi32(u[0], bit); + + u[1] = _mm_mullo_epi32(in[15], cospi60); + x = _mm_mullo_epi32(in[1], cospi4); + u[1] = _mm_sub_epi32(u[1], x); + u[1] = _mm_add_epi32(u[1], rnding); + u[1] = _mm_srai_epi32(u[1], bit); + + // (2) + u[2] = _mm_mullo_epi32(in[11], cospi20); + x = _mm_mullo_epi32(in[5], cospi44); + u[2] = _mm_add_epi32(u[2], x); + u[2] = _mm_add_epi32(u[2], rnding); + u[2] = _mm_srai_epi32(u[2], bit); + + u[3] = _mm_mullo_epi32(in[11], cospi44); + x = _mm_mullo_epi32(in[5], cospi20); + u[3] = _mm_sub_epi32(u[3], x); + u[3] = _mm_add_epi32(u[3], rnding); + u[3] = _mm_srai_epi32(u[3], bit); + + // (3) + u[4] = _mm_mullo_epi32(in[7], cospi36); + x = _mm_mullo_epi32(in[9], cospi28); + u[4] = _mm_add_epi32(u[4], x); + u[4] = _mm_add_epi32(u[4], rnding); + u[4] = _mm_srai_epi32(u[4], bit); + + u[5] = _mm_mullo_epi32(in[7], cospi28); + x = _mm_mullo_epi32(in[9], cospi36); + u[5] = _mm_sub_epi32(u[5], x); + u[5] = _mm_add_epi32(u[5], rnding); + u[5] = _mm_srai_epi32(u[5], bit); + + // (4) + u[6] = _mm_mullo_epi32(in[3], cospi52); + x = _mm_mullo_epi32(in[13], cospi12); + u[6] = _mm_add_epi32(u[6], x); + u[6] = _mm_add_epi32(u[6], rnding); + u[6] = _mm_srai_epi32(u[6], bit); + + u[7] = _mm_mullo_epi32(in[3], cospi12); + x = _mm_mullo_epi32(in[13], cospi52); + u[7] = _mm_sub_epi32(u[7], x); + u[7] = _mm_add_epi32(u[7], rnding); + u[7] = _mm_srai_epi32(u[7], bit); + + // stage 3 + addsub_sse4_1(u[0], u[4], &v[0], &v[4], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[1], u[5], &v[1], &v[5], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[2], u[6], &v[2], &v[6], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[3], u[7], &v[3], &v[7], &clamp_lo, &clamp_hi); + + // stage 4 + u[0] = v[0]; + u[1] = v[1]; + u[2] = v[2]; + u[3] = v[3]; + + u[4] = _mm_mullo_epi32(v[4], cospi16); + x = _mm_mullo_epi32(v[5], cospi48); + u[4] = _mm_add_epi32(u[4], x); + u[4] = _mm_add_epi32(u[4], rnding); + u[4] = _mm_srai_epi32(u[4], bit); + + u[5] = _mm_mullo_epi32(v[4], cospi48); + x = _mm_mullo_epi32(v[5], cospi16); + u[5] = _mm_sub_epi32(u[5], x); + u[5] = _mm_add_epi32(u[5], rnding); + u[5] = _mm_srai_epi32(u[5], bit); + + u[6] = _mm_mullo_epi32(v[6], cospim48); + x = _mm_mullo_epi32(v[7], cospi16); + u[6] = _mm_add_epi32(u[6], x); + u[6] = _mm_add_epi32(u[6], rnding); + u[6] = _mm_srai_epi32(u[6], bit); + + u[7] = _mm_mullo_epi32(v[6], cospi16); + x = _mm_mullo_epi32(v[7], cospim48); + u[7] = _mm_sub_epi32(u[7], x); + u[7] = _mm_add_epi32(u[7], rnding); + u[7] = _mm_srai_epi32(u[7], bit); + + // stage 5 + addsub_sse4_1(u[0], u[2], &v[0], &v[2], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[1], u[3], &v[1], &v[3], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[4], u[6], &v[4], &v[6], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[5], u[7], &v[5], &v[7], &clamp_lo, &clamp_hi); + + // stage 6 + u[0] = v[0]; + u[1] = v[1]; + u[4] = v[4]; + u[5] = v[5]; + + v[0] = _mm_mullo_epi32(v[2], cospi32); + x = _mm_mullo_epi32(v[3], cospi32); + u[2] = _mm_add_epi32(v[0], x); + u[2] = _mm_add_epi32(u[2], rnding); + u[2] = _mm_srai_epi32(u[2], bit); + + u[3] = _mm_sub_epi32(v[0], x); + u[3] = _mm_add_epi32(u[3], rnding); + u[3] = _mm_srai_epi32(u[3], bit); + + v[0] = _mm_mullo_epi32(v[6], cospi32); + x = _mm_mullo_epi32(v[7], cospi32); + u[6] = _mm_add_epi32(v[0], x); + u[6] = _mm_add_epi32(u[6], rnding); + u[6] = _mm_srai_epi32(u[6], bit); + + u[7] = _mm_sub_epi32(v[0], x); + u[7] = _mm_add_epi32(u[7], rnding); + u[7] = _mm_srai_epi32(u[7], bit); + + // stage 7 + if (do_cols) { + out[1] = u[0]; + out[3] = _mm_sub_epi32(kZero, u[4]); + out[5] = u[6]; + out[7] = _mm_sub_epi32(kZero, u[2]); + out[9] = u[3]; + out[11] = _mm_sub_epi32(kZero, u[7]); + out[13] = u[5]; + out[15] = _mm_sub_epi32(kZero, u[1]); + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m128i clamp_lo_out = _mm_set1_epi32(-(1 << (log_range_out - 1))); + const __m128i clamp_hi_out = _mm_set1_epi32((1 << (log_range_out - 1)) - 1); + + neg_shift_sse4_1(u[0], u[4], out + 1, out + 3, &clamp_lo_out, &clamp_hi_out, + out_shift); + neg_shift_sse4_1(u[6], u[2], out + 5, out + 7, &clamp_lo_out, &clamp_hi_out, + out_shift); + neg_shift_sse4_1(u[3], u[7], out + 9, out + 11, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(u[5], u[1], out + 13, out + 15, &clamp_lo_out, + &clamp_hi_out, out_shift); + } +} + +static void round_shift_8x8(__m128i *in, int shift) { + round_shift_4x4(&in[0], shift); + round_shift_4x4(&in[4], shift); + round_shift_4x4(&in[8], shift); + round_shift_4x4(&in[12], shift); +} + +static __m128i get_recon_8x8(const __m128i pred, __m128i res_lo, __m128i res_hi, + int fliplr, int bd) { + __m128i x0, x1; + const __m128i zero = _mm_setzero_si128(); + + x0 = _mm_unpacklo_epi16(pred, zero); + x1 = _mm_unpackhi_epi16(pred, zero); + + if (fliplr) { + res_lo = _mm_shuffle_epi32(res_lo, 0x1B); + res_hi = _mm_shuffle_epi32(res_hi, 0x1B); + x0 = _mm_add_epi32(res_hi, x0); + x1 = _mm_add_epi32(res_lo, x1); + + } else { + x0 = _mm_add_epi32(res_lo, x0); + x1 = _mm_add_epi32(res_hi, x1); + } + + x0 = _mm_packus_epi32(x0, x1); + return highbd_clamp_epi16(x0, bd); +} + +static void write_buffer_8x8(__m128i *in, uint16_t *output, int stride, + int fliplr, int flipud, int shift, int bd) { + __m128i u0, u1, u2, u3, u4, u5, u6, u7; + __m128i v0, v1, v2, v3, v4, v5, v6, v7; + + round_shift_8x8(in, shift); + + v0 = _mm_load_si128((__m128i const *)(output + 0 * stride)); + v1 = _mm_load_si128((__m128i const *)(output + 1 * stride)); + v2 = _mm_load_si128((__m128i const *)(output + 2 * stride)); + v3 = _mm_load_si128((__m128i const *)(output + 3 * stride)); + v4 = _mm_load_si128((__m128i const *)(output + 4 * stride)); + v5 = _mm_load_si128((__m128i const *)(output + 5 * stride)); + v6 = _mm_load_si128((__m128i const *)(output + 6 * stride)); + v7 = _mm_load_si128((__m128i const *)(output + 7 * stride)); + + if (flipud) { + u0 = get_recon_8x8(v0, in[14], in[15], fliplr, bd); + u1 = get_recon_8x8(v1, in[12], in[13], fliplr, bd); + u2 = get_recon_8x8(v2, in[10], in[11], fliplr, bd); + u3 = get_recon_8x8(v3, in[8], in[9], fliplr, bd); + u4 = get_recon_8x8(v4, in[6], in[7], fliplr, bd); + u5 = get_recon_8x8(v5, in[4], in[5], fliplr, bd); + u6 = get_recon_8x8(v6, in[2], in[3], fliplr, bd); + u7 = get_recon_8x8(v7, in[0], in[1], fliplr, bd); + } else { + u0 = get_recon_8x8(v0, in[0], in[1], fliplr, bd); + u1 = get_recon_8x8(v1, in[2], in[3], fliplr, bd); + u2 = get_recon_8x8(v2, in[4], in[5], fliplr, bd); + u3 = get_recon_8x8(v3, in[6], in[7], fliplr, bd); + u4 = get_recon_8x8(v4, in[8], in[9], fliplr, bd); + u5 = get_recon_8x8(v5, in[10], in[11], fliplr, bd); + u6 = get_recon_8x8(v6, in[12], in[13], fliplr, bd); + u7 = get_recon_8x8(v7, in[14], in[15], fliplr, bd); + } + + _mm_store_si128((__m128i *)(output + 0 * stride), u0); + _mm_store_si128((__m128i *)(output + 1 * stride), u1); + _mm_store_si128((__m128i *)(output + 2 * stride), u2); + _mm_store_si128((__m128i *)(output + 3 * stride), u3); + _mm_store_si128((__m128i *)(output + 4 * stride), u4); + _mm_store_si128((__m128i *)(output + 5 * stride), u5); + _mm_store_si128((__m128i *)(output + 6 * stride), u6); + _mm_store_si128((__m128i *)(output + 7 * stride), u7); +} + +void av1_inv_txfm2d_add_8x8_sse4_1(const int32_t *coeff, uint16_t *output, + int stride, TX_TYPE tx_type, int bd) { + __m128i in[16], out[16]; + const int8_t *shift = inv_txfm_shift_ls[TX_8X8]; + const int txw_idx = get_txw_idx(TX_8X8); + const int txh_idx = get_txh_idx(TX_8X8); + + switch (tx_type) { + case DCT_DCT: + load_buffer_8x8(coeff, in); + transpose_8x8(in, out); + idct8x8_sse4_1(out, in, inv_cos_bit_row[txw_idx][txh_idx], 0, bd, + -shift[0]); + transpose_8x8(in, out); + idct8x8_sse4_1(out, in, inv_cos_bit_col[txw_idx][txh_idx], 1, bd, 0); + write_buffer_8x8(in, output, stride, 0, 0, -shift[1], bd); + break; + case DCT_ADST: + load_buffer_8x8(coeff, in); + transpose_8x8(in, out); + iadst8x8_sse4_1(out, in, inv_cos_bit_row[txw_idx][txh_idx], 0, bd, + -shift[0]); + transpose_8x8(in, out); + idct8x8_sse4_1(out, in, inv_cos_bit_col[txw_idx][txh_idx], 1, bd, 0); + write_buffer_8x8(in, output, stride, 0, 0, -shift[1], bd); + break; + case ADST_DCT: + load_buffer_8x8(coeff, in); + transpose_8x8(in, out); + idct8x8_sse4_1(out, in, inv_cos_bit_row[txw_idx][txh_idx], 0, bd, + -shift[0]); + transpose_8x8(in, out); + iadst8x8_sse4_1(out, in, inv_cos_bit_col[txw_idx][txh_idx], 1, bd, 0); + write_buffer_8x8(in, output, stride, 0, 0, -shift[1], bd); + break; + case ADST_ADST: + load_buffer_8x8(coeff, in); + transpose_8x8(in, out); + iadst8x8_sse4_1(out, in, inv_cos_bit_row[txw_idx][txh_idx], 0, bd, + -shift[0]); + transpose_8x8(in, out); + iadst8x8_sse4_1(out, in, inv_cos_bit_col[txw_idx][txh_idx], 1, bd, 0); + write_buffer_8x8(in, output, stride, 0, 0, -shift[1], bd); + break; + case FLIPADST_DCT: + load_buffer_8x8(coeff, in); + transpose_8x8(in, out); + idct8x8_sse4_1(out, in, inv_cos_bit_row[txw_idx][txh_idx], 0, bd, + -shift[0]); + transpose_8x8(in, out); + iadst8x8_sse4_1(out, in, inv_cos_bit_col[txw_idx][txh_idx], 1, bd, 0); + write_buffer_8x8(in, output, stride, 0, 1, -shift[1], bd); + break; + case DCT_FLIPADST: + load_buffer_8x8(coeff, in); + transpose_8x8(in, out); + iadst8x8_sse4_1(out, in, inv_cos_bit_row[txw_idx][txh_idx], 0, bd, + -shift[0]); + transpose_8x8(in, out); + idct8x8_sse4_1(out, in, inv_cos_bit_col[txw_idx][txh_idx], 1, bd, 0); + write_buffer_8x8(in, output, stride, 1, 0, -shift[1], bd); + break; + case ADST_FLIPADST: + load_buffer_8x8(coeff, in); + transpose_8x8(in, out); + iadst8x8_sse4_1(out, in, inv_cos_bit_row[txw_idx][txh_idx], 0, bd, + -shift[0]); + transpose_8x8(in, out); + iadst8x8_sse4_1(out, in, inv_cos_bit_col[txw_idx][txh_idx], 1, bd, 0); + write_buffer_8x8(in, output, stride, 1, 0, -shift[1], bd); + break; + case FLIPADST_FLIPADST: + load_buffer_8x8(coeff, in); + transpose_8x8(in, out); + iadst8x8_sse4_1(out, in, inv_cos_bit_row[txw_idx][txh_idx], 0, bd, + -shift[0]); + transpose_8x8(in, out); + iadst8x8_sse4_1(out, in, inv_cos_bit_col[txw_idx][txh_idx], 1, bd, 0); + write_buffer_8x8(in, output, stride, 1, 1, -shift[1], bd); + break; + case FLIPADST_ADST: + load_buffer_8x8(coeff, in); + transpose_8x8(in, out); + iadst8x8_sse4_1(out, in, inv_cos_bit_row[txw_idx][txh_idx], 0, bd, + -shift[0]); + transpose_8x8(in, out); + iadst8x8_sse4_1(out, in, inv_cos_bit_col[txw_idx][txh_idx], 1, bd, 0); + write_buffer_8x8(in, output, stride, 0, 1, -shift[1], bd); + break; + default: assert(0); + } +} + +static void idct8x8_low1_sse4_1(__m128i *in, __m128i *out, int bit, int do_cols, + int bd, int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + const __m128i rnding = _mm_set1_epi32(1 << (bit - 1)); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + __m128i x; + + // stage 0 + // stage 1 + // stage 2 + // stage 3 + x = _mm_mullo_epi32(in[0], cospi32); + x = _mm_add_epi32(x, rnding); + x = _mm_srai_epi32(x, bit); + + // stage 4 + // stage 5 + if (!do_cols) { + const int log_range_out = AOMMAX(16, bd + 6); + const __m128i clamp_lo_out = _mm_set1_epi32(AOMMAX( + -(1 << (log_range_out - 1)), -(1 << (log_range - 1 - out_shift)))); + const __m128i clamp_hi_out = _mm_set1_epi32(AOMMIN( + (1 << (log_range_out - 1)) - 1, (1 << (log_range - 1 - out_shift)))); + + __m128i offset = _mm_set1_epi32((1 << out_shift) >> 1); + x = _mm_add_epi32(x, offset); + x = _mm_sra_epi32(x, _mm_cvtsi32_si128(out_shift)); + x = _mm_max_epi32(x, clamp_lo_out); + x = _mm_min_epi32(x, clamp_hi_out); + } + + out[0] = x; + out[1] = x; + out[2] = x; + out[3] = x; + out[4] = x; + out[5] = x; + out[6] = x; + out[7] = x; +} + +static void idct8x8_new_sse4_1(__m128i *in, __m128i *out, int bit, int do_cols, + int bd, int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m128i cospi56 = _mm_set1_epi32(cospi[56]); + const __m128i cospim8 = _mm_set1_epi32(-cospi[8]); + const __m128i cospi24 = _mm_set1_epi32(cospi[24]); + const __m128i cospim40 = _mm_set1_epi32(-cospi[40]); + const __m128i cospi40 = _mm_set1_epi32(cospi[40]); + const __m128i cospi8 = _mm_set1_epi32(cospi[8]); + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + const __m128i cospi48 = _mm_set1_epi32(cospi[48]); + const __m128i cospim16 = _mm_set1_epi32(-cospi[16]); + const __m128i cospi16 = _mm_set1_epi32(cospi[16]); + const __m128i rnding = _mm_set1_epi32(1 << (bit - 1)); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m128i clamp_lo = _mm_set1_epi32(-(1 << (log_range - 1))); + const __m128i clamp_hi = _mm_set1_epi32((1 << (log_range - 1)) - 1); + __m128i u0, u1, u2, u3, u4, u5, u6, u7; + __m128i v0, v1, v2, v3, v4, v5, v6, v7; + __m128i x, y; + + // stage 0 + // stage 1 + // stage 2 + u0 = in[0]; + u1 = in[4]; + u2 = in[2]; + u3 = in[6]; + + x = _mm_mullo_epi32(in[1], cospi56); + y = _mm_mullo_epi32(in[7], cospim8); + u4 = _mm_add_epi32(x, y); + u4 = _mm_add_epi32(u4, rnding); + u4 = _mm_srai_epi32(u4, bit); + + x = _mm_mullo_epi32(in[1], cospi8); + y = _mm_mullo_epi32(in[7], cospi56); + u7 = _mm_add_epi32(x, y); + u7 = _mm_add_epi32(u7, rnding); + u7 = _mm_srai_epi32(u7, bit); + + x = _mm_mullo_epi32(in[5], cospi24); + y = _mm_mullo_epi32(in[3], cospim40); + u5 = _mm_add_epi32(x, y); + u5 = _mm_add_epi32(u5, rnding); + u5 = _mm_srai_epi32(u5, bit); + + x = _mm_mullo_epi32(in[5], cospi40); + y = _mm_mullo_epi32(in[3], cospi24); + u6 = _mm_add_epi32(x, y); + u6 = _mm_add_epi32(u6, rnding); + u6 = _mm_srai_epi32(u6, bit); + + // stage 3 + x = _mm_mullo_epi32(u0, cospi32); + y = _mm_mullo_epi32(u1, cospi32); + v0 = _mm_add_epi32(x, y); + v0 = _mm_add_epi32(v0, rnding); + v0 = _mm_srai_epi32(v0, bit); + + v1 = _mm_sub_epi32(x, y); + v1 = _mm_add_epi32(v1, rnding); + v1 = _mm_srai_epi32(v1, bit); + + x = _mm_mullo_epi32(u2, cospi48); + y = _mm_mullo_epi32(u3, cospim16); + v2 = _mm_add_epi32(x, y); + v2 = _mm_add_epi32(v2, rnding); + v2 = _mm_srai_epi32(v2, bit); + + x = _mm_mullo_epi32(u2, cospi16); + y = _mm_mullo_epi32(u3, cospi48); + v3 = _mm_add_epi32(x, y); + v3 = _mm_add_epi32(v3, rnding); + v3 = _mm_srai_epi32(v3, bit); + + addsub_sse4_1(u4, u5, &v4, &v5, &clamp_lo, &clamp_hi); + addsub_sse4_1(u7, u6, &v7, &v6, &clamp_lo, &clamp_hi); + + // stage 4 + addsub_sse4_1(v0, v3, &u0, &u3, &clamp_lo, &clamp_hi); + addsub_sse4_1(v1, v2, &u1, &u2, &clamp_lo, &clamp_hi); + u4 = v4; + u7 = v7; + + x = _mm_mullo_epi32(v5, cospi32); + y = _mm_mullo_epi32(v6, cospi32); + u6 = _mm_add_epi32(y, x); + u6 = _mm_add_epi32(u6, rnding); + u6 = _mm_srai_epi32(u6, bit); + + u5 = _mm_sub_epi32(y, x); + u5 = _mm_add_epi32(u5, rnding); + u5 = _mm_srai_epi32(u5, bit); + + // stage 5 + if (do_cols) { + addsub_no_clamp_sse4_1(u0, u7, out + 0, out + 7); + addsub_no_clamp_sse4_1(u1, u6, out + 1, out + 6); + addsub_no_clamp_sse4_1(u2, u5, out + 2, out + 5); + addsub_no_clamp_sse4_1(u3, u4, out + 3, out + 4); + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m128i clamp_lo_out = _mm_set1_epi32(AOMMAX( + -(1 << (log_range_out - 1)), -(1 << (log_range - 1 - out_shift)))); + const __m128i clamp_hi_out = _mm_set1_epi32(AOMMIN( + (1 << (log_range_out - 1)) - 1, (1 << (log_range - 1 - out_shift)))); + addsub_shift_sse4_1(u0, u7, out + 0, out + 7, &clamp_lo_out, &clamp_hi_out, + out_shift); + addsub_shift_sse4_1(u1, u6, out + 1, out + 6, &clamp_lo_out, &clamp_hi_out, + out_shift); + addsub_shift_sse4_1(u2, u5, out + 2, out + 5, &clamp_lo_out, &clamp_hi_out, + out_shift); + addsub_shift_sse4_1(u3, u4, out + 3, out + 4, &clamp_lo_out, &clamp_hi_out, + out_shift); + } +} + +static void iadst8x8_low1_sse4_1(__m128i *in, __m128i *out, int bit, + int do_cols, int bd, int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m128i cospi4 = _mm_set1_epi32(cospi[4]); + const __m128i cospi60 = _mm_set1_epi32(cospi[60]); + const __m128i cospi16 = _mm_set1_epi32(cospi[16]); + const __m128i cospi48 = _mm_set1_epi32(cospi[48]); + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + const __m128i rnding = _mm_set1_epi32(1 << (bit - 1)); + const __m128i kZero = _mm_setzero_si128(); + __m128i u[8], x; + + // stage 0 + // stage 1 + // stage 2 + + x = _mm_mullo_epi32(in[0], cospi60); + u[0] = _mm_add_epi32(x, rnding); + u[0] = _mm_srai_epi32(u[0], bit); + + x = _mm_mullo_epi32(in[0], cospi4); + u[1] = _mm_sub_epi32(kZero, x); + u[1] = _mm_add_epi32(u[1], rnding); + u[1] = _mm_srai_epi32(u[1], bit); + + // stage 3 + // stage 4 + __m128i temp1, temp2; + temp1 = _mm_mullo_epi32(u[0], cospi16); + x = _mm_mullo_epi32(u[1], cospi48); + temp1 = _mm_add_epi32(temp1, x); + temp1 = _mm_add_epi32(temp1, rnding); + temp1 = _mm_srai_epi32(temp1, bit); + u[4] = temp1; + + temp2 = _mm_mullo_epi32(u[0], cospi48); + x = _mm_mullo_epi32(u[1], cospi16); + u[5] = _mm_sub_epi32(temp2, x); + u[5] = _mm_add_epi32(u[5], rnding); + u[5] = _mm_srai_epi32(u[5], bit); + + // stage 5 + // stage 6 + temp1 = _mm_mullo_epi32(u[0], cospi32); + x = _mm_mullo_epi32(u[1], cospi32); + u[2] = _mm_add_epi32(temp1, x); + u[2] = _mm_add_epi32(u[2], rnding); + u[2] = _mm_srai_epi32(u[2], bit); + + u[3] = _mm_sub_epi32(temp1, x); + u[3] = _mm_add_epi32(u[3], rnding); + u[3] = _mm_srai_epi32(u[3], bit); + + temp1 = _mm_mullo_epi32(u[4], cospi32); + x = _mm_mullo_epi32(u[5], cospi32); + u[6] = _mm_add_epi32(temp1, x); + u[6] = _mm_add_epi32(u[6], rnding); + u[6] = _mm_srai_epi32(u[6], bit); + + u[7] = _mm_sub_epi32(temp1, x); + u[7] = _mm_add_epi32(u[7], rnding); + u[7] = _mm_srai_epi32(u[7], bit); + + // stage 7 + if (do_cols) { + out[0] = u[0]; + out[1] = _mm_sub_epi32(kZero, u[4]); + out[2] = u[6]; + out[3] = _mm_sub_epi32(kZero, u[2]); + out[4] = u[3]; + out[5] = _mm_sub_epi32(kZero, u[7]); + out[6] = u[5]; + out[7] = _mm_sub_epi32(kZero, u[1]); + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m128i clamp_lo_out = _mm_set1_epi32(-(1 << (log_range_out - 1))); + const __m128i clamp_hi_out = _mm_set1_epi32((1 << (log_range_out - 1)) - 1); + + neg_shift_sse4_1(u[0], u[4], out + 0, out + 1, &clamp_lo_out, &clamp_hi_out, + out_shift); + neg_shift_sse4_1(u[6], u[2], out + 2, out + 3, &clamp_lo_out, &clamp_hi_out, + out_shift); + neg_shift_sse4_1(u[3], u[7], out + 4, out + 5, &clamp_lo_out, &clamp_hi_out, + out_shift); + neg_shift_sse4_1(u[5], u[1], out + 6, out + 7, &clamp_lo_out, &clamp_hi_out, + out_shift); + } +} + +static void iadst8x8_new_sse4_1(__m128i *in, __m128i *out, int bit, int do_cols, + int bd, int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m128i cospi4 = _mm_set1_epi32(cospi[4]); + const __m128i cospi60 = _mm_set1_epi32(cospi[60]); + const __m128i cospi20 = _mm_set1_epi32(cospi[20]); + const __m128i cospi44 = _mm_set1_epi32(cospi[44]); + const __m128i cospi36 = _mm_set1_epi32(cospi[36]); + const __m128i cospi28 = _mm_set1_epi32(cospi[28]); + const __m128i cospi52 = _mm_set1_epi32(cospi[52]); + const __m128i cospi12 = _mm_set1_epi32(cospi[12]); + const __m128i cospi16 = _mm_set1_epi32(cospi[16]); + const __m128i cospi48 = _mm_set1_epi32(cospi[48]); + const __m128i cospim48 = _mm_set1_epi32(-cospi[48]); + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + const __m128i rnding = _mm_set1_epi32(1 << (bit - 1)); + const __m128i kZero = _mm_setzero_si128(); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m128i clamp_lo = _mm_set1_epi32(-(1 << (log_range - 1))); + const __m128i clamp_hi = _mm_set1_epi32((1 << (log_range - 1)) - 1); + __m128i u[8], v[8], x; + + // stage 0 + // stage 1 + // stage 2 + + u[0] = _mm_mullo_epi32(in[7], cospi4); + x = _mm_mullo_epi32(in[0], cospi60); + u[0] = _mm_add_epi32(u[0], x); + u[0] = _mm_add_epi32(u[0], rnding); + u[0] = _mm_srai_epi32(u[0], bit); + + u[1] = _mm_mullo_epi32(in[7], cospi60); + x = _mm_mullo_epi32(in[0], cospi4); + u[1] = _mm_sub_epi32(u[1], x); + u[1] = _mm_add_epi32(u[1], rnding); + u[1] = _mm_srai_epi32(u[1], bit); + + // (2) + u[2] = _mm_mullo_epi32(in[5], cospi20); + x = _mm_mullo_epi32(in[2], cospi44); + u[2] = _mm_add_epi32(u[2], x); + u[2] = _mm_add_epi32(u[2], rnding); + u[2] = _mm_srai_epi32(u[2], bit); + + u[3] = _mm_mullo_epi32(in[5], cospi44); + x = _mm_mullo_epi32(in[2], cospi20); + u[3] = _mm_sub_epi32(u[3], x); + u[3] = _mm_add_epi32(u[3], rnding); + u[3] = _mm_srai_epi32(u[3], bit); + + // (3) + u[4] = _mm_mullo_epi32(in[3], cospi36); + x = _mm_mullo_epi32(in[4], cospi28); + u[4] = _mm_add_epi32(u[4], x); + u[4] = _mm_add_epi32(u[4], rnding); + u[4] = _mm_srai_epi32(u[4], bit); + + u[5] = _mm_mullo_epi32(in[3], cospi28); + x = _mm_mullo_epi32(in[4], cospi36); + u[5] = _mm_sub_epi32(u[5], x); + u[5] = _mm_add_epi32(u[5], rnding); + u[5] = _mm_srai_epi32(u[5], bit); + + // (4) + u[6] = _mm_mullo_epi32(in[1], cospi52); + x = _mm_mullo_epi32(in[6], cospi12); + u[6] = _mm_add_epi32(u[6], x); + u[6] = _mm_add_epi32(u[6], rnding); + u[6] = _mm_srai_epi32(u[6], bit); + + u[7] = _mm_mullo_epi32(in[1], cospi12); + x = _mm_mullo_epi32(in[6], cospi52); + u[7] = _mm_sub_epi32(u[7], x); + u[7] = _mm_add_epi32(u[7], rnding); + u[7] = _mm_srai_epi32(u[7], bit); + + // stage 3 + addsub_sse4_1(u[0], u[4], &v[0], &v[4], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[1], u[5], &v[1], &v[5], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[2], u[6], &v[2], &v[6], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[3], u[7], &v[3], &v[7], &clamp_lo, &clamp_hi); + + // stage 4 + u[0] = v[0]; + u[1] = v[1]; + u[2] = v[2]; + u[3] = v[3]; + + u[4] = _mm_mullo_epi32(v[4], cospi16); + x = _mm_mullo_epi32(v[5], cospi48); + u[4] = _mm_add_epi32(u[4], x); + u[4] = _mm_add_epi32(u[4], rnding); + u[4] = _mm_srai_epi32(u[4], bit); + + u[5] = _mm_mullo_epi32(v[4], cospi48); + x = _mm_mullo_epi32(v[5], cospi16); + u[5] = _mm_sub_epi32(u[5], x); + u[5] = _mm_add_epi32(u[5], rnding); + u[5] = _mm_srai_epi32(u[5], bit); + + u[6] = _mm_mullo_epi32(v[6], cospim48); + x = _mm_mullo_epi32(v[7], cospi16); + u[6] = _mm_add_epi32(u[6], x); + u[6] = _mm_add_epi32(u[6], rnding); + u[6] = _mm_srai_epi32(u[6], bit); + + u[7] = _mm_mullo_epi32(v[6], cospi16); + x = _mm_mullo_epi32(v[7], cospim48); + u[7] = _mm_sub_epi32(u[7], x); + u[7] = _mm_add_epi32(u[7], rnding); + u[7] = _mm_srai_epi32(u[7], bit); + + // stage 5 + addsub_sse4_1(u[0], u[2], &v[0], &v[2], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[1], u[3], &v[1], &v[3], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[4], u[6], &v[4], &v[6], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[5], u[7], &v[5], &v[7], &clamp_lo, &clamp_hi); + + // stage 6 + u[0] = v[0]; + u[1] = v[1]; + u[4] = v[4]; + u[5] = v[5]; + + v[0] = _mm_mullo_epi32(v[2], cospi32); + x = _mm_mullo_epi32(v[3], cospi32); + u[2] = _mm_add_epi32(v[0], x); + u[2] = _mm_add_epi32(u[2], rnding); + u[2] = _mm_srai_epi32(u[2], bit); + + u[3] = _mm_sub_epi32(v[0], x); + u[3] = _mm_add_epi32(u[3], rnding); + u[3] = _mm_srai_epi32(u[3], bit); + + v[0] = _mm_mullo_epi32(v[6], cospi32); + x = _mm_mullo_epi32(v[7], cospi32); + u[6] = _mm_add_epi32(v[0], x); + u[6] = _mm_add_epi32(u[6], rnding); + u[6] = _mm_srai_epi32(u[6], bit); + + u[7] = _mm_sub_epi32(v[0], x); + u[7] = _mm_add_epi32(u[7], rnding); + u[7] = _mm_srai_epi32(u[7], bit); + + // stage 7 + if (do_cols) { + out[0] = u[0]; + out[1] = _mm_sub_epi32(kZero, u[4]); + out[2] = u[6]; + out[3] = _mm_sub_epi32(kZero, u[2]); + out[4] = u[3]; + out[5] = _mm_sub_epi32(kZero, u[7]); + out[6] = u[5]; + out[7] = _mm_sub_epi32(kZero, u[1]); + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m128i clamp_lo_out = _mm_set1_epi32(-(1 << (log_range_out - 1))); + const __m128i clamp_hi_out = _mm_set1_epi32((1 << (log_range_out - 1)) - 1); + + neg_shift_sse4_1(u[0], u[4], out + 0, out + 1, &clamp_lo_out, &clamp_hi_out, + out_shift); + neg_shift_sse4_1(u[6], u[2], out + 2, out + 3, &clamp_lo_out, &clamp_hi_out, + out_shift); + neg_shift_sse4_1(u[3], u[7], out + 4, out + 5, &clamp_lo_out, &clamp_hi_out, + out_shift); + neg_shift_sse4_1(u[5], u[1], out + 6, out + 7, &clamp_lo_out, &clamp_hi_out, + out_shift); + } +} + +static void idct16x16_low1_sse4_1(__m128i *in, __m128i *out, int bit, + int do_cols, int bd, int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + const __m128i rnding = _mm_set1_epi32(1 << (bit - 1)); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m128i clamp_lo = _mm_set1_epi32(-(1 << (log_range - 1))); + const __m128i clamp_hi = _mm_set1_epi32((1 << (log_range - 1)) - 1); + + { + // stage 0 + // stage 1 + // stage 2 + // stage 3 + // stage 4 + in[0] = _mm_mullo_epi32(in[0], cospi32); + in[0] = _mm_add_epi32(in[0], rnding); + in[0] = _mm_srai_epi32(in[0], bit); + + // stage 5 + // stage 6 + // stage 7 + if (do_cols) { + in[0] = _mm_max_epi32(in[0], clamp_lo); + in[0] = _mm_min_epi32(in[0], clamp_hi); + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m128i clamp_lo_out = _mm_set1_epi32(AOMMAX( + -(1 << (log_range_out - 1)), -(1 << (log_range - 1 - out_shift)))); + const __m128i clamp_hi_out = _mm_set1_epi32(AOMMIN( + (1 << (log_range_out - 1)) - 1, (1 << (log_range - 1 - out_shift)))); + __m128i offset = _mm_set1_epi32((1 << out_shift) >> 1); + in[0] = _mm_add_epi32(in[0], offset); + in[0] = _mm_sra_epi32(in[0], _mm_cvtsi32_si128(out_shift)); + in[0] = _mm_max_epi32(in[0], clamp_lo_out); + in[0] = _mm_min_epi32(in[0], clamp_hi_out); + } + + out[0] = in[0]; + out[1] = in[0]; + out[2] = in[0]; + out[3] = in[0]; + out[4] = in[0]; + out[5] = in[0]; + out[6] = in[0]; + out[7] = in[0]; + out[8] = in[0]; + out[9] = in[0]; + out[10] = in[0]; + out[11] = in[0]; + out[12] = in[0]; + out[13] = in[0]; + out[14] = in[0]; + out[15] = in[0]; + } +} + +static void idct16x16_low8_sse4_1(__m128i *in, __m128i *out, int bit, + int do_cols, int bd, int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m128i cospi60 = _mm_set1_epi32(cospi[60]); + const __m128i cospi28 = _mm_set1_epi32(cospi[28]); + const __m128i cospi44 = _mm_set1_epi32(cospi[44]); + const __m128i cospi20 = _mm_set1_epi32(cospi[20]); + const __m128i cospi12 = _mm_set1_epi32(cospi[12]); + const __m128i cospi4 = _mm_set1_epi32(cospi[4]); + const __m128i cospi56 = _mm_set1_epi32(cospi[56]); + const __m128i cospi24 = _mm_set1_epi32(cospi[24]); + const __m128i cospim40 = _mm_set1_epi32(-cospi[40]); + const __m128i cospi8 = _mm_set1_epi32(cospi[8]); + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + const __m128i cospi48 = _mm_set1_epi32(cospi[48]); + const __m128i cospi16 = _mm_set1_epi32(cospi[16]); + const __m128i cospim16 = _mm_set1_epi32(-cospi[16]); + const __m128i cospim48 = _mm_set1_epi32(-cospi[48]); + const __m128i cospim36 = _mm_set1_epi32(-cospi[36]); + const __m128i cospim52 = _mm_set1_epi32(-cospi[52]); + const __m128i rnding = _mm_set1_epi32(1 << (bit - 1)); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m128i clamp_lo = _mm_set1_epi32(-(1 << (log_range - 1))); + const __m128i clamp_hi = _mm_set1_epi32((1 << (log_range - 1)) - 1); + __m128i u[16], x, y; + + { + // stage 0 + // stage 1 + u[0] = in[0]; + u[2] = in[4]; + u[4] = in[2]; + u[6] = in[6]; + u[8] = in[1]; + u[10] = in[5]; + u[12] = in[3]; + u[14] = in[7]; + + // stage 2 + u[15] = half_btf_0_sse4_1(&cospi4, &u[8], &rnding, bit); + u[8] = half_btf_0_sse4_1(&cospi60, &u[8], &rnding, bit); + + u[9] = half_btf_0_sse4_1(&cospim36, &u[14], &rnding, bit); + u[14] = half_btf_0_sse4_1(&cospi28, &u[14], &rnding, bit); + + u[13] = half_btf_0_sse4_1(&cospi20, &u[10], &rnding, bit); + u[10] = half_btf_0_sse4_1(&cospi44, &u[10], &rnding, bit); + + u[11] = half_btf_0_sse4_1(&cospim52, &u[12], &rnding, bit); + u[12] = half_btf_0_sse4_1(&cospi12, &u[12], &rnding, bit); + + // stage 3 + u[7] = half_btf_0_sse4_1(&cospi8, &u[4], &rnding, bit); + u[4] = half_btf_0_sse4_1(&cospi56, &u[4], &rnding, bit); + u[5] = half_btf_0_sse4_1(&cospim40, &u[6], &rnding, bit); + u[6] = half_btf_0_sse4_1(&cospi24, &u[6], &rnding, bit); + + addsub_sse4_1(u[8], u[9], &u[8], &u[9], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[11], u[10], &u[11], &u[10], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[12], u[13], &u[12], &u[13], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[15], u[14], &u[15], &u[14], &clamp_lo, &clamp_hi); + + // stage 4 + x = _mm_mullo_epi32(u[0], cospi32); + u[0] = _mm_add_epi32(x, rnding); + u[0] = _mm_srai_epi32(u[0], bit); + u[1] = u[0]; + + u[3] = half_btf_0_sse4_1(&cospi16, &u[2], &rnding, bit); + u[2] = half_btf_0_sse4_1(&cospi48, &u[2], &rnding, bit); + + addsub_sse4_1(u[4], u[5], &u[4], &u[5], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[7], u[6], &u[7], &u[6], &clamp_lo, &clamp_hi); + + x = half_btf_sse4_1(&cospim16, &u[9], &cospi48, &u[14], &rnding, bit); + u[14] = half_btf_sse4_1(&cospi48, &u[9], &cospi16, &u[14], &rnding, bit); + u[9] = x; + y = half_btf_sse4_1(&cospim48, &u[10], &cospim16, &u[13], &rnding, bit); + u[13] = half_btf_sse4_1(&cospim16, &u[10], &cospi48, &u[13], &rnding, bit); + u[10] = y; + + // stage 5 + addsub_sse4_1(u[0], u[3], &u[0], &u[3], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[1], u[2], &u[1], &u[2], &clamp_lo, &clamp_hi); + + x = _mm_mullo_epi32(u[5], cospi32); + y = _mm_mullo_epi32(u[6], cospi32); + u[5] = _mm_sub_epi32(y, x); + u[5] = _mm_add_epi32(u[5], rnding); + u[5] = _mm_srai_epi32(u[5], bit); + + u[6] = _mm_add_epi32(y, x); + u[6] = _mm_add_epi32(u[6], rnding); + u[6] = _mm_srai_epi32(u[6], bit); + + addsub_sse4_1(u[8], u[11], &u[8], &u[11], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[9], u[10], &u[9], &u[10], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[15], u[12], &u[15], &u[12], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[14], u[13], &u[14], &u[13], &clamp_lo, &clamp_hi); + + // stage 6 + addsub_sse4_1(u[0], u[7], &u[0], &u[7], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[1], u[6], &u[1], &u[6], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[2], u[5], &u[2], &u[5], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[3], u[4], &u[3], &u[4], &clamp_lo, &clamp_hi); + + x = _mm_mullo_epi32(u[10], cospi32); + y = _mm_mullo_epi32(u[13], cospi32); + u[10] = _mm_sub_epi32(y, x); + u[10] = _mm_add_epi32(u[10], rnding); + u[10] = _mm_srai_epi32(u[10], bit); + + u[13] = _mm_add_epi32(x, y); + u[13] = _mm_add_epi32(u[13], rnding); + u[13] = _mm_srai_epi32(u[13], bit); + + x = _mm_mullo_epi32(u[11], cospi32); + y = _mm_mullo_epi32(u[12], cospi32); + u[11] = _mm_sub_epi32(y, x); + u[11] = _mm_add_epi32(u[11], rnding); + u[11] = _mm_srai_epi32(u[11], bit); + + u[12] = _mm_add_epi32(x, y); + u[12] = _mm_add_epi32(u[12], rnding); + u[12] = _mm_srai_epi32(u[12], bit); + // stage 7 + if (do_cols) { + addsub_no_clamp_sse4_1(u[0], u[15], out + 0, out + 15); + addsub_no_clamp_sse4_1(u[1], u[14], out + 1, out + 14); + addsub_no_clamp_sse4_1(u[2], u[13], out + 2, out + 13); + addsub_no_clamp_sse4_1(u[3], u[12], out + 3, out + 12); + addsub_no_clamp_sse4_1(u[4], u[11], out + 4, out + 11); + addsub_no_clamp_sse4_1(u[5], u[10], out + 5, out + 10); + addsub_no_clamp_sse4_1(u[6], u[9], out + 6, out + 9); + addsub_no_clamp_sse4_1(u[7], u[8], out + 7, out + 8); + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m128i clamp_lo_out = _mm_set1_epi32(AOMMAX( + -(1 << (log_range_out - 1)), -(1 << (log_range - 1 - out_shift)))); + const __m128i clamp_hi_out = _mm_set1_epi32(AOMMIN( + (1 << (log_range_out - 1)) - 1, (1 << (log_range - 1 - out_shift)))); + + addsub_shift_sse4_1(u[0], u[15], out + 0, out + 15, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(u[1], u[14], out + 1, out + 14, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(u[2], u[13], out + 2, out + 13, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(u[3], u[12], out + 3, out + 12, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(u[4], u[11], out + 4, out + 11, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(u[5], u[10], out + 5, out + 10, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(u[6], u[9], out + 6, out + 9, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(u[7], u[8], out + 7, out + 8, &clamp_lo_out, + &clamp_hi_out, out_shift); + } + } +} + +static void iadst16x16_low1_sse4_1(__m128i *in, __m128i *out, int bit, + int do_cols, int bd, int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m128i cospi2 = _mm_set1_epi32(cospi[2]); + const __m128i cospi62 = _mm_set1_epi32(cospi[62]); + const __m128i cospi8 = _mm_set1_epi32(cospi[8]); + const __m128i cospi56 = _mm_set1_epi32(cospi[56]); + const __m128i cospi48 = _mm_set1_epi32(cospi[48]); + const __m128i cospi16 = _mm_set1_epi32(cospi[16]); + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + const __m128i rnding = _mm_set1_epi32(1 << (bit - 1)); + const __m128i zero = _mm_setzero_si128(); + __m128i v[16], x, y, temp1, temp2; + + // Calculate the column 0, 1, 2, 3 + { + // stage 0 + // stage 1 + // stage 2 + x = _mm_mullo_epi32(in[0], cospi62); + v[0] = _mm_add_epi32(x, rnding); + v[0] = _mm_srai_epi32(v[0], bit); + + x = _mm_mullo_epi32(in[0], cospi2); + v[1] = _mm_sub_epi32(zero, x); + v[1] = _mm_add_epi32(v[1], rnding); + v[1] = _mm_srai_epi32(v[1], bit); + + // stage 3 + v[8] = v[0]; + v[9] = v[1]; + + // stage 4 + temp1 = _mm_mullo_epi32(v[8], cospi8); + x = _mm_mullo_epi32(v[9], cospi56); + temp1 = _mm_add_epi32(temp1, x); + temp1 = _mm_add_epi32(temp1, rnding); + temp1 = _mm_srai_epi32(temp1, bit); + + temp2 = _mm_mullo_epi32(v[8], cospi56); + x = _mm_mullo_epi32(v[9], cospi8); + temp2 = _mm_sub_epi32(temp2, x); + temp2 = _mm_add_epi32(temp2, rnding); + temp2 = _mm_srai_epi32(temp2, bit); + v[8] = temp1; + v[9] = temp2; + + // stage 5 + v[4] = v[0]; + v[5] = v[1]; + v[12] = v[8]; + v[13] = v[9]; + + // stage 6 + temp1 = _mm_mullo_epi32(v[4], cospi16); + x = _mm_mullo_epi32(v[5], cospi48); + temp1 = _mm_add_epi32(temp1, x); + temp1 = _mm_add_epi32(temp1, rnding); + temp1 = _mm_srai_epi32(temp1, bit); + + temp2 = _mm_mullo_epi32(v[4], cospi48); + x = _mm_mullo_epi32(v[5], cospi16); + temp2 = _mm_sub_epi32(temp2, x); + temp2 = _mm_add_epi32(temp2, rnding); + temp2 = _mm_srai_epi32(temp2, bit); + v[4] = temp1; + v[5] = temp2; + + temp1 = _mm_mullo_epi32(v[12], cospi16); + x = _mm_mullo_epi32(v[13], cospi48); + temp1 = _mm_add_epi32(temp1, x); + temp1 = _mm_add_epi32(temp1, rnding); + temp1 = _mm_srai_epi32(temp1, bit); + + temp2 = _mm_mullo_epi32(v[12], cospi48); + x = _mm_mullo_epi32(v[13], cospi16); + temp2 = _mm_sub_epi32(temp2, x); + temp2 = _mm_add_epi32(temp2, rnding); + temp2 = _mm_srai_epi32(temp2, bit); + v[12] = temp1; + v[13] = temp2; + + // stage 7 + v[2] = v[0]; + v[3] = v[1]; + v[6] = v[4]; + v[7] = v[5]; + v[10] = v[8]; + v[11] = v[9]; + v[14] = v[12]; + v[15] = v[13]; + + // stage 8 + y = _mm_mullo_epi32(v[2], cospi32); + x = _mm_mullo_epi32(v[3], cospi32); + v[2] = _mm_add_epi32(y, x); + v[2] = _mm_add_epi32(v[2], rnding); + v[2] = _mm_srai_epi32(v[2], bit); + + v[3] = _mm_sub_epi32(y, x); + v[3] = _mm_add_epi32(v[3], rnding); + v[3] = _mm_srai_epi32(v[3], bit); + + y = _mm_mullo_epi32(v[6], cospi32); + x = _mm_mullo_epi32(v[7], cospi32); + v[6] = _mm_add_epi32(y, x); + v[6] = _mm_add_epi32(v[6], rnding); + v[6] = _mm_srai_epi32(v[6], bit); + + v[7] = _mm_sub_epi32(y, x); + v[7] = _mm_add_epi32(v[7], rnding); + v[7] = _mm_srai_epi32(v[7], bit); + + y = _mm_mullo_epi32(v[10], cospi32); + x = _mm_mullo_epi32(v[11], cospi32); + v[10] = _mm_add_epi32(y, x); + v[10] = _mm_add_epi32(v[10], rnding); + v[10] = _mm_srai_epi32(v[10], bit); + + v[11] = _mm_sub_epi32(y, x); + v[11] = _mm_add_epi32(v[11], rnding); + v[11] = _mm_srai_epi32(v[11], bit); + + y = _mm_mullo_epi32(v[14], cospi32); + x = _mm_mullo_epi32(v[15], cospi32); + v[14] = _mm_add_epi32(y, x); + v[14] = _mm_add_epi32(v[14], rnding); + v[14] = _mm_srai_epi32(v[14], bit); + + v[15] = _mm_sub_epi32(y, x); + v[15] = _mm_add_epi32(v[15], rnding); + v[15] = _mm_srai_epi32(v[15], bit); + + // stage 9 + if (do_cols) { + out[0] = v[0]; + out[1] = _mm_sub_epi32(_mm_setzero_si128(), v[8]); + out[2] = v[12]; + out[3] = _mm_sub_epi32(_mm_setzero_si128(), v[4]); + out[4] = v[6]; + out[5] = _mm_sub_epi32(_mm_setzero_si128(), v[14]); + out[6] = v[10]; + out[7] = _mm_sub_epi32(_mm_setzero_si128(), v[2]); + out[8] = v[3]; + out[9] = _mm_sub_epi32(_mm_setzero_si128(), v[11]); + out[10] = v[15]; + out[11] = _mm_sub_epi32(_mm_setzero_si128(), v[7]); + out[12] = v[5]; + out[13] = _mm_sub_epi32(_mm_setzero_si128(), v[13]); + out[14] = v[9]; + out[15] = _mm_sub_epi32(_mm_setzero_si128(), v[1]); + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m128i clamp_lo_out = _mm_set1_epi32(-(1 << (log_range_out - 1))); + const __m128i clamp_hi_out = + _mm_set1_epi32((1 << (log_range_out - 1)) - 1); + + neg_shift_sse4_1(v[0], v[8], out + 0, out + 1, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(v[12], v[4], out + 2, out + 3, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(v[6], v[14], out + 4, out + 5, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(v[10], v[2], out + 6, out + 7, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(v[3], v[11], out + 8, out + 9, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(v[15], v[7], out + 10, out + 11, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(v[5], v[13], out + 12, out + 13, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(v[9], v[1], out + 14, out + 15, &clamp_lo_out, + &clamp_hi_out, out_shift); + } + } +} + +static void iadst16x16_low8_sse4_1(__m128i *in, __m128i *out, int bit, + int do_cols, int bd, int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m128i cospi2 = _mm_set1_epi32(cospi[2]); + const __m128i cospi62 = _mm_set1_epi32(cospi[62]); + const __m128i cospi10 = _mm_set1_epi32(cospi[10]); + const __m128i cospi54 = _mm_set1_epi32(cospi[54]); + const __m128i cospi18 = _mm_set1_epi32(cospi[18]); + const __m128i cospi46 = _mm_set1_epi32(cospi[46]); + const __m128i cospi26 = _mm_set1_epi32(cospi[26]); + const __m128i cospi38 = _mm_set1_epi32(cospi[38]); + const __m128i cospi34 = _mm_set1_epi32(cospi[34]); + const __m128i cospi30 = _mm_set1_epi32(cospi[30]); + const __m128i cospi42 = _mm_set1_epi32(cospi[42]); + const __m128i cospi22 = _mm_set1_epi32(cospi[22]); + const __m128i cospi50 = _mm_set1_epi32(cospi[50]); + const __m128i cospi14 = _mm_set1_epi32(cospi[14]); + const __m128i cospi58 = _mm_set1_epi32(cospi[58]); + const __m128i cospi6 = _mm_set1_epi32(cospi[6]); + const __m128i cospi8 = _mm_set1_epi32(cospi[8]); + const __m128i cospi56 = _mm_set1_epi32(cospi[56]); + const __m128i cospi40 = _mm_set1_epi32(cospi[40]); + const __m128i cospi24 = _mm_set1_epi32(cospi[24]); + const __m128i cospim56 = _mm_set1_epi32(-cospi[56]); + const __m128i cospim24 = _mm_set1_epi32(-cospi[24]); + const __m128i cospi48 = _mm_set1_epi32(cospi[48]); + const __m128i cospi16 = _mm_set1_epi32(cospi[16]); + const __m128i cospim48 = _mm_set1_epi32(-cospi[48]); + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + const __m128i rnding = _mm_set1_epi32(1 << (bit - 1)); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m128i clamp_lo = _mm_set1_epi32(-(1 << (log_range - 1))); + const __m128i clamp_hi = _mm_set1_epi32((1 << (log_range - 1)) - 1); + __m128i u[16], x, y; + + // Calculate the column 0, 1, 2, 3 + { + // stage 0 + // stage 1 + // stage 2 + __m128i zero = _mm_setzero_si128(); + x = _mm_mullo_epi32(in[0], cospi62); + u[0] = _mm_add_epi32(x, rnding); + u[0] = _mm_srai_epi32(u[0], bit); + + x = _mm_mullo_epi32(in[0], cospi2); + u[1] = _mm_sub_epi32(zero, x); + u[1] = _mm_add_epi32(u[1], rnding); + u[1] = _mm_srai_epi32(u[1], bit); + + x = _mm_mullo_epi32(in[2], cospi54); + u[2] = _mm_add_epi32(x, rnding); + u[2] = _mm_srai_epi32(u[2], bit); + + x = _mm_mullo_epi32(in[2], cospi10); + u[3] = _mm_sub_epi32(zero, x); + u[3] = _mm_add_epi32(u[3], rnding); + u[3] = _mm_srai_epi32(u[3], bit); + + x = _mm_mullo_epi32(in[4], cospi46); + u[4] = _mm_add_epi32(x, rnding); + u[4] = _mm_srai_epi32(u[4], bit); + + x = _mm_mullo_epi32(in[4], cospi18); + u[5] = _mm_sub_epi32(zero, x); + u[5] = _mm_add_epi32(u[5], rnding); + u[5] = _mm_srai_epi32(u[5], bit); + + x = _mm_mullo_epi32(in[6], cospi38); + u[6] = _mm_add_epi32(x, rnding); + u[6] = _mm_srai_epi32(u[6], bit); + + x = _mm_mullo_epi32(in[6], cospi26); + u[7] = _mm_sub_epi32(zero, x); + u[7] = _mm_add_epi32(u[7], rnding); + u[7] = _mm_srai_epi32(u[7], bit); + + u[8] = _mm_mullo_epi32(in[7], cospi34); + u[8] = _mm_add_epi32(u[8], rnding); + u[8] = _mm_srai_epi32(u[8], bit); + + u[9] = _mm_mullo_epi32(in[7], cospi30); + u[9] = _mm_add_epi32(u[9], rnding); + u[9] = _mm_srai_epi32(u[9], bit); + + u[10] = _mm_mullo_epi32(in[5], cospi42); + u[10] = _mm_add_epi32(u[10], rnding); + u[10] = _mm_srai_epi32(u[10], bit); + + u[11] = _mm_mullo_epi32(in[5], cospi22); + u[11] = _mm_add_epi32(u[11], rnding); + u[11] = _mm_srai_epi32(u[11], bit); + + u[12] = _mm_mullo_epi32(in[3], cospi50); + u[12] = _mm_add_epi32(u[12], rnding); + u[12] = _mm_srai_epi32(u[12], bit); + + u[13] = _mm_mullo_epi32(in[3], cospi14); + u[13] = _mm_add_epi32(u[13], rnding); + u[13] = _mm_srai_epi32(u[13], bit); + + u[14] = _mm_mullo_epi32(in[1], cospi58); + u[14] = _mm_add_epi32(u[14], rnding); + u[14] = _mm_srai_epi32(u[14], bit); + + u[15] = _mm_mullo_epi32(in[1], cospi6); + u[15] = _mm_add_epi32(u[15], rnding); + u[15] = _mm_srai_epi32(u[15], bit); + + // stage 3 + addsub_sse4_1(u[0], u[8], &u[0], &u[8], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[1], u[9], &u[1], &u[9], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[2], u[10], &u[2], &u[10], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[3], u[11], &u[3], &u[11], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[4], u[12], &u[4], &u[12], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[5], u[13], &u[5], &u[13], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[6], u[14], &u[6], &u[14], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[7], u[15], &u[7], &u[15], &clamp_lo, &clamp_hi); + + // stage 4 + y = _mm_mullo_epi32(u[8], cospi56); + x = _mm_mullo_epi32(u[9], cospi56); + u[8] = _mm_mullo_epi32(u[8], cospi8); + u[8] = _mm_add_epi32(u[8], x); + u[8] = _mm_add_epi32(u[8], rnding); + u[8] = _mm_srai_epi32(u[8], bit); + + x = _mm_mullo_epi32(u[9], cospi8); + u[9] = _mm_sub_epi32(y, x); + u[9] = _mm_add_epi32(u[9], rnding); + u[9] = _mm_srai_epi32(u[9], bit); + + x = _mm_mullo_epi32(u[11], cospi24); + y = _mm_mullo_epi32(u[10], cospi24); + u[10] = _mm_mullo_epi32(u[10], cospi40); + u[10] = _mm_add_epi32(u[10], x); + u[10] = _mm_add_epi32(u[10], rnding); + u[10] = _mm_srai_epi32(u[10], bit); + + x = _mm_mullo_epi32(u[11], cospi40); + u[11] = _mm_sub_epi32(y, x); + u[11] = _mm_add_epi32(u[11], rnding); + u[11] = _mm_srai_epi32(u[11], bit); + + x = _mm_mullo_epi32(u[13], cospi8); + y = _mm_mullo_epi32(u[12], cospi8); + u[12] = _mm_mullo_epi32(u[12], cospim56); + u[12] = _mm_add_epi32(u[12], x); + u[12] = _mm_add_epi32(u[12], rnding); + u[12] = _mm_srai_epi32(u[12], bit); + + x = _mm_mullo_epi32(u[13], cospim56); + u[13] = _mm_sub_epi32(y, x); + u[13] = _mm_add_epi32(u[13], rnding); + u[13] = _mm_srai_epi32(u[13], bit); + + x = _mm_mullo_epi32(u[15], cospi40); + y = _mm_mullo_epi32(u[14], cospi40); + u[14] = _mm_mullo_epi32(u[14], cospim24); + u[14] = _mm_add_epi32(u[14], x); + u[14] = _mm_add_epi32(u[14], rnding); + u[14] = _mm_srai_epi32(u[14], bit); + + x = _mm_mullo_epi32(u[15], cospim24); + u[15] = _mm_sub_epi32(y, x); + u[15] = _mm_add_epi32(u[15], rnding); + u[15] = _mm_srai_epi32(u[15], bit); + + // stage 5 + addsub_sse4_1(u[0], u[4], &u[0], &u[4], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[1], u[5], &u[1], &u[5], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[2], u[6], &u[2], &u[6], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[3], u[7], &u[3], &u[7], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[8], u[12], &u[8], &u[12], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[9], u[13], &u[9], &u[13], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[10], u[14], &u[10], &u[14], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[11], u[15], &u[11], &u[15], &clamp_lo, &clamp_hi); + + // stage 6 + x = _mm_mullo_epi32(u[5], cospi48); + y = _mm_mullo_epi32(u[4], cospi48); + u[4] = _mm_mullo_epi32(u[4], cospi16); + u[4] = _mm_add_epi32(u[4], x); + u[4] = _mm_add_epi32(u[4], rnding); + u[4] = _mm_srai_epi32(u[4], bit); + + x = _mm_mullo_epi32(u[5], cospi16); + u[5] = _mm_sub_epi32(y, x); + u[5] = _mm_add_epi32(u[5], rnding); + u[5] = _mm_srai_epi32(u[5], bit); + + x = _mm_mullo_epi32(u[7], cospi16); + y = _mm_mullo_epi32(u[6], cospi16); + u[6] = _mm_mullo_epi32(u[6], cospim48); + u[6] = _mm_add_epi32(u[6], x); + u[6] = _mm_add_epi32(u[6], rnding); + u[6] = _mm_srai_epi32(u[6], bit); + + x = _mm_mullo_epi32(u[7], cospim48); + u[7] = _mm_sub_epi32(y, x); + u[7] = _mm_add_epi32(u[7], rnding); + u[7] = _mm_srai_epi32(u[7], bit); + + x = _mm_mullo_epi32(u[13], cospi48); + y = _mm_mullo_epi32(u[12], cospi48); + u[12] = _mm_mullo_epi32(u[12], cospi16); + u[12] = _mm_add_epi32(u[12], x); + u[12] = _mm_add_epi32(u[12], rnding); + u[12] = _mm_srai_epi32(u[12], bit); + + x = _mm_mullo_epi32(u[13], cospi16); + u[13] = _mm_sub_epi32(y, x); + u[13] = _mm_add_epi32(u[13], rnding); + u[13] = _mm_srai_epi32(u[13], bit); + + x = _mm_mullo_epi32(u[15], cospi16); + y = _mm_mullo_epi32(u[14], cospi16); + u[14] = _mm_mullo_epi32(u[14], cospim48); + u[14] = _mm_add_epi32(u[14], x); + u[14] = _mm_add_epi32(u[14], rnding); + u[14] = _mm_srai_epi32(u[14], bit); + + x = _mm_mullo_epi32(u[15], cospim48); + u[15] = _mm_sub_epi32(y, x); + u[15] = _mm_add_epi32(u[15], rnding); + u[15] = _mm_srai_epi32(u[15], bit); + + // stage 7 + addsub_sse4_1(u[0], u[2], &u[0], &u[2], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[1], u[3], &u[1], &u[3], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[4], u[6], &u[4], &u[6], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[5], u[7], &u[5], &u[7], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[8], u[10], &u[8], &u[10], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[9], u[11], &u[9], &u[11], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[12], u[14], &u[12], &u[14], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[13], u[15], &u[13], &u[15], &clamp_lo, &clamp_hi); + + // stage 8 + y = _mm_mullo_epi32(u[2], cospi32); + x = _mm_mullo_epi32(u[3], cospi32); + u[2] = _mm_add_epi32(y, x); + u[2] = _mm_add_epi32(u[2], rnding); + u[2] = _mm_srai_epi32(u[2], bit); + + u[3] = _mm_sub_epi32(y, x); + u[3] = _mm_add_epi32(u[3], rnding); + u[3] = _mm_srai_epi32(u[3], bit); + y = _mm_mullo_epi32(u[6], cospi32); + x = _mm_mullo_epi32(u[7], cospi32); + u[6] = _mm_add_epi32(y, x); + u[6] = _mm_add_epi32(u[6], rnding); + u[6] = _mm_srai_epi32(u[6], bit); + + u[7] = _mm_sub_epi32(y, x); + u[7] = _mm_add_epi32(u[7], rnding); + u[7] = _mm_srai_epi32(u[7], bit); + + y = _mm_mullo_epi32(u[10], cospi32); + x = _mm_mullo_epi32(u[11], cospi32); + u[10] = _mm_add_epi32(y, x); + u[10] = _mm_add_epi32(u[10], rnding); + u[10] = _mm_srai_epi32(u[10], bit); + + u[11] = _mm_sub_epi32(y, x); + u[11] = _mm_add_epi32(u[11], rnding); + u[11] = _mm_srai_epi32(u[11], bit); + + y = _mm_mullo_epi32(u[14], cospi32); + x = _mm_mullo_epi32(u[15], cospi32); + u[14] = _mm_add_epi32(y, x); + u[14] = _mm_add_epi32(u[14], rnding); + u[14] = _mm_srai_epi32(u[14], bit); + + u[15] = _mm_sub_epi32(y, x); + u[15] = _mm_add_epi32(u[15], rnding); + u[15] = _mm_srai_epi32(u[15], bit); + + // stage 9 + if (do_cols) { + out[0] = u[0]; + out[1] = _mm_sub_epi32(_mm_setzero_si128(), u[8]); + out[2] = u[12]; + out[3] = _mm_sub_epi32(_mm_setzero_si128(), u[4]); + out[4] = u[6]; + out[5] = _mm_sub_epi32(_mm_setzero_si128(), u[14]); + out[6] = u[10]; + out[7] = _mm_sub_epi32(_mm_setzero_si128(), u[2]); + out[8] = u[3]; + out[9] = _mm_sub_epi32(_mm_setzero_si128(), u[11]); + out[10] = u[15]; + out[11] = _mm_sub_epi32(_mm_setzero_si128(), u[7]); + out[12] = u[5]; + out[13] = _mm_sub_epi32(_mm_setzero_si128(), u[13]); + out[14] = u[9]; + out[15] = _mm_sub_epi32(_mm_setzero_si128(), u[1]); + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m128i clamp_lo_out = _mm_set1_epi32(-(1 << (log_range_out - 1))); + const __m128i clamp_hi_out = + _mm_set1_epi32((1 << (log_range_out - 1)) - 1); + + neg_shift_sse4_1(u[0], u[8], out + 0, out + 1, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(u[12], u[4], out + 2, out + 3, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(u[6], u[14], out + 4, out + 5, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(u[10], u[2], out + 6, out + 7, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(u[3], u[11], out + 8, out + 9, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(u[15], u[7], out + 10, out + 11, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(u[5], u[13], out + 12, out + 13, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(u[9], u[1], out + 14, out + 15, &clamp_lo_out, + &clamp_hi_out, out_shift); + } + } +} + +static void idct16x16_sse4_1(__m128i *in, __m128i *out, int bit, int do_cols, + int bd, int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m128i cospi60 = _mm_set1_epi32(cospi[60]); + const __m128i cospim4 = _mm_set1_epi32(-cospi[4]); + const __m128i cospi28 = _mm_set1_epi32(cospi[28]); + const __m128i cospim36 = _mm_set1_epi32(-cospi[36]); + const __m128i cospi44 = _mm_set1_epi32(cospi[44]); + const __m128i cospi20 = _mm_set1_epi32(cospi[20]); + const __m128i cospim20 = _mm_set1_epi32(-cospi[20]); + const __m128i cospi12 = _mm_set1_epi32(cospi[12]); + const __m128i cospim52 = _mm_set1_epi32(-cospi[52]); + const __m128i cospi52 = _mm_set1_epi32(cospi[52]); + const __m128i cospi36 = _mm_set1_epi32(cospi[36]); + const __m128i cospi4 = _mm_set1_epi32(cospi[4]); + const __m128i cospi56 = _mm_set1_epi32(cospi[56]); + const __m128i cospim8 = _mm_set1_epi32(-cospi[8]); + const __m128i cospi24 = _mm_set1_epi32(cospi[24]); + const __m128i cospim40 = _mm_set1_epi32(-cospi[40]); + const __m128i cospi40 = _mm_set1_epi32(cospi[40]); + const __m128i cospi8 = _mm_set1_epi32(cospi[8]); + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + const __m128i cospi48 = _mm_set1_epi32(cospi[48]); + const __m128i cospi16 = _mm_set1_epi32(cospi[16]); + const __m128i cospim16 = _mm_set1_epi32(-cospi[16]); + const __m128i cospim48 = _mm_set1_epi32(-cospi[48]); + const __m128i rnding = _mm_set1_epi32(1 << (bit - 1)); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m128i clamp_lo = _mm_set1_epi32(-(1 << (log_range - 1))); + const __m128i clamp_hi = _mm_set1_epi32((1 << (log_range - 1)) - 1); + __m128i u[16], v[16], x, y; + + { + // stage 0 + // stage 1 + u[0] = in[0]; + u[1] = in[8]; + u[2] = in[4]; + u[3] = in[12]; + u[4] = in[2]; + u[5] = in[10]; + u[6] = in[6]; + u[7] = in[14]; + u[8] = in[1]; + u[9] = in[9]; + u[10] = in[5]; + u[11] = in[13]; + u[12] = in[3]; + u[13] = in[11]; + u[14] = in[7]; + u[15] = in[15]; + + // stage 2 + v[0] = u[0]; + v[1] = u[1]; + v[2] = u[2]; + v[3] = u[3]; + v[4] = u[4]; + v[5] = u[5]; + v[6] = u[6]; + v[7] = u[7]; + + v[8] = half_btf_sse4_1(&cospi60, &u[8], &cospim4, &u[15], &rnding, bit); + v[9] = half_btf_sse4_1(&cospi28, &u[9], &cospim36, &u[14], &rnding, bit); + v[10] = half_btf_sse4_1(&cospi44, &u[10], &cospim20, &u[13], &rnding, bit); + v[11] = half_btf_sse4_1(&cospi12, &u[11], &cospim52, &u[12], &rnding, bit); + v[12] = half_btf_sse4_1(&cospi52, &u[11], &cospi12, &u[12], &rnding, bit); + v[13] = half_btf_sse4_1(&cospi20, &u[10], &cospi44, &u[13], &rnding, bit); + v[14] = half_btf_sse4_1(&cospi36, &u[9], &cospi28, &u[14], &rnding, bit); + v[15] = half_btf_sse4_1(&cospi4, &u[8], &cospi60, &u[15], &rnding, bit); + + // stage 3 + u[0] = v[0]; + u[1] = v[1]; + u[2] = v[2]; + u[3] = v[3]; + u[4] = half_btf_sse4_1(&cospi56, &v[4], &cospim8, &v[7], &rnding, bit); + u[5] = half_btf_sse4_1(&cospi24, &v[5], &cospim40, &v[6], &rnding, bit); + u[6] = half_btf_sse4_1(&cospi40, &v[5], &cospi24, &v[6], &rnding, bit); + u[7] = half_btf_sse4_1(&cospi8, &v[4], &cospi56, &v[7], &rnding, bit); + addsub_sse4_1(v[8], v[9], &u[8], &u[9], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[11], v[10], &u[11], &u[10], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[12], v[13], &u[12], &u[13], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[15], v[14], &u[15], &u[14], &clamp_lo, &clamp_hi); + + // stage 4 + x = _mm_mullo_epi32(u[0], cospi32); + y = _mm_mullo_epi32(u[1], cospi32); + v[0] = _mm_add_epi32(x, y); + v[0] = _mm_add_epi32(v[0], rnding); + v[0] = _mm_srai_epi32(v[0], bit); + + v[1] = _mm_sub_epi32(x, y); + v[1] = _mm_add_epi32(v[1], rnding); + v[1] = _mm_srai_epi32(v[1], bit); + + v[2] = half_btf_sse4_1(&cospi48, &u[2], &cospim16, &u[3], &rnding, bit); + v[3] = half_btf_sse4_1(&cospi16, &u[2], &cospi48, &u[3], &rnding, bit); + addsub_sse4_1(u[4], u[5], &v[4], &v[5], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[7], u[6], &v[7], &v[6], &clamp_lo, &clamp_hi); + v[8] = u[8]; + v[9] = half_btf_sse4_1(&cospim16, &u[9], &cospi48, &u[14], &rnding, bit); + v[10] = half_btf_sse4_1(&cospim48, &u[10], &cospim16, &u[13], &rnding, bit); + v[11] = u[11]; + v[12] = u[12]; + v[13] = half_btf_sse4_1(&cospim16, &u[10], &cospi48, &u[13], &rnding, bit); + v[14] = half_btf_sse4_1(&cospi48, &u[9], &cospi16, &u[14], &rnding, bit); + v[15] = u[15]; + + // stage 5 + addsub_sse4_1(v[0], v[3], &u[0], &u[3], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[1], v[2], &u[1], &u[2], &clamp_lo, &clamp_hi); + u[4] = v[4]; + + x = _mm_mullo_epi32(v[5], cospi32); + y = _mm_mullo_epi32(v[6], cospi32); + u[5] = _mm_sub_epi32(y, x); + u[5] = _mm_add_epi32(u[5], rnding); + u[5] = _mm_srai_epi32(u[5], bit); + + u[6] = _mm_add_epi32(y, x); + u[6] = _mm_add_epi32(u[6], rnding); + u[6] = _mm_srai_epi32(u[6], bit); + + u[7] = v[7]; + addsub_sse4_1(v[8], v[11], &u[8], &u[11], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[9], v[10], &u[9], &u[10], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[15], v[12], &u[15], &u[12], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[14], v[13], &u[14], &u[13], &clamp_lo, &clamp_hi); + + // stage 6 + addsub_sse4_1(u[0], u[7], &v[0], &v[7], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[1], u[6], &v[1], &v[6], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[2], u[5], &v[2], &v[5], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[3], u[4], &v[3], &v[4], &clamp_lo, &clamp_hi); + v[8] = u[8]; + v[9] = u[9]; + + x = _mm_mullo_epi32(u[10], cospi32); + y = _mm_mullo_epi32(u[13], cospi32); + v[10] = _mm_sub_epi32(y, x); + v[10] = _mm_add_epi32(v[10], rnding); + v[10] = _mm_srai_epi32(v[10], bit); + + v[13] = _mm_add_epi32(x, y); + v[13] = _mm_add_epi32(v[13], rnding); + v[13] = _mm_srai_epi32(v[13], bit); + + x = _mm_mullo_epi32(u[11], cospi32); + y = _mm_mullo_epi32(u[12], cospi32); + v[11] = _mm_sub_epi32(y, x); + v[11] = _mm_add_epi32(v[11], rnding); + v[11] = _mm_srai_epi32(v[11], bit); + + v[12] = _mm_add_epi32(x, y); + v[12] = _mm_add_epi32(v[12], rnding); + v[12] = _mm_srai_epi32(v[12], bit); + + v[14] = u[14]; + v[15] = u[15]; + + // stage 7 + if (do_cols) { + addsub_no_clamp_sse4_1(v[0], v[15], out + 0, out + 15); + addsub_no_clamp_sse4_1(v[1], v[14], out + 1, out + 14); + addsub_no_clamp_sse4_1(v[2], v[13], out + 2, out + 13); + addsub_no_clamp_sse4_1(v[3], v[12], out + 3, out + 12); + addsub_no_clamp_sse4_1(v[4], v[11], out + 4, out + 11); + addsub_no_clamp_sse4_1(v[5], v[10], out + 5, out + 10); + addsub_no_clamp_sse4_1(v[6], v[9], out + 6, out + 9); + addsub_no_clamp_sse4_1(v[7], v[8], out + 7, out + 8); + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m128i clamp_lo_out = _mm_set1_epi32(AOMMAX( + -(1 << (log_range_out - 1)), -(1 << (log_range - 1 - out_shift)))); + const __m128i clamp_hi_out = _mm_set1_epi32(AOMMIN( + (1 << (log_range_out - 1)) - 1, (1 << (log_range - 1 - out_shift)))); + + addsub_shift_sse4_1(v[0], v[15], out + 0, out + 15, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(v[1], v[14], out + 1, out + 14, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(v[2], v[13], out + 2, out + 13, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(v[3], v[12], out + 3, out + 12, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(v[4], v[11], out + 4, out + 11, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(v[5], v[10], out + 5, out + 10, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(v[6], v[9], out + 6, out + 9, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(v[7], v[8], out + 7, out + 8, &clamp_lo_out, + &clamp_hi_out, out_shift); + } + } +} + +static void iadst16x16_sse4_1(__m128i *in, __m128i *out, int bit, int do_cols, + int bd, int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m128i cospi2 = _mm_set1_epi32(cospi[2]); + const __m128i cospi62 = _mm_set1_epi32(cospi[62]); + const __m128i cospi10 = _mm_set1_epi32(cospi[10]); + const __m128i cospi54 = _mm_set1_epi32(cospi[54]); + const __m128i cospi18 = _mm_set1_epi32(cospi[18]); + const __m128i cospi46 = _mm_set1_epi32(cospi[46]); + const __m128i cospi26 = _mm_set1_epi32(cospi[26]); + const __m128i cospi38 = _mm_set1_epi32(cospi[38]); + const __m128i cospi34 = _mm_set1_epi32(cospi[34]); + const __m128i cospi30 = _mm_set1_epi32(cospi[30]); + const __m128i cospi42 = _mm_set1_epi32(cospi[42]); + const __m128i cospi22 = _mm_set1_epi32(cospi[22]); + const __m128i cospi50 = _mm_set1_epi32(cospi[50]); + const __m128i cospi14 = _mm_set1_epi32(cospi[14]); + const __m128i cospi58 = _mm_set1_epi32(cospi[58]); + const __m128i cospi6 = _mm_set1_epi32(cospi[6]); + const __m128i cospi8 = _mm_set1_epi32(cospi[8]); + const __m128i cospi56 = _mm_set1_epi32(cospi[56]); + const __m128i cospi40 = _mm_set1_epi32(cospi[40]); + const __m128i cospi24 = _mm_set1_epi32(cospi[24]); + const __m128i cospim56 = _mm_set1_epi32(-cospi[56]); + const __m128i cospim24 = _mm_set1_epi32(-cospi[24]); + const __m128i cospi48 = _mm_set1_epi32(cospi[48]); + const __m128i cospi16 = _mm_set1_epi32(cospi[16]); + const __m128i cospim48 = _mm_set1_epi32(-cospi[48]); + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + const __m128i rnding = _mm_set1_epi32(1 << (bit - 1)); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m128i clamp_lo = _mm_set1_epi32(-(1 << (log_range - 1))); + const __m128i clamp_hi = _mm_set1_epi32((1 << (log_range - 1)) - 1); + __m128i u[16], v[16], x, y; + + // Calculate the column 0, 1, 2, 3 + { + // stage 0 + // stage 1 + // stage 2 + v[0] = _mm_mullo_epi32(in[15], cospi2); + x = _mm_mullo_epi32(in[0], cospi62); + v[0] = _mm_add_epi32(v[0], x); + v[0] = _mm_add_epi32(v[0], rnding); + v[0] = _mm_srai_epi32(v[0], bit); + + v[1] = _mm_mullo_epi32(in[15], cospi62); + x = _mm_mullo_epi32(in[0], cospi2); + v[1] = _mm_sub_epi32(v[1], x); + v[1] = _mm_add_epi32(v[1], rnding); + v[1] = _mm_srai_epi32(v[1], bit); + + v[2] = _mm_mullo_epi32(in[13], cospi10); + x = _mm_mullo_epi32(in[2], cospi54); + v[2] = _mm_add_epi32(v[2], x); + v[2] = _mm_add_epi32(v[2], rnding); + v[2] = _mm_srai_epi32(v[2], bit); + + v[3] = _mm_mullo_epi32(in[13], cospi54); + x = _mm_mullo_epi32(in[2], cospi10); + v[3] = _mm_sub_epi32(v[3], x); + v[3] = _mm_add_epi32(v[3], rnding); + v[3] = _mm_srai_epi32(v[3], bit); + + v[4] = _mm_mullo_epi32(in[11], cospi18); + x = _mm_mullo_epi32(in[4], cospi46); + v[4] = _mm_add_epi32(v[4], x); + v[4] = _mm_add_epi32(v[4], rnding); + v[4] = _mm_srai_epi32(v[4], bit); + + v[5] = _mm_mullo_epi32(in[11], cospi46); + x = _mm_mullo_epi32(in[4], cospi18); + v[5] = _mm_sub_epi32(v[5], x); + v[5] = _mm_add_epi32(v[5], rnding); + v[5] = _mm_srai_epi32(v[5], bit); + + v[6] = _mm_mullo_epi32(in[9], cospi26); + x = _mm_mullo_epi32(in[6], cospi38); + v[6] = _mm_add_epi32(v[6], x); + v[6] = _mm_add_epi32(v[6], rnding); + v[6] = _mm_srai_epi32(v[6], bit); + + v[7] = _mm_mullo_epi32(in[9], cospi38); + x = _mm_mullo_epi32(in[6], cospi26); + v[7] = _mm_sub_epi32(v[7], x); + v[7] = _mm_add_epi32(v[7], rnding); + v[7] = _mm_srai_epi32(v[7], bit); + + v[8] = _mm_mullo_epi32(in[7], cospi34); + x = _mm_mullo_epi32(in[8], cospi30); + v[8] = _mm_add_epi32(v[8], x); + v[8] = _mm_add_epi32(v[8], rnding); + v[8] = _mm_srai_epi32(v[8], bit); + + v[9] = _mm_mullo_epi32(in[7], cospi30); + x = _mm_mullo_epi32(in[8], cospi34); + v[9] = _mm_sub_epi32(v[9], x); + v[9] = _mm_add_epi32(v[9], rnding); + v[9] = _mm_srai_epi32(v[9], bit); + + v[10] = _mm_mullo_epi32(in[5], cospi42); + x = _mm_mullo_epi32(in[10], cospi22); + v[10] = _mm_add_epi32(v[10], x); + v[10] = _mm_add_epi32(v[10], rnding); + v[10] = _mm_srai_epi32(v[10], bit); + + v[11] = _mm_mullo_epi32(in[5], cospi22); + x = _mm_mullo_epi32(in[10], cospi42); + v[11] = _mm_sub_epi32(v[11], x); + v[11] = _mm_add_epi32(v[11], rnding); + v[11] = _mm_srai_epi32(v[11], bit); + + v[12] = _mm_mullo_epi32(in[3], cospi50); + x = _mm_mullo_epi32(in[12], cospi14); + v[12] = _mm_add_epi32(v[12], x); + v[12] = _mm_add_epi32(v[12], rnding); + v[12] = _mm_srai_epi32(v[12], bit); + + v[13] = _mm_mullo_epi32(in[3], cospi14); + x = _mm_mullo_epi32(in[12], cospi50); + v[13] = _mm_sub_epi32(v[13], x); + v[13] = _mm_add_epi32(v[13], rnding); + v[13] = _mm_srai_epi32(v[13], bit); + + v[14] = _mm_mullo_epi32(in[1], cospi58); + x = _mm_mullo_epi32(in[14], cospi6); + v[14] = _mm_add_epi32(v[14], x); + v[14] = _mm_add_epi32(v[14], rnding); + v[14] = _mm_srai_epi32(v[14], bit); + + v[15] = _mm_mullo_epi32(in[1], cospi6); + x = _mm_mullo_epi32(in[14], cospi58); + v[15] = _mm_sub_epi32(v[15], x); + v[15] = _mm_add_epi32(v[15], rnding); + v[15] = _mm_srai_epi32(v[15], bit); + + // stage 3 + addsub_sse4_1(v[0], v[8], &u[0], &u[8], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[1], v[9], &u[1], &u[9], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[2], v[10], &u[2], &u[10], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[3], v[11], &u[3], &u[11], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[4], v[12], &u[4], &u[12], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[5], v[13], &u[5], &u[13], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[6], v[14], &u[6], &u[14], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[7], v[15], &u[7], &u[15], &clamp_lo, &clamp_hi); + + // stage 4 + v[0] = u[0]; + v[1] = u[1]; + v[2] = u[2]; + v[3] = u[3]; + v[4] = u[4]; + v[5] = u[5]; + v[6] = u[6]; + v[7] = u[7]; + + v[8] = _mm_mullo_epi32(u[8], cospi8); + x = _mm_mullo_epi32(u[9], cospi56); + v[8] = _mm_add_epi32(v[8], x); + v[8] = _mm_add_epi32(v[8], rnding); + v[8] = _mm_srai_epi32(v[8], bit); + + v[9] = _mm_mullo_epi32(u[8], cospi56); + x = _mm_mullo_epi32(u[9], cospi8); + v[9] = _mm_sub_epi32(v[9], x); + v[9] = _mm_add_epi32(v[9], rnding); + v[9] = _mm_srai_epi32(v[9], bit); + + v[10] = _mm_mullo_epi32(u[10], cospi40); + x = _mm_mullo_epi32(u[11], cospi24); + v[10] = _mm_add_epi32(v[10], x); + v[10] = _mm_add_epi32(v[10], rnding); + v[10] = _mm_srai_epi32(v[10], bit); + + v[11] = _mm_mullo_epi32(u[10], cospi24); + x = _mm_mullo_epi32(u[11], cospi40); + v[11] = _mm_sub_epi32(v[11], x); + v[11] = _mm_add_epi32(v[11], rnding); + v[11] = _mm_srai_epi32(v[11], bit); + + v[12] = _mm_mullo_epi32(u[12], cospim56); + x = _mm_mullo_epi32(u[13], cospi8); + v[12] = _mm_add_epi32(v[12], x); + v[12] = _mm_add_epi32(v[12], rnding); + v[12] = _mm_srai_epi32(v[12], bit); + + v[13] = _mm_mullo_epi32(u[12], cospi8); + x = _mm_mullo_epi32(u[13], cospim56); + v[13] = _mm_sub_epi32(v[13], x); + v[13] = _mm_add_epi32(v[13], rnding); + v[13] = _mm_srai_epi32(v[13], bit); + + v[14] = _mm_mullo_epi32(u[14], cospim24); + x = _mm_mullo_epi32(u[15], cospi40); + v[14] = _mm_add_epi32(v[14], x); + v[14] = _mm_add_epi32(v[14], rnding); + v[14] = _mm_srai_epi32(v[14], bit); + + v[15] = _mm_mullo_epi32(u[14], cospi40); + x = _mm_mullo_epi32(u[15], cospim24); + v[15] = _mm_sub_epi32(v[15], x); + v[15] = _mm_add_epi32(v[15], rnding); + v[15] = _mm_srai_epi32(v[15], bit); + + // stage 5 + addsub_sse4_1(v[0], v[4], &u[0], &u[4], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[1], v[5], &u[1], &u[5], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[2], v[6], &u[2], &u[6], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[3], v[7], &u[3], &u[7], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[8], v[12], &u[8], &u[12], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[9], v[13], &u[9], &u[13], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[10], v[14], &u[10], &u[14], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[11], v[15], &u[11], &u[15], &clamp_lo, &clamp_hi); + + // stage 6 + v[0] = u[0]; + v[1] = u[1]; + v[2] = u[2]; + v[3] = u[3]; + + v[4] = _mm_mullo_epi32(u[4], cospi16); + x = _mm_mullo_epi32(u[5], cospi48); + v[4] = _mm_add_epi32(v[4], x); + v[4] = _mm_add_epi32(v[4], rnding); + v[4] = _mm_srai_epi32(v[4], bit); + + v[5] = _mm_mullo_epi32(u[4], cospi48); + x = _mm_mullo_epi32(u[5], cospi16); + v[5] = _mm_sub_epi32(v[5], x); + v[5] = _mm_add_epi32(v[5], rnding); + v[5] = _mm_srai_epi32(v[5], bit); + + v[6] = _mm_mullo_epi32(u[6], cospim48); + x = _mm_mullo_epi32(u[7], cospi16); + v[6] = _mm_add_epi32(v[6], x); + v[6] = _mm_add_epi32(v[6], rnding); + v[6] = _mm_srai_epi32(v[6], bit); + + v[7] = _mm_mullo_epi32(u[6], cospi16); + x = _mm_mullo_epi32(u[7], cospim48); + v[7] = _mm_sub_epi32(v[7], x); + v[7] = _mm_add_epi32(v[7], rnding); + v[7] = _mm_srai_epi32(v[7], bit); + + v[8] = u[8]; + v[9] = u[9]; + v[10] = u[10]; + v[11] = u[11]; + + v[12] = _mm_mullo_epi32(u[12], cospi16); + x = _mm_mullo_epi32(u[13], cospi48); + v[12] = _mm_add_epi32(v[12], x); + v[12] = _mm_add_epi32(v[12], rnding); + v[12] = _mm_srai_epi32(v[12], bit); + + v[13] = _mm_mullo_epi32(u[12], cospi48); + x = _mm_mullo_epi32(u[13], cospi16); + v[13] = _mm_sub_epi32(v[13], x); + v[13] = _mm_add_epi32(v[13], rnding); + v[13] = _mm_srai_epi32(v[13], bit); + + v[14] = _mm_mullo_epi32(u[14], cospim48); + x = _mm_mullo_epi32(u[15], cospi16); + v[14] = _mm_add_epi32(v[14], x); + v[14] = _mm_add_epi32(v[14], rnding); + v[14] = _mm_srai_epi32(v[14], bit); + + v[15] = _mm_mullo_epi32(u[14], cospi16); + x = _mm_mullo_epi32(u[15], cospim48); + v[15] = _mm_sub_epi32(v[15], x); + v[15] = _mm_add_epi32(v[15], rnding); + v[15] = _mm_srai_epi32(v[15], bit); + + // stage 7 + addsub_sse4_1(v[0], v[2], &u[0], &u[2], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[1], v[3], &u[1], &u[3], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[4], v[6], &u[4], &u[6], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[5], v[7], &u[5], &u[7], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[8], v[10], &u[8], &u[10], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[9], v[11], &u[9], &u[11], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[12], v[14], &u[12], &u[14], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[13], v[15], &u[13], &u[15], &clamp_lo, &clamp_hi); + + // stage 8 + v[0] = u[0]; + v[1] = u[1]; + + y = _mm_mullo_epi32(u[2], cospi32); + x = _mm_mullo_epi32(u[3], cospi32); + v[2] = _mm_add_epi32(y, x); + v[2] = _mm_add_epi32(v[2], rnding); + v[2] = _mm_srai_epi32(v[2], bit); + + v[3] = _mm_sub_epi32(y, x); + v[3] = _mm_add_epi32(v[3], rnding); + v[3] = _mm_srai_epi32(v[3], bit); + + v[4] = u[4]; + v[5] = u[5]; + + y = _mm_mullo_epi32(u[6], cospi32); + x = _mm_mullo_epi32(u[7], cospi32); + v[6] = _mm_add_epi32(y, x); + v[6] = _mm_add_epi32(v[6], rnding); + v[6] = _mm_srai_epi32(v[6], bit); + + v[7] = _mm_sub_epi32(y, x); + v[7] = _mm_add_epi32(v[7], rnding); + v[7] = _mm_srai_epi32(v[7], bit); + + v[8] = u[8]; + v[9] = u[9]; + + y = _mm_mullo_epi32(u[10], cospi32); + x = _mm_mullo_epi32(u[11], cospi32); + v[10] = _mm_add_epi32(y, x); + v[10] = _mm_add_epi32(v[10], rnding); + v[10] = _mm_srai_epi32(v[10], bit); + + v[11] = _mm_sub_epi32(y, x); + v[11] = _mm_add_epi32(v[11], rnding); + v[11] = _mm_srai_epi32(v[11], bit); + + v[12] = u[12]; + v[13] = u[13]; + + y = _mm_mullo_epi32(u[14], cospi32); + x = _mm_mullo_epi32(u[15], cospi32); + v[14] = _mm_add_epi32(y, x); + v[14] = _mm_add_epi32(v[14], rnding); + v[14] = _mm_srai_epi32(v[14], bit); + + v[15] = _mm_sub_epi32(y, x); + v[15] = _mm_add_epi32(v[15], rnding); + v[15] = _mm_srai_epi32(v[15], bit); + + // stage 9 + if (do_cols) { + out[0] = v[0]; + out[1] = _mm_sub_epi32(_mm_setzero_si128(), v[8]); + out[2] = v[12]; + out[3] = _mm_sub_epi32(_mm_setzero_si128(), v[4]); + out[4] = v[6]; + out[5] = _mm_sub_epi32(_mm_setzero_si128(), v[14]); + out[6] = v[10]; + out[7] = _mm_sub_epi32(_mm_setzero_si128(), v[2]); + out[8] = v[3]; + out[9] = _mm_sub_epi32(_mm_setzero_si128(), v[11]); + out[10] = v[15]; + out[11] = _mm_sub_epi32(_mm_setzero_si128(), v[7]); + out[12] = v[5]; + out[13] = _mm_sub_epi32(_mm_setzero_si128(), v[13]); + out[14] = v[9]; + out[15] = _mm_sub_epi32(_mm_setzero_si128(), v[1]); + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m128i clamp_lo_out = _mm_set1_epi32(-(1 << (log_range_out - 1))); + const __m128i clamp_hi_out = + _mm_set1_epi32((1 << (log_range_out - 1)) - 1); + + neg_shift_sse4_1(v[0], v[8], out + 0, out + 1, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(v[12], v[4], out + 2, out + 3, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(v[6], v[14], out + 4, out + 5, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(v[10], v[2], out + 6, out + 7, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(v[3], v[11], out + 8, out + 9, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(v[15], v[7], out + 10, out + 11, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(v[5], v[13], out + 12, out + 13, &clamp_lo_out, + &clamp_hi_out, out_shift); + neg_shift_sse4_1(v[9], v[1], out + 14, out + 15, &clamp_lo_out, + &clamp_hi_out, out_shift); + } + } +} + +static INLINE void idct64_stage8_sse4_1( + __m128i *u, const __m128i *cospim32, const __m128i *cospi32, + const __m128i *cospim16, const __m128i *cospi48, const __m128i *cospi16, + const __m128i *cospim48, const __m128i *clamp_lo, const __m128i *clamp_hi, + const __m128i *rnding, int bit) { + int i; + __m128i temp1, temp2, temp3, temp4; + temp1 = half_btf_sse4_1(cospim32, &u[10], cospi32, &u[13], rnding, bit); + u[13] = half_btf_sse4_1(cospi32, &u[10], cospi32, &u[13], rnding, bit); + u[10] = temp1; + temp2 = half_btf_sse4_1(cospim32, &u[11], cospi32, &u[12], rnding, bit); + u[12] = half_btf_sse4_1(cospi32, &u[11], cospi32, &u[12], rnding, bit); + u[11] = temp2; + + for (i = 16; i < 20; ++i) { + addsub_sse4_1(u[i], u[i ^ 7], &u[i], &u[i ^ 7], clamp_lo, clamp_hi); + addsub_sse4_1(u[i ^ 15], u[i ^ 8], &u[i ^ 15], &u[i ^ 8], clamp_lo, + clamp_hi); + } + + temp1 = half_btf_sse4_1(cospim16, &u[36], cospi48, &u[59], rnding, bit); + temp2 = half_btf_sse4_1(cospim16, &u[37], cospi48, &u[58], rnding, bit); + temp3 = half_btf_sse4_1(cospim16, &u[38], cospi48, &u[57], rnding, bit); + temp4 = half_btf_sse4_1(cospim16, &u[39], cospi48, &u[56], rnding, bit); + u[56] = half_btf_sse4_1(cospi48, &u[39], cospi16, &u[56], rnding, bit); + u[57] = half_btf_sse4_1(cospi48, &u[38], cospi16, &u[57], rnding, bit); + u[58] = half_btf_sse4_1(cospi48, &u[37], cospi16, &u[58], rnding, bit); + u[59] = half_btf_sse4_1(cospi48, &u[36], cospi16, &u[59], rnding, bit); + u[36] = temp1; + u[37] = temp2; + u[38] = temp3; + u[39] = temp4; + + temp1 = half_btf_sse4_1(cospim48, &u[40], cospim16, &u[55], rnding, bit); + temp2 = half_btf_sse4_1(cospim48, &u[41], cospim16, &u[54], rnding, bit); + temp3 = half_btf_sse4_1(cospim48, &u[42], cospim16, &u[53], rnding, bit); + temp4 = half_btf_sse4_1(cospim48, &u[43], cospim16, &u[52], rnding, bit); + u[52] = half_btf_sse4_1(cospim16, &u[43], cospi48, &u[52], rnding, bit); + u[53] = half_btf_sse4_1(cospim16, &u[42], cospi48, &u[53], rnding, bit); + u[54] = half_btf_sse4_1(cospim16, &u[41], cospi48, &u[54], rnding, bit); + u[55] = half_btf_sse4_1(cospim16, &u[40], cospi48, &u[55], rnding, bit); + u[40] = temp1; + u[41] = temp2; + u[42] = temp3; + u[43] = temp4; +} + +static INLINE void idct64_stage9_sse4_1(__m128i *u, const __m128i *cospim32, + const __m128i *cospi32, + const __m128i *clamp_lo, + const __m128i *clamp_hi, + const __m128i *rnding, int bit) { + int i; + __m128i temp1, temp2, temp3, temp4; + for (i = 0; i < 8; ++i) { + addsub_sse4_1(u[i], u[15 - i], &u[i], &u[15 - i], clamp_lo, clamp_hi); + } + + temp1 = half_btf_sse4_1(cospim32, &u[20], cospi32, &u[27], rnding, bit); + temp2 = half_btf_sse4_1(cospim32, &u[21], cospi32, &u[26], rnding, bit); + temp3 = half_btf_sse4_1(cospim32, &u[22], cospi32, &u[25], rnding, bit); + temp4 = half_btf_sse4_1(cospim32, &u[23], cospi32, &u[24], rnding, bit); + u[24] = half_btf_sse4_1(cospi32, &u[23], cospi32, &u[24], rnding, bit); + u[25] = half_btf_sse4_1(cospi32, &u[22], cospi32, &u[25], rnding, bit); + u[26] = half_btf_sse4_1(cospi32, &u[21], cospi32, &u[26], rnding, bit); + u[27] = half_btf_sse4_1(cospi32, &u[20], cospi32, &u[27], rnding, bit); + u[20] = temp1; + u[21] = temp2; + u[22] = temp3; + u[23] = temp4; + for (i = 32; i < 40; i++) { + addsub_sse4_1(u[i], u[i ^ 15], &u[i], &u[i ^ 15], clamp_lo, clamp_hi); + } + + for (i = 48; i < 56; i++) { + addsub_sse4_1(u[i ^ 15], u[i], &u[i ^ 15], &u[i], clamp_lo, clamp_hi); + } +} + +static INLINE void idct64_stage10_sse4_1(__m128i *u, const __m128i *cospim32, + const __m128i *cospi32, + const __m128i *clamp_lo, + const __m128i *clamp_hi, + const __m128i *rnding, int bit) { + __m128i temp1, temp2, temp3, temp4; + for (int i = 0; i < 16; i++) { + addsub_sse4_1(u[i], u[31 - i], &u[i], &u[31 - i], clamp_lo, clamp_hi); + } + + temp1 = half_btf_sse4_1(cospim32, &u[40], cospi32, &u[55], rnding, bit); + temp2 = half_btf_sse4_1(cospim32, &u[41], cospi32, &u[54], rnding, bit); + temp3 = half_btf_sse4_1(cospim32, &u[42], cospi32, &u[53], rnding, bit); + temp4 = half_btf_sse4_1(cospim32, &u[43], cospi32, &u[52], rnding, bit); + u[52] = half_btf_sse4_1(cospi32, &u[43], cospi32, &u[52], rnding, bit); + u[53] = half_btf_sse4_1(cospi32, &u[42], cospi32, &u[53], rnding, bit); + u[54] = half_btf_sse4_1(cospi32, &u[41], cospi32, &u[54], rnding, bit); + u[55] = half_btf_sse4_1(cospi32, &u[40], cospi32, &u[55], rnding, bit); + u[40] = temp1; + u[41] = temp2; + u[42] = temp3; + u[43] = temp4; + + temp1 = half_btf_sse4_1(cospim32, &u[44], cospi32, &u[51], rnding, bit); + temp2 = half_btf_sse4_1(cospim32, &u[45], cospi32, &u[50], rnding, bit); + temp3 = half_btf_sse4_1(cospim32, &u[46], cospi32, &u[49], rnding, bit); + temp4 = half_btf_sse4_1(cospim32, &u[47], cospi32, &u[48], rnding, bit); + u[48] = half_btf_sse4_1(cospi32, &u[47], cospi32, &u[48], rnding, bit); + u[49] = half_btf_sse4_1(cospi32, &u[46], cospi32, &u[49], rnding, bit); + u[50] = half_btf_sse4_1(cospi32, &u[45], cospi32, &u[50], rnding, bit); + u[51] = half_btf_sse4_1(cospi32, &u[44], cospi32, &u[51], rnding, bit); + u[44] = temp1; + u[45] = temp2; + u[46] = temp3; + u[47] = temp4; +} + +static INLINE void idct64_stage11_sse4_1(__m128i *u, __m128i *out, int do_cols, + int bd, int out_shift, + const int log_range) { + if (do_cols) { + for (int i = 0; i < 32; i++) { + addsub_no_clamp_sse4_1(u[i], u[63 - i], &out[(i)], &out[(63 - i)]); + } + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m128i clamp_lo_out = _mm_set1_epi32(AOMMAX( + -(1 << (log_range_out - 1)), -(1 << (log_range - 1 - out_shift)))); + const __m128i clamp_hi_out = _mm_set1_epi32(AOMMIN( + (1 << (log_range_out - 1)) - 1, (1 << (log_range - 1 - out_shift)))); + + for (int i = 0; i < 32; i++) { + addsub_shift_sse4_1(u[i], u[63 - i], &out[(i)], &out[(63 - i)], + &clamp_lo_out, &clamp_hi_out, out_shift); + } + } +} + +static void idct64x64_low1_sse4_1(__m128i *in, __m128i *out, int bit, + int do_cols, int bd, int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m128i rnding = _mm_set1_epi32(1 << (bit - 1)); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m128i clamp_lo = _mm_set1_epi32(-(1 << (log_range - 1))); + const __m128i clamp_hi = _mm_set1_epi32((1 << (log_range - 1)) - 1); + + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + + { + __m128i x; + + // stage 1 + // stage 2 + // stage 3 + // stage 4 + // stage 5 + // stage 6 + x = half_btf_0_sse4_1(&cospi32, &in[0], &rnding, bit); + + // stage 8 + // stage 9 + // stage 10 + // stage 11 + if (do_cols) { + x = _mm_max_epi32(x, clamp_lo); + x = _mm_min_epi32(x, clamp_hi); + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m128i clamp_lo_out = _mm_set1_epi32(AOMMAX( + -(1 << (log_range_out - 1)), -(1 << (log_range - 1 - out_shift)))); + const __m128i clamp_hi_out = _mm_set1_epi32(AOMMIN( + (1 << (log_range_out - 1)) - 1, (1 << (log_range - 1 - out_shift)))); + + __m128i offset = _mm_set1_epi32((1 << out_shift) >> 1); + x = _mm_add_epi32(x, offset); + x = _mm_sra_epi32(x, _mm_cvtsi32_si128(out_shift)); + + x = _mm_max_epi32(x, clamp_lo_out); + x = _mm_min_epi32(x, clamp_hi_out); + } + + out[0] = x; + out[63] = x; + out[1] = x; + out[62] = x; + out[2] = x; + out[61] = x; + out[3] = x; + out[60] = x; + out[4] = x; + out[59] = x; + out[5] = x; + out[58] = x; + out[6] = x; + out[57] = x; + out[7] = x; + out[56] = x; + out[8] = x; + out[55] = x; + out[9] = x; + out[54] = x; + out[10] = x; + out[53] = x; + out[11] = x; + out[52] = x; + out[12] = x; + out[51] = x; + out[13] = x; + out[50] = x; + out[14] = x; + out[49] = x; + out[15] = x; + out[48] = x; + out[16] = x; + out[47] = x; + out[17] = x; + out[46] = x; + out[18] = x; + out[45] = x; + out[19] = x; + out[44] = x; + out[20] = x; + out[43] = x; + out[21] = x; + out[42] = x; + out[22] = x; + out[41] = x; + out[23] = x; + out[40] = x; + out[24] = x; + out[39] = x; + out[25] = x; + out[38] = x; + out[26] = x; + out[37] = x; + out[27] = x; + out[36] = x; + out[28] = x; + out[35] = x; + out[29] = x; + out[34] = x; + out[30] = x; + out[33] = x; + out[31] = x; + out[32] = x; + } +} + +static void idct64x64_low8_sse4_1(__m128i *in, __m128i *out, int bit, + int do_cols, int bd, int out_shift) { + int i, j; + const int32_t *cospi = cospi_arr(bit); + const __m128i rnding = _mm_set1_epi32(1 << (bit - 1)); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m128i clamp_lo = _mm_set1_epi32(-(1 << (log_range - 1))); + const __m128i clamp_hi = _mm_set1_epi32((1 << (log_range - 1)) - 1); + + const __m128i cospi1 = _mm_set1_epi32(cospi[1]); + const __m128i cospi2 = _mm_set1_epi32(cospi[2]); + const __m128i cospi3 = _mm_set1_epi32(cospi[3]); + const __m128i cospi4 = _mm_set1_epi32(cospi[4]); + const __m128i cospi6 = _mm_set1_epi32(cospi[6]); + const __m128i cospi8 = _mm_set1_epi32(cospi[8]); + const __m128i cospi12 = _mm_set1_epi32(cospi[12]); + const __m128i cospi16 = _mm_set1_epi32(cospi[16]); + const __m128i cospi20 = _mm_set1_epi32(cospi[20]); + const __m128i cospi24 = _mm_set1_epi32(cospi[24]); + const __m128i cospi28 = _mm_set1_epi32(cospi[28]); + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + const __m128i cospi40 = _mm_set1_epi32(cospi[40]); + const __m128i cospi44 = _mm_set1_epi32(cospi[44]); + const __m128i cospi48 = _mm_set1_epi32(cospi[48]); + const __m128i cospi56 = _mm_set1_epi32(cospi[56]); + const __m128i cospi60 = _mm_set1_epi32(cospi[60]); + const __m128i cospim4 = _mm_set1_epi32(-cospi[4]); + const __m128i cospim8 = _mm_set1_epi32(-cospi[8]); + const __m128i cospim12 = _mm_set1_epi32(-cospi[12]); + const __m128i cospim16 = _mm_set1_epi32(-cospi[16]); + const __m128i cospim20 = _mm_set1_epi32(-cospi[20]); + const __m128i cospim24 = _mm_set1_epi32(-cospi[24]); + const __m128i cospim28 = _mm_set1_epi32(-cospi[28]); + const __m128i cospim32 = _mm_set1_epi32(-cospi[32]); + const __m128i cospim36 = _mm_set1_epi32(-cospi[36]); + const __m128i cospim40 = _mm_set1_epi32(-cospi[40]); + const __m128i cospim48 = _mm_set1_epi32(-cospi[48]); + const __m128i cospim52 = _mm_set1_epi32(-cospi[52]); + const __m128i cospim56 = _mm_set1_epi32(-cospi[56]); + const __m128i cospi63 = _mm_set1_epi32(cospi[63]); + const __m128i cospim57 = _mm_set1_epi32(-cospi[57]); + const __m128i cospi7 = _mm_set1_epi32(cospi[7]); + const __m128i cospi5 = _mm_set1_epi32(cospi[5]); + const __m128i cospi59 = _mm_set1_epi32(cospi[59]); + const __m128i cospim61 = _mm_set1_epi32(-cospi[61]); + const __m128i cospim58 = _mm_set1_epi32(-cospi[58]); + const __m128i cospi62 = _mm_set1_epi32(cospi[62]); + + { + __m128i u[64]; + + // stage 1 + u[0] = in[0]; + u[8] = in[4]; + u[16] = in[2]; + u[24] = in[6]; + u[32] = in[1]; + u[40] = in[5]; + u[48] = in[3]; + u[56] = in[7]; + + // stage 2 + u[63] = half_btf_0_sse4_1(&cospi1, &u[32], &rnding, bit); + u[32] = half_btf_0_sse4_1(&cospi63, &u[32], &rnding, bit); + u[39] = half_btf_0_sse4_1(&cospim57, &u[56], &rnding, bit); + u[56] = half_btf_0_sse4_1(&cospi7, &u[56], &rnding, bit); + u[55] = half_btf_0_sse4_1(&cospi5, &u[40], &rnding, bit); + u[40] = half_btf_0_sse4_1(&cospi59, &u[40], &rnding, bit); + u[47] = half_btf_0_sse4_1(&cospim61, &u[48], &rnding, bit); + u[48] = half_btf_0_sse4_1(&cospi3, &u[48], &rnding, bit); + + // stage 3 + u[31] = half_btf_0_sse4_1(&cospi2, &u[16], &rnding, bit); + u[16] = half_btf_0_sse4_1(&cospi62, &u[16], &rnding, bit); + u[23] = half_btf_0_sse4_1(&cospim58, &u[24], &rnding, bit); + u[24] = half_btf_0_sse4_1(&cospi6, &u[24], &rnding, bit); + u[33] = u[32]; + u[38] = u[39]; + u[41] = u[40]; + u[46] = u[47]; + u[49] = u[48]; + u[54] = u[55]; + u[57] = u[56]; + u[62] = u[63]; + + // stage 4 + __m128i temp1, temp2; + u[15] = half_btf_0_sse4_1(&cospi4, &u[8], &rnding, bit); + u[8] = half_btf_0_sse4_1(&cospi60, &u[8], &rnding, bit); + u[17] = u[16]; + u[22] = u[23]; + u[25] = u[24]; + u[30] = u[31]; + + temp1 = half_btf_sse4_1(&cospim4, &u[33], &cospi60, &u[62], &rnding, bit); + u[62] = half_btf_sse4_1(&cospi60, &u[33], &cospi4, &u[62], &rnding, bit); + u[33] = temp1; + + temp2 = half_btf_sse4_1(&cospim36, &u[38], &cospi28, &u[57], &rnding, bit); + u[38] = half_btf_sse4_1(&cospim28, &u[38], &cospim36, &u[57], &rnding, bit); + u[57] = temp2; + + temp1 = half_btf_sse4_1(&cospim20, &u[41], &cospi44, &u[54], &rnding, bit); + u[54] = half_btf_sse4_1(&cospi44, &u[41], &cospi20, &u[54], &rnding, bit); + u[41] = temp1; + + temp2 = half_btf_sse4_1(&cospim12, &u[46], &cospim52, &u[49], &rnding, bit); + u[49] = half_btf_sse4_1(&cospim52, &u[46], &cospi12, &u[49], &rnding, bit); + u[46] = temp2; + + // stage 5 + u[9] = u[8]; + u[14] = u[15]; + + temp1 = half_btf_sse4_1(&cospim8, &u[17], &cospi56, &u[30], &rnding, bit); + u[30] = half_btf_sse4_1(&cospi56, &u[17], &cospi8, &u[30], &rnding, bit); + u[17] = temp1; + + temp2 = half_btf_sse4_1(&cospim24, &u[22], &cospim40, &u[25], &rnding, bit); + u[25] = half_btf_sse4_1(&cospim40, &u[22], &cospi24, &u[25], &rnding, bit); + u[22] = temp2; + + u[35] = u[32]; + u[34] = u[33]; + u[36] = u[39]; + u[37] = u[38]; + u[43] = u[40]; + u[42] = u[41]; + u[44] = u[47]; + u[45] = u[46]; + u[51] = u[48]; + u[50] = u[49]; + u[52] = u[55]; + u[53] = u[54]; + u[59] = u[56]; + u[58] = u[57]; + u[60] = u[63]; + u[61] = u[62]; + + // stage 6 + temp1 = half_btf_0_sse4_1(&cospi32, &u[0], &rnding, bit); + u[1] = half_btf_0_sse4_1(&cospi32, &u[0], &rnding, bit); + u[0] = temp1; + + temp2 = half_btf_sse4_1(&cospim16, &u[9], &cospi48, &u[14], &rnding, bit); + u[14] = half_btf_sse4_1(&cospi48, &u[9], &cospi16, &u[14], &rnding, bit); + u[9] = temp2; + u[19] = u[16]; + u[18] = u[17]; + u[20] = u[23]; + u[21] = u[22]; + u[27] = u[24]; + u[26] = u[25]; + u[28] = u[31]; + u[29] = u[30]; + + temp1 = half_btf_sse4_1(&cospim8, &u[34], &cospi56, &u[61], &rnding, bit); + u[61] = half_btf_sse4_1(&cospi56, &u[34], &cospi8, &u[61], &rnding, bit); + u[34] = temp1; + temp2 = half_btf_sse4_1(&cospim8, &u[35], &cospi56, &u[60], &rnding, bit); + u[60] = half_btf_sse4_1(&cospi56, &u[35], &cospi8, &u[60], &rnding, bit); + u[35] = temp2; + temp1 = half_btf_sse4_1(&cospim56, &u[36], &cospim8, &u[59], &rnding, bit); + u[59] = half_btf_sse4_1(&cospim8, &u[36], &cospi56, &u[59], &rnding, bit); + u[36] = temp1; + temp2 = half_btf_sse4_1(&cospim56, &u[37], &cospim8, &u[58], &rnding, bit); + u[58] = half_btf_sse4_1(&cospim8, &u[37], &cospi56, &u[58], &rnding, bit); + u[37] = temp2; + temp1 = half_btf_sse4_1(&cospim40, &u[42], &cospi24, &u[53], &rnding, bit); + u[53] = half_btf_sse4_1(&cospi24, &u[42], &cospi40, &u[53], &rnding, bit); + u[42] = temp1; + temp2 = half_btf_sse4_1(&cospim40, &u[43], &cospi24, &u[52], &rnding, bit); + u[52] = half_btf_sse4_1(&cospi24, &u[43], &cospi40, &u[52], &rnding, bit); + u[43] = temp2; + temp1 = half_btf_sse4_1(&cospim24, &u[44], &cospim40, &u[51], &rnding, bit); + u[51] = half_btf_sse4_1(&cospim40, &u[44], &cospi24, &u[51], &rnding, bit); + u[44] = temp1; + temp2 = half_btf_sse4_1(&cospim24, &u[45], &cospim40, &u[50], &rnding, bit); + u[50] = half_btf_sse4_1(&cospim40, &u[45], &cospi24, &u[50], &rnding, bit); + u[45] = temp2; + + // stage 7 + u[3] = u[0]; + u[2] = u[1]; + u[11] = u[8]; + u[10] = u[9]; + u[12] = u[15]; + u[13] = u[14]; + + temp1 = half_btf_sse4_1(&cospim16, &u[18], &cospi48, &u[29], &rnding, bit); + u[29] = half_btf_sse4_1(&cospi48, &u[18], &cospi16, &u[29], &rnding, bit); + u[18] = temp1; + temp2 = half_btf_sse4_1(&cospim16, &u[19], &cospi48, &u[28], &rnding, bit); + u[28] = half_btf_sse4_1(&cospi48, &u[19], &cospi16, &u[28], &rnding, bit); + u[19] = temp2; + temp1 = half_btf_sse4_1(&cospim48, &u[20], &cospim16, &u[27], &rnding, bit); + u[27] = half_btf_sse4_1(&cospim16, &u[20], &cospi48, &u[27], &rnding, bit); + u[20] = temp1; + temp2 = half_btf_sse4_1(&cospim48, &u[21], &cospim16, &u[26], &rnding, bit); + u[26] = half_btf_sse4_1(&cospim16, &u[21], &cospi48, &u[26], &rnding, bit); + u[21] = temp2; + for (i = 32; i < 64; i += 16) { + for (j = i; j < i + 4; j++) { + addsub_sse4_1(u[j], u[j ^ 7], &u[j], &u[j ^ 7], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[j ^ 15], u[j ^ 8], &u[j ^ 15], &u[j ^ 8], &clamp_lo, + &clamp_hi); + } + } + + // stage 8 + u[7] = u[0]; + u[6] = u[1]; + u[5] = u[2]; + u[4] = u[3]; + u[9] = u[9]; + + idct64_stage8_sse4_1(u, &cospim32, &cospi32, &cospim16, &cospi48, &cospi16, + &cospim48, &clamp_lo, &clamp_hi, &rnding, bit); + + // stage 9 + idct64_stage9_sse4_1(u, &cospim32, &cospi32, &clamp_lo, &clamp_hi, &rnding, + bit); + + // stage 10 + idct64_stage10_sse4_1(u, &cospim32, &cospi32, &clamp_lo, &clamp_hi, &rnding, + bit); + + // stage 11 + idct64_stage11_sse4_1(u, out, do_cols, bd, out_shift, log_range); + } +} + +static void idct64x64_low16_sse4_1(__m128i *in, __m128i *out, int bit, + int do_cols, int bd, int out_shift) { + int i, j; + const int32_t *cospi = cospi_arr(bit); + const __m128i rnding = _mm_set1_epi32(1 << (bit - 1)); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m128i clamp_lo = _mm_set1_epi32(-(1 << (log_range - 1))); + const __m128i clamp_hi = _mm_set1_epi32((1 << (log_range - 1)) - 1); + + const __m128i cospi1 = _mm_set1_epi32(cospi[1]); + const __m128i cospi2 = _mm_set1_epi32(cospi[2]); + const __m128i cospi3 = _mm_set1_epi32(cospi[3]); + const __m128i cospi4 = _mm_set1_epi32(cospi[4]); + const __m128i cospi5 = _mm_set1_epi32(cospi[5]); + const __m128i cospi6 = _mm_set1_epi32(cospi[6]); + const __m128i cospi7 = _mm_set1_epi32(cospi[7]); + const __m128i cospi8 = _mm_set1_epi32(cospi[8]); + const __m128i cospi9 = _mm_set1_epi32(cospi[9]); + const __m128i cospi10 = _mm_set1_epi32(cospi[10]); + const __m128i cospi11 = _mm_set1_epi32(cospi[11]); + const __m128i cospi12 = _mm_set1_epi32(cospi[12]); + const __m128i cospi13 = _mm_set1_epi32(cospi[13]); + const __m128i cospi14 = _mm_set1_epi32(cospi[14]); + const __m128i cospi15 = _mm_set1_epi32(cospi[15]); + const __m128i cospi16 = _mm_set1_epi32(cospi[16]); + const __m128i cospi20 = _mm_set1_epi32(cospi[20]); + const __m128i cospi24 = _mm_set1_epi32(cospi[24]); + const __m128i cospi28 = _mm_set1_epi32(cospi[28]); + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + const __m128i cospi36 = _mm_set1_epi32(cospi[36]); + const __m128i cospi40 = _mm_set1_epi32(cospi[40]); + const __m128i cospi44 = _mm_set1_epi32(cospi[44]); + const __m128i cospi48 = _mm_set1_epi32(cospi[48]); + const __m128i cospi51 = _mm_set1_epi32(cospi[51]); + const __m128i cospi52 = _mm_set1_epi32(cospi[52]); + const __m128i cospi54 = _mm_set1_epi32(cospi[54]); + const __m128i cospi55 = _mm_set1_epi32(cospi[55]); + const __m128i cospi56 = _mm_set1_epi32(cospi[56]); + const __m128i cospi59 = _mm_set1_epi32(cospi[59]); + const __m128i cospi60 = _mm_set1_epi32(cospi[60]); + const __m128i cospi62 = _mm_set1_epi32(cospi[62]); + const __m128i cospi63 = _mm_set1_epi32(cospi[63]); + + const __m128i cospim4 = _mm_set1_epi32(-cospi[4]); + const __m128i cospim8 = _mm_set1_epi32(-cospi[8]); + const __m128i cospim12 = _mm_set1_epi32(-cospi[12]); + const __m128i cospim16 = _mm_set1_epi32(-cospi[16]); + const __m128i cospim20 = _mm_set1_epi32(-cospi[20]); + const __m128i cospim24 = _mm_set1_epi32(-cospi[24]); + const __m128i cospim28 = _mm_set1_epi32(-cospi[28]); + const __m128i cospim32 = _mm_set1_epi32(-cospi[32]); + const __m128i cospim36 = _mm_set1_epi32(-cospi[36]); + const __m128i cospim40 = _mm_set1_epi32(-cospi[40]); + const __m128i cospim44 = _mm_set1_epi32(-cospi[44]); + const __m128i cospim48 = _mm_set1_epi32(-cospi[48]); + const __m128i cospim49 = _mm_set1_epi32(-cospi[49]); + const __m128i cospim50 = _mm_set1_epi32(-cospi[50]); + const __m128i cospim52 = _mm_set1_epi32(-cospi[52]); + const __m128i cospim53 = _mm_set1_epi32(-cospi[53]); + const __m128i cospim56 = _mm_set1_epi32(-cospi[56]); + const __m128i cospim57 = _mm_set1_epi32(-cospi[57]); + const __m128i cospim58 = _mm_set1_epi32(-cospi[58]); + const __m128i cospim60 = _mm_set1_epi32(-cospi[60]); + const __m128i cospim61 = _mm_set1_epi32(-cospi[61]); + + { + __m128i u[64]; + __m128i tmp1, tmp2, tmp3, tmp4; + // stage 1 + u[0] = in[0]; + u[32] = in[1]; + u[36] = in[9]; + u[40] = in[5]; + u[44] = in[13]; + u[48] = in[3]; + u[52] = in[11]; + u[56] = in[7]; + u[60] = in[15]; + u[16] = in[2]; + u[20] = in[10]; + u[24] = in[6]; + u[28] = in[14]; + u[4] = in[8]; + u[8] = in[4]; + u[12] = in[12]; + + // stage 2 + u[63] = half_btf_0_sse4_1(&cospi1, &u[32], &rnding, bit); + u[32] = half_btf_0_sse4_1(&cospi63, &u[32], &rnding, bit); + u[35] = half_btf_0_sse4_1(&cospim49, &u[60], &rnding, bit); + u[60] = half_btf_0_sse4_1(&cospi15, &u[60], &rnding, bit); + u[59] = half_btf_0_sse4_1(&cospi9, &u[36], &rnding, bit); + u[36] = half_btf_0_sse4_1(&cospi55, &u[36], &rnding, bit); + u[39] = half_btf_0_sse4_1(&cospim57, &u[56], &rnding, bit); + u[56] = half_btf_0_sse4_1(&cospi7, &u[56], &rnding, bit); + u[55] = half_btf_0_sse4_1(&cospi5, &u[40], &rnding, bit); + u[40] = half_btf_0_sse4_1(&cospi59, &u[40], &rnding, bit); + u[43] = half_btf_0_sse4_1(&cospim53, &u[52], &rnding, bit); + u[52] = half_btf_0_sse4_1(&cospi11, &u[52], &rnding, bit); + u[47] = half_btf_0_sse4_1(&cospim61, &u[48], &rnding, bit); + u[48] = half_btf_0_sse4_1(&cospi3, &u[48], &rnding, bit); + u[51] = half_btf_0_sse4_1(&cospi13, &u[44], &rnding, bit); + u[44] = half_btf_0_sse4_1(&cospi51, &u[44], &rnding, bit); + + // stage 3 + u[31] = half_btf_0_sse4_1(&cospi2, &u[16], &rnding, bit); + u[16] = half_btf_0_sse4_1(&cospi62, &u[16], &rnding, bit); + u[19] = half_btf_0_sse4_1(&cospim50, &u[28], &rnding, bit); + u[28] = half_btf_0_sse4_1(&cospi14, &u[28], &rnding, bit); + u[27] = half_btf_0_sse4_1(&cospi10, &u[20], &rnding, bit); + u[20] = half_btf_0_sse4_1(&cospi54, &u[20], &rnding, bit); + u[23] = half_btf_0_sse4_1(&cospim58, &u[24], &rnding, bit); + u[24] = half_btf_0_sse4_1(&cospi6, &u[24], &rnding, bit); + u[33] = u[32]; + u[34] = u[35]; + u[37] = u[36]; + u[38] = u[39]; + u[41] = u[40]; + u[42] = u[43]; + u[45] = u[44]; + u[46] = u[47]; + u[49] = u[48]; + u[50] = u[51]; + u[53] = u[52]; + u[54] = u[55]; + u[57] = u[56]; + u[58] = u[59]; + u[61] = u[60]; + u[62] = u[63]; + + // stage 4 + u[15] = half_btf_0_sse4_1(&cospi4, &u[8], &rnding, bit); + u[8] = half_btf_0_sse4_1(&cospi60, &u[8], &rnding, bit); + u[11] = half_btf_0_sse4_1(&cospim52, &u[12], &rnding, bit); + u[12] = half_btf_0_sse4_1(&cospi12, &u[12], &rnding, bit); + + u[17] = u[16]; + u[18] = u[19]; + u[21] = u[20]; + u[22] = u[23]; + u[25] = u[24]; + u[26] = u[27]; + u[29] = u[28]; + u[30] = u[31]; + + tmp1 = half_btf_sse4_1(&cospim4, &u[33], &cospi60, &u[62], &rnding, bit); + tmp2 = half_btf_sse4_1(&cospim60, &u[34], &cospim4, &u[61], &rnding, bit); + tmp3 = half_btf_sse4_1(&cospim36, &u[37], &cospi28, &u[58], &rnding, bit); + tmp4 = half_btf_sse4_1(&cospim28, &u[38], &cospim36, &u[57], &rnding, bit); + u[57] = half_btf_sse4_1(&cospim36, &u[38], &cospi28, &u[57], &rnding, bit); + u[58] = half_btf_sse4_1(&cospi28, &u[37], &cospi36, &u[58], &rnding, bit); + u[61] = half_btf_sse4_1(&cospim4, &u[34], &cospi60, &u[61], &rnding, bit); + u[62] = half_btf_sse4_1(&cospi60, &u[33], &cospi4, &u[62], &rnding, bit); + u[33] = tmp1; + u[34] = tmp2; + u[37] = tmp3; + u[38] = tmp4; + + tmp1 = half_btf_sse4_1(&cospim20, &u[41], &cospi44, &u[54], &rnding, bit); + tmp2 = half_btf_sse4_1(&cospim44, &u[42], &cospim20, &u[53], &rnding, bit); + tmp3 = half_btf_sse4_1(&cospim52, &u[45], &cospi12, &u[50], &rnding, bit); + tmp4 = half_btf_sse4_1(&cospim12, &u[46], &cospim52, &u[49], &rnding, bit); + u[49] = half_btf_sse4_1(&cospim52, &u[46], &cospi12, &u[49], &rnding, bit); + u[50] = half_btf_sse4_1(&cospi12, &u[45], &cospi52, &u[50], &rnding, bit); + u[53] = half_btf_sse4_1(&cospim20, &u[42], &cospi44, &u[53], &rnding, bit); + u[54] = half_btf_sse4_1(&cospi44, &u[41], &cospi20, &u[54], &rnding, bit); + u[41] = tmp1; + u[42] = tmp2; + u[45] = tmp3; + u[46] = tmp4; + + // stage 5 + u[7] = half_btf_0_sse4_1(&cospi8, &u[4], &rnding, bit); + u[4] = half_btf_0_sse4_1(&cospi56, &u[4], &rnding, bit); + + u[9] = u[8]; + u[10] = u[11]; + u[13] = u[12]; + u[14] = u[15]; + + tmp1 = half_btf_sse4_1(&cospim8, &u[17], &cospi56, &u[30], &rnding, bit); + tmp2 = half_btf_sse4_1(&cospim56, &u[18], &cospim8, &u[29], &rnding, bit); + tmp3 = half_btf_sse4_1(&cospim40, &u[21], &cospi24, &u[26], &rnding, bit); + tmp4 = half_btf_sse4_1(&cospim24, &u[22], &cospim40, &u[25], &rnding, bit); + u[25] = half_btf_sse4_1(&cospim40, &u[22], &cospi24, &u[25], &rnding, bit); + u[26] = half_btf_sse4_1(&cospi24, &u[21], &cospi40, &u[26], &rnding, bit); + u[29] = half_btf_sse4_1(&cospim8, &u[18], &cospi56, &u[29], &rnding, bit); + u[30] = half_btf_sse4_1(&cospi56, &u[17], &cospi8, &u[30], &rnding, bit); + u[17] = tmp1; + u[18] = tmp2; + u[21] = tmp3; + u[22] = tmp4; + + for (i = 32; i < 64; i += 8) { + addsub_sse4_1(u[i + 0], u[i + 3], &u[i + 0], &u[i + 3], &clamp_lo, + &clamp_hi); + addsub_sse4_1(u[i + 1], u[i + 2], &u[i + 1], &u[i + 2], &clamp_lo, + &clamp_hi); + + addsub_sse4_1(u[i + 7], u[i + 4], &u[i + 7], &u[i + 4], &clamp_lo, + &clamp_hi); + addsub_sse4_1(u[i + 6], u[i + 5], &u[i + 6], &u[i + 5], &clamp_lo, + &clamp_hi); + } + + // stage 6 + tmp1 = half_btf_0_sse4_1(&cospi32, &u[0], &rnding, bit); + u[1] = half_btf_0_sse4_1(&cospi32, &u[0], &rnding, bit); + u[0] = tmp1; + u[5] = u[4]; + u[6] = u[7]; + + tmp1 = half_btf_sse4_1(&cospim16, &u[9], &cospi48, &u[14], &rnding, bit); + u[14] = half_btf_sse4_1(&cospi48, &u[9], &cospi16, &u[14], &rnding, bit); + u[9] = tmp1; + tmp2 = half_btf_sse4_1(&cospim48, &u[10], &cospim16, &u[13], &rnding, bit); + u[13] = half_btf_sse4_1(&cospim16, &u[10], &cospi48, &u[13], &rnding, bit); + u[10] = tmp2; + + for (i = 16; i < 32; i += 8) { + addsub_sse4_1(u[i + 0], u[i + 3], &u[i + 0], &u[i + 3], &clamp_lo, + &clamp_hi); + addsub_sse4_1(u[i + 1], u[i + 2], &u[i + 1], &u[i + 2], &clamp_lo, + &clamp_hi); + + addsub_sse4_1(u[i + 7], u[i + 4], &u[i + 7], &u[i + 4], &clamp_lo, + &clamp_hi); + addsub_sse4_1(u[i + 6], u[i + 5], &u[i + 6], &u[i + 5], &clamp_lo, + &clamp_hi); + } + + tmp1 = half_btf_sse4_1(&cospim8, &u[34], &cospi56, &u[61], &rnding, bit); + tmp2 = half_btf_sse4_1(&cospim8, &u[35], &cospi56, &u[60], &rnding, bit); + tmp3 = half_btf_sse4_1(&cospim56, &u[36], &cospim8, &u[59], &rnding, bit); + tmp4 = half_btf_sse4_1(&cospim56, &u[37], &cospim8, &u[58], &rnding, bit); + u[58] = half_btf_sse4_1(&cospim8, &u[37], &cospi56, &u[58], &rnding, bit); + u[59] = half_btf_sse4_1(&cospim8, &u[36], &cospi56, &u[59], &rnding, bit); + u[60] = half_btf_sse4_1(&cospi56, &u[35], &cospi8, &u[60], &rnding, bit); + u[61] = half_btf_sse4_1(&cospi56, &u[34], &cospi8, &u[61], &rnding, bit); + u[34] = tmp1; + u[35] = tmp2; + u[36] = tmp3; + u[37] = tmp4; + + tmp1 = half_btf_sse4_1(&cospim40, &u[42], &cospi24, &u[53], &rnding, bit); + tmp2 = half_btf_sse4_1(&cospim40, &u[43], &cospi24, &u[52], &rnding, bit); + tmp3 = half_btf_sse4_1(&cospim24, &u[44], &cospim40, &u[51], &rnding, bit); + tmp4 = half_btf_sse4_1(&cospim24, &u[45], &cospim40, &u[50], &rnding, bit); + u[50] = half_btf_sse4_1(&cospim40, &u[45], &cospi24, &u[50], &rnding, bit); + u[51] = half_btf_sse4_1(&cospim40, &u[44], &cospi24, &u[51], &rnding, bit); + u[52] = half_btf_sse4_1(&cospi24, &u[43], &cospi40, &u[52], &rnding, bit); + u[53] = half_btf_sse4_1(&cospi24, &u[42], &cospi40, &u[53], &rnding, bit); + u[42] = tmp1; + u[43] = tmp2; + u[44] = tmp3; + u[45] = tmp4; + + // stage 7 + u[3] = u[0]; + u[2] = u[1]; + tmp1 = half_btf_sse4_1(&cospim32, &u[5], &cospi32, &u[6], &rnding, bit); + u[6] = half_btf_sse4_1(&cospi32, &u[5], &cospi32, &u[6], &rnding, bit); + u[5] = tmp1; + addsub_sse4_1(u[8], u[11], &u[8], &u[11], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[9], u[10], &u[9], &u[10], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[15], u[12], &u[15], &u[12], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[14], u[13], &u[14], &u[13], &clamp_lo, &clamp_hi); + + tmp1 = half_btf_sse4_1(&cospim16, &u[18], &cospi48, &u[29], &rnding, bit); + tmp2 = half_btf_sse4_1(&cospim16, &u[19], &cospi48, &u[28], &rnding, bit); + tmp3 = half_btf_sse4_1(&cospim48, &u[20], &cospim16, &u[27], &rnding, bit); + tmp4 = half_btf_sse4_1(&cospim48, &u[21], &cospim16, &u[26], &rnding, bit); + u[26] = half_btf_sse4_1(&cospim16, &u[21], &cospi48, &u[26], &rnding, bit); + u[27] = half_btf_sse4_1(&cospim16, &u[20], &cospi48, &u[27], &rnding, bit); + u[28] = half_btf_sse4_1(&cospi48, &u[19], &cospi16, &u[28], &rnding, bit); + u[29] = half_btf_sse4_1(&cospi48, &u[18], &cospi16, &u[29], &rnding, bit); + u[18] = tmp1; + u[19] = tmp2; + u[20] = tmp3; + u[21] = tmp4; + + for (i = 32; i < 64; i += 16) { + for (j = i; j < i + 4; j++) { + addsub_sse4_1(u[j], u[j ^ 7], &u[j], &u[j ^ 7], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[j ^ 15], u[j ^ 8], &u[j ^ 15], &u[j ^ 8], &clamp_lo, + &clamp_hi); + } + } + + // stage 8 + for (i = 0; i < 4; ++i) { + addsub_sse4_1(u[i], u[7 - i], &u[i], &u[7 - i], &clamp_lo, &clamp_hi); + } + + idct64_stage8_sse4_1(u, &cospim32, &cospi32, &cospim16, &cospi48, &cospi16, + &cospim48, &clamp_lo, &clamp_hi, &rnding, bit); + + // stage 9 + idct64_stage9_sse4_1(u, &cospim32, &cospi32, &clamp_lo, &clamp_hi, &rnding, + bit); + + // stage 10 + idct64_stage10_sse4_1(u, &cospim32, &cospi32, &clamp_lo, &clamp_hi, &rnding, + bit); + + // stage 11 + idct64_stage11_sse4_1(u, out, do_cols, bd, out_shift, log_range); + } +} + +static void idct64x64_sse4_1(__m128i *in, __m128i *out, int bit, int do_cols, + int bd, int out_shift) { + int i, j; + const int32_t *cospi = cospi_arr(bit); + const __m128i rnding = _mm_set1_epi32(1 << (bit - 1)); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m128i clamp_lo = _mm_set1_epi32(-(1 << (log_range - 1))); + const __m128i clamp_hi = _mm_set1_epi32((1 << (log_range - 1)) - 1); + + const __m128i cospi1 = _mm_set1_epi32(cospi[1]); + const __m128i cospi2 = _mm_set1_epi32(cospi[2]); + const __m128i cospi3 = _mm_set1_epi32(cospi[3]); + const __m128i cospi4 = _mm_set1_epi32(cospi[4]); + const __m128i cospi5 = _mm_set1_epi32(cospi[5]); + const __m128i cospi6 = _mm_set1_epi32(cospi[6]); + const __m128i cospi7 = _mm_set1_epi32(cospi[7]); + const __m128i cospi8 = _mm_set1_epi32(cospi[8]); + const __m128i cospi9 = _mm_set1_epi32(cospi[9]); + const __m128i cospi10 = _mm_set1_epi32(cospi[10]); + const __m128i cospi11 = _mm_set1_epi32(cospi[11]); + const __m128i cospi12 = _mm_set1_epi32(cospi[12]); + const __m128i cospi13 = _mm_set1_epi32(cospi[13]); + const __m128i cospi14 = _mm_set1_epi32(cospi[14]); + const __m128i cospi15 = _mm_set1_epi32(cospi[15]); + const __m128i cospi16 = _mm_set1_epi32(cospi[16]); + const __m128i cospi17 = _mm_set1_epi32(cospi[17]); + const __m128i cospi18 = _mm_set1_epi32(cospi[18]); + const __m128i cospi19 = _mm_set1_epi32(cospi[19]); + const __m128i cospi20 = _mm_set1_epi32(cospi[20]); + const __m128i cospi21 = _mm_set1_epi32(cospi[21]); + const __m128i cospi22 = _mm_set1_epi32(cospi[22]); + const __m128i cospi23 = _mm_set1_epi32(cospi[23]); + const __m128i cospi24 = _mm_set1_epi32(cospi[24]); + const __m128i cospi25 = _mm_set1_epi32(cospi[25]); + const __m128i cospi26 = _mm_set1_epi32(cospi[26]); + const __m128i cospi27 = _mm_set1_epi32(cospi[27]); + const __m128i cospi28 = _mm_set1_epi32(cospi[28]); + const __m128i cospi29 = _mm_set1_epi32(cospi[29]); + const __m128i cospi30 = _mm_set1_epi32(cospi[30]); + const __m128i cospi31 = _mm_set1_epi32(cospi[31]); + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + const __m128i cospi35 = _mm_set1_epi32(cospi[35]); + const __m128i cospi36 = _mm_set1_epi32(cospi[36]); + const __m128i cospi38 = _mm_set1_epi32(cospi[38]); + const __m128i cospi39 = _mm_set1_epi32(cospi[39]); + const __m128i cospi40 = _mm_set1_epi32(cospi[40]); + const __m128i cospi43 = _mm_set1_epi32(cospi[43]); + const __m128i cospi44 = _mm_set1_epi32(cospi[44]); + const __m128i cospi46 = _mm_set1_epi32(cospi[46]); + const __m128i cospi47 = _mm_set1_epi32(cospi[47]); + const __m128i cospi48 = _mm_set1_epi32(cospi[48]); + const __m128i cospi51 = _mm_set1_epi32(cospi[51]); + const __m128i cospi52 = _mm_set1_epi32(cospi[52]); + const __m128i cospi54 = _mm_set1_epi32(cospi[54]); + const __m128i cospi55 = _mm_set1_epi32(cospi[55]); + const __m128i cospi56 = _mm_set1_epi32(cospi[56]); + const __m128i cospi59 = _mm_set1_epi32(cospi[59]); + const __m128i cospi60 = _mm_set1_epi32(cospi[60]); + const __m128i cospi62 = _mm_set1_epi32(cospi[62]); + const __m128i cospi63 = _mm_set1_epi32(cospi[63]); + + const __m128i cospim4 = _mm_set1_epi32(-cospi[4]); + const __m128i cospim8 = _mm_set1_epi32(-cospi[8]); + const __m128i cospim12 = _mm_set1_epi32(-cospi[12]); + const __m128i cospim16 = _mm_set1_epi32(-cospi[16]); + const __m128i cospim20 = _mm_set1_epi32(-cospi[20]); + const __m128i cospim24 = _mm_set1_epi32(-cospi[24]); + const __m128i cospim28 = _mm_set1_epi32(-cospi[28]); + const __m128i cospim32 = _mm_set1_epi32(-cospi[32]); + const __m128i cospim33 = _mm_set1_epi32(-cospi[33]); + const __m128i cospim34 = _mm_set1_epi32(-cospi[34]); + const __m128i cospim36 = _mm_set1_epi32(-cospi[36]); + const __m128i cospim37 = _mm_set1_epi32(-cospi[37]); + const __m128i cospim40 = _mm_set1_epi32(-cospi[40]); + const __m128i cospim41 = _mm_set1_epi32(-cospi[41]); + const __m128i cospim42 = _mm_set1_epi32(-cospi[42]); + const __m128i cospim44 = _mm_set1_epi32(-cospi[44]); + const __m128i cospim45 = _mm_set1_epi32(-cospi[45]); + const __m128i cospim48 = _mm_set1_epi32(-cospi[48]); + const __m128i cospim49 = _mm_set1_epi32(-cospi[49]); + const __m128i cospim50 = _mm_set1_epi32(-cospi[50]); + const __m128i cospim52 = _mm_set1_epi32(-cospi[52]); + const __m128i cospim53 = _mm_set1_epi32(-cospi[53]); + const __m128i cospim56 = _mm_set1_epi32(-cospi[56]); + const __m128i cospim57 = _mm_set1_epi32(-cospi[57]); + const __m128i cospim58 = _mm_set1_epi32(-cospi[58]); + const __m128i cospim60 = _mm_set1_epi32(-cospi[60]); + const __m128i cospim61 = _mm_set1_epi32(-cospi[61]); + + { + __m128i u[64], v[64]; + + // stage 1 + u[32] = in[1]; + u[34] = in[17]; + u[36] = in[9]; + u[38] = in[25]; + u[40] = in[5]; + u[42] = in[21]; + u[44] = in[13]; + u[46] = in[29]; + u[48] = in[3]; + u[50] = in[19]; + u[52] = in[11]; + u[54] = in[27]; + u[56] = in[7]; + u[58] = in[23]; + u[60] = in[15]; + u[62] = in[31]; + + v[16] = in[2]; + v[18] = in[18]; + v[20] = in[10]; + v[22] = in[26]; + v[24] = in[6]; + v[26] = in[22]; + v[28] = in[14]; + v[30] = in[30]; + + u[8] = in[4]; + u[10] = in[20]; + u[12] = in[12]; + u[14] = in[28]; + + v[4] = in[8]; + v[6] = in[24]; + + u[0] = in[0]; + u[2] = in[16]; + + // stage 2 + v[32] = half_btf_0_sse4_1(&cospi63, &u[32], &rnding, bit); + v[33] = half_btf_0_sse4_1(&cospim33, &u[62], &rnding, bit); + v[34] = half_btf_0_sse4_1(&cospi47, &u[34], &rnding, bit); + v[35] = half_btf_0_sse4_1(&cospim49, &u[60], &rnding, bit); + v[36] = half_btf_0_sse4_1(&cospi55, &u[36], &rnding, bit); + v[37] = half_btf_0_sse4_1(&cospim41, &u[58], &rnding, bit); + v[38] = half_btf_0_sse4_1(&cospi39, &u[38], &rnding, bit); + v[39] = half_btf_0_sse4_1(&cospim57, &u[56], &rnding, bit); + v[40] = half_btf_0_sse4_1(&cospi59, &u[40], &rnding, bit); + v[41] = half_btf_0_sse4_1(&cospim37, &u[54], &rnding, bit); + v[42] = half_btf_0_sse4_1(&cospi43, &u[42], &rnding, bit); + v[43] = half_btf_0_sse4_1(&cospim53, &u[52], &rnding, bit); + v[44] = half_btf_0_sse4_1(&cospi51, &u[44], &rnding, bit); + v[45] = half_btf_0_sse4_1(&cospim45, &u[50], &rnding, bit); + v[46] = half_btf_0_sse4_1(&cospi35, &u[46], &rnding, bit); + v[47] = half_btf_0_sse4_1(&cospim61, &u[48], &rnding, bit); + v[48] = half_btf_0_sse4_1(&cospi3, &u[48], &rnding, bit); + v[49] = half_btf_0_sse4_1(&cospi29, &u[46], &rnding, bit); + v[50] = half_btf_0_sse4_1(&cospi19, &u[50], &rnding, bit); + v[51] = half_btf_0_sse4_1(&cospi13, &u[44], &rnding, bit); + v[52] = half_btf_0_sse4_1(&cospi11, &u[52], &rnding, bit); + v[53] = half_btf_0_sse4_1(&cospi21, &u[42], &rnding, bit); + v[54] = half_btf_0_sse4_1(&cospi27, &u[54], &rnding, bit); + v[55] = half_btf_0_sse4_1(&cospi5, &u[40], &rnding, bit); + v[56] = half_btf_0_sse4_1(&cospi7, &u[56], &rnding, bit); + v[57] = half_btf_0_sse4_1(&cospi25, &u[38], &rnding, bit); + v[58] = half_btf_0_sse4_1(&cospi23, &u[58], &rnding, bit); + v[59] = half_btf_0_sse4_1(&cospi9, &u[36], &rnding, bit); + v[60] = half_btf_0_sse4_1(&cospi15, &u[60], &rnding, bit); + v[61] = half_btf_0_sse4_1(&cospi17, &u[34], &rnding, bit); + v[62] = half_btf_0_sse4_1(&cospi31, &u[62], &rnding, bit); + v[63] = half_btf_0_sse4_1(&cospi1, &u[32], &rnding, bit); + + // stage 3 + u[16] = half_btf_0_sse4_1(&cospi62, &v[16], &rnding, bit); + u[17] = half_btf_0_sse4_1(&cospim34, &v[30], &rnding, bit); + u[18] = half_btf_0_sse4_1(&cospi46, &v[18], &rnding, bit); + u[19] = half_btf_0_sse4_1(&cospim50, &v[28], &rnding, bit); + u[20] = half_btf_0_sse4_1(&cospi54, &v[20], &rnding, bit); + u[21] = half_btf_0_sse4_1(&cospim42, &v[26], &rnding, bit); + u[22] = half_btf_0_sse4_1(&cospi38, &v[22], &rnding, bit); + u[23] = half_btf_0_sse4_1(&cospim58, &v[24], &rnding, bit); + u[24] = half_btf_0_sse4_1(&cospi6, &v[24], &rnding, bit); + u[25] = half_btf_0_sse4_1(&cospi26, &v[22], &rnding, bit); + u[26] = half_btf_0_sse4_1(&cospi22, &v[26], &rnding, bit); + u[27] = half_btf_0_sse4_1(&cospi10, &v[20], &rnding, bit); + u[28] = half_btf_0_sse4_1(&cospi14, &v[28], &rnding, bit); + u[29] = half_btf_0_sse4_1(&cospi18, &v[18], &rnding, bit); + u[30] = half_btf_0_sse4_1(&cospi30, &v[30], &rnding, bit); + u[31] = half_btf_0_sse4_1(&cospi2, &v[16], &rnding, bit); + + for (i = 32; i < 64; i += 4) { + addsub_sse4_1(v[i + 0], v[i + 1], &u[i + 0], &u[i + 1], &clamp_lo, + &clamp_hi); + addsub_sse4_1(v[i + 3], v[i + 2], &u[i + 3], &u[i + 2], &clamp_lo, + &clamp_hi); + } + + // stage 4 + v[8] = half_btf_0_sse4_1(&cospi60, &u[8], &rnding, bit); + v[9] = half_btf_0_sse4_1(&cospim36, &u[14], &rnding, bit); + v[10] = half_btf_0_sse4_1(&cospi44, &u[10], &rnding, bit); + v[11] = half_btf_0_sse4_1(&cospim52, &u[12], &rnding, bit); + v[12] = half_btf_0_sse4_1(&cospi12, &u[12], &rnding, bit); + v[13] = half_btf_0_sse4_1(&cospi20, &u[10], &rnding, bit); + v[14] = half_btf_0_sse4_1(&cospi28, &u[14], &rnding, bit); + v[15] = half_btf_0_sse4_1(&cospi4, &u[8], &rnding, bit); + + for (i = 16; i < 32; i += 4) { + addsub_sse4_1(u[i + 0], u[i + 1], &v[i + 0], &v[i + 1], &clamp_lo, + &clamp_hi); + addsub_sse4_1(u[i + 3], u[i + 2], &v[i + 3], &v[i + 2], &clamp_lo, + &clamp_hi); + } + + for (i = 32; i < 64; i += 4) { + v[i + 0] = u[i + 0]; + v[i + 3] = u[i + 3]; + } + + v[33] = half_btf_sse4_1(&cospim4, &u[33], &cospi60, &u[62], &rnding, bit); + v[34] = half_btf_sse4_1(&cospim60, &u[34], &cospim4, &u[61], &rnding, bit); + v[37] = half_btf_sse4_1(&cospim36, &u[37], &cospi28, &u[58], &rnding, bit); + v[38] = half_btf_sse4_1(&cospim28, &u[38], &cospim36, &u[57], &rnding, bit); + v[41] = half_btf_sse4_1(&cospim20, &u[41], &cospi44, &u[54], &rnding, bit); + v[42] = half_btf_sse4_1(&cospim44, &u[42], &cospim20, &u[53], &rnding, bit); + v[45] = half_btf_sse4_1(&cospim52, &u[45], &cospi12, &u[50], &rnding, bit); + v[46] = half_btf_sse4_1(&cospim12, &u[46], &cospim52, &u[49], &rnding, bit); + v[49] = half_btf_sse4_1(&cospim52, &u[46], &cospi12, &u[49], &rnding, bit); + v[50] = half_btf_sse4_1(&cospi12, &u[45], &cospi52, &u[50], &rnding, bit); + v[53] = half_btf_sse4_1(&cospim20, &u[42], &cospi44, &u[53], &rnding, bit); + v[54] = half_btf_sse4_1(&cospi44, &u[41], &cospi20, &u[54], &rnding, bit); + v[57] = half_btf_sse4_1(&cospim36, &u[38], &cospi28, &u[57], &rnding, bit); + v[58] = half_btf_sse4_1(&cospi28, &u[37], &cospi36, &u[58], &rnding, bit); + v[61] = half_btf_sse4_1(&cospim4, &u[34], &cospi60, &u[61], &rnding, bit); + v[62] = half_btf_sse4_1(&cospi60, &u[33], &cospi4, &u[62], &rnding, bit); + + // stage 5 + u[4] = half_btf_0_sse4_1(&cospi56, &v[4], &rnding, bit); + u[5] = half_btf_0_sse4_1(&cospim40, &v[6], &rnding, bit); + u[6] = half_btf_0_sse4_1(&cospi24, &v[6], &rnding, bit); + u[7] = half_btf_0_sse4_1(&cospi8, &v[4], &rnding, bit); + + for (i = 8; i < 16; i += 4) { + addsub_sse4_1(v[i + 0], v[i + 1], &u[i + 0], &u[i + 1], &clamp_lo, + &clamp_hi); + addsub_sse4_1(v[i + 3], v[i + 2], &u[i + 3], &u[i + 2], &clamp_lo, + &clamp_hi); + } + + for (i = 16; i < 32; i += 4) { + u[i + 0] = v[i + 0]; + u[i + 3] = v[i + 3]; + } + + u[17] = half_btf_sse4_1(&cospim8, &v[17], &cospi56, &v[30], &rnding, bit); + u[18] = half_btf_sse4_1(&cospim56, &v[18], &cospim8, &v[29], &rnding, bit); + u[21] = half_btf_sse4_1(&cospim40, &v[21], &cospi24, &v[26], &rnding, bit); + u[22] = half_btf_sse4_1(&cospim24, &v[22], &cospim40, &v[25], &rnding, bit); + u[25] = half_btf_sse4_1(&cospim40, &v[22], &cospi24, &v[25], &rnding, bit); + u[26] = half_btf_sse4_1(&cospi24, &v[21], &cospi40, &v[26], &rnding, bit); + u[29] = half_btf_sse4_1(&cospim8, &v[18], &cospi56, &v[29], &rnding, bit); + u[30] = half_btf_sse4_1(&cospi56, &v[17], &cospi8, &v[30], &rnding, bit); + + for (i = 32; i < 64; i += 8) { + addsub_sse4_1(v[i + 0], v[i + 3], &u[i + 0], &u[i + 3], &clamp_lo, + &clamp_hi); + addsub_sse4_1(v[i + 1], v[i + 2], &u[i + 1], &u[i + 2], &clamp_lo, + &clamp_hi); + + addsub_sse4_1(v[i + 7], v[i + 4], &u[i + 7], &u[i + 4], &clamp_lo, + &clamp_hi); + addsub_sse4_1(v[i + 6], v[i + 5], &u[i + 6], &u[i + 5], &clamp_lo, + &clamp_hi); + } + + // stage 6 + v[0] = half_btf_0_sse4_1(&cospi32, &u[0], &rnding, bit); + v[1] = half_btf_0_sse4_1(&cospi32, &u[0], &rnding, bit); + v[2] = half_btf_0_sse4_1(&cospi48, &u[2], &rnding, bit); + v[3] = half_btf_0_sse4_1(&cospi16, &u[2], &rnding, bit); + + addsub_sse4_1(u[4], u[5], &v[4], &v[5], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[7], u[6], &v[7], &v[6], &clamp_lo, &clamp_hi); + + for (i = 8; i < 16; i += 4) { + v[i + 0] = u[i + 0]; + v[i + 3] = u[i + 3]; + } + + v[9] = half_btf_sse4_1(&cospim16, &u[9], &cospi48, &u[14], &rnding, bit); + v[10] = half_btf_sse4_1(&cospim48, &u[10], &cospim16, &u[13], &rnding, bit); + v[13] = half_btf_sse4_1(&cospim16, &u[10], &cospi48, &u[13], &rnding, bit); + v[14] = half_btf_sse4_1(&cospi48, &u[9], &cospi16, &u[14], &rnding, bit); + + for (i = 16; i < 32; i += 8) { + addsub_sse4_1(u[i + 0], u[i + 3], &v[i + 0], &v[i + 3], &clamp_lo, + &clamp_hi); + addsub_sse4_1(u[i + 1], u[i + 2], &v[i + 1], &v[i + 2], &clamp_lo, + &clamp_hi); + + addsub_sse4_1(u[i + 7], u[i + 4], &v[i + 7], &v[i + 4], &clamp_lo, + &clamp_hi); + addsub_sse4_1(u[i + 6], u[i + 5], &v[i + 6], &v[i + 5], &clamp_lo, + &clamp_hi); + } + + for (i = 32; i < 64; i += 8) { + v[i + 0] = u[i + 0]; + v[i + 1] = u[i + 1]; + v[i + 6] = u[i + 6]; + v[i + 7] = u[i + 7]; + } + + v[34] = half_btf_sse4_1(&cospim8, &u[34], &cospi56, &u[61], &rnding, bit); + v[35] = half_btf_sse4_1(&cospim8, &u[35], &cospi56, &u[60], &rnding, bit); + v[36] = half_btf_sse4_1(&cospim56, &u[36], &cospim8, &u[59], &rnding, bit); + v[37] = half_btf_sse4_1(&cospim56, &u[37], &cospim8, &u[58], &rnding, bit); + v[42] = half_btf_sse4_1(&cospim40, &u[42], &cospi24, &u[53], &rnding, bit); + v[43] = half_btf_sse4_1(&cospim40, &u[43], &cospi24, &u[52], &rnding, bit); + v[44] = half_btf_sse4_1(&cospim24, &u[44], &cospim40, &u[51], &rnding, bit); + v[45] = half_btf_sse4_1(&cospim24, &u[45], &cospim40, &u[50], &rnding, bit); + v[50] = half_btf_sse4_1(&cospim40, &u[45], &cospi24, &u[50], &rnding, bit); + v[51] = half_btf_sse4_1(&cospim40, &u[44], &cospi24, &u[51], &rnding, bit); + v[52] = half_btf_sse4_1(&cospi24, &u[43], &cospi40, &u[52], &rnding, bit); + v[53] = half_btf_sse4_1(&cospi24, &u[42], &cospi40, &u[53], &rnding, bit); + v[58] = half_btf_sse4_1(&cospim8, &u[37], &cospi56, &u[58], &rnding, bit); + v[59] = half_btf_sse4_1(&cospim8, &u[36], &cospi56, &u[59], &rnding, bit); + v[60] = half_btf_sse4_1(&cospi56, &u[35], &cospi8, &u[60], &rnding, bit); + v[61] = half_btf_sse4_1(&cospi56, &u[34], &cospi8, &u[61], &rnding, bit); + + // stage 7 + addsub_sse4_1(v[0], v[3], &u[0], &u[3], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[1], v[2], &u[1], &u[2], &clamp_lo, &clamp_hi); + + u[4] = v[4]; + u[7] = v[7]; + u[5] = half_btf_sse4_1(&cospim32, &v[5], &cospi32, &v[6], &rnding, bit); + u[6] = half_btf_sse4_1(&cospi32, &v[5], &cospi32, &v[6], &rnding, bit); + + addsub_sse4_1(v[8], v[11], &u[8], &u[11], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[9], v[10], &u[9], &u[10], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[15], v[12], &u[15], &u[12], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[14], v[13], &u[14], &u[13], &clamp_lo, &clamp_hi); + + for (i = 16; i < 32; i += 8) { + u[i + 0] = v[i + 0]; + u[i + 1] = v[i + 1]; + u[i + 6] = v[i + 6]; + u[i + 7] = v[i + 7]; + } + + u[18] = half_btf_sse4_1(&cospim16, &v[18], &cospi48, &v[29], &rnding, bit); + u[19] = half_btf_sse4_1(&cospim16, &v[19], &cospi48, &v[28], &rnding, bit); + u[20] = half_btf_sse4_1(&cospim48, &v[20], &cospim16, &v[27], &rnding, bit); + u[21] = half_btf_sse4_1(&cospim48, &v[21], &cospim16, &v[26], &rnding, bit); + u[26] = half_btf_sse4_1(&cospim16, &v[21], &cospi48, &v[26], &rnding, bit); + u[27] = half_btf_sse4_1(&cospim16, &v[20], &cospi48, &v[27], &rnding, bit); + u[28] = half_btf_sse4_1(&cospi48, &v[19], &cospi16, &v[28], &rnding, bit); + u[29] = half_btf_sse4_1(&cospi48, &v[18], &cospi16, &v[29], &rnding, bit); + + for (i = 32; i < 64; i += 16) { + for (j = i; j < i + 4; j++) { + addsub_sse4_1(v[j], v[j ^ 7], &u[j], &u[j ^ 7], &clamp_lo, &clamp_hi); + addsub_sse4_1(v[j ^ 15], v[j ^ 8], &u[j ^ 15], &u[j ^ 8], &clamp_lo, + &clamp_hi); + } + } + + // stage 8 + for (i = 0; i < 4; ++i) { + addsub_sse4_1(u[i], u[7 - i], &v[i], &v[7 - i], &clamp_lo, &clamp_hi); + } + + v[8] = u[8]; + v[9] = u[9]; + v[14] = u[14]; + v[15] = u[15]; + + v[10] = half_btf_sse4_1(&cospim32, &u[10], &cospi32, &u[13], &rnding, bit); + v[11] = half_btf_sse4_1(&cospim32, &u[11], &cospi32, &u[12], &rnding, bit); + v[12] = half_btf_sse4_1(&cospi32, &u[11], &cospi32, &u[12], &rnding, bit); + v[13] = half_btf_sse4_1(&cospi32, &u[10], &cospi32, &u[13], &rnding, bit); + + for (i = 16; i < 20; ++i) { + addsub_sse4_1(u[i], u[i ^ 7], &v[i], &v[i ^ 7], &clamp_lo, &clamp_hi); + addsub_sse4_1(u[i ^ 15], u[i ^ 8], &v[i ^ 15], &v[i ^ 8], &clamp_lo, + &clamp_hi); + } + + for (i = 32; i < 36; ++i) { + v[i] = u[i]; + v[i + 12] = u[i + 12]; + v[i + 16] = u[i + 16]; + v[i + 28] = u[i + 28]; + } + + v[36] = half_btf_sse4_1(&cospim16, &u[36], &cospi48, &u[59], &rnding, bit); + v[37] = half_btf_sse4_1(&cospim16, &u[37], &cospi48, &u[58], &rnding, bit); + v[38] = half_btf_sse4_1(&cospim16, &u[38], &cospi48, &u[57], &rnding, bit); + v[39] = half_btf_sse4_1(&cospim16, &u[39], &cospi48, &u[56], &rnding, bit); + v[40] = half_btf_sse4_1(&cospim48, &u[40], &cospim16, &u[55], &rnding, bit); + v[41] = half_btf_sse4_1(&cospim48, &u[41], &cospim16, &u[54], &rnding, bit); + v[42] = half_btf_sse4_1(&cospim48, &u[42], &cospim16, &u[53], &rnding, bit); + v[43] = half_btf_sse4_1(&cospim48, &u[43], &cospim16, &u[52], &rnding, bit); + v[52] = half_btf_sse4_1(&cospim16, &u[43], &cospi48, &u[52], &rnding, bit); + v[53] = half_btf_sse4_1(&cospim16, &u[42], &cospi48, &u[53], &rnding, bit); + v[54] = half_btf_sse4_1(&cospim16, &u[41], &cospi48, &u[54], &rnding, bit); + v[55] = half_btf_sse4_1(&cospim16, &u[40], &cospi48, &u[55], &rnding, bit); + v[56] = half_btf_sse4_1(&cospi48, &u[39], &cospi16, &u[56], &rnding, bit); + v[57] = half_btf_sse4_1(&cospi48, &u[38], &cospi16, &u[57], &rnding, bit); + v[58] = half_btf_sse4_1(&cospi48, &u[37], &cospi16, &u[58], &rnding, bit); + v[59] = half_btf_sse4_1(&cospi48, &u[36], &cospi16, &u[59], &rnding, bit); + + // stage 9 + for (i = 0; i < 8; ++i) { + addsub_sse4_1(v[i], v[15 - i], &u[i], &u[15 - i], &clamp_lo, &clamp_hi); + } + + for (i = 16; i < 20; ++i) { + u[i] = v[i]; + u[i + 12] = v[i + 12]; + } + + u[20] = half_btf_sse4_1(&cospim32, &v[20], &cospi32, &v[27], &rnding, bit); + u[21] = half_btf_sse4_1(&cospim32, &v[21], &cospi32, &v[26], &rnding, bit); + u[22] = half_btf_sse4_1(&cospim32, &v[22], &cospi32, &v[25], &rnding, bit); + u[23] = half_btf_sse4_1(&cospim32, &v[23], &cospi32, &v[24], &rnding, bit); + u[24] = half_btf_sse4_1(&cospi32, &v[23], &cospi32, &v[24], &rnding, bit); + u[25] = half_btf_sse4_1(&cospi32, &v[22], &cospi32, &v[25], &rnding, bit); + u[26] = half_btf_sse4_1(&cospi32, &v[21], &cospi32, &v[26], &rnding, bit); + u[27] = half_btf_sse4_1(&cospi32, &v[20], &cospi32, &v[27], &rnding, bit); + + for (i = 32; i < 40; i++) { + addsub_sse4_1(v[i], v[i ^ 15], &u[i], &u[i ^ 15], &clamp_lo, &clamp_hi); + } + + for (i = 48; i < 56; i++) { + addsub_sse4_1(v[i ^ 15], v[i], &u[i ^ 15], &u[i], &clamp_lo, &clamp_hi); + } + + // stage 10 + for (i = 0; i < 16; i++) { + addsub_sse4_1(u[i], u[31 - i], &v[i], &v[31 - i], &clamp_lo, &clamp_hi); + } + + for (i = 32; i < 40; i++) v[i] = u[i]; + + v[40] = half_btf_sse4_1(&cospim32, &u[40], &cospi32, &u[55], &rnding, bit); + v[41] = half_btf_sse4_1(&cospim32, &u[41], &cospi32, &u[54], &rnding, bit); + v[42] = half_btf_sse4_1(&cospim32, &u[42], &cospi32, &u[53], &rnding, bit); + v[43] = half_btf_sse4_1(&cospim32, &u[43], &cospi32, &u[52], &rnding, bit); + v[44] = half_btf_sse4_1(&cospim32, &u[44], &cospi32, &u[51], &rnding, bit); + v[45] = half_btf_sse4_1(&cospim32, &u[45], &cospi32, &u[50], &rnding, bit); + v[46] = half_btf_sse4_1(&cospim32, &u[46], &cospi32, &u[49], &rnding, bit); + v[47] = half_btf_sse4_1(&cospim32, &u[47], &cospi32, &u[48], &rnding, bit); + v[48] = half_btf_sse4_1(&cospi32, &u[47], &cospi32, &u[48], &rnding, bit); + v[49] = half_btf_sse4_1(&cospi32, &u[46], &cospi32, &u[49], &rnding, bit); + v[50] = half_btf_sse4_1(&cospi32, &u[45], &cospi32, &u[50], &rnding, bit); + v[51] = half_btf_sse4_1(&cospi32, &u[44], &cospi32, &u[51], &rnding, bit); + v[52] = half_btf_sse4_1(&cospi32, &u[43], &cospi32, &u[52], &rnding, bit); + v[53] = half_btf_sse4_1(&cospi32, &u[42], &cospi32, &u[53], &rnding, bit); + v[54] = half_btf_sse4_1(&cospi32, &u[41], &cospi32, &u[54], &rnding, bit); + v[55] = half_btf_sse4_1(&cospi32, &u[40], &cospi32, &u[55], &rnding, bit); + + for (i = 56; i < 64; i++) v[i] = u[i]; + + // stage 11 + if (do_cols) { + for (i = 0; i < 32; i++) { + addsub_no_clamp_sse4_1(v[i], v[63 - i], &out[(i)], &out[(63 - i)]); + } + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m128i clamp_lo_out = _mm_set1_epi32(AOMMAX( + -(1 << (log_range_out - 1)), -(1 << (log_range - 1 - out_shift)))); + const __m128i clamp_hi_out = _mm_set1_epi32(AOMMIN( + (1 << (log_range_out - 1)) - 1, (1 << (log_range - 1 - out_shift)))); + + for (i = 0; i < 32; i++) { + addsub_shift_sse4_1(v[i], v[63 - i], &out[(i)], &out[(63 - i)], + &clamp_lo_out, &clamp_hi_out, out_shift); + } + } + } +} + +static void idct32x32_low1_sse4_1(__m128i *in, __m128i *out, int bit, + int do_cols, int bd, int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + const __m128i rounding = _mm_set1_epi32(1 << (bit - 1)); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m128i clamp_lo = _mm_set1_epi32(-(1 << (log_range - 1))); + const __m128i clamp_hi = _mm_set1_epi32((1 << (log_range - 1)) - 1); + __m128i bf1; + + // stage 0 + // stage 1 + bf1 = in[0]; + + // stage 2 + // stage 3 + // stage 4 + // stage 5 + bf1 = half_btf_0_sse4_1(&cospi32, &bf1, &rounding, bit); + + // stage 6 + // stage 7 + // stage 8 + // stage 9 + if (do_cols) { + bf1 = _mm_max_epi32(bf1, clamp_lo); + bf1 = _mm_min_epi32(bf1, clamp_hi); + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m128i clamp_lo_out = _mm_set1_epi32(AOMMAX( + -(1 << (log_range_out - 1)), -(1 << (log_range - 1 - out_shift)))); + const __m128i clamp_hi_out = _mm_set1_epi32(AOMMIN( + (1 << (log_range_out - 1)) - 1, (1 << (log_range - 1 - out_shift)))); + + __m128i offset = _mm_set1_epi32((1 << out_shift) >> 1); + bf1 = _mm_add_epi32(bf1, offset); + bf1 = _mm_sra_epi32(bf1, _mm_cvtsi32_si128(out_shift)); + bf1 = _mm_max_epi32(bf1, clamp_lo_out); + bf1 = _mm_min_epi32(bf1, clamp_hi_out); + } + out[0] = bf1; + out[1] = bf1; + out[2] = bf1; + out[3] = bf1; + out[4] = bf1; + out[5] = bf1; + out[6] = bf1; + out[7] = bf1; + out[8] = bf1; + out[9] = bf1; + out[10] = bf1; + out[11] = bf1; + out[12] = bf1; + out[13] = bf1; + out[14] = bf1; + out[15] = bf1; + out[16] = bf1; + out[17] = bf1; + out[18] = bf1; + out[19] = bf1; + out[20] = bf1; + out[21] = bf1; + out[22] = bf1; + out[23] = bf1; + out[24] = bf1; + out[25] = bf1; + out[26] = bf1; + out[27] = bf1; + out[28] = bf1; + out[29] = bf1; + out[30] = bf1; + out[31] = bf1; +} + +static void idct32x32_low8_sse4_1(__m128i *in, __m128i *out, int bit, + int do_cols, int bd, int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m128i cospi62 = _mm_set1_epi32(cospi[62]); + const __m128i cospi14 = _mm_set1_epi32(cospi[14]); + const __m128i cospi54 = _mm_set1_epi32(cospi[54]); + const __m128i cospi6 = _mm_set1_epi32(cospi[6]); + const __m128i cospi10 = _mm_set1_epi32(cospi[10]); + const __m128i cospi2 = _mm_set1_epi32(cospi[2]); + const __m128i cospim58 = _mm_set1_epi32(-cospi[58]); + const __m128i cospim50 = _mm_set1_epi32(-cospi[50]); + const __m128i cospi60 = _mm_set1_epi32(cospi[60]); + const __m128i cospi12 = _mm_set1_epi32(cospi[12]); + const __m128i cospi4 = _mm_set1_epi32(cospi[4]); + const __m128i cospim52 = _mm_set1_epi32(-cospi[52]); + const __m128i cospi56 = _mm_set1_epi32(cospi[56]); + const __m128i cospi24 = _mm_set1_epi32(cospi[24]); + const __m128i cospi40 = _mm_set1_epi32(cospi[40]); + const __m128i cospi8 = _mm_set1_epi32(cospi[8]); + const __m128i cospim40 = _mm_set1_epi32(-cospi[40]); + const __m128i cospim8 = _mm_set1_epi32(-cospi[8]); + const __m128i cospim56 = _mm_set1_epi32(-cospi[56]); + const __m128i cospim24 = _mm_set1_epi32(-cospi[24]); + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + const __m128i cospim32 = _mm_set1_epi32(-cospi[32]); + const __m128i cospi48 = _mm_set1_epi32(cospi[48]); + const __m128i cospim48 = _mm_set1_epi32(-cospi[48]); + const __m128i cospi16 = _mm_set1_epi32(cospi[16]); + const __m128i cospim16 = _mm_set1_epi32(-cospi[16]); + const __m128i rounding = _mm_set1_epi32(1 << (bit - 1)); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m128i clamp_lo = _mm_set1_epi32(-(1 << (log_range - 1))); + const __m128i clamp_hi = _mm_set1_epi32((1 << (log_range - 1)) - 1); + __m128i bf1[32]; + + // stage 0 + // stage 1 + bf1[0] = in[0]; + bf1[4] = in[4]; + bf1[8] = in[2]; + bf1[12] = in[6]; + bf1[16] = in[1]; + bf1[20] = in[5]; + bf1[24] = in[3]; + bf1[28] = in[7]; + + // stage 2 + bf1[31] = half_btf_0_sse4_1(&cospi2, &bf1[16], &rounding, bit); + bf1[16] = half_btf_0_sse4_1(&cospi62, &bf1[16], &rounding, bit); + bf1[19] = half_btf_0_sse4_1(&cospim50, &bf1[28], &rounding, bit); + bf1[28] = half_btf_0_sse4_1(&cospi14, &bf1[28], &rounding, bit); + bf1[27] = half_btf_0_sse4_1(&cospi10, &bf1[20], &rounding, bit); + bf1[20] = half_btf_0_sse4_1(&cospi54, &bf1[20], &rounding, bit); + bf1[23] = half_btf_0_sse4_1(&cospim58, &bf1[24], &rounding, bit); + bf1[24] = half_btf_0_sse4_1(&cospi6, &bf1[24], &rounding, bit); + + // stage 3 + bf1[15] = half_btf_0_sse4_1(&cospi4, &bf1[8], &rounding, bit); + bf1[8] = half_btf_0_sse4_1(&cospi60, &bf1[8], &rounding, bit); + + bf1[11] = half_btf_0_sse4_1(&cospim52, &bf1[12], &rounding, bit); + bf1[12] = half_btf_0_sse4_1(&cospi12, &bf1[12], &rounding, bit); + bf1[17] = bf1[16]; + bf1[18] = bf1[19]; + bf1[21] = bf1[20]; + bf1[22] = bf1[23]; + bf1[25] = bf1[24]; + bf1[26] = bf1[27]; + bf1[29] = bf1[28]; + bf1[30] = bf1[31]; + + // stage 4 : + bf1[7] = half_btf_0_sse4_1(&cospi8, &bf1[4], &rounding, bit); + bf1[4] = half_btf_0_sse4_1(&cospi56, &bf1[4], &rounding, bit); + + bf1[9] = bf1[8]; + bf1[10] = bf1[11]; + bf1[13] = bf1[12]; + bf1[14] = bf1[15]; + + idct32_stage4_sse4_1(bf1, &cospim8, &cospi56, &cospi8, &cospim56, &cospim40, + &cospi24, &cospi40, &cospim24, &rounding, bit); + + // stage 5 + bf1[0] = half_btf_0_sse4_1(&cospi32, &bf1[0], &rounding, bit); + bf1[1] = bf1[0]; + bf1[5] = bf1[4]; + bf1[6] = bf1[7]; + + idct32_stage5_sse4_1(bf1, &cospim16, &cospi48, &cospi16, &cospim48, &clamp_lo, + &clamp_hi, &rounding, bit); + + // stage 6 + bf1[3] = bf1[0]; + bf1[2] = bf1[1]; + + idct32_stage6_sse4_1(bf1, &cospim32, &cospi32, &cospim16, &cospi48, &cospi16, + &cospim48, &clamp_lo, &clamp_hi, &rounding, bit); + + // stage 7 + idct32_stage7_sse4_1(bf1, &cospim32, &cospi32, &clamp_lo, &clamp_hi, + &rounding, bit); + + // stage 8 + idct32_stage8_sse4_1(bf1, &cospim32, &cospi32, &clamp_lo, &clamp_hi, + &rounding, bit); + + // stage 9 + idct32_stage9_sse4_1(bf1, out, do_cols, bd, out_shift, log_range); +} + +static void idct32x32_low16_sse4_1(__m128i *in, __m128i *out, int bit, + int do_cols, int bd, int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m128i cospi62 = _mm_set1_epi32(cospi[62]); + const __m128i cospi30 = _mm_set1_epi32(cospi[30]); + const __m128i cospi46 = _mm_set1_epi32(cospi[46]); + const __m128i cospi14 = _mm_set1_epi32(cospi[14]); + const __m128i cospi54 = _mm_set1_epi32(cospi[54]); + const __m128i cospi22 = _mm_set1_epi32(cospi[22]); + const __m128i cospi38 = _mm_set1_epi32(cospi[38]); + const __m128i cospi6 = _mm_set1_epi32(cospi[6]); + const __m128i cospi26 = _mm_set1_epi32(cospi[26]); + const __m128i cospi10 = _mm_set1_epi32(cospi[10]); + const __m128i cospi18 = _mm_set1_epi32(cospi[18]); + const __m128i cospi2 = _mm_set1_epi32(cospi[2]); + const __m128i cospim58 = _mm_set1_epi32(-cospi[58]); + const __m128i cospim42 = _mm_set1_epi32(-cospi[42]); + const __m128i cospim50 = _mm_set1_epi32(-cospi[50]); + const __m128i cospim34 = _mm_set1_epi32(-cospi[34]); + const __m128i cospi60 = _mm_set1_epi32(cospi[60]); + const __m128i cospi28 = _mm_set1_epi32(cospi[28]); + const __m128i cospi44 = _mm_set1_epi32(cospi[44]); + const __m128i cospi12 = _mm_set1_epi32(cospi[12]); + const __m128i cospi20 = _mm_set1_epi32(cospi[20]); + const __m128i cospi4 = _mm_set1_epi32(cospi[4]); + const __m128i cospim52 = _mm_set1_epi32(-cospi[52]); + const __m128i cospim36 = _mm_set1_epi32(-cospi[36]); + const __m128i cospi56 = _mm_set1_epi32(cospi[56]); + const __m128i cospi24 = _mm_set1_epi32(cospi[24]); + const __m128i cospi40 = _mm_set1_epi32(cospi[40]); + const __m128i cospi8 = _mm_set1_epi32(cospi[8]); + const __m128i cospim40 = _mm_set1_epi32(-cospi[40]); + const __m128i cospim8 = _mm_set1_epi32(-cospi[8]); + const __m128i cospim56 = _mm_set1_epi32(-cospi[56]); + const __m128i cospim24 = _mm_set1_epi32(-cospi[24]); + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + const __m128i cospim32 = _mm_set1_epi32(-cospi[32]); + const __m128i cospi48 = _mm_set1_epi32(cospi[48]); + const __m128i cospim48 = _mm_set1_epi32(-cospi[48]); + const __m128i cospi16 = _mm_set1_epi32(cospi[16]); + const __m128i cospim16 = _mm_set1_epi32(-cospi[16]); + const __m128i rounding = _mm_set1_epi32(1 << (bit - 1)); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m128i clamp_lo = _mm_set1_epi32(-(1 << (log_range - 1))); + const __m128i clamp_hi = _mm_set1_epi32((1 << (log_range - 1)) - 1); + __m128i bf1[32]; + + // stage 0 + // stage 1 + + bf1[0] = in[0]; + bf1[2] = in[8]; + bf1[4] = in[4]; + bf1[6] = in[12]; + bf1[8] = in[2]; + bf1[10] = in[10]; + bf1[12] = in[6]; + bf1[14] = in[14]; + bf1[16] = in[1]; + bf1[18] = in[9]; + bf1[20] = in[5]; + bf1[22] = in[13]; + bf1[24] = in[3]; + bf1[26] = in[11]; + bf1[28] = in[7]; + bf1[30] = in[15]; + + // stage 2 + bf1[31] = half_btf_0_sse4_1(&cospi2, &bf1[16], &rounding, bit); + bf1[16] = half_btf_0_sse4_1(&cospi62, &bf1[16], &rounding, bit); + bf1[17] = half_btf_0_sse4_1(&cospim34, &bf1[30], &rounding, bit); + bf1[30] = half_btf_0_sse4_1(&cospi30, &bf1[30], &rounding, bit); + bf1[29] = half_btf_0_sse4_1(&cospi18, &bf1[18], &rounding, bit); + bf1[18] = half_btf_0_sse4_1(&cospi46, &bf1[18], &rounding, bit); + bf1[19] = half_btf_0_sse4_1(&cospim50, &bf1[28], &rounding, bit); + bf1[28] = half_btf_0_sse4_1(&cospi14, &bf1[28], &rounding, bit); + bf1[27] = half_btf_0_sse4_1(&cospi10, &bf1[20], &rounding, bit); + bf1[20] = half_btf_0_sse4_1(&cospi54, &bf1[20], &rounding, bit); + bf1[21] = half_btf_0_sse4_1(&cospim42, &bf1[26], &rounding, bit); + bf1[26] = half_btf_0_sse4_1(&cospi22, &bf1[26], &rounding, bit); + bf1[25] = half_btf_0_sse4_1(&cospi26, &bf1[22], &rounding, bit); + bf1[22] = half_btf_0_sse4_1(&cospi38, &bf1[22], &rounding, bit); + bf1[23] = half_btf_0_sse4_1(&cospim58, &bf1[24], &rounding, bit); + bf1[24] = half_btf_0_sse4_1(&cospi6, &bf1[24], &rounding, bit); + + // stage 3 + bf1[15] = half_btf_0_sse4_1(&cospi4, &bf1[8], &rounding, bit); + bf1[8] = half_btf_0_sse4_1(&cospi60, &bf1[8], &rounding, bit); + bf1[9] = half_btf_0_sse4_1(&cospim36, &bf1[14], &rounding, bit); + bf1[14] = half_btf_0_sse4_1(&cospi28, &bf1[14], &rounding, bit); + bf1[13] = half_btf_0_sse4_1(&cospi20, &bf1[10], &rounding, bit); + bf1[10] = half_btf_0_sse4_1(&cospi44, &bf1[10], &rounding, bit); + bf1[11] = half_btf_0_sse4_1(&cospim52, &bf1[12], &rounding, bit); + bf1[12] = half_btf_0_sse4_1(&cospi12, &bf1[12], &rounding, bit); + + addsub_sse4_1(bf1[16], bf1[17], bf1 + 16, bf1 + 17, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[19], bf1[18], bf1 + 19, bf1 + 18, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[20], bf1[21], bf1 + 20, bf1 + 21, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[23], bf1[22], bf1 + 23, bf1 + 22, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[24], bf1[25], bf1 + 24, bf1 + 25, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[27], bf1[26], bf1 + 27, bf1 + 26, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[28], bf1[29], bf1 + 28, bf1 + 29, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[31], bf1[30], bf1 + 31, bf1 + 30, &clamp_lo, &clamp_hi); + // stage 4 + bf1[7] = half_btf_0_sse4_1(&cospi8, &bf1[4], &rounding, bit); + bf1[4] = half_btf_0_sse4_1(&cospi56, &bf1[4], &rounding, bit); + bf1[5] = half_btf_0_sse4_1(&cospim40, &bf1[6], &rounding, bit); + bf1[6] = half_btf_0_sse4_1(&cospi24, &bf1[6], &rounding, bit); + + addsub_sse4_1(bf1[8], bf1[9], bf1 + 8, bf1 + 9, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[11], bf1[10], bf1 + 11, bf1 + 10, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[12], bf1[13], bf1 + 12, bf1 + 13, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[15], bf1[14], bf1 + 15, bf1 + 14, &clamp_lo, &clamp_hi); + + idct32_stage4_sse4_1(bf1, &cospim8, &cospi56, &cospi8, &cospim56, &cospim40, + &cospi24, &cospi40, &cospim24, &rounding, bit); + + // stage 5 + bf1[0] = half_btf_0_sse4_1(&cospi32, &bf1[0], &rounding, bit); + bf1[1] = bf1[0]; + bf1[3] = half_btf_0_sse4_1(&cospi16, &bf1[2], &rounding, bit); + bf1[2] = half_btf_0_sse4_1(&cospi48, &bf1[2], &rounding, bit); + + addsub_sse4_1(bf1[4], bf1[5], bf1 + 4, bf1 + 5, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[7], bf1[6], bf1 + 7, bf1 + 6, &clamp_lo, &clamp_hi); + + idct32_stage5_sse4_1(bf1, &cospim16, &cospi48, &cospi16, &cospim48, &clamp_lo, + &clamp_hi, &rounding, bit); + + // stage 6 + addsub_sse4_1(bf1[0], bf1[3], bf1 + 0, bf1 + 3, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[1], bf1[2], bf1 + 1, bf1 + 2, &clamp_lo, &clamp_hi); + + idct32_stage6_sse4_1(bf1, &cospim32, &cospi32, &cospim16, &cospi48, &cospi16, + &cospim48, &clamp_lo, &clamp_hi, &rounding, bit); + + // stage 7 + idct32_stage7_sse4_1(bf1, &cospim32, &cospi32, &clamp_lo, &clamp_hi, + &rounding, bit); + + // stage 8 + idct32_stage8_sse4_1(bf1, &cospim32, &cospi32, &clamp_lo, &clamp_hi, + &rounding, bit); + + // stage 9 + idct32_stage9_sse4_1(bf1, out, do_cols, bd, out_shift, log_range); +} + +static void idct32x32_sse4_1(__m128i *in, __m128i *out, int bit, int do_cols, + int bd, int out_shift) { + const int32_t *cospi = cospi_arr(bit); + const __m128i cospi62 = _mm_set1_epi32(cospi[62]); + const __m128i cospi30 = _mm_set1_epi32(cospi[30]); + const __m128i cospi46 = _mm_set1_epi32(cospi[46]); + const __m128i cospi14 = _mm_set1_epi32(cospi[14]); + const __m128i cospi54 = _mm_set1_epi32(cospi[54]); + const __m128i cospi22 = _mm_set1_epi32(cospi[22]); + const __m128i cospi38 = _mm_set1_epi32(cospi[38]); + const __m128i cospi6 = _mm_set1_epi32(cospi[6]); + const __m128i cospi58 = _mm_set1_epi32(cospi[58]); + const __m128i cospi26 = _mm_set1_epi32(cospi[26]); + const __m128i cospi42 = _mm_set1_epi32(cospi[42]); + const __m128i cospi10 = _mm_set1_epi32(cospi[10]); + const __m128i cospi50 = _mm_set1_epi32(cospi[50]); + const __m128i cospi18 = _mm_set1_epi32(cospi[18]); + const __m128i cospi34 = _mm_set1_epi32(cospi[34]); + const __m128i cospi2 = _mm_set1_epi32(cospi[2]); + const __m128i cospim58 = _mm_set1_epi32(-cospi[58]); + const __m128i cospim26 = _mm_set1_epi32(-cospi[26]); + const __m128i cospim42 = _mm_set1_epi32(-cospi[42]); + const __m128i cospim10 = _mm_set1_epi32(-cospi[10]); + const __m128i cospim50 = _mm_set1_epi32(-cospi[50]); + const __m128i cospim18 = _mm_set1_epi32(-cospi[18]); + const __m128i cospim34 = _mm_set1_epi32(-cospi[34]); + const __m128i cospim2 = _mm_set1_epi32(-cospi[2]); + const __m128i cospi60 = _mm_set1_epi32(cospi[60]); + const __m128i cospi28 = _mm_set1_epi32(cospi[28]); + const __m128i cospi44 = _mm_set1_epi32(cospi[44]); + const __m128i cospi12 = _mm_set1_epi32(cospi[12]); + const __m128i cospi52 = _mm_set1_epi32(cospi[52]); + const __m128i cospi20 = _mm_set1_epi32(cospi[20]); + const __m128i cospi36 = _mm_set1_epi32(cospi[36]); + const __m128i cospi4 = _mm_set1_epi32(cospi[4]); + const __m128i cospim52 = _mm_set1_epi32(-cospi[52]); + const __m128i cospim20 = _mm_set1_epi32(-cospi[20]); + const __m128i cospim36 = _mm_set1_epi32(-cospi[36]); + const __m128i cospim4 = _mm_set1_epi32(-cospi[4]); + const __m128i cospi56 = _mm_set1_epi32(cospi[56]); + const __m128i cospi24 = _mm_set1_epi32(cospi[24]); + const __m128i cospi40 = _mm_set1_epi32(cospi[40]); + const __m128i cospi8 = _mm_set1_epi32(cospi[8]); + const __m128i cospim40 = _mm_set1_epi32(-cospi[40]); + const __m128i cospim8 = _mm_set1_epi32(-cospi[8]); + const __m128i cospim56 = _mm_set1_epi32(-cospi[56]); + const __m128i cospim24 = _mm_set1_epi32(-cospi[24]); + const __m128i cospi32 = _mm_set1_epi32(cospi[32]); + const __m128i cospim32 = _mm_set1_epi32(-cospi[32]); + const __m128i cospi48 = _mm_set1_epi32(cospi[48]); + const __m128i cospim48 = _mm_set1_epi32(-cospi[48]); + const __m128i cospi16 = _mm_set1_epi32(cospi[16]); + const __m128i cospim16 = _mm_set1_epi32(-cospi[16]); + const __m128i rounding = _mm_set1_epi32(1 << (bit - 1)); + const int log_range = AOMMAX(16, bd + (do_cols ? 6 : 8)); + const __m128i clamp_lo = _mm_set1_epi32(-(1 << (log_range - 1))); + const __m128i clamp_hi = _mm_set1_epi32((1 << (log_range - 1)) - 1); + __m128i bf1[32], bf0[32]; + + // stage 0 + // stage 1 + bf1[0] = in[0]; + bf1[1] = in[16]; + bf1[2] = in[8]; + bf1[3] = in[24]; + bf1[4] = in[4]; + bf1[5] = in[20]; + bf1[6] = in[12]; + bf1[7] = in[28]; + bf1[8] = in[2]; + bf1[9] = in[18]; + bf1[10] = in[10]; + bf1[11] = in[26]; + bf1[12] = in[6]; + bf1[13] = in[22]; + bf1[14] = in[14]; + bf1[15] = in[30]; + bf1[16] = in[1]; + bf1[17] = in[17]; + bf1[18] = in[9]; + bf1[19] = in[25]; + bf1[20] = in[5]; + bf1[21] = in[21]; + bf1[22] = in[13]; + bf1[23] = in[29]; + bf1[24] = in[3]; + bf1[25] = in[19]; + bf1[26] = in[11]; + bf1[27] = in[27]; + bf1[28] = in[7]; + bf1[29] = in[23]; + bf1[30] = in[15]; + bf1[31] = in[31]; + + // stage 2 + bf0[0] = bf1[0]; + bf0[1] = bf1[1]; + bf0[2] = bf1[2]; + bf0[3] = bf1[3]; + bf0[4] = bf1[4]; + bf0[5] = bf1[5]; + bf0[6] = bf1[6]; + bf0[7] = bf1[7]; + bf0[8] = bf1[8]; + bf0[9] = bf1[9]; + bf0[10] = bf1[10]; + bf0[11] = bf1[11]; + bf0[12] = bf1[12]; + bf0[13] = bf1[13]; + bf0[14] = bf1[14]; + bf0[15] = bf1[15]; + bf0[16] = + half_btf_sse4_1(&cospi62, &bf1[16], &cospim2, &bf1[31], &rounding, bit); + bf0[17] = + half_btf_sse4_1(&cospi30, &bf1[17], &cospim34, &bf1[30], &rounding, bit); + bf0[18] = + half_btf_sse4_1(&cospi46, &bf1[18], &cospim18, &bf1[29], &rounding, bit); + bf0[19] = + half_btf_sse4_1(&cospi14, &bf1[19], &cospim50, &bf1[28], &rounding, bit); + bf0[20] = + half_btf_sse4_1(&cospi54, &bf1[20], &cospim10, &bf1[27], &rounding, bit); + bf0[21] = + half_btf_sse4_1(&cospi22, &bf1[21], &cospim42, &bf1[26], &rounding, bit); + bf0[22] = + half_btf_sse4_1(&cospi38, &bf1[22], &cospim26, &bf1[25], &rounding, bit); + bf0[23] = + half_btf_sse4_1(&cospi6, &bf1[23], &cospim58, &bf1[24], &rounding, bit); + bf0[24] = + half_btf_sse4_1(&cospi58, &bf1[23], &cospi6, &bf1[24], &rounding, bit); + bf0[25] = + half_btf_sse4_1(&cospi26, &bf1[22], &cospi38, &bf1[25], &rounding, bit); + bf0[26] = + half_btf_sse4_1(&cospi42, &bf1[21], &cospi22, &bf1[26], &rounding, bit); + bf0[27] = + half_btf_sse4_1(&cospi10, &bf1[20], &cospi54, &bf1[27], &rounding, bit); + bf0[28] = + half_btf_sse4_1(&cospi50, &bf1[19], &cospi14, &bf1[28], &rounding, bit); + bf0[29] = + half_btf_sse4_1(&cospi18, &bf1[18], &cospi46, &bf1[29], &rounding, bit); + bf0[30] = + half_btf_sse4_1(&cospi34, &bf1[17], &cospi30, &bf1[30], &rounding, bit); + bf0[31] = + half_btf_sse4_1(&cospi2, &bf1[16], &cospi62, &bf1[31], &rounding, bit); + + // stage 3 + bf1[0] = bf0[0]; + bf1[1] = bf0[1]; + bf1[2] = bf0[2]; + bf1[3] = bf0[3]; + bf1[4] = bf0[4]; + bf1[5] = bf0[5]; + bf1[6] = bf0[6]; + bf1[7] = bf0[7]; + bf1[8] = + half_btf_sse4_1(&cospi60, &bf0[8], &cospim4, &bf0[15], &rounding, bit); + bf1[9] = + half_btf_sse4_1(&cospi28, &bf0[9], &cospim36, &bf0[14], &rounding, bit); + bf1[10] = + half_btf_sse4_1(&cospi44, &bf0[10], &cospim20, &bf0[13], &rounding, bit); + bf1[11] = + half_btf_sse4_1(&cospi12, &bf0[11], &cospim52, &bf0[12], &rounding, bit); + bf1[12] = + half_btf_sse4_1(&cospi52, &bf0[11], &cospi12, &bf0[12], &rounding, bit); + bf1[13] = + half_btf_sse4_1(&cospi20, &bf0[10], &cospi44, &bf0[13], &rounding, bit); + bf1[14] = + half_btf_sse4_1(&cospi36, &bf0[9], &cospi28, &bf0[14], &rounding, bit); + bf1[15] = + half_btf_sse4_1(&cospi4, &bf0[8], &cospi60, &bf0[15], &rounding, bit); + + addsub_sse4_1(bf0[16], bf0[17], bf1 + 16, bf1 + 17, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[19], bf0[18], bf1 + 19, bf1 + 18, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[20], bf0[21], bf1 + 20, bf1 + 21, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[23], bf0[22], bf1 + 23, bf1 + 22, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[24], bf0[25], bf1 + 24, bf1 + 25, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[27], bf0[26], bf1 + 27, bf1 + 26, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[28], bf0[29], bf1 + 28, bf1 + 29, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[31], bf0[30], bf1 + 31, bf1 + 30, &clamp_lo, &clamp_hi); + + // stage 4 + bf0[0] = bf1[0]; + bf0[1] = bf1[1]; + bf0[2] = bf1[2]; + bf0[3] = bf1[3]; + bf0[4] = + half_btf_sse4_1(&cospi56, &bf1[4], &cospim8, &bf1[7], &rounding, bit); + bf0[5] = + half_btf_sse4_1(&cospi24, &bf1[5], &cospim40, &bf1[6], &rounding, bit); + bf0[6] = + half_btf_sse4_1(&cospi40, &bf1[5], &cospi24, &bf1[6], &rounding, bit); + bf0[7] = half_btf_sse4_1(&cospi8, &bf1[4], &cospi56, &bf1[7], &rounding, bit); + + addsub_sse4_1(bf1[8], bf1[9], bf0 + 8, bf0 + 9, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[11], bf1[10], bf0 + 11, bf0 + 10, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[12], bf1[13], bf0 + 12, bf0 + 13, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[15], bf1[14], bf0 + 15, bf0 + 14, &clamp_lo, &clamp_hi); + + bf0[16] = bf1[16]; + bf0[17] = + half_btf_sse4_1(&cospim8, &bf1[17], &cospi56, &bf1[30], &rounding, bit); + bf0[18] = + half_btf_sse4_1(&cospim56, &bf1[18], &cospim8, &bf1[29], &rounding, bit); + bf0[19] = bf1[19]; + bf0[20] = bf1[20]; + bf0[21] = + half_btf_sse4_1(&cospim40, &bf1[21], &cospi24, &bf1[26], &rounding, bit); + bf0[22] = + half_btf_sse4_1(&cospim24, &bf1[22], &cospim40, &bf1[25], &rounding, bit); + bf0[23] = bf1[23]; + bf0[24] = bf1[24]; + bf0[25] = + half_btf_sse4_1(&cospim40, &bf1[22], &cospi24, &bf1[25], &rounding, bit); + bf0[26] = + half_btf_sse4_1(&cospi24, &bf1[21], &cospi40, &bf1[26], &rounding, bit); + bf0[27] = bf1[27]; + bf0[28] = bf1[28]; + bf0[29] = + half_btf_sse4_1(&cospim8, &bf1[18], &cospi56, &bf1[29], &rounding, bit); + bf0[30] = + half_btf_sse4_1(&cospi56, &bf1[17], &cospi8, &bf1[30], &rounding, bit); + bf0[31] = bf1[31]; + + // stage 5 + bf1[0] = + half_btf_sse4_1(&cospi32, &bf0[0], &cospi32, &bf0[1], &rounding, bit); + bf1[1] = + half_btf_sse4_1(&cospi32, &bf0[0], &cospim32, &bf0[1], &rounding, bit); + bf1[2] = + half_btf_sse4_1(&cospi48, &bf0[2], &cospim16, &bf0[3], &rounding, bit); + bf1[3] = + half_btf_sse4_1(&cospi16, &bf0[2], &cospi48, &bf0[3], &rounding, bit); + addsub_sse4_1(bf0[4], bf0[5], bf1 + 4, bf1 + 5, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[7], bf0[6], bf1 + 7, bf1 + 6, &clamp_lo, &clamp_hi); + bf1[8] = bf0[8]; + bf1[9] = + half_btf_sse4_1(&cospim16, &bf0[9], &cospi48, &bf0[14], &rounding, bit); + bf1[10] = + half_btf_sse4_1(&cospim48, &bf0[10], &cospim16, &bf0[13], &rounding, bit); + bf1[11] = bf0[11]; + bf1[12] = bf0[12]; + bf1[13] = + half_btf_sse4_1(&cospim16, &bf0[10], &cospi48, &bf0[13], &rounding, bit); + bf1[14] = + half_btf_sse4_1(&cospi48, &bf0[9], &cospi16, &bf0[14], &rounding, bit); + bf1[15] = bf0[15]; + addsub_sse4_1(bf0[16], bf0[19], bf1 + 16, bf1 + 19, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[17], bf0[18], bf1 + 17, bf1 + 18, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[23], bf0[20], bf1 + 23, bf1 + 20, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[22], bf0[21], bf1 + 22, bf1 + 21, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[24], bf0[27], bf1 + 24, bf1 + 27, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[25], bf0[26], bf1 + 25, bf1 + 26, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[31], bf0[28], bf1 + 31, bf1 + 28, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[30], bf0[29], bf1 + 30, bf1 + 29, &clamp_lo, &clamp_hi); + + // stage 6 + addsub_sse4_1(bf1[0], bf1[3], bf0 + 0, bf0 + 3, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[1], bf1[2], bf0 + 1, bf0 + 2, &clamp_lo, &clamp_hi); + bf0[4] = bf1[4]; + bf0[5] = + half_btf_sse4_1(&cospim32, &bf1[5], &cospi32, &bf1[6], &rounding, bit); + bf0[6] = + half_btf_sse4_1(&cospi32, &bf1[5], &cospi32, &bf1[6], &rounding, bit); + bf0[7] = bf1[7]; + addsub_sse4_1(bf1[8], bf1[11], bf0 + 8, bf0 + 11, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[9], bf1[10], bf0 + 9, bf0 + 10, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[15], bf1[12], bf0 + 15, bf0 + 12, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[14], bf1[13], bf0 + 14, bf0 + 13, &clamp_lo, &clamp_hi); + bf0[16] = bf1[16]; + bf0[17] = bf1[17]; + bf0[18] = + half_btf_sse4_1(&cospim16, &bf1[18], &cospi48, &bf1[29], &rounding, bit); + bf0[19] = + half_btf_sse4_1(&cospim16, &bf1[19], &cospi48, &bf1[28], &rounding, bit); + bf0[20] = + half_btf_sse4_1(&cospim48, &bf1[20], &cospim16, &bf1[27], &rounding, bit); + bf0[21] = + half_btf_sse4_1(&cospim48, &bf1[21], &cospim16, &bf1[26], &rounding, bit); + bf0[22] = bf1[22]; + bf0[23] = bf1[23]; + bf0[24] = bf1[24]; + bf0[25] = bf1[25]; + bf0[26] = + half_btf_sse4_1(&cospim16, &bf1[21], &cospi48, &bf1[26], &rounding, bit); + bf0[27] = + half_btf_sse4_1(&cospim16, &bf1[20], &cospi48, &bf1[27], &rounding, bit); + bf0[28] = + half_btf_sse4_1(&cospi48, &bf1[19], &cospi16, &bf1[28], &rounding, bit); + bf0[29] = + half_btf_sse4_1(&cospi48, &bf1[18], &cospi16, &bf1[29], &rounding, bit); + bf0[30] = bf1[30]; + bf0[31] = bf1[31]; + + // stage 7 + addsub_sse4_1(bf0[0], bf0[7], bf1 + 0, bf1 + 7, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[1], bf0[6], bf1 + 1, bf1 + 6, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[2], bf0[5], bf1 + 2, bf1 + 5, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[3], bf0[4], bf1 + 3, bf1 + 4, &clamp_lo, &clamp_hi); + bf1[8] = bf0[8]; + bf1[9] = bf0[9]; + bf1[10] = + half_btf_sse4_1(&cospim32, &bf0[10], &cospi32, &bf0[13], &rounding, bit); + bf1[11] = + half_btf_sse4_1(&cospim32, &bf0[11], &cospi32, &bf0[12], &rounding, bit); + bf1[12] = + half_btf_sse4_1(&cospi32, &bf0[11], &cospi32, &bf0[12], &rounding, bit); + bf1[13] = + half_btf_sse4_1(&cospi32, &bf0[10], &cospi32, &bf0[13], &rounding, bit); + bf1[14] = bf0[14]; + bf1[15] = bf0[15]; + addsub_sse4_1(bf0[16], bf0[23], bf1 + 16, bf1 + 23, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[17], bf0[22], bf1 + 17, bf1 + 22, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[18], bf0[21], bf1 + 18, bf1 + 21, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[19], bf0[20], bf1 + 19, bf1 + 20, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[31], bf0[24], bf1 + 31, bf1 + 24, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[30], bf0[25], bf1 + 30, bf1 + 25, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[29], bf0[26], bf1 + 29, bf1 + 26, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf0[28], bf0[27], bf1 + 28, bf1 + 27, &clamp_lo, &clamp_hi); + + // stage 8 + addsub_sse4_1(bf1[0], bf1[15], bf0 + 0, bf0 + 15, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[1], bf1[14], bf0 + 1, bf0 + 14, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[2], bf1[13], bf0 + 2, bf0 + 13, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[3], bf1[12], bf0 + 3, bf0 + 12, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[4], bf1[11], bf0 + 4, bf0 + 11, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[5], bf1[10], bf0 + 5, bf0 + 10, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[6], bf1[9], bf0 + 6, bf0 + 9, &clamp_lo, &clamp_hi); + addsub_sse4_1(bf1[7], bf1[8], bf0 + 7, bf0 + 8, &clamp_lo, &clamp_hi); + bf0[16] = bf1[16]; + bf0[17] = bf1[17]; + bf0[18] = bf1[18]; + bf0[19] = bf1[19]; + bf0[20] = + half_btf_sse4_1(&cospim32, &bf1[20], &cospi32, &bf1[27], &rounding, bit); + bf0[21] = + half_btf_sse4_1(&cospim32, &bf1[21], &cospi32, &bf1[26], &rounding, bit); + bf0[22] = + half_btf_sse4_1(&cospim32, &bf1[22], &cospi32, &bf1[25], &rounding, bit); + bf0[23] = + half_btf_sse4_1(&cospim32, &bf1[23], &cospi32, &bf1[24], &rounding, bit); + bf0[24] = + half_btf_sse4_1(&cospi32, &bf1[23], &cospi32, &bf1[24], &rounding, bit); + bf0[25] = + half_btf_sse4_1(&cospi32, &bf1[22], &cospi32, &bf1[25], &rounding, bit); + bf0[26] = + half_btf_sse4_1(&cospi32, &bf1[21], &cospi32, &bf1[26], &rounding, bit); + bf0[27] = + half_btf_sse4_1(&cospi32, &bf1[20], &cospi32, &bf1[27], &rounding, bit); + bf0[28] = bf1[28]; + bf0[29] = bf1[29]; + bf0[30] = bf1[30]; + bf0[31] = bf1[31]; + + // stage 9 + if (do_cols) { + addsub_no_clamp_sse4_1(bf0[0], bf0[31], out + 0, out + 31); + addsub_no_clamp_sse4_1(bf0[1], bf0[30], out + 1, out + 30); + addsub_no_clamp_sse4_1(bf0[2], bf0[29], out + 2, out + 29); + addsub_no_clamp_sse4_1(bf0[3], bf0[28], out + 3, out + 28); + addsub_no_clamp_sse4_1(bf0[4], bf0[27], out + 4, out + 27); + addsub_no_clamp_sse4_1(bf0[5], bf0[26], out + 5, out + 26); + addsub_no_clamp_sse4_1(bf0[6], bf0[25], out + 6, out + 25); + addsub_no_clamp_sse4_1(bf0[7], bf0[24], out + 7, out + 24); + addsub_no_clamp_sse4_1(bf0[8], bf0[23], out + 8, out + 23); + addsub_no_clamp_sse4_1(bf0[9], bf0[22], out + 9, out + 22); + addsub_no_clamp_sse4_1(bf0[10], bf0[21], out + 10, out + 21); + addsub_no_clamp_sse4_1(bf0[11], bf0[20], out + 11, out + 20); + addsub_no_clamp_sse4_1(bf0[12], bf0[19], out + 12, out + 19); + addsub_no_clamp_sse4_1(bf0[13], bf0[18], out + 13, out + 18); + addsub_no_clamp_sse4_1(bf0[14], bf0[17], out + 14, out + 17); + addsub_no_clamp_sse4_1(bf0[15], bf0[16], out + 15, out + 16); + } else { + const int log_range_out = AOMMAX(16, bd + 6); + const __m128i clamp_lo_out = _mm_set1_epi32(AOMMAX( + -(1 << (log_range_out - 1)), -(1 << (log_range - 1 - out_shift)))); + const __m128i clamp_hi_out = _mm_set1_epi32(AOMMIN( + (1 << (log_range_out - 1)) - 1, (1 << (log_range - 1 - out_shift)))); + + addsub_shift_sse4_1(bf0[0], bf0[31], out + 0, out + 31, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf0[1], bf0[30], out + 1, out + 30, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf0[2], bf0[29], out + 2, out + 29, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf0[3], bf0[28], out + 3, out + 28, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf0[4], bf0[27], out + 4, out + 27, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf0[5], bf0[26], out + 5, out + 26, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf0[6], bf0[25], out + 6, out + 25, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf0[7], bf0[24], out + 7, out + 24, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf0[8], bf0[23], out + 8, out + 23, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf0[9], bf0[22], out + 9, out + 22, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf0[10], bf0[21], out + 10, out + 21, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf0[11], bf0[20], out + 11, out + 20, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf0[12], bf0[19], out + 12, out + 19, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf0[13], bf0[18], out + 13, out + 18, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf0[14], bf0[17], out + 14, out + 17, &clamp_lo_out, + &clamp_hi_out, out_shift); + addsub_shift_sse4_1(bf0[15], bf0[16], out + 15, out + 16, &clamp_lo_out, + &clamp_hi_out, out_shift); + } +} + +void av1_highbd_inv_txfm_add_8x8_sse4_1(const tran_low_t *input, uint8_t *dest, + int stride, + const TxfmParam *txfm_param) { + int bd = txfm_param->bd; + const TX_TYPE tx_type = txfm_param->tx_type; + const int32_t *src = cast_to_int32(input); + switch (tx_type) { + // Assembly version doesn't support some transform types, so use C version + // for those. + case V_DCT: + case H_DCT: + case V_ADST: + case H_ADST: + case V_FLIPADST: + case H_FLIPADST: + case IDTX: + av1_inv_txfm2d_add_8x8_c(src, CONVERT_TO_SHORTPTR(dest), stride, tx_type, + bd); + break; + default: + av1_inv_txfm2d_add_8x8_sse4_1(src, CONVERT_TO_SHORTPTR(dest), stride, + tx_type, bd); + break; + } +} + +void av1_highbd_inv_txfm_add_16x8_sse4_1(const tran_low_t *input, uint8_t *dest, + int stride, + const TxfmParam *txfm_param) { + int bd = txfm_param->bd; + const TX_TYPE tx_type = txfm_param->tx_type; + const int32_t *src = cast_to_int32(input); + switch (tx_type) { + // Assembly version doesn't support some transform types, so use C version + // for those. + case V_DCT: + case H_DCT: + case V_ADST: + case H_ADST: + case V_FLIPADST: + case H_FLIPADST: + case IDTX: + av1_inv_txfm2d_add_16x8_c(src, CONVERT_TO_SHORTPTR(dest), stride, + txfm_param->tx_type, txfm_param->bd); + break; + default: + av1_highbd_inv_txfm2d_add_universe_sse4_1(input, dest, stride, tx_type, + txfm_param->tx_size, + txfm_param->eob, bd); + break; + } +} + +void av1_highbd_inv_txfm_add_8x16_sse4_1(const tran_low_t *input, uint8_t *dest, + int stride, + const TxfmParam *txfm_param) { + int bd = txfm_param->bd; + const TX_TYPE tx_type = txfm_param->tx_type; + const int32_t *src = cast_to_int32(input); + switch (tx_type) { + // Assembly version doesn't support some transform types, so use C version + // for those. + case V_DCT: + case H_DCT: + case V_ADST: + case H_ADST: + case V_FLIPADST: + case H_FLIPADST: + case IDTX: + av1_inv_txfm2d_add_8x16_c(src, CONVERT_TO_SHORTPTR(dest), stride, + txfm_param->tx_type, txfm_param->bd); + break; + default: + av1_highbd_inv_txfm2d_add_universe_sse4_1(input, dest, stride, tx_type, + txfm_param->tx_size, + txfm_param->eob, bd); + break; + } +} + +void av1_highbd_inv_txfm_add_16x16_sse4_1(const tran_low_t *input, + uint8_t *dest, int stride, + const TxfmParam *txfm_param) { + int bd = txfm_param->bd; + const TX_TYPE tx_type = txfm_param->tx_type; + const int32_t *src = cast_to_int32(input); + switch (tx_type) { + // Assembly version doesn't support some transform types, so use C version + // for those. + case V_DCT: + case H_DCT: + case V_ADST: + case H_ADST: + case V_FLIPADST: + case H_FLIPADST: + case IDTX: + av1_inv_txfm2d_add_16x16_c(src, CONVERT_TO_SHORTPTR(dest), stride, + tx_type, bd); + break; + default: + av1_highbd_inv_txfm2d_add_universe_sse4_1(input, dest, stride, tx_type, + txfm_param->tx_size, + txfm_param->eob, bd); + break; + } +} + +void av1_highbd_inv_txfm_add_32x32_sse4_1(const tran_low_t *input, + uint8_t *dest, int stride, + const TxfmParam *txfm_param) { + int bd = txfm_param->bd; + const TX_TYPE tx_type = txfm_param->tx_type; + const int32_t *src = cast_to_int32(input); + switch (tx_type) { + case DCT_DCT: + av1_highbd_inv_txfm2d_add_universe_sse4_1(input, dest, stride, tx_type, + txfm_param->tx_size, + txfm_param->eob, bd); + break; + // Assembly version doesn't support IDTX, so use C version for it. + case IDTX: + av1_inv_txfm2d_add_32x32_c(src, CONVERT_TO_SHORTPTR(dest), stride, + tx_type, bd); + break; + default: assert(0); + } +} + +void av1_highbd_inv_txfm_add_4x4_sse4_1(const tran_low_t *input, uint8_t *dest, + int stride, + const TxfmParam *txfm_param) { + assert(av1_ext_tx_used[txfm_param->tx_set_type][txfm_param->tx_type]); + int eob = txfm_param->eob; + int bd = txfm_param->bd; + int lossless = txfm_param->lossless; + const int32_t *src = cast_to_int32(input); + const TX_TYPE tx_type = txfm_param->tx_type; + if (lossless) { + assert(tx_type == DCT_DCT); + av1_highbd_iwht4x4_add(input, dest, stride, eob, bd); + return; + } + switch (tx_type) { + // Assembly version doesn't support some transform types, so use C version + // for those. + case V_DCT: + case H_DCT: + case V_ADST: + case H_ADST: + case V_FLIPADST: + case H_FLIPADST: + case IDTX: + av1_inv_txfm2d_add_4x4_c(src, CONVERT_TO_SHORTPTR(dest), stride, tx_type, + bd); + break; + default: + av1_inv_txfm2d_add_4x4_sse4_1(src, CONVERT_TO_SHORTPTR(dest), stride, + tx_type, bd); + break; + } +} + +static const transform_1d_sse4_1 + highbd_txfm_all_1d_zeros_w8_arr[TX_SIZES][ITX_TYPES_1D][4] = { + { + { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL }, + }, + { { idct8x8_low1_sse4_1, idct8x8_new_sse4_1, NULL, NULL }, + { iadst8x8_low1_sse4_1, iadst8x8_new_sse4_1, NULL, NULL }, + { NULL, NULL, NULL, NULL } }, + { + { idct16x16_low1_sse4_1, idct16x16_low8_sse4_1, idct16x16_sse4_1, + NULL }, + { iadst16x16_low1_sse4_1, iadst16x16_low8_sse4_1, iadst16x16_sse4_1, + NULL }, + { NULL, NULL, NULL, NULL }, + }, + { { idct32x32_low1_sse4_1, idct32x32_low8_sse4_1, idct32x32_low16_sse4_1, + idct32x32_sse4_1 }, + { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL } }, + { { idct64x64_low1_sse4_1, idct64x64_low8_sse4_1, idct64x64_low16_sse4_1, + idct64x64_sse4_1 }, + { NULL, NULL, NULL, NULL }, + { NULL, NULL, NULL, NULL } } + }; + +static void highbd_inv_txfm2d_add_no_identity_sse41(const int32_t *input, + uint16_t *output, + int stride, TX_TYPE tx_type, + TX_SIZE tx_size, int eob, + const int bd) { + __m128i buf1[64 * 16]; + int eobx, eoby; + get_eobx_eoby_scan_default(&eobx, &eoby, tx_size, eob); + const int8_t *shift = inv_txfm_shift_ls[tx_size]; + const int txw_idx = get_txw_idx(tx_size); + const int txh_idx = get_txh_idx(tx_size); + const int txfm_size_col = tx_size_wide[tx_size]; + const int txfm_size_row = tx_size_high[tx_size]; + const int buf_size_w_div8 = txfm_size_col >> 2; + const int buf_size_nonzero_w_div8 = (eobx + 8) >> 3; + const int buf_size_nonzero_h_div8 = (eoby + 8) >> 3; + const int input_stride = AOMMIN(32, txfm_size_col); + const int rect_type = get_rect_tx_log_ratio(txfm_size_col, txfm_size_row); + + const int fun_idx_x = lowbd_txfm_all_1d_zeros_idx[eobx]; + const int fun_idx_y = lowbd_txfm_all_1d_zeros_idx[eoby]; + const transform_1d_sse4_1 row_txfm = + highbd_txfm_all_1d_zeros_w8_arr[txw_idx][hitx_1d_tab[tx_type]][fun_idx_x]; + const transform_1d_sse4_1 col_txfm = + highbd_txfm_all_1d_zeros_w8_arr[txh_idx][vitx_1d_tab[tx_type]][fun_idx_y]; + + assert(col_txfm != NULL); + assert(row_txfm != NULL); + int ud_flip, lr_flip; + get_flip_cfg(tx_type, &ud_flip, &lr_flip); + + // 1st stage: column transform + for (int i = 0; i < buf_size_nonzero_h_div8 << 1; i++) { + __m128i buf0[64]; + const int32_t *input_row = input + i * input_stride * 4; + for (int j = 0; j < buf_size_nonzero_w_div8 << 1; ++j) { + __m128i *buf0_cur = buf0 + j * 4; + load_buffer_32bit_input(input_row + j * 4, input_stride, buf0_cur, 4); + + TRANSPOSE_4X4(buf0_cur[0], buf0_cur[1], buf0_cur[2], buf0_cur[3], + buf0_cur[0], buf0_cur[1], buf0_cur[2], buf0_cur[3]); + } + if (rect_type == 1 || rect_type == -1) { + av1_round_shift_rect_array_32_sse4_1( + buf0, buf0, buf_size_nonzero_w_div8 << 3, 0, NewInvSqrt2); + } + row_txfm(buf0, buf0, inv_cos_bit_row[txw_idx][txh_idx], 0, bd, -shift[0]); + + __m128i *_buf1 = buf1 + i * 4; + if (lr_flip) { + for (int j = 0; j < buf_size_w_div8; ++j) { + TRANSPOSE_4X4(buf0[4 * j + 3], buf0[4 * j + 2], buf0[4 * j + 1], + buf0[4 * j], + _buf1[txfm_size_row * (buf_size_w_div8 - 1 - j) + 0], + _buf1[txfm_size_row * (buf_size_w_div8 - 1 - j) + 1], + _buf1[txfm_size_row * (buf_size_w_div8 - 1 - j) + 2], + _buf1[txfm_size_row * (buf_size_w_div8 - 1 - j) + 3]); + } + } else { + for (int j = 0; j < buf_size_w_div8; ++j) { + TRANSPOSE_4X4( + buf0[j * 4 + 0], buf0[j * 4 + 1], buf0[j * 4 + 2], buf0[j * 4 + 3], + _buf1[j * txfm_size_row + 0], _buf1[j * txfm_size_row + 1], + _buf1[j * txfm_size_row + 2], _buf1[j * txfm_size_row + 3]); + } + } + } + // 2nd stage: column transform + for (int i = 0; i < buf_size_w_div8; i++) { + col_txfm(buf1 + i * txfm_size_row, buf1 + i * txfm_size_row, + inv_cos_bit_col[txw_idx][txh_idx], 1, bd, 0); + + av1_round_shift_array_32_sse4_1(buf1 + i * txfm_size_row, + buf1 + i * txfm_size_row, txfm_size_row, + -shift[1]); + } + + // write to buffer + { + for (int i = 0; i < (txfm_size_col >> 3); i++) { + highbd_write_buffer_8xn_sse4_1(buf1 + i * txfm_size_row * 2, + output + 8 * i, stride, ud_flip, + txfm_size_row, bd); + } + } +} + +void av1_highbd_inv_txfm2d_add_universe_sse4_1(const int32_t *input, + uint8_t *output, int stride, + TX_TYPE tx_type, TX_SIZE tx_size, + int eob, const int bd) { + switch (tx_type) { + case DCT_DCT: + case ADST_DCT: + case DCT_ADST: + case ADST_ADST: + case FLIPADST_DCT: + case DCT_FLIPADST: + case FLIPADST_FLIPADST: + case ADST_FLIPADST: + case FLIPADST_ADST: + highbd_inv_txfm2d_add_no_identity_sse41( + input, CONVERT_TO_SHORTPTR(output), stride, tx_type, tx_size, eob, + bd); + break; + default: assert(0); break; + } +} + +void av1_highbd_inv_txfm_add_sse4_1(const tran_low_t *input, uint8_t *dest, + int stride, const TxfmParam *txfm_param) { + assert(av1_ext_tx_used[txfm_param->tx_set_type][txfm_param->tx_type]); + const TX_SIZE tx_size = txfm_param->tx_size; + switch (tx_size) { + case TX_32X32: + av1_highbd_inv_txfm_add_32x32_sse4_1(input, dest, stride, txfm_param); + break; + case TX_16X16: + av1_highbd_inv_txfm_add_16x16_sse4_1(input, dest, stride, txfm_param); + break; + case TX_8X8: + av1_highbd_inv_txfm_add_8x8_sse4_1(input, dest, stride, txfm_param); + break; + case TX_4X8: + av1_highbd_inv_txfm_add_4x8(input, dest, stride, txfm_param); + break; + case TX_8X4: + av1_highbd_inv_txfm_add_8x4(input, dest, stride, txfm_param); + break; + case TX_8X16: + av1_highbd_inv_txfm_add_8x16_sse4_1(input, dest, stride, txfm_param); + break; + case TX_16X8: + av1_highbd_inv_txfm_add_16x8_sse4_1(input, dest, stride, txfm_param); + break; + case TX_16X32: + av1_highbd_inv_txfm_add_16x32(input, dest, stride, txfm_param); + break; + case TX_32X16: + av1_highbd_inv_txfm_add_32x16(input, dest, stride, txfm_param); + break; + case TX_32X64: + av1_highbd_inv_txfm_add_32x64(input, dest, stride, txfm_param); + break; + case TX_64X32: + av1_highbd_inv_txfm_add_64x32(input, dest, stride, txfm_param); + break; + case TX_4X4: + av1_highbd_inv_txfm_add_4x4_sse4_1(input, dest, stride, txfm_param); + break; + case TX_16X4: + av1_highbd_inv_txfm_add_16x4(input, dest, stride, txfm_param); + break; + case TX_4X16: + av1_highbd_inv_txfm_add_4x16(input, dest, stride, txfm_param); + break; + case TX_8X32: + av1_highbd_inv_txfm_add_8x32(input, dest, stride, txfm_param); + break; + case TX_32X8: + av1_highbd_inv_txfm_add_32x8(input, dest, stride, txfm_param); + break; + case TX_64X64: + case TX_16X64: + case TX_64X16: + av1_highbd_inv_txfm2d_add_universe_sse4_1( + input, dest, stride, txfm_param->tx_type, txfm_param->tx_size, + txfm_param->eob, txfm_param->bd); + break; + default: assert(0 && "Invalid transform size"); break; + } +} diff --git a/third_party/aom/av1/common/x86/highbd_jnt_convolve_avx2.c b/third_party/aom/av1/common/x86/highbd_jnt_convolve_avx2.c new file mode 100644 index 0000000000..e298cf6531 --- /dev/null +++ b/third_party/aom/av1/common/x86/highbd_jnt_convolve_avx2.c @@ -0,0 +1,846 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <immintrin.h> +#include <assert.h> + +#include "config/aom_dsp_rtcd.h" + +#include "aom_dsp/x86/convolve_avx2.h" +#include "aom_dsp/x86/convolve_common_intrin.h" +#include "aom_dsp/x86/convolve_sse4_1.h" +#include "aom_dsp/x86/synonyms.h" +#include "aom_dsp/aom_dsp_common.h" +#include "aom_dsp/aom_filter.h" +#include "av1/common/convolve.h" + +void av1_highbd_jnt_convolve_2d_copy_avx2( + const uint16_t *src, int src_stride, uint16_t *dst0, int dst_stride0, int w, + int h, const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, const int subpel_x_q4, + const int subpel_y_q4, ConvolveParams *conv_params, int bd) { + CONV_BUF_TYPE *dst = conv_params->dst; + int dst_stride = conv_params->dst_stride; + (void)filter_params_x; + (void)filter_params_y; + (void)subpel_x_q4; + (void)subpel_y_q4; + + const int bits = + FILTER_BITS * 2 - conv_params->round_1 - conv_params->round_0; + const __m128i left_shift = _mm_cvtsi32_si128(bits); + const int do_average = conv_params->do_average; + const int use_jnt_comp_avg = conv_params->use_jnt_comp_avg; + const int w0 = conv_params->fwd_offset; + const int w1 = conv_params->bck_offset; + const __m256i wt0 = _mm256_set1_epi32(w0); + const __m256i wt1 = _mm256_set1_epi32(w1); + const __m256i zero = _mm256_setzero_si256(); + int i, j; + + const int offset_0 = + bd + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const int offset = (1 << offset_0) + (1 << (offset_0 - 1)); + const __m256i offset_const = _mm256_set1_epi32(offset); + const __m256i offset_const_16b = _mm256_set1_epi16(offset); + const int rounding_shift = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const __m256i rounding_const = _mm256_set1_epi32((1 << rounding_shift) >> 1); + const __m256i clip_pixel_to_bd = + _mm256_set1_epi16(bd == 10 ? 1023 : (bd == 12 ? 4095 : 255)); + + assert(bits <= 4); + + if (!(w % 16)) { + for (i = 0; i < h; i += 1) { + for (j = 0; j < w; j += 16) { + const __m256i src_16bit = + _mm256_loadu_si256((__m256i *)(&src[i * src_stride + j])); + + const __m256i res = _mm256_sll_epi16(src_16bit, left_shift); + + if (do_average) { + const __m256i data_0 = + _mm256_loadu_si256((__m256i *)(&dst[i * dst_stride + j])); + + const __m256i data_ref_0_lo = _mm256_unpacklo_epi16(data_0, zero); + const __m256i data_ref_0_hi = _mm256_unpackhi_epi16(data_0, zero); + + const __m256i res_32b_lo = _mm256_unpacklo_epi16(res, zero); + const __m256i res_unsigned_lo = + _mm256_add_epi32(res_32b_lo, offset_const); + + const __m256i comp_avg_res_lo = highbd_comp_avg( + &data_ref_0_lo, &res_unsigned_lo, &wt0, &wt1, use_jnt_comp_avg); + + const __m256i res_32b_hi = _mm256_unpackhi_epi16(res, zero); + const __m256i res_unsigned_hi = + _mm256_add_epi32(res_32b_hi, offset_const); + + const __m256i comp_avg_res_hi = highbd_comp_avg( + &data_ref_0_hi, &res_unsigned_hi, &wt0, &wt1, use_jnt_comp_avg); + + const __m256i round_result_lo = highbd_convolve_rounding( + &comp_avg_res_lo, &offset_const, &rounding_const, rounding_shift); + const __m256i round_result_hi = highbd_convolve_rounding( + &comp_avg_res_hi, &offset_const, &rounding_const, rounding_shift); + + const __m256i res_16b = + _mm256_packus_epi32(round_result_lo, round_result_hi); + const __m256i res_clip = _mm256_min_epi16(res_16b, clip_pixel_to_bd); + + _mm256_store_si256((__m256i *)(&dst0[i * dst_stride0 + j]), res_clip); + } else { + const __m256i res_unsigned_16b = + _mm256_adds_epu16(res, offset_const_16b); + + _mm256_store_si256((__m256i *)(&dst[i * dst_stride + j]), + res_unsigned_16b); + } + } + } + } else if (!(w % 4)) { + for (i = 0; i < h; i += 2) { + for (j = 0; j < w; j += 8) { + const __m128i src_row_0 = + _mm_loadu_si128((__m128i *)(&src[i * src_stride + j])); + const __m128i src_row_1 = + _mm_loadu_si128((__m128i *)(&src[i * src_stride + j + src_stride])); + // since not all compilers yet support _mm256_set_m128i() + const __m256i src_10 = _mm256_insertf128_si256( + _mm256_castsi128_si256(src_row_0), src_row_1, 1); + + const __m256i res = _mm256_sll_epi16(src_10, left_shift); + + if (w - j < 8) { + if (do_average) { + const __m256i data_0 = _mm256_castsi128_si256( + _mm_loadl_epi64((__m128i *)(&dst[i * dst_stride + j]))); + const __m256i data_1 = _mm256_castsi128_si256(_mm_loadl_epi64( + (__m128i *)(&dst[i * dst_stride + j + dst_stride]))); + const __m256i data_01 = + _mm256_permute2x128_si256(data_0, data_1, 0x20); + + const __m256i data_ref_0 = _mm256_unpacklo_epi16(data_01, zero); + + const __m256i res_32b = _mm256_unpacklo_epi16(res, zero); + const __m256i res_unsigned_lo = + _mm256_add_epi32(res_32b, offset_const); + + const __m256i comp_avg_res = highbd_comp_avg( + &data_ref_0, &res_unsigned_lo, &wt0, &wt1, use_jnt_comp_avg); + + const __m256i round_result = highbd_convolve_rounding( + &comp_avg_res, &offset_const, &rounding_const, rounding_shift); + + const __m256i res_16b = + _mm256_packus_epi32(round_result, round_result); + const __m256i res_clip = + _mm256_min_epi16(res_16b, clip_pixel_to_bd); + + const __m128i res_0 = _mm256_castsi256_si128(res_clip); + const __m128i res_1 = _mm256_extracti128_si256(res_clip, 1); + + _mm_storel_epi64((__m128i *)(&dst0[i * dst_stride0 + j]), res_0); + _mm_storel_epi64( + (__m128i *)(&dst0[i * dst_stride0 + j + dst_stride0]), res_1); + } else { + const __m256i res_unsigned_16b = + _mm256_adds_epu16(res, offset_const_16b); + + const __m128i res_0 = _mm256_castsi256_si128(res_unsigned_16b); + const __m128i res_1 = _mm256_extracti128_si256(res_unsigned_16b, 1); + + _mm_storel_epi64((__m128i *)(&dst[i * dst_stride + j]), res_0); + _mm_storel_epi64((__m128i *)(&dst[i * dst_stride + j + dst_stride]), + res_1); + } + } else { + if (do_average) { + const __m256i data_0 = _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(&dst[i * dst_stride + j]))); + const __m256i data_1 = _mm256_castsi128_si256(_mm_loadu_si128( + (__m128i *)(&dst[i * dst_stride + j + dst_stride]))); + const __m256i data_01 = + _mm256_permute2x128_si256(data_0, data_1, 0x20); + + const __m256i data_ref_0_lo = _mm256_unpacklo_epi16(data_01, zero); + const __m256i data_ref_0_hi = _mm256_unpackhi_epi16(data_01, zero); + + const __m256i res_32b_lo = _mm256_unpacklo_epi16(res, zero); + const __m256i res_unsigned_lo = + _mm256_add_epi32(res_32b_lo, offset_const); + + const __m256i comp_avg_res_lo = highbd_comp_avg( + &data_ref_0_lo, &res_unsigned_lo, &wt0, &wt1, use_jnt_comp_avg); + + const __m256i res_32b_hi = _mm256_unpackhi_epi16(res, zero); + const __m256i res_unsigned_hi = + _mm256_add_epi32(res_32b_hi, offset_const); + + const __m256i comp_avg_res_hi = highbd_comp_avg( + &data_ref_0_hi, &res_unsigned_hi, &wt0, &wt1, use_jnt_comp_avg); + + const __m256i round_result_lo = + highbd_convolve_rounding(&comp_avg_res_lo, &offset_const, + &rounding_const, rounding_shift); + const __m256i round_result_hi = + highbd_convolve_rounding(&comp_avg_res_hi, &offset_const, + &rounding_const, rounding_shift); + + const __m256i res_16b = + _mm256_packus_epi32(round_result_lo, round_result_hi); + const __m256i res_clip = + _mm256_min_epi16(res_16b, clip_pixel_to_bd); + + const __m128i res_0 = _mm256_castsi256_si128(res_clip); + const __m128i res_1 = _mm256_extracti128_si256(res_clip, 1); + + _mm_store_si128((__m128i *)(&dst0[i * dst_stride0 + j]), res_0); + _mm_store_si128( + (__m128i *)(&dst0[i * dst_stride0 + j + dst_stride0]), res_1); + } else { + const __m256i res_unsigned_16b = + _mm256_adds_epu16(res, offset_const_16b); + const __m128i res_0 = _mm256_castsi256_si128(res_unsigned_16b); + const __m128i res_1 = _mm256_extracti128_si256(res_unsigned_16b, 1); + + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j]), res_0); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j + dst_stride]), + res_1); + } + } + } + } + } +} + +void av1_highbd_jnt_convolve_2d_avx2( + const uint16_t *src, int src_stride, uint16_t *dst0, int dst_stride0, int w, + int h, const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, const int subpel_x_q4, + const int subpel_y_q4, ConvolveParams *conv_params, int bd) { + DECLARE_ALIGNED(32, int16_t, im_block[(MAX_SB_SIZE + MAX_FILTER_TAP) * 8]); + CONV_BUF_TYPE *dst = conv_params->dst; + int dst_stride = conv_params->dst_stride; + int im_h = h + filter_params_y->taps - 1; + int im_stride = 8; + int i, j; + const int fo_vert = filter_params_y->taps / 2 - 1; + const int fo_horiz = filter_params_x->taps / 2 - 1; + const uint16_t *const src_ptr = src - fo_vert * src_stride - fo_horiz; + + // Check that, even with 12-bit input, the intermediate values will fit + // into an unsigned 16-bit intermediate array. + assert(bd + FILTER_BITS + 2 - conv_params->round_0 <= 16); + + __m256i s[8], coeffs_y[4], coeffs_x[4]; + const int do_average = conv_params->do_average; + const int use_jnt_comp_avg = conv_params->use_jnt_comp_avg; + + const int w0 = conv_params->fwd_offset; + const int w1 = conv_params->bck_offset; + const __m256i wt0 = _mm256_set1_epi32(w0); + const __m256i wt1 = _mm256_set1_epi32(w1); + const __m256i zero = _mm256_setzero_si256(); + + const __m256i round_const_x = _mm256_set1_epi32( + ((1 << conv_params->round_0) >> 1) + (1 << (bd + FILTER_BITS - 1))); + const __m128i round_shift_x = _mm_cvtsi32_si128(conv_params->round_0); + + const __m256i round_const_y = _mm256_set1_epi32( + ((1 << conv_params->round_1) >> 1) - + (1 << (bd + 2 * FILTER_BITS - conv_params->round_0 - 1))); + const __m128i round_shift_y = _mm_cvtsi32_si128(conv_params->round_1); + + const int offset_0 = + bd + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const int offset = (1 << offset_0) + (1 << (offset_0 - 1)); + const __m256i offset_const = _mm256_set1_epi32(offset); + const int rounding_shift = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const __m256i rounding_const = _mm256_set1_epi32((1 << rounding_shift) >> 1); + + const __m256i clip_pixel_to_bd = + _mm256_set1_epi16(bd == 10 ? 1023 : (bd == 12 ? 4095 : 255)); + + prepare_coeffs(filter_params_x, subpel_x_q4, coeffs_x); + prepare_coeffs(filter_params_y, subpel_y_q4, coeffs_y); + + for (j = 0; j < w; j += 8) { + /* Horizontal filter */ + { + for (i = 0; i < im_h; i += 2) { + const __m256i row0 = + _mm256_loadu_si256((__m256i *)&src_ptr[i * src_stride + j]); + __m256i row1 = _mm256_set1_epi16(0); + if (i + 1 < im_h) + row1 = + _mm256_loadu_si256((__m256i *)&src_ptr[(i + 1) * src_stride + j]); + + const __m256i r0 = _mm256_permute2x128_si256(row0, row1, 0x20); + const __m256i r1 = _mm256_permute2x128_si256(row0, row1, 0x31); + + // even pixels + s[0] = _mm256_alignr_epi8(r1, r0, 0); + s[1] = _mm256_alignr_epi8(r1, r0, 4); + s[2] = _mm256_alignr_epi8(r1, r0, 8); + s[3] = _mm256_alignr_epi8(r1, r0, 12); + + __m256i res_even = convolve(s, coeffs_x); + res_even = _mm256_sra_epi32(_mm256_add_epi32(res_even, round_const_x), + round_shift_x); + + // odd pixels + s[0] = _mm256_alignr_epi8(r1, r0, 2); + s[1] = _mm256_alignr_epi8(r1, r0, 6); + s[2] = _mm256_alignr_epi8(r1, r0, 10); + s[3] = _mm256_alignr_epi8(r1, r0, 14); + + __m256i res_odd = convolve(s, coeffs_x); + res_odd = _mm256_sra_epi32(_mm256_add_epi32(res_odd, round_const_x), + round_shift_x); + + __m256i res_even1 = _mm256_packs_epi32(res_even, res_even); + __m256i res_odd1 = _mm256_packs_epi32(res_odd, res_odd); + __m256i res = _mm256_unpacklo_epi16(res_even1, res_odd1); + + _mm256_store_si256((__m256i *)&im_block[i * im_stride], res); + } + } + + /* Vertical filter */ + { + __m256i s0 = _mm256_loadu_si256((__m256i *)(im_block + 0 * im_stride)); + __m256i s1 = _mm256_loadu_si256((__m256i *)(im_block + 1 * im_stride)); + __m256i s2 = _mm256_loadu_si256((__m256i *)(im_block + 2 * im_stride)); + __m256i s3 = _mm256_loadu_si256((__m256i *)(im_block + 3 * im_stride)); + __m256i s4 = _mm256_loadu_si256((__m256i *)(im_block + 4 * im_stride)); + __m256i s5 = _mm256_loadu_si256((__m256i *)(im_block + 5 * im_stride)); + + s[0] = _mm256_unpacklo_epi16(s0, s1); + s[1] = _mm256_unpacklo_epi16(s2, s3); + s[2] = _mm256_unpacklo_epi16(s4, s5); + + s[4] = _mm256_unpackhi_epi16(s0, s1); + s[5] = _mm256_unpackhi_epi16(s2, s3); + s[6] = _mm256_unpackhi_epi16(s4, s5); + + for (i = 0; i < h; i += 2) { + const int16_t *data = &im_block[i * im_stride]; + + const __m256i s6 = + _mm256_loadu_si256((__m256i *)(data + 6 * im_stride)); + const __m256i s7 = + _mm256_loadu_si256((__m256i *)(data + 7 * im_stride)); + + s[3] = _mm256_unpacklo_epi16(s6, s7); + s[7] = _mm256_unpackhi_epi16(s6, s7); + + const __m256i res_a = convolve(s, coeffs_y); + + const __m256i res_a_round = _mm256_sra_epi32( + _mm256_add_epi32(res_a, round_const_y), round_shift_y); + + const __m256i res_unsigned_lo = + _mm256_add_epi32(res_a_round, offset_const); + + if (w - j < 8) { + if (do_average) { + const __m256i data_0 = _mm256_castsi128_si256( + _mm_loadl_epi64((__m128i *)(&dst[i * dst_stride + j]))); + const __m256i data_1 = _mm256_castsi128_si256(_mm_loadl_epi64( + (__m128i *)(&dst[i * dst_stride + j + dst_stride]))); + const __m256i data_01 = + _mm256_permute2x128_si256(data_0, data_1, 0x20); + + const __m256i data_ref_0 = _mm256_unpacklo_epi16(data_01, zero); + + const __m256i comp_avg_res = highbd_comp_avg( + &data_ref_0, &res_unsigned_lo, &wt0, &wt1, use_jnt_comp_avg); + + const __m256i round_result = highbd_convolve_rounding( + &comp_avg_res, &offset_const, &rounding_const, rounding_shift); + + const __m256i res_16b = + _mm256_packus_epi32(round_result, round_result); + const __m256i res_clip = + _mm256_min_epi16(res_16b, clip_pixel_to_bd); + + const __m128i res_0 = _mm256_castsi256_si128(res_clip); + const __m128i res_1 = _mm256_extracti128_si256(res_clip, 1); + + _mm_storel_epi64((__m128i *)(&dst0[i * dst_stride0 + j]), res_0); + _mm_storel_epi64( + (__m128i *)(&dst0[i * dst_stride0 + j + dst_stride0]), res_1); + } else { + __m256i res_16b = + _mm256_packus_epi32(res_unsigned_lo, res_unsigned_lo); + const __m128i res_0 = _mm256_castsi256_si128(res_16b); + const __m128i res_1 = _mm256_extracti128_si256(res_16b, 1); + + _mm_storel_epi64((__m128i *)(&dst[i * dst_stride + j]), res_0); + _mm_storel_epi64((__m128i *)(&dst[i * dst_stride + j + dst_stride]), + res_1); + } + } else { + const __m256i res_b = convolve(s + 4, coeffs_y); + const __m256i res_b_round = _mm256_sra_epi32( + _mm256_add_epi32(res_b, round_const_y), round_shift_y); + + __m256i res_unsigned_hi = _mm256_add_epi32(res_b_round, offset_const); + + if (do_average) { + const __m256i data_0 = _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(&dst[i * dst_stride + j]))); + const __m256i data_1 = _mm256_castsi128_si256(_mm_loadu_si128( + (__m128i *)(&dst[i * dst_stride + j + dst_stride]))); + const __m256i data_01 = + _mm256_permute2x128_si256(data_0, data_1, 0x20); + + const __m256i data_ref_0_lo = _mm256_unpacklo_epi16(data_01, zero); + const __m256i data_ref_0_hi = _mm256_unpackhi_epi16(data_01, zero); + + const __m256i comp_avg_res_lo = highbd_comp_avg( + &data_ref_0_lo, &res_unsigned_lo, &wt0, &wt1, use_jnt_comp_avg); + const __m256i comp_avg_res_hi = highbd_comp_avg( + &data_ref_0_hi, &res_unsigned_hi, &wt0, &wt1, use_jnt_comp_avg); + + const __m256i round_result_lo = + highbd_convolve_rounding(&comp_avg_res_lo, &offset_const, + &rounding_const, rounding_shift); + const __m256i round_result_hi = + highbd_convolve_rounding(&comp_avg_res_hi, &offset_const, + &rounding_const, rounding_shift); + + const __m256i res_16b = + _mm256_packus_epi32(round_result_lo, round_result_hi); + const __m256i res_clip = + _mm256_min_epi16(res_16b, clip_pixel_to_bd); + + const __m128i res_0 = _mm256_castsi256_si128(res_clip); + const __m128i res_1 = _mm256_extracti128_si256(res_clip, 1); + + _mm_store_si128((__m128i *)(&dst0[i * dst_stride0 + j]), res_0); + _mm_store_si128( + (__m128i *)(&dst0[i * dst_stride0 + j + dst_stride0]), res_1); + } else { + __m256i res_16b = + _mm256_packus_epi32(res_unsigned_lo, res_unsigned_hi); + const __m128i res_0 = _mm256_castsi256_si128(res_16b); + const __m128i res_1 = _mm256_extracti128_si256(res_16b, 1); + + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j]), res_0); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j + dst_stride]), + res_1); + } + } + + s[0] = s[1]; + s[1] = s[2]; + s[2] = s[3]; + + s[4] = s[5]; + s[5] = s[6]; + s[6] = s[7]; + } + } + } +} + +void av1_highbd_jnt_convolve_x_avx2( + const uint16_t *src, int src_stride, uint16_t *dst0, int dst_stride0, int w, + int h, const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, const int subpel_x_q4, + const int subpel_y_q4, ConvolveParams *conv_params, int bd) { + CONV_BUF_TYPE *dst = conv_params->dst; + int dst_stride = conv_params->dst_stride; + const int fo_horiz = filter_params_x->taps / 2 - 1; + const uint16_t *const src_ptr = src - fo_horiz; + const int bits = FILTER_BITS - conv_params->round_1; + (void)filter_params_y; + (void)subpel_y_q4; + + int i, j; + __m256i s[4], coeffs_x[4]; + + const int do_average = conv_params->do_average; + const int use_jnt_comp_avg = conv_params->use_jnt_comp_avg; + const int w0 = conv_params->fwd_offset; + const int w1 = conv_params->bck_offset; + const __m256i wt0 = _mm256_set1_epi32(w0); + const __m256i wt1 = _mm256_set1_epi32(w1); + const __m256i zero = _mm256_setzero_si256(); + + const __m256i round_const_x = + _mm256_set1_epi32(((1 << conv_params->round_0) >> 1)); + const __m128i round_shift_x = _mm_cvtsi32_si128(conv_params->round_0); + const __m128i round_shift_bits = _mm_cvtsi32_si128(bits); + + const int offset_0 = + bd + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const int offset = (1 << offset_0) + (1 << (offset_0 - 1)); + const __m256i offset_const = _mm256_set1_epi32(offset); + const int rounding_shift = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const __m256i rounding_const = _mm256_set1_epi32((1 << rounding_shift) >> 1); + const __m256i clip_pixel_to_bd = + _mm256_set1_epi16(bd == 10 ? 1023 : (bd == 12 ? 4095 : 255)); + + assert(bits >= 0); + prepare_coeffs(filter_params_x, subpel_x_q4, coeffs_x); + + for (j = 0; j < w; j += 8) { + /* Horizontal filter */ + for (i = 0; i < h; i += 2) { + const __m256i row0 = + _mm256_loadu_si256((__m256i *)&src_ptr[i * src_stride + j]); + __m256i row1 = + _mm256_loadu_si256((__m256i *)&src_ptr[(i + 1) * src_stride + j]); + + const __m256i r0 = _mm256_permute2x128_si256(row0, row1, 0x20); + const __m256i r1 = _mm256_permute2x128_si256(row0, row1, 0x31); + + // even pixels + s[0] = _mm256_alignr_epi8(r1, r0, 0); + s[1] = _mm256_alignr_epi8(r1, r0, 4); + s[2] = _mm256_alignr_epi8(r1, r0, 8); + s[3] = _mm256_alignr_epi8(r1, r0, 12); + + __m256i res_even = convolve(s, coeffs_x); + res_even = _mm256_sra_epi32(_mm256_add_epi32(res_even, round_const_x), + round_shift_x); + + // odd pixels + s[0] = _mm256_alignr_epi8(r1, r0, 2); + s[1] = _mm256_alignr_epi8(r1, r0, 6); + s[2] = _mm256_alignr_epi8(r1, r0, 10); + s[3] = _mm256_alignr_epi8(r1, r0, 14); + + __m256i res_odd = convolve(s, coeffs_x); + res_odd = _mm256_sra_epi32(_mm256_add_epi32(res_odd, round_const_x), + round_shift_x); + + res_even = _mm256_sll_epi32(res_even, round_shift_bits); + res_odd = _mm256_sll_epi32(res_odd, round_shift_bits); + + __m256i res1 = _mm256_unpacklo_epi32(res_even, res_odd); + + __m256i res_unsigned_lo = _mm256_add_epi32(res1, offset_const); + + if (w - j < 8) { + if (do_average) { + const __m256i data_0 = _mm256_castsi128_si256( + _mm_loadl_epi64((__m128i *)(&dst[i * dst_stride + j]))); + const __m256i data_1 = _mm256_castsi128_si256(_mm_loadl_epi64( + (__m128i *)(&dst[i * dst_stride + j + dst_stride]))); + const __m256i data_01 = + _mm256_permute2x128_si256(data_0, data_1, 0x20); + + const __m256i data_ref_0 = _mm256_unpacklo_epi16(data_01, zero); + + const __m256i comp_avg_res = highbd_comp_avg( + &data_ref_0, &res_unsigned_lo, &wt0, &wt1, use_jnt_comp_avg); + + const __m256i round_result = highbd_convolve_rounding( + &comp_avg_res, &offset_const, &rounding_const, rounding_shift); + + const __m256i res_16b = + _mm256_packus_epi32(round_result, round_result); + const __m256i res_clip = _mm256_min_epi16(res_16b, clip_pixel_to_bd); + + const __m128i res_0 = _mm256_castsi256_si128(res_clip); + const __m128i res_1 = _mm256_extracti128_si256(res_clip, 1); + + _mm_storel_epi64((__m128i *)(&dst0[i * dst_stride0 + j]), res_0); + _mm_storel_epi64( + (__m128i *)(&dst0[i * dst_stride0 + j + dst_stride0]), res_1); + } else { + __m256i res_16b = + _mm256_packus_epi32(res_unsigned_lo, res_unsigned_lo); + const __m128i res_0 = _mm256_castsi256_si128(res_16b); + const __m128i res_1 = _mm256_extracti128_si256(res_16b, 1); + + _mm_storel_epi64((__m128i *)(&dst[i * dst_stride + j]), res_0); + _mm_storel_epi64((__m128i *)(&dst[i * dst_stride + j + dst_stride]), + res_1); + } + } else { + __m256i res2 = _mm256_unpackhi_epi32(res_even, res_odd); + __m256i res_unsigned_hi = _mm256_add_epi32(res2, offset_const); + + if (do_average) { + const __m256i data_0 = _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(&dst[i * dst_stride + j]))); + const __m256i data_1 = _mm256_castsi128_si256(_mm_loadu_si128( + (__m128i *)(&dst[i * dst_stride + j + dst_stride]))); + const __m256i data_01 = + _mm256_permute2x128_si256(data_0, data_1, 0x20); + + const __m256i data_ref_0_lo = _mm256_unpacklo_epi16(data_01, zero); + const __m256i data_ref_0_hi = _mm256_unpackhi_epi16(data_01, zero); + + const __m256i comp_avg_res_lo = highbd_comp_avg( + &data_ref_0_lo, &res_unsigned_lo, &wt0, &wt1, use_jnt_comp_avg); + const __m256i comp_avg_res_hi = highbd_comp_avg( + &data_ref_0_hi, &res_unsigned_hi, &wt0, &wt1, use_jnt_comp_avg); + + const __m256i round_result_lo = highbd_convolve_rounding( + &comp_avg_res_lo, &offset_const, &rounding_const, rounding_shift); + const __m256i round_result_hi = highbd_convolve_rounding( + &comp_avg_res_hi, &offset_const, &rounding_const, rounding_shift); + + const __m256i res_16b = + _mm256_packus_epi32(round_result_lo, round_result_hi); + const __m256i res_clip = _mm256_min_epi16(res_16b, clip_pixel_to_bd); + + const __m128i res_0 = _mm256_castsi256_si128(res_clip); + const __m128i res_1 = _mm256_extracti128_si256(res_clip, 1); + + _mm_store_si128((__m128i *)(&dst0[i * dst_stride0 + j]), res_0); + _mm_store_si128((__m128i *)(&dst0[i * dst_stride0 + j + dst_stride0]), + res_1); + } else { + __m256i res_16b = + _mm256_packus_epi32(res_unsigned_lo, res_unsigned_hi); + const __m128i res_0 = _mm256_castsi256_si128(res_16b); + const __m128i res_1 = _mm256_extracti128_si256(res_16b, 1); + + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j]), res_0); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j + dst_stride]), + res_1); + } + } + } + } +} + +void av1_highbd_jnt_convolve_y_avx2( + const uint16_t *src, int src_stride, uint16_t *dst0, int dst_stride0, int w, + int h, const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, const int subpel_x_q4, + const int subpel_y_q4, ConvolveParams *conv_params, int bd) { + CONV_BUF_TYPE *dst = conv_params->dst; + int dst_stride = conv_params->dst_stride; + const int fo_vert = filter_params_y->taps / 2 - 1; + const uint16_t *const src_ptr = src - fo_vert * src_stride; + const int bits = FILTER_BITS - conv_params->round_0; + (void)filter_params_x; + (void)subpel_x_q4; + + assert(bits >= 0); + int i, j; + __m256i s[8], coeffs_y[4]; + const int do_average = conv_params->do_average; + const int use_jnt_comp_avg = conv_params->use_jnt_comp_avg; + + const int w0 = conv_params->fwd_offset; + const int w1 = conv_params->bck_offset; + const __m256i wt0 = _mm256_set1_epi32(w0); + const __m256i wt1 = _mm256_set1_epi32(w1); + const __m256i round_const_y = + _mm256_set1_epi32(((1 << conv_params->round_1) >> 1)); + const __m128i round_shift_y = _mm_cvtsi32_si128(conv_params->round_1); + const __m128i round_shift_bits = _mm_cvtsi32_si128(bits); + + const int offset_0 = + bd + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const int offset = (1 << offset_0) + (1 << (offset_0 - 1)); + const __m256i offset_const = _mm256_set1_epi32(offset); + const int rounding_shift = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const __m256i rounding_const = _mm256_set1_epi32((1 << rounding_shift) >> 1); + const __m256i clip_pixel_to_bd = + _mm256_set1_epi16(bd == 10 ? 1023 : (bd == 12 ? 4095 : 255)); + const __m256i zero = _mm256_setzero_si256(); + + prepare_coeffs(filter_params_y, subpel_y_q4, coeffs_y); + + for (j = 0; j < w; j += 8) { + const uint16_t *data = &src_ptr[j]; + /* Vertical filter */ + { + __m256i src6; + __m256i s01 = _mm256_permute2x128_si256( + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 0 * src_stride))), + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 1 * src_stride))), + 0x20); + __m256i s12 = _mm256_permute2x128_si256( + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 1 * src_stride))), + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 2 * src_stride))), + 0x20); + __m256i s23 = _mm256_permute2x128_si256( + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 2 * src_stride))), + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 3 * src_stride))), + 0x20); + __m256i s34 = _mm256_permute2x128_si256( + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 3 * src_stride))), + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 4 * src_stride))), + 0x20); + __m256i s45 = _mm256_permute2x128_si256( + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 4 * src_stride))), + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 5 * src_stride))), + 0x20); + src6 = _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 6 * src_stride))); + __m256i s56 = _mm256_permute2x128_si256( + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 5 * src_stride))), + src6, 0x20); + + s[0] = _mm256_unpacklo_epi16(s01, s12); + s[1] = _mm256_unpacklo_epi16(s23, s34); + s[2] = _mm256_unpacklo_epi16(s45, s56); + + s[4] = _mm256_unpackhi_epi16(s01, s12); + s[5] = _mm256_unpackhi_epi16(s23, s34); + s[6] = _mm256_unpackhi_epi16(s45, s56); + + for (i = 0; i < h; i += 2) { + data = &src_ptr[i * src_stride + j]; + + const __m256i s67 = _mm256_permute2x128_si256( + src6, + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 7 * src_stride))), + 0x20); + + src6 = _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 8 * src_stride))); + + const __m256i s78 = _mm256_permute2x128_si256( + _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + 7 * src_stride))), + src6, 0x20); + + s[3] = _mm256_unpacklo_epi16(s67, s78); + s[7] = _mm256_unpackhi_epi16(s67, s78); + + const __m256i res_a = convolve(s, coeffs_y); + + __m256i res_a_round = _mm256_sll_epi32(res_a, round_shift_bits); + res_a_round = _mm256_sra_epi32( + _mm256_add_epi32(res_a_round, round_const_y), round_shift_y); + + __m256i res_unsigned_lo = _mm256_add_epi32(res_a_round, offset_const); + + if (w - j < 8) { + if (do_average) { + const __m256i data_0 = _mm256_castsi128_si256( + _mm_loadl_epi64((__m128i *)(&dst[i * dst_stride + j]))); + const __m256i data_1 = _mm256_castsi128_si256(_mm_loadl_epi64( + (__m128i *)(&dst[i * dst_stride + j + dst_stride]))); + const __m256i data_01 = + _mm256_permute2x128_si256(data_0, data_1, 0x20); + + const __m256i data_ref_0 = _mm256_unpacklo_epi16(data_01, zero); + + const __m256i comp_avg_res = highbd_comp_avg( + &data_ref_0, &res_unsigned_lo, &wt0, &wt1, use_jnt_comp_avg); + + const __m256i round_result = highbd_convolve_rounding( + &comp_avg_res, &offset_const, &rounding_const, rounding_shift); + + const __m256i res_16b = + _mm256_packus_epi32(round_result, round_result); + const __m256i res_clip = + _mm256_min_epi16(res_16b, clip_pixel_to_bd); + + const __m128i res_0 = _mm256_castsi256_si128(res_clip); + const __m128i res_1 = _mm256_extracti128_si256(res_clip, 1); + + _mm_storel_epi64((__m128i *)(&dst0[i * dst_stride0 + j]), res_0); + _mm_storel_epi64( + (__m128i *)(&dst0[i * dst_stride0 + j + dst_stride0]), res_1); + } else { + __m256i res_16b = + _mm256_packus_epi32(res_unsigned_lo, res_unsigned_lo); + const __m128i res_0 = _mm256_castsi256_si128(res_16b); + const __m128i res_1 = _mm256_extracti128_si256(res_16b, 1); + + _mm_storel_epi64((__m128i *)(&dst[i * dst_stride + j]), res_0); + _mm_storel_epi64((__m128i *)(&dst[i * dst_stride + j + dst_stride]), + res_1); + } + } else { + const __m256i res_b = convolve(s + 4, coeffs_y); + __m256i res_b_round = _mm256_sll_epi32(res_b, round_shift_bits); + res_b_round = _mm256_sra_epi32( + _mm256_add_epi32(res_b_round, round_const_y), round_shift_y); + + __m256i res_unsigned_hi = _mm256_add_epi32(res_b_round, offset_const); + + if (do_average) { + const __m256i data_0 = _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(&dst[i * dst_stride + j]))); + const __m256i data_1 = _mm256_castsi128_si256(_mm_loadu_si128( + (__m128i *)(&dst[i * dst_stride + j + dst_stride]))); + const __m256i data_01 = + _mm256_permute2x128_si256(data_0, data_1, 0x20); + + const __m256i data_ref_0_lo = _mm256_unpacklo_epi16(data_01, zero); + const __m256i data_ref_0_hi = _mm256_unpackhi_epi16(data_01, zero); + + const __m256i comp_avg_res_lo = highbd_comp_avg( + &data_ref_0_lo, &res_unsigned_lo, &wt0, &wt1, use_jnt_comp_avg); + const __m256i comp_avg_res_hi = highbd_comp_avg( + &data_ref_0_hi, &res_unsigned_hi, &wt0, &wt1, use_jnt_comp_avg); + + const __m256i round_result_lo = + highbd_convolve_rounding(&comp_avg_res_lo, &offset_const, + &rounding_const, rounding_shift); + const __m256i round_result_hi = + highbd_convolve_rounding(&comp_avg_res_hi, &offset_const, + &rounding_const, rounding_shift); + + const __m256i res_16b = + _mm256_packus_epi32(round_result_lo, round_result_hi); + const __m256i res_clip = + _mm256_min_epi16(res_16b, clip_pixel_to_bd); + + const __m128i res_0 = _mm256_castsi256_si128(res_clip); + const __m128i res_1 = _mm256_extracti128_si256(res_clip, 1); + + _mm_store_si128((__m128i *)(&dst0[i * dst_stride0 + j]), res_0); + _mm_store_si128( + (__m128i *)(&dst0[i * dst_stride0 + j + dst_stride0]), res_1); + } else { + __m256i res_16b = + _mm256_packus_epi32(res_unsigned_lo, res_unsigned_hi); + const __m128i res_0 = _mm256_castsi256_si128(res_16b); + const __m128i res_1 = _mm256_extracti128_si256(res_16b, 1); + + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j]), res_0); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j + dst_stride]), + res_1); + } + } + s[0] = s[1]; + s[1] = s[2]; + s[2] = s[3]; + + s[4] = s[5]; + s[5] = s[6]; + s[6] = s[7]; + } + } + } +} diff --git a/third_party/aom/av1/common/x86/highbd_jnt_convolve_sse4.c b/third_party/aom/av1/common/x86/highbd_jnt_convolve_sse4.c new file mode 100644 index 0000000000..1a29985b56 --- /dev/null +++ b/third_party/aom/av1/common/x86/highbd_jnt_convolve_sse4.c @@ -0,0 +1,383 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <smmintrin.h> +#include <assert.h> + +#include "config/aom_dsp_rtcd.h" + +#include "aom_dsp/x86/convolve_sse2.h" +#include "aom_dsp/x86/convolve_sse4_1.h" + +void av1_highbd_jnt_convolve_y_sse4_1( + const uint16_t *src, int src_stride, uint16_t *dst0, int dst_stride0, int w, + int h, const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, const int subpel_x_q4, + const int subpel_y_q4, ConvolveParams *conv_params, int bd) { + CONV_BUF_TYPE *dst = conv_params->dst; + int dst_stride = conv_params->dst_stride; + const int fo_vert = filter_params_y->taps / 2 - 1; + const uint16_t *const src_ptr = src - fo_vert * src_stride; + const int bits = FILTER_BITS - conv_params->round_0; + (void)filter_params_x; + (void)subpel_x_q4; + + assert(bits >= 0); + int i, j; + const int do_average = conv_params->do_average; + const int use_jnt_comp_avg = conv_params->use_jnt_comp_avg; + + const int w0 = conv_params->fwd_offset; + const int w1 = conv_params->bck_offset; + const __m128i wt0 = _mm_set1_epi32(w0); + const __m128i wt1 = _mm_set1_epi32(w1); + const __m128i round_const_y = + _mm_set1_epi32(((1 << conv_params->round_1) >> 1)); + const __m128i round_shift_y = _mm_cvtsi32_si128(conv_params->round_1); + const __m128i round_shift_bits = _mm_cvtsi32_si128(bits); + + const int offset_0 = + bd + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const int offset = (1 << offset_0) + (1 << (offset_0 - 1)); + const __m128i offset_const = _mm_set1_epi32(offset); + const int rounding_shift = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const __m128i rounding_const = _mm_set1_epi32((1 << rounding_shift) >> 1); + const __m128i clip_pixel_to_bd = + _mm_set1_epi16(bd == 10 ? 1023 : (bd == 12 ? 4095 : 255)); + const __m128i zero = _mm_setzero_si128(); + __m128i s[16], coeffs_y[4]; + + prepare_coeffs(filter_params_y, subpel_y_q4, coeffs_y); + + for (j = 0; j < w; j += 8) { + const uint16_t *data = &src_ptr[j]; + /* Vertical filter */ + { + __m128i s0 = _mm_loadu_si128((__m128i *)(data + 0 * src_stride)); + __m128i s1 = _mm_loadu_si128((__m128i *)(data + 1 * src_stride)); + __m128i s2 = _mm_loadu_si128((__m128i *)(data + 2 * src_stride)); + __m128i s3 = _mm_loadu_si128((__m128i *)(data + 3 * src_stride)); + __m128i s4 = _mm_loadu_si128((__m128i *)(data + 4 * src_stride)); + __m128i s5 = _mm_loadu_si128((__m128i *)(data + 5 * src_stride)); + __m128i s6 = _mm_loadu_si128((__m128i *)(data + 6 * src_stride)); + + s[0] = _mm_unpacklo_epi16(s0, s1); + s[1] = _mm_unpacklo_epi16(s2, s3); + s[2] = _mm_unpacklo_epi16(s4, s5); + + s[4] = _mm_unpackhi_epi16(s0, s1); + s[5] = _mm_unpackhi_epi16(s2, s3); + s[6] = _mm_unpackhi_epi16(s4, s5); + + s[0 + 8] = _mm_unpacklo_epi16(s1, s2); + s[1 + 8] = _mm_unpacklo_epi16(s3, s4); + s[2 + 8] = _mm_unpacklo_epi16(s5, s6); + + s[4 + 8] = _mm_unpackhi_epi16(s1, s2); + s[5 + 8] = _mm_unpackhi_epi16(s3, s4); + s[6 + 8] = _mm_unpackhi_epi16(s5, s6); + + for (i = 0; i < h; i += 2) { + data = &src_ptr[i * src_stride + j]; + + __m128i s7 = _mm_loadu_si128((__m128i *)(data + 7 * src_stride)); + __m128i s8 = _mm_loadu_si128((__m128i *)(data + 8 * src_stride)); + + s[3] = _mm_unpacklo_epi16(s6, s7); + s[7] = _mm_unpackhi_epi16(s6, s7); + + s[3 + 8] = _mm_unpacklo_epi16(s7, s8); + s[7 + 8] = _mm_unpackhi_epi16(s7, s8); + + const __m128i res_a0 = convolve(s, coeffs_y); + __m128i res_a_round0 = _mm_sll_epi32(res_a0, round_shift_bits); + res_a_round0 = _mm_sra_epi32(_mm_add_epi32(res_a_round0, round_const_y), + round_shift_y); + + const __m128i res_a1 = convolve(s + 8, coeffs_y); + __m128i res_a_round1 = _mm_sll_epi32(res_a1, round_shift_bits); + res_a_round1 = _mm_sra_epi32(_mm_add_epi32(res_a_round1, round_const_y), + round_shift_y); + + __m128i res_unsigned_lo_0 = _mm_add_epi32(res_a_round0, offset_const); + __m128i res_unsigned_lo_1 = _mm_add_epi32(res_a_round1, offset_const); + + if (w - j < 8) { + if (do_average) { + const __m128i data_0 = + _mm_loadl_epi64((__m128i *)(&dst[i * dst_stride + j])); + const __m128i data_1 = _mm_loadl_epi64( + (__m128i *)(&dst[i * dst_stride + j + dst_stride])); + + const __m128i data_ref_0 = _mm_unpacklo_epi16(data_0, zero); + const __m128i data_ref_1 = _mm_unpacklo_epi16(data_1, zero); + + const __m128i comp_avg_res_0 = highbd_comp_avg_sse4_1( + &data_ref_0, &res_unsigned_lo_0, &wt0, &wt1, use_jnt_comp_avg); + const __m128i comp_avg_res_1 = highbd_comp_avg_sse4_1( + &data_ref_1, &res_unsigned_lo_1, &wt0, &wt1, use_jnt_comp_avg); + + const __m128i round_result_0 = + highbd_convolve_rounding_sse2(&comp_avg_res_0, &offset_const, + &rounding_const, rounding_shift); + const __m128i round_result_1 = + highbd_convolve_rounding_sse2(&comp_avg_res_1, &offset_const, + &rounding_const, rounding_shift); + + const __m128i res_16b_0 = + _mm_packus_epi32(round_result_0, round_result_0); + const __m128i res_clip_0 = + _mm_min_epi16(res_16b_0, clip_pixel_to_bd); + const __m128i res_16b_1 = + _mm_packus_epi32(round_result_1, round_result_1); + const __m128i res_clip_1 = + _mm_min_epi16(res_16b_1, clip_pixel_to_bd); + + _mm_storel_epi64((__m128i *)(&dst0[i * dst_stride0 + j]), + res_clip_0); + _mm_storel_epi64( + (__m128i *)(&dst0[i * dst_stride0 + j + dst_stride0]), + res_clip_1); + + } else { + __m128i res_16b_0 = + _mm_packus_epi32(res_unsigned_lo_0, res_unsigned_lo_0); + + __m128i res_16b_1 = + _mm_packus_epi32(res_unsigned_lo_1, res_unsigned_lo_1); + + _mm_storel_epi64((__m128i *)&dst[i * dst_stride + j], res_16b_0); + _mm_storel_epi64((__m128i *)&dst[i * dst_stride + j + dst_stride], + res_16b_1); + } + } else { + const __m128i res_b0 = convolve(s + 4, coeffs_y); + __m128i res_b_round0 = _mm_sll_epi32(res_b0, round_shift_bits); + res_b_round0 = _mm_sra_epi32( + _mm_add_epi32(res_b_round0, round_const_y), round_shift_y); + + const __m128i res_b1 = convolve(s + 4 + 8, coeffs_y); + __m128i res_b_round1 = _mm_sll_epi32(res_b1, round_shift_bits); + res_b_round1 = _mm_sra_epi32( + _mm_add_epi32(res_b_round1, round_const_y), round_shift_y); + + __m128i res_unsigned_hi_0 = _mm_add_epi32(res_b_round0, offset_const); + __m128i res_unsigned_hi_1 = _mm_add_epi32(res_b_round1, offset_const); + + if (do_average) { + const __m128i data_0 = + _mm_loadu_si128((__m128i *)(&dst[i * dst_stride + j])); + const __m128i data_1 = _mm_loadu_si128( + (__m128i *)(&dst[i * dst_stride + j + dst_stride])); + const __m128i data_ref_0_lo_0 = _mm_unpacklo_epi16(data_0, zero); + const __m128i data_ref_0_lo_1 = _mm_unpacklo_epi16(data_1, zero); + + const __m128i data_ref_0_hi_0 = _mm_unpackhi_epi16(data_0, zero); + const __m128i data_ref_0_hi_1 = _mm_unpackhi_epi16(data_1, zero); + + const __m128i comp_avg_res_lo_0 = + highbd_comp_avg_sse4_1(&data_ref_0_lo_0, &res_unsigned_lo_0, + &wt0, &wt1, use_jnt_comp_avg); + const __m128i comp_avg_res_lo_1 = + highbd_comp_avg_sse4_1(&data_ref_0_lo_1, &res_unsigned_lo_1, + &wt0, &wt1, use_jnt_comp_avg); + const __m128i comp_avg_res_hi_0 = + highbd_comp_avg_sse4_1(&data_ref_0_hi_0, &res_unsigned_hi_0, + &wt0, &wt1, use_jnt_comp_avg); + const __m128i comp_avg_res_hi_1 = + highbd_comp_avg_sse4_1(&data_ref_0_hi_1, &res_unsigned_hi_1, + &wt0, &wt1, use_jnt_comp_avg); + + const __m128i round_result_lo_0 = + highbd_convolve_rounding_sse2(&comp_avg_res_lo_0, &offset_const, + &rounding_const, rounding_shift); + const __m128i round_result_lo_1 = + highbd_convolve_rounding_sse2(&comp_avg_res_lo_1, &offset_const, + &rounding_const, rounding_shift); + const __m128i round_result_hi_0 = + highbd_convolve_rounding_sse2(&comp_avg_res_hi_0, &offset_const, + &rounding_const, rounding_shift); + const __m128i round_result_hi_1 = + highbd_convolve_rounding_sse2(&comp_avg_res_hi_1, &offset_const, + &rounding_const, rounding_shift); + + const __m128i res_16b_0 = + _mm_packus_epi32(round_result_lo_0, round_result_hi_0); + const __m128i res_clip_0 = + _mm_min_epi16(res_16b_0, clip_pixel_to_bd); + + const __m128i res_16b_1 = + _mm_packus_epi32(round_result_lo_1, round_result_hi_1); + const __m128i res_clip_1 = + _mm_min_epi16(res_16b_1, clip_pixel_to_bd); + + _mm_store_si128((__m128i *)(&dst0[i * dst_stride0 + j]), + res_clip_0); + _mm_store_si128( + (__m128i *)(&dst0[i * dst_stride0 + j + dst_stride0]), + res_clip_1); + } else { + __m128i res_16bit0 = + _mm_packus_epi32(res_unsigned_lo_0, res_unsigned_hi_0); + __m128i res_16bit1 = + _mm_packus_epi32(res_unsigned_lo_1, res_unsigned_hi_1); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j]), res_16bit0); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j + dst_stride]), + res_16bit1); + } + } + s[0] = s[1]; + s[1] = s[2]; + s[2] = s[3]; + + s[4] = s[5]; + s[5] = s[6]; + s[6] = s[7]; + + s[0 + 8] = s[1 + 8]; + s[1 + 8] = s[2 + 8]; + s[2 + 8] = s[3 + 8]; + + s[4 + 8] = s[5 + 8]; + s[5 + 8] = s[6 + 8]; + s[6 + 8] = s[7 + 8]; + + s6 = s8; + } + } + } +} + +void av1_highbd_jnt_convolve_x_sse4_1( + const uint16_t *src, int src_stride, uint16_t *dst0, int dst_stride0, int w, + int h, const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, const int subpel_x_q4, + const int subpel_y_q4, ConvolveParams *conv_params, int bd) { + CONV_BUF_TYPE *dst = conv_params->dst; + int dst_stride = conv_params->dst_stride; + const int fo_horiz = filter_params_x->taps / 2 - 1; + const uint16_t *const src_ptr = src - fo_horiz; + const int bits = FILTER_BITS - conv_params->round_1; + (void)filter_params_y; + (void)subpel_y_q4; + + int i, j; + __m128i s[4], coeffs_x[4]; + + const int do_average = conv_params->do_average; + const int use_jnt_comp_avg = conv_params->use_jnt_comp_avg; + const int w0 = conv_params->fwd_offset; + const int w1 = conv_params->bck_offset; + const __m128i wt0 = _mm_set1_epi32(w0); + const __m128i wt1 = _mm_set1_epi32(w1); + const __m128i zero = _mm_setzero_si128(); + + const __m128i round_const_x = + _mm_set1_epi32(((1 << conv_params->round_0) >> 1)); + const __m128i round_shift_x = _mm_cvtsi32_si128(conv_params->round_0); + const __m128i round_shift_bits = _mm_cvtsi32_si128(bits); + + const int offset_0 = + bd + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const int offset = (1 << offset_0) + (1 << (offset_0 - 1)); + const __m128i offset_const = _mm_set1_epi32(offset); + const int rounding_shift = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const __m128i rounding_const = _mm_set1_epi32((1 << rounding_shift) >> 1); + const __m128i clip_pixel_to_bd = + _mm_set1_epi16(bd == 10 ? 1023 : (bd == 12 ? 4095 : 255)); + + assert(bits >= 0); + prepare_coeffs(filter_params_x, subpel_x_q4, coeffs_x); + + for (j = 0; j < w; j += 8) { + /* Horizontal filter */ + for (i = 0; i < h; i += 1) { + const __m128i row00 = + _mm_loadu_si128((__m128i *)&src_ptr[i * src_stride + j]); + const __m128i row01 = + _mm_loadu_si128((__m128i *)&src_ptr[i * src_stride + (j + 8)]); + + // even pixels + s[0] = _mm_alignr_epi8(row01, row00, 0); + s[1] = _mm_alignr_epi8(row01, row00, 4); + s[2] = _mm_alignr_epi8(row01, row00, 8); + s[3] = _mm_alignr_epi8(row01, row00, 12); + + __m128i res_even = convolve(s, coeffs_x); + res_even = + _mm_sra_epi32(_mm_add_epi32(res_even, round_const_x), round_shift_x); + + // odd pixels + s[0] = _mm_alignr_epi8(row01, row00, 2); + s[1] = _mm_alignr_epi8(row01, row00, 6); + s[2] = _mm_alignr_epi8(row01, row00, 10); + s[3] = _mm_alignr_epi8(row01, row00, 14); + + __m128i res_odd = convolve(s, coeffs_x); + res_odd = + _mm_sra_epi32(_mm_add_epi32(res_odd, round_const_x), round_shift_x); + + res_even = _mm_sll_epi32(res_even, round_shift_bits); + res_odd = _mm_sll_epi32(res_odd, round_shift_bits); + + __m128i res1 = _mm_unpacklo_epi32(res_even, res_odd); + __m128i res_unsigned_lo = _mm_add_epi32(res1, offset_const); + if (w - j < 8) { + if (do_average) { + const __m128i data_0 = + _mm_loadl_epi64((__m128i *)(&dst[i * dst_stride + j])); + const __m128i data_ref_0 = _mm_unpacklo_epi16(data_0, zero); + + const __m128i comp_avg_res = highbd_comp_avg_sse4_1( + &data_ref_0, &res_unsigned_lo, &wt0, &wt1, use_jnt_comp_avg); + const __m128i round_result = highbd_convolve_rounding_sse2( + &comp_avg_res, &offset_const, &rounding_const, rounding_shift); + + const __m128i res_16b = _mm_packus_epi32(round_result, round_result); + const __m128i res_clip = _mm_min_epi16(res_16b, clip_pixel_to_bd); + _mm_storel_epi64((__m128i *)(&dst0[i * dst_stride0 + j]), res_clip); + } else { + __m128i res_16b = _mm_packus_epi32(res_unsigned_lo, res_unsigned_lo); + _mm_storel_epi64((__m128i *)&dst[i * dst_stride + j], res_16b); + } + } else { + __m128i res2 = _mm_unpackhi_epi32(res_even, res_odd); + __m128i res_unsigned_hi = _mm_add_epi32(res2, offset_const); + if (do_average) { + const __m128i data_0 = + _mm_loadu_si128((__m128i *)(&dst[i * dst_stride + j])); + const __m128i data_ref_0_lo = _mm_unpacklo_epi16(data_0, zero); + const __m128i data_ref_0_hi = _mm_unpackhi_epi16(data_0, zero); + + const __m128i comp_avg_res_lo = highbd_comp_avg_sse4_1( + &data_ref_0_lo, &res_unsigned_lo, &wt0, &wt1, use_jnt_comp_avg); + const __m128i comp_avg_res_hi = highbd_comp_avg_sse4_1( + &data_ref_0_hi, &res_unsigned_hi, &wt0, &wt1, use_jnt_comp_avg); + + const __m128i round_result_lo = highbd_convolve_rounding_sse2( + &comp_avg_res_lo, &offset_const, &rounding_const, rounding_shift); + const __m128i round_result_hi = highbd_convolve_rounding_sse2( + &comp_avg_res_hi, &offset_const, &rounding_const, rounding_shift); + + const __m128i res_16b = + _mm_packus_epi32(round_result_lo, round_result_hi); + const __m128i res_clip = _mm_min_epi16(res_16b, clip_pixel_to_bd); + _mm_store_si128((__m128i *)(&dst0[i * dst_stride0 + j]), res_clip); + } else { + __m128i res_16b = _mm_packus_epi32(res_unsigned_lo, res_unsigned_hi); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j]), res_16b); + } + } + } + } +} diff --git a/third_party/aom/av1/common/x86/highbd_txfm_utility_sse4.h b/third_party/aom/av1/common/x86/highbd_txfm_utility_sse4.h new file mode 100644 index 0000000000..6f24e5948a --- /dev/null +++ b/third_party/aom/av1/common/x86/highbd_txfm_utility_sse4.h @@ -0,0 +1,125 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#ifndef AOM_AV1_COMMON_X86_HIGHBD_TXFM_UTILITY_SSE4_H_ +#define AOM_AV1_COMMON_X86_HIGHBD_TXFM_UTILITY_SSE4_H_ + +#include <smmintrin.h> /* SSE4.1 */ + +#define TRANSPOSE_4X4(x0, x1, x2, x3, y0, y1, y2, y3) \ + do { \ + __m128i u0, u1, u2, u3; \ + u0 = _mm_unpacklo_epi32(x0, x1); \ + u1 = _mm_unpackhi_epi32(x0, x1); \ + u2 = _mm_unpacklo_epi32(x2, x3); \ + u3 = _mm_unpackhi_epi32(x2, x3); \ + y0 = _mm_unpacklo_epi64(u0, u2); \ + y1 = _mm_unpackhi_epi64(u0, u2); \ + y2 = _mm_unpacklo_epi64(u1, u3); \ + y3 = _mm_unpackhi_epi64(u1, u3); \ + } while (0) + +static INLINE void transpose_8x8(const __m128i *in, __m128i *out) { + TRANSPOSE_4X4(in[0], in[2], in[4], in[6], out[0], out[2], out[4], out[6]); + TRANSPOSE_4X4(in[1], in[3], in[5], in[7], out[8], out[10], out[12], out[14]); + TRANSPOSE_4X4(in[8], in[10], in[12], in[14], out[1], out[3], out[5], out[7]); + TRANSPOSE_4X4(in[9], in[11], in[13], in[15], out[9], out[11], out[13], + out[15]); +} + +static INLINE void transpose_16x16(const __m128i *in, __m128i *out) { + // Upper left 8x8 + TRANSPOSE_4X4(in[0], in[4], in[8], in[12], out[0], out[4], out[8], out[12]); + TRANSPOSE_4X4(in[1], in[5], in[9], in[13], out[16], out[20], out[24], + out[28]); + TRANSPOSE_4X4(in[16], in[20], in[24], in[28], out[1], out[5], out[9], + out[13]); + TRANSPOSE_4X4(in[17], in[21], in[25], in[29], out[17], out[21], out[25], + out[29]); + + // Upper right 8x8 + TRANSPOSE_4X4(in[2], in[6], in[10], in[14], out[32], out[36], out[40], + out[44]); + TRANSPOSE_4X4(in[3], in[7], in[11], in[15], out[48], out[52], out[56], + out[60]); + TRANSPOSE_4X4(in[18], in[22], in[26], in[30], out[33], out[37], out[41], + out[45]); + TRANSPOSE_4X4(in[19], in[23], in[27], in[31], out[49], out[53], out[57], + out[61]); + + // Lower left 8x8 + TRANSPOSE_4X4(in[32], in[36], in[40], in[44], out[2], out[6], out[10], + out[14]); + TRANSPOSE_4X4(in[33], in[37], in[41], in[45], out[18], out[22], out[26], + out[30]); + TRANSPOSE_4X4(in[48], in[52], in[56], in[60], out[3], out[7], out[11], + out[15]); + TRANSPOSE_4X4(in[49], in[53], in[57], in[61], out[19], out[23], out[27], + out[31]); + // Lower right 8x8 + TRANSPOSE_4X4(in[34], in[38], in[42], in[46], out[34], out[38], out[42], + out[46]); + TRANSPOSE_4X4(in[35], in[39], in[43], in[47], out[50], out[54], out[58], + out[62]); + TRANSPOSE_4X4(in[50], in[54], in[58], in[62], out[35], out[39], out[43], + out[47]); + TRANSPOSE_4X4(in[51], in[55], in[59], in[63], out[51], out[55], out[59], + out[63]); +} + +static INLINE void transpose_32x32(const __m128i *input, __m128i *output) { + for (int j = 0; j < 8; j++) { + for (int i = 0; i < 8; i++) { + TRANSPOSE_4X4(input[i * 32 + j + 0], input[i * 32 + j + 8], + input[i * 32 + j + 16], input[i * 32 + j + 24], + output[j * 32 + i + 0], output[j * 32 + i + 8], + output[j * 32 + i + 16], output[j * 32 + i + 24]); + } + } +} + +// Note: +// rounding = 1 << (bit - 1) +static INLINE __m128i half_btf_sse4_1(const __m128i *w0, const __m128i *n0, + const __m128i *w1, const __m128i *n1, + const __m128i *rounding, int bit) { + __m128i x, y; + + x = _mm_mullo_epi32(*w0, *n0); + y = _mm_mullo_epi32(*w1, *n1); + x = _mm_add_epi32(x, y); + x = _mm_add_epi32(x, *rounding); + x = _mm_srai_epi32(x, bit); + return x; +} + +static INLINE __m128i half_btf_0_sse4_1(const __m128i *w0, const __m128i *n0, + const __m128i *rounding, int bit) { + __m128i x; + + x = _mm_mullo_epi32(*w0, *n0); + x = _mm_add_epi32(x, *rounding); + x = _mm_srai_epi32(x, bit); + return x; +} + +typedef void (*transform_1d_sse4_1)(__m128i *in, __m128i *out, int bit, + int do_cols, int bd, int out_shift); + +typedef void (*fwd_transform_1d_sse4_1)(__m128i *in, __m128i *out, int bit, + const int num_cols); + +void av1_highbd_inv_txfm2d_add_universe_sse4_1(const int32_t *input, + uint8_t *output, int stride, + TX_TYPE tx_type, TX_SIZE tx_size, + int eob, const int bd); + +#endif // AOM_AV1_COMMON_X86_HIGHBD_TXFM_UTILITY_SSE4_H_ diff --git a/third_party/aom/av1/common/x86/highbd_warp_plane_sse4.c b/third_party/aom/av1/common/x86/highbd_warp_plane_sse4.c new file mode 100644 index 0000000000..4bcab05645 --- /dev/null +++ b/third_party/aom/av1/common/x86/highbd_warp_plane_sse4.c @@ -0,0 +1,624 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <smmintrin.h> + +#include "config/av1_rtcd.h" + +#include "av1/common/warped_motion.h" + +static const uint8_t warp_highbd_arrange_bytes[16] = { + 0, 2, 4, 6, 8, 10, 12, 14, 1, 3, 5, 7, 9, 11, 13, 15 +}; + +static const uint8_t highbd_shuffle_alpha0_mask0[16] = { + 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3 +}; +static const uint8_t highbd_shuffle_alpha0_mask1[16] = { + 4, 5, 6, 7, 4, 5, 6, 7, 4, 5, 6, 7, 4, 5, 6, 7 +}; +static const uint8_t highbd_shuffle_alpha0_mask2[16] = { + 8, 9, 10, 11, 8, 9, 10, 11, 8, 9, 10, 11, 8, 9, 10, 11 +}; +static const uint8_t highbd_shuffle_alpha0_mask3[16] = { + 12, 13, 14, 15, 12, 13, 14, 15, 12, 13, 14, 15, 12, 13, 14, 15 +}; + +static INLINE void highbd_prepare_horizontal_filter_coeff(int alpha, int sx, + __m128i *coeff) { + // Filter even-index pixels + const __m128i tmp_0 = _mm_loadu_si128( + (__m128i *)(warped_filter + ((sx + 0 * alpha) >> WARPEDDIFF_PREC_BITS))); + const __m128i tmp_2 = _mm_loadu_si128( + (__m128i *)(warped_filter + ((sx + 2 * alpha) >> WARPEDDIFF_PREC_BITS))); + const __m128i tmp_4 = _mm_loadu_si128( + (__m128i *)(warped_filter + ((sx + 4 * alpha) >> WARPEDDIFF_PREC_BITS))); + const __m128i tmp_6 = _mm_loadu_si128( + (__m128i *)(warped_filter + ((sx + 6 * alpha) >> WARPEDDIFF_PREC_BITS))); + + // coeffs 0 1 0 1 2 3 2 3 for pixels 0, 2 + const __m128i tmp_8 = _mm_unpacklo_epi32(tmp_0, tmp_2); + // coeffs 0 1 0 1 2 3 2 3 for pixels 4, 6 + const __m128i tmp_10 = _mm_unpacklo_epi32(tmp_4, tmp_6); + // coeffs 4 5 4 5 6 7 6 7 for pixels 0, 2 + const __m128i tmp_12 = _mm_unpackhi_epi32(tmp_0, tmp_2); + // coeffs 4 5 4 5 6 7 6 7 for pixels 4, 6 + const __m128i tmp_14 = _mm_unpackhi_epi32(tmp_4, tmp_6); + + // coeffs 0 1 0 1 0 1 0 1 for pixels 0, 2, 4, 6 + coeff[0] = _mm_unpacklo_epi64(tmp_8, tmp_10); + // coeffs 2 3 2 3 2 3 2 3 for pixels 0, 2, 4, 6 + coeff[2] = _mm_unpackhi_epi64(tmp_8, tmp_10); + // coeffs 4 5 4 5 4 5 4 5 for pixels 0, 2, 4, 6 + coeff[4] = _mm_unpacklo_epi64(tmp_12, tmp_14); + // coeffs 6 7 6 7 6 7 6 7 for pixels 0, 2, 4, 6 + coeff[6] = _mm_unpackhi_epi64(tmp_12, tmp_14); + + // Filter odd-index pixels + const __m128i tmp_1 = _mm_loadu_si128( + (__m128i *)(warped_filter + ((sx + 1 * alpha) >> WARPEDDIFF_PREC_BITS))); + const __m128i tmp_3 = _mm_loadu_si128( + (__m128i *)(warped_filter + ((sx + 3 * alpha) >> WARPEDDIFF_PREC_BITS))); + const __m128i tmp_5 = _mm_loadu_si128( + (__m128i *)(warped_filter + ((sx + 5 * alpha) >> WARPEDDIFF_PREC_BITS))); + const __m128i tmp_7 = _mm_loadu_si128( + (__m128i *)(warped_filter + ((sx + 7 * alpha) >> WARPEDDIFF_PREC_BITS))); + + const __m128i tmp_9 = _mm_unpacklo_epi32(tmp_1, tmp_3); + const __m128i tmp_11 = _mm_unpacklo_epi32(tmp_5, tmp_7); + const __m128i tmp_13 = _mm_unpackhi_epi32(tmp_1, tmp_3); + const __m128i tmp_15 = _mm_unpackhi_epi32(tmp_5, tmp_7); + + coeff[1] = _mm_unpacklo_epi64(tmp_9, tmp_11); + coeff[3] = _mm_unpackhi_epi64(tmp_9, tmp_11); + coeff[5] = _mm_unpacklo_epi64(tmp_13, tmp_15); + coeff[7] = _mm_unpackhi_epi64(tmp_13, tmp_15); +} + +static INLINE void highbd_prepare_horizontal_filter_coeff_alpha0( + int sx, __m128i *coeff) { + // Filter coeff + const __m128i tmp_0 = _mm_loadu_si128( + (__m128i *)(warped_filter + (sx >> WARPEDDIFF_PREC_BITS))); + + coeff[0] = _mm_shuffle_epi8( + tmp_0, _mm_loadu_si128((__m128i *)highbd_shuffle_alpha0_mask0)); + coeff[2] = _mm_shuffle_epi8( + tmp_0, _mm_loadu_si128((__m128i *)highbd_shuffle_alpha0_mask1)); + coeff[4] = _mm_shuffle_epi8( + tmp_0, _mm_loadu_si128((__m128i *)highbd_shuffle_alpha0_mask2)); + coeff[6] = _mm_shuffle_epi8( + tmp_0, _mm_loadu_si128((__m128i *)highbd_shuffle_alpha0_mask3)); + + coeff[1] = coeff[0]; + coeff[3] = coeff[2]; + coeff[5] = coeff[4]; + coeff[7] = coeff[6]; +} + +static INLINE void highbd_filter_src_pixels( + const __m128i *src, const __m128i *src2, __m128i *tmp, __m128i *coeff, + const int offset_bits_horiz, const int reduce_bits_horiz, int k) { + const __m128i src_1 = *src; + const __m128i src2_1 = *src2; + + const __m128i round_const = _mm_set1_epi32((1 << offset_bits_horiz) + + ((1 << reduce_bits_horiz) >> 1)); + + const __m128i res_0 = _mm_madd_epi16(src_1, coeff[0]); + const __m128i res_2 = + _mm_madd_epi16(_mm_alignr_epi8(src2_1, src_1, 4), coeff[2]); + const __m128i res_4 = + _mm_madd_epi16(_mm_alignr_epi8(src2_1, src_1, 8), coeff[4]); + const __m128i res_6 = + _mm_madd_epi16(_mm_alignr_epi8(src2_1, src_1, 12), coeff[6]); + + __m128i res_even = + _mm_add_epi32(_mm_add_epi32(res_0, res_4), _mm_add_epi32(res_2, res_6)); + res_even = _mm_sra_epi32(_mm_add_epi32(res_even, round_const), + _mm_cvtsi32_si128(reduce_bits_horiz)); + + const __m128i res_1 = + _mm_madd_epi16(_mm_alignr_epi8(src2_1, src_1, 2), coeff[1]); + const __m128i res_3 = + _mm_madd_epi16(_mm_alignr_epi8(src2_1, src_1, 6), coeff[3]); + const __m128i res_5 = + _mm_madd_epi16(_mm_alignr_epi8(src2_1, src_1, 10), coeff[5]); + const __m128i res_7 = + _mm_madd_epi16(_mm_alignr_epi8(src2_1, src_1, 14), coeff[7]); + + __m128i res_odd = + _mm_add_epi32(_mm_add_epi32(res_1, res_5), _mm_add_epi32(res_3, res_7)); + res_odd = _mm_sra_epi32(_mm_add_epi32(res_odd, round_const), + _mm_cvtsi32_si128(reduce_bits_horiz)); + + // Combine results into one register. + // We store the columns in the order 0, 2, 4, 6, 1, 3, 5, 7 + // as this order helps with the vertical filter. + tmp[k + 7] = _mm_packs_epi32(res_even, res_odd); +} + +static INLINE void highbd_horiz_filter(const __m128i *src, const __m128i *src2, + __m128i *tmp, int sx, int alpha, int k, + const int offset_bits_horiz, + const int reduce_bits_horiz) { + __m128i coeff[8]; + highbd_prepare_horizontal_filter_coeff(alpha, sx, coeff); + highbd_filter_src_pixels(src, src2, tmp, coeff, offset_bits_horiz, + reduce_bits_horiz, k); +} + +static INLINE void highbd_warp_horizontal_filter_alpha0_beta0( + const uint16_t *ref, __m128i *tmp, int stride, int32_t ix4, int32_t iy4, + int32_t sx4, int alpha, int beta, int p_height, int height, int i, + const int offset_bits_horiz, const int reduce_bits_horiz) { + (void)beta; + (void)alpha; + int k; + + __m128i coeff[8]; + highbd_prepare_horizontal_filter_coeff_alpha0(sx4, coeff); + + for (k = -7; k < AOMMIN(8, p_height - i); ++k) { + int iy = iy4 + k; + if (iy < 0) + iy = 0; + else if (iy > height - 1) + iy = height - 1; + + // Load source pixels + const __m128i src = + _mm_loadu_si128((__m128i *)(ref + iy * stride + ix4 - 7)); + const __m128i src2 = + _mm_loadu_si128((__m128i *)(ref + iy * stride + ix4 + 1)); + highbd_filter_src_pixels(&src, &src2, tmp, coeff, offset_bits_horiz, + reduce_bits_horiz, k); + } +} + +static INLINE void highbd_warp_horizontal_filter_alpha0( + const uint16_t *ref, __m128i *tmp, int stride, int32_t ix4, int32_t iy4, + int32_t sx4, int alpha, int beta, int p_height, int height, int i, + const int offset_bits_horiz, const int reduce_bits_horiz) { + (void)alpha; + int k; + for (k = -7; k < AOMMIN(8, p_height - i); ++k) { + int iy = iy4 + k; + if (iy < 0) + iy = 0; + else if (iy > height - 1) + iy = height - 1; + int sx = sx4 + beta * (k + 4); + + // Load source pixels + const __m128i src = + _mm_loadu_si128((__m128i *)(ref + iy * stride + ix4 - 7)); + const __m128i src2 = + _mm_loadu_si128((__m128i *)(ref + iy * stride + ix4 + 1)); + + __m128i coeff[8]; + highbd_prepare_horizontal_filter_coeff_alpha0(sx, coeff); + highbd_filter_src_pixels(&src, &src2, tmp, coeff, offset_bits_horiz, + reduce_bits_horiz, k); + } +} + +static INLINE void highbd_warp_horizontal_filter_beta0( + const uint16_t *ref, __m128i *tmp, int stride, int32_t ix4, int32_t iy4, + int32_t sx4, int alpha, int beta, int p_height, int height, int i, + const int offset_bits_horiz, const int reduce_bits_horiz) { + (void)beta; + int k; + __m128i coeff[8]; + highbd_prepare_horizontal_filter_coeff(alpha, sx4, coeff); + + for (k = -7; k < AOMMIN(8, p_height - i); ++k) { + int iy = iy4 + k; + if (iy < 0) + iy = 0; + else if (iy > height - 1) + iy = height - 1; + + // Load source pixels + const __m128i src = + _mm_loadu_si128((__m128i *)(ref + iy * stride + ix4 - 7)); + const __m128i src2 = + _mm_loadu_si128((__m128i *)(ref + iy * stride + ix4 + 1)); + highbd_filter_src_pixels(&src, &src2, tmp, coeff, offset_bits_horiz, + reduce_bits_horiz, k); + } +} + +static INLINE void highbd_warp_horizontal_filter( + const uint16_t *ref, __m128i *tmp, int stride, int32_t ix4, int32_t iy4, + int32_t sx4, int alpha, int beta, int p_height, int height, int i, + const int offset_bits_horiz, const int reduce_bits_horiz) { + int k; + for (k = -7; k < AOMMIN(8, p_height - i); ++k) { + int iy = iy4 + k; + if (iy < 0) + iy = 0; + else if (iy > height - 1) + iy = height - 1; + int sx = sx4 + beta * (k + 4); + + // Load source pixels + const __m128i src = + _mm_loadu_si128((__m128i *)(ref + iy * stride + ix4 - 7)); + const __m128i src2 = + _mm_loadu_si128((__m128i *)(ref + iy * stride + ix4 + 1)); + + highbd_horiz_filter(&src, &src2, tmp, sx, alpha, k, offset_bits_horiz, + reduce_bits_horiz); + } +} + +static INLINE void highbd_prepare_warp_horizontal_filter( + const uint16_t *ref, __m128i *tmp, int stride, int32_t ix4, int32_t iy4, + int32_t sx4, int alpha, int beta, int p_height, int height, int i, + const int offset_bits_horiz, const int reduce_bits_horiz) { + if (alpha == 0 && beta == 0) + highbd_warp_horizontal_filter_alpha0_beta0( + ref, tmp, stride, ix4, iy4, sx4, alpha, beta, p_height, height, i, + offset_bits_horiz, reduce_bits_horiz); + + else if (alpha == 0 && beta != 0) + highbd_warp_horizontal_filter_alpha0(ref, tmp, stride, ix4, iy4, sx4, alpha, + beta, p_height, height, i, + offset_bits_horiz, reduce_bits_horiz); + + else if (alpha != 0 && beta == 0) + highbd_warp_horizontal_filter_beta0(ref, tmp, stride, ix4, iy4, sx4, alpha, + beta, p_height, height, i, + offset_bits_horiz, reduce_bits_horiz); + else + highbd_warp_horizontal_filter(ref, tmp, stride, ix4, iy4, sx4, alpha, beta, + p_height, height, i, offset_bits_horiz, + reduce_bits_horiz); +} + +void av1_highbd_warp_affine_sse4_1(const int32_t *mat, const uint16_t *ref, + int width, int height, int stride, + uint16_t *pred, int p_col, int p_row, + int p_width, int p_height, int p_stride, + int subsampling_x, int subsampling_y, int bd, + ConvolveParams *conv_params, int16_t alpha, + int16_t beta, int16_t gamma, int16_t delta) { + __m128i tmp[15]; + int i, j, k; + const int reduce_bits_horiz = + conv_params->round_0 + + AOMMAX(bd + FILTER_BITS - conv_params->round_0 - 14, 0); + const int reduce_bits_vert = conv_params->is_compound + ? conv_params->round_1 + : 2 * FILTER_BITS - reduce_bits_horiz; + const int offset_bits_horiz = bd + FILTER_BITS - 1; + assert(IMPLIES(conv_params->is_compound, conv_params->dst != NULL)); + assert(!(bd == 12 && reduce_bits_horiz < 5)); + assert(IMPLIES(conv_params->do_average, conv_params->is_compound)); + + const int offset_bits_vert = bd + 2 * FILTER_BITS - reduce_bits_horiz; + const __m128i clip_pixel = + _mm_set1_epi16(bd == 10 ? 1023 : (bd == 12 ? 4095 : 255)); + const __m128i reduce_bits_vert_shift = _mm_cvtsi32_si128(reduce_bits_vert); + const __m128i reduce_bits_vert_const = + _mm_set1_epi32(((1 << reduce_bits_vert) >> 1)); + const __m128i res_add_const = _mm_set1_epi32(1 << offset_bits_vert); + const int round_bits = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const int offset_bits = bd + 2 * FILTER_BITS - conv_params->round_0; + const __m128i res_sub_const = + _mm_set1_epi32(-(1 << (offset_bits - conv_params->round_1)) - + (1 << (offset_bits - conv_params->round_1 - 1))); + __m128i round_bits_shift = _mm_cvtsi32_si128(round_bits); + __m128i round_bits_const = _mm_set1_epi32(((1 << round_bits) >> 1)); + + const int w0 = conv_params->fwd_offset; + const int w1 = conv_params->bck_offset; + const __m128i wt0 = _mm_set1_epi32(w0); + const __m128i wt1 = _mm_set1_epi32(w1); + + /* Note: For this code to work, the left/right frame borders need to be + extended by at least 13 pixels each. By the time we get here, other + code will have set up this border, but we allow an explicit check + for debugging purposes. + */ + /*for (i = 0; i < height; ++i) { + for (j = 0; j < 13; ++j) { + assert(ref[i * stride - 13 + j] == ref[i * stride]); + assert(ref[i * stride + width + j] == ref[i * stride + (width - 1)]); + } + }*/ + + for (i = 0; i < p_height; i += 8) { + for (j = 0; j < p_width; j += 8) { + const int32_t src_x = (p_col + j + 4) << subsampling_x; + const int32_t src_y = (p_row + i + 4) << subsampling_y; + const int32_t dst_x = mat[2] * src_x + mat[3] * src_y + mat[0]; + const int32_t dst_y = mat[4] * src_x + mat[5] * src_y + mat[1]; + const int32_t x4 = dst_x >> subsampling_x; + const int32_t y4 = dst_y >> subsampling_y; + + int32_t ix4 = x4 >> WARPEDMODEL_PREC_BITS; + int32_t sx4 = x4 & ((1 << WARPEDMODEL_PREC_BITS) - 1); + int32_t iy4 = y4 >> WARPEDMODEL_PREC_BITS; + int32_t sy4 = y4 & ((1 << WARPEDMODEL_PREC_BITS) - 1); + + // Add in all the constant terms, including rounding and offset + sx4 += alpha * (-4) + beta * (-4) + (1 << (WARPEDDIFF_PREC_BITS - 1)) + + (WARPEDPIXEL_PREC_SHIFTS << WARPEDDIFF_PREC_BITS); + sy4 += gamma * (-4) + delta * (-4) + (1 << (WARPEDDIFF_PREC_BITS - 1)) + + (WARPEDPIXEL_PREC_SHIFTS << WARPEDDIFF_PREC_BITS); + + sx4 &= ~((1 << WARP_PARAM_REDUCE_BITS) - 1); + sy4 &= ~((1 << WARP_PARAM_REDUCE_BITS) - 1); + + // Horizontal filter + // If the block is aligned such that, after clamping, every sample + // would be taken from the leftmost/rightmost column, then we can + // skip the expensive horizontal filter. + if (ix4 <= -7) { + for (k = -7; k < AOMMIN(8, p_height - i); ++k) { + int iy = iy4 + k; + if (iy < 0) + iy = 0; + else if (iy > height - 1) + iy = height - 1; + tmp[k + 7] = _mm_set1_epi16( + (1 << (bd + FILTER_BITS - reduce_bits_horiz - 1)) + + ref[iy * stride] * (1 << (FILTER_BITS - reduce_bits_horiz))); + } + } else if (ix4 >= width + 6) { + for (k = -7; k < AOMMIN(8, p_height - i); ++k) { + int iy = iy4 + k; + if (iy < 0) + iy = 0; + else if (iy > height - 1) + iy = height - 1; + tmp[k + 7] = + _mm_set1_epi16((1 << (bd + FILTER_BITS - reduce_bits_horiz - 1)) + + ref[iy * stride + (width - 1)] * + (1 << (FILTER_BITS - reduce_bits_horiz))); + } + } else if (((ix4 - 7) < 0) || ((ix4 + 9) > width)) { + const int out_of_boundary_left = -(ix4 - 6); + const int out_of_boundary_right = (ix4 + 8) - width; + + for (k = -7; k < AOMMIN(8, p_height - i); ++k) { + int iy = iy4 + k; + if (iy < 0) + iy = 0; + else if (iy > height - 1) + iy = height - 1; + int sx = sx4 + beta * (k + 4); + + // Load source pixels + const __m128i src = + _mm_loadu_si128((__m128i *)(ref + iy * stride + ix4 - 7)); + const __m128i src2 = + _mm_loadu_si128((__m128i *)(ref + iy * stride + ix4 + 1)); + + const __m128i src_01 = _mm_shuffle_epi8( + src, _mm_loadu_si128((__m128i *)warp_highbd_arrange_bytes)); + const __m128i src2_01 = _mm_shuffle_epi8( + src2, _mm_loadu_si128((__m128i *)warp_highbd_arrange_bytes)); + + __m128i src_lo = _mm_unpacklo_epi64(src_01, src2_01); + __m128i src_hi = _mm_unpackhi_epi64(src_01, src2_01); + + if (out_of_boundary_left >= 0) { + const __m128i shuffle_reg_left = + _mm_loadu_si128((__m128i *)warp_pad_left[out_of_boundary_left]); + src_lo = _mm_shuffle_epi8(src_lo, shuffle_reg_left); + src_hi = _mm_shuffle_epi8(src_hi, shuffle_reg_left); + } + + if (out_of_boundary_right >= 0) { + const __m128i shuffle_reg_right = _mm_loadu_si128( + (__m128i *)warp_pad_right[out_of_boundary_right]); + src_lo = _mm_shuffle_epi8(src_lo, shuffle_reg_right); + src_hi = _mm_shuffle_epi8(src_hi, shuffle_reg_right); + } + + const __m128i src_padded = _mm_unpacklo_epi8(src_lo, src_hi); + const __m128i src2_padded = _mm_unpackhi_epi8(src_lo, src_hi); + + highbd_horiz_filter(&src_padded, &src2_padded, tmp, sx, alpha, k, + offset_bits_horiz, reduce_bits_horiz); + } + } else { + highbd_prepare_warp_horizontal_filter( + ref, tmp, stride, ix4, iy4, sx4, alpha, beta, p_height, height, i, + offset_bits_horiz, reduce_bits_horiz); + } + + // Vertical filter + for (k = -4; k < AOMMIN(4, p_height - i - 4); ++k) { + int sy = sy4 + delta * (k + 4); + + // Load from tmp and rearrange pairs of consecutive rows into the + // column order 0 0 2 2 4 4 6 6; 1 1 3 3 5 5 7 7 + const __m128i *src = tmp + (k + 4); + const __m128i src_0 = _mm_unpacklo_epi16(src[0], src[1]); + const __m128i src_2 = _mm_unpacklo_epi16(src[2], src[3]); + const __m128i src_4 = _mm_unpacklo_epi16(src[4], src[5]); + const __m128i src_6 = _mm_unpacklo_epi16(src[6], src[7]); + + // Filter even-index pixels + const __m128i tmp_0 = _mm_loadu_si128( + (__m128i *)(warped_filter + + ((sy + 0 * gamma) >> WARPEDDIFF_PREC_BITS))); + const __m128i tmp_2 = _mm_loadu_si128( + (__m128i *)(warped_filter + + ((sy + 2 * gamma) >> WARPEDDIFF_PREC_BITS))); + const __m128i tmp_4 = _mm_loadu_si128( + (__m128i *)(warped_filter + + ((sy + 4 * gamma) >> WARPEDDIFF_PREC_BITS))); + const __m128i tmp_6 = _mm_loadu_si128( + (__m128i *)(warped_filter + + ((sy + 6 * gamma) >> WARPEDDIFF_PREC_BITS))); + + const __m128i tmp_8 = _mm_unpacklo_epi32(tmp_0, tmp_2); + const __m128i tmp_10 = _mm_unpacklo_epi32(tmp_4, tmp_6); + const __m128i tmp_12 = _mm_unpackhi_epi32(tmp_0, tmp_2); + const __m128i tmp_14 = _mm_unpackhi_epi32(tmp_4, tmp_6); + + const __m128i coeff_0 = _mm_unpacklo_epi64(tmp_8, tmp_10); + const __m128i coeff_2 = _mm_unpackhi_epi64(tmp_8, tmp_10); + const __m128i coeff_4 = _mm_unpacklo_epi64(tmp_12, tmp_14); + const __m128i coeff_6 = _mm_unpackhi_epi64(tmp_12, tmp_14); + + const __m128i res_0 = _mm_madd_epi16(src_0, coeff_0); + const __m128i res_2 = _mm_madd_epi16(src_2, coeff_2); + const __m128i res_4 = _mm_madd_epi16(src_4, coeff_4); + const __m128i res_6 = _mm_madd_epi16(src_6, coeff_6); + + const __m128i res_even = _mm_add_epi32(_mm_add_epi32(res_0, res_2), + _mm_add_epi32(res_4, res_6)); + + // Filter odd-index pixels + const __m128i src_1 = _mm_unpackhi_epi16(src[0], src[1]); + const __m128i src_3 = _mm_unpackhi_epi16(src[2], src[3]); + const __m128i src_5 = _mm_unpackhi_epi16(src[4], src[5]); + const __m128i src_7 = _mm_unpackhi_epi16(src[6], src[7]); + + const __m128i tmp_1 = _mm_loadu_si128( + (__m128i *)(warped_filter + + ((sy + 1 * gamma) >> WARPEDDIFF_PREC_BITS))); + const __m128i tmp_3 = _mm_loadu_si128( + (__m128i *)(warped_filter + + ((sy + 3 * gamma) >> WARPEDDIFF_PREC_BITS))); + const __m128i tmp_5 = _mm_loadu_si128( + (__m128i *)(warped_filter + + ((sy + 5 * gamma) >> WARPEDDIFF_PREC_BITS))); + const __m128i tmp_7 = _mm_loadu_si128( + (__m128i *)(warped_filter + + ((sy + 7 * gamma) >> WARPEDDIFF_PREC_BITS))); + + const __m128i tmp_9 = _mm_unpacklo_epi32(tmp_1, tmp_3); + const __m128i tmp_11 = _mm_unpacklo_epi32(tmp_5, tmp_7); + const __m128i tmp_13 = _mm_unpackhi_epi32(tmp_1, tmp_3); + const __m128i tmp_15 = _mm_unpackhi_epi32(tmp_5, tmp_7); + + const __m128i coeff_1 = _mm_unpacklo_epi64(tmp_9, tmp_11); + const __m128i coeff_3 = _mm_unpackhi_epi64(tmp_9, tmp_11); + const __m128i coeff_5 = _mm_unpacklo_epi64(tmp_13, tmp_15); + const __m128i coeff_7 = _mm_unpackhi_epi64(tmp_13, tmp_15); + + const __m128i res_1 = _mm_madd_epi16(src_1, coeff_1); + const __m128i res_3 = _mm_madd_epi16(src_3, coeff_3); + const __m128i res_5 = _mm_madd_epi16(src_5, coeff_5); + const __m128i res_7 = _mm_madd_epi16(src_7, coeff_7); + + const __m128i res_odd = _mm_add_epi32(_mm_add_epi32(res_1, res_3), + _mm_add_epi32(res_5, res_7)); + + // Rearrange pixels back into the order 0 ... 7 + __m128i res_lo = _mm_unpacklo_epi32(res_even, res_odd); + __m128i res_hi = _mm_unpackhi_epi32(res_even, res_odd); + + if (conv_params->is_compound) { + __m128i *const p = + (__m128i *)&conv_params + ->dst[(i + k + 4) * conv_params->dst_stride + j]; + res_lo = _mm_add_epi32(res_lo, res_add_const); + res_lo = _mm_sra_epi32(_mm_add_epi32(res_lo, reduce_bits_vert_const), + reduce_bits_vert_shift); + + if (conv_params->do_average) { + __m128i *const dst16 = (__m128i *)&pred[(i + k + 4) * p_stride + j]; + __m128i p_32 = _mm_cvtepu16_epi32(_mm_loadl_epi64(p)); + + if (conv_params->use_jnt_comp_avg) { + res_lo = _mm_add_epi32(_mm_mullo_epi32(p_32, wt0), + _mm_mullo_epi32(res_lo, wt1)); + res_lo = _mm_srai_epi32(res_lo, DIST_PRECISION_BITS); + } else { + res_lo = _mm_srai_epi32(_mm_add_epi32(p_32, res_lo), 1); + } + + __m128i res32_lo = _mm_add_epi32(res_lo, res_sub_const); + res32_lo = _mm_sra_epi32(_mm_add_epi32(res32_lo, round_bits_const), + round_bits_shift); + + __m128i res16_lo = _mm_packus_epi32(res32_lo, res32_lo); + res16_lo = _mm_min_epi16(res16_lo, clip_pixel); + _mm_storel_epi64(dst16, res16_lo); + } else { + res_lo = _mm_packus_epi32(res_lo, res_lo); + _mm_storel_epi64(p, res_lo); + } + if (p_width > 4) { + __m128i *const p4 = + (__m128i *)&conv_params + ->dst[(i + k + 4) * conv_params->dst_stride + j + 4]; + + res_hi = _mm_add_epi32(res_hi, res_add_const); + res_hi = + _mm_sra_epi32(_mm_add_epi32(res_hi, reduce_bits_vert_const), + reduce_bits_vert_shift); + if (conv_params->do_average) { + __m128i *const dst16_4 = + (__m128i *)&pred[(i + k + 4) * p_stride + j + 4]; + __m128i p4_32 = _mm_cvtepu16_epi32(_mm_loadl_epi64(p4)); + + if (conv_params->use_jnt_comp_avg) { + res_hi = _mm_add_epi32(_mm_mullo_epi32(p4_32, wt0), + _mm_mullo_epi32(res_hi, wt1)); + res_hi = _mm_srai_epi32(res_hi, DIST_PRECISION_BITS); + } else { + res_hi = _mm_srai_epi32(_mm_add_epi32(p4_32, res_hi), 1); + } + + __m128i res32_hi = _mm_add_epi32(res_hi, res_sub_const); + res32_hi = _mm_sra_epi32( + _mm_add_epi32(res32_hi, round_bits_const), round_bits_shift); + __m128i res16_hi = _mm_packus_epi32(res32_hi, res32_hi); + res16_hi = _mm_min_epi16(res16_hi, clip_pixel); + _mm_storel_epi64(dst16_4, res16_hi); + } else { + res_hi = _mm_packus_epi32(res_hi, res_hi); + _mm_storel_epi64(p4, res_hi); + } + } + } else { + // Round and pack into 8 bits + const __m128i round_const = + _mm_set1_epi32(-(1 << (bd + reduce_bits_vert - 1)) + + ((1 << reduce_bits_vert) >> 1)); + + const __m128i res_lo_round = _mm_srai_epi32( + _mm_add_epi32(res_lo, round_const), reduce_bits_vert); + const __m128i res_hi_round = _mm_srai_epi32( + _mm_add_epi32(res_hi, round_const), reduce_bits_vert); + + __m128i res_16bit = _mm_packs_epi32(res_lo_round, res_hi_round); + // Clamp res_16bit to the range [0, 2^bd - 1] + const __m128i max_val = _mm_set1_epi16((1 << bd) - 1); + const __m128i zero = _mm_setzero_si128(); + res_16bit = _mm_max_epi16(_mm_min_epi16(res_16bit, max_val), zero); + + // Store, blending with 'pred' if needed + __m128i *const p = (__m128i *)&pred[(i + k + 4) * p_stride + j]; + + // Note: If we're outputting a 4x4 block, we need to be very careful + // to only output 4 pixels at this point, to avoid encode/decode + // mismatches when encoding with multiple threads. + if (p_width == 4) { + _mm_storel_epi64(p, res_16bit); + } else { + _mm_storeu_si128(p, res_16bit); + } + } + } + } + } +} diff --git a/third_party/aom/av1/common/x86/highbd_wiener_convolve_avx2.c b/third_party/aom/av1/common/x86/highbd_wiener_convolve_avx2.c new file mode 100644 index 0000000000..0c8a8505b0 --- /dev/null +++ b/third_party/aom/av1/common/x86/highbd_wiener_convolve_avx2.c @@ -0,0 +1,245 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <immintrin.h> +#include <assert.h> + +#include "config/aom_dsp_rtcd.h" + +#include "av1/common/convolve.h" +#include "aom_dsp/aom_dsp_common.h" +#include "aom_dsp/aom_filter.h" +#include "aom_dsp/x86/synonyms.h" +#include "aom_dsp/x86/synonyms_avx2.h" + +// 128-bit xmmwords are written as [ ... ] with the MSB on the left. +// 256-bit ymmwords are written as two xmmwords, [ ... ][ ... ] with the MSB +// on the left. +// A row of, say, 16-bit pixels with values p0, p1, p2, ..., p14, p15 will be +// loaded and stored as [ p15 ... p9 p8 ][ p7 ... p1 p0 ]. +void av1_highbd_wiener_convolve_add_src_avx2( + const uint8_t *src8, ptrdiff_t src_stride, uint8_t *dst8, + ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, + const int16_t *filter_y, int y_step_q4, int w, int h, + const ConvolveParams *conv_params, int bd) { + assert(x_step_q4 == 16 && y_step_q4 == 16); + assert(!(w & 7)); + assert(bd + FILTER_BITS - conv_params->round_0 + 2 <= 16); + (void)x_step_q4; + (void)y_step_q4; + + const uint16_t *const src = CONVERT_TO_SHORTPTR(src8); + uint16_t *const dst = CONVERT_TO_SHORTPTR(dst8); + + DECLARE_ALIGNED(32, uint16_t, + temp[(MAX_SB_SIZE + SUBPEL_TAPS - 1) * MAX_SB_SIZE]); + int intermediate_height = h + SUBPEL_TAPS - 1; + const int center_tap = ((SUBPEL_TAPS - 1) / 2); + const uint16_t *const src_ptr = src - center_tap * src_stride - center_tap; + + const __m128i zero_128 = _mm_setzero_si128(); + const __m256i zero_256 = _mm256_setzero_si256(); + + // Add an offset to account for the "add_src" part of the convolve function. + const __m128i offset = _mm_insert_epi16(zero_128, 1 << FILTER_BITS, 3); + + const __m256i clamp_low = zero_256; + + /* Horizontal filter */ + { + const __m256i clamp_high_ep = + _mm256_set1_epi16(WIENER_CLAMP_LIMIT(conv_params->round_0, bd) - 1); + + // coeffs [ f7 f6 f5 f4 f3 f2 f1 f0 ] + const __m128i coeffs_x = _mm_add_epi16(xx_loadu_128(filter_x), offset); + + // coeffs [ f3 f2 f3 f2 f1 f0 f1 f0 ] + const __m128i coeffs_0123 = _mm_unpacklo_epi32(coeffs_x, coeffs_x); + // coeffs [ f7 f6 f7 f6 f5 f4 f5 f4 ] + const __m128i coeffs_4567 = _mm_unpackhi_epi32(coeffs_x, coeffs_x); + + // coeffs [ f1 f0 f1 f0 f1 f0 f1 f0 ] + const __m128i coeffs_01_128 = _mm_unpacklo_epi64(coeffs_0123, coeffs_0123); + // coeffs [ f3 f2 f3 f2 f3 f2 f3 f2 ] + const __m128i coeffs_23_128 = _mm_unpackhi_epi64(coeffs_0123, coeffs_0123); + // coeffs [ f5 f4 f5 f4 f5 f4 f5 f4 ] + const __m128i coeffs_45_128 = _mm_unpacklo_epi64(coeffs_4567, coeffs_4567); + // coeffs [ f7 f6 f7 f6 f7 f6 f7 f6 ] + const __m128i coeffs_67_128 = _mm_unpackhi_epi64(coeffs_4567, coeffs_4567); + + // coeffs [ f1 f0 f1 f0 f1 f0 f1 f0 ][ f1 f0 f1 f0 f1 f0 f1 f0 ] + const __m256i coeffs_01 = yy_set_m128i(coeffs_01_128, coeffs_01_128); + // coeffs [ f3 f2 f3 f2 f3 f2 f3 f2 ][ f3 f2 f3 f2 f3 f2 f3 f2 ] + const __m256i coeffs_23 = yy_set_m128i(coeffs_23_128, coeffs_23_128); + // coeffs [ f5 f4 f5 f4 f5 f4 f5 f4 ][ f5 f4 f5 f4 f5 f4 f5 f4 ] + const __m256i coeffs_45 = yy_set_m128i(coeffs_45_128, coeffs_45_128); + // coeffs [ f7 f6 f7 f6 f7 f6 f7 f6 ][ f7 f6 f7 f6 f7 f6 f7 f6 ] + const __m256i coeffs_67 = yy_set_m128i(coeffs_67_128, coeffs_67_128); + + const __m256i round_const = _mm256_set1_epi32( + (1 << (conv_params->round_0 - 1)) + (1 << (bd + FILTER_BITS - 1))); + + for (int i = 0; i < intermediate_height; ++i) { + for (int j = 0; j < w; j += 16) { + const uint16_t *src_ij = src_ptr + i * src_stride + j; + + // Load 16-bit src data + const __m256i src_0 = yy_loadu_256(src_ij + 0); + const __m256i src_1 = yy_loadu_256(src_ij + 1); + const __m256i src_2 = yy_loadu_256(src_ij + 2); + const __m256i src_3 = yy_loadu_256(src_ij + 3); + const __m256i src_4 = yy_loadu_256(src_ij + 4); + const __m256i src_5 = yy_loadu_256(src_ij + 5); + const __m256i src_6 = yy_loadu_256(src_ij + 6); + const __m256i src_7 = yy_loadu_256(src_ij + 7); + + // Multiply src data by filter coeffs and sum pairs + const __m256i res_0 = _mm256_madd_epi16(src_0, coeffs_01); + const __m256i res_1 = _mm256_madd_epi16(src_1, coeffs_01); + const __m256i res_2 = _mm256_madd_epi16(src_2, coeffs_23); + const __m256i res_3 = _mm256_madd_epi16(src_3, coeffs_23); + const __m256i res_4 = _mm256_madd_epi16(src_4, coeffs_45); + const __m256i res_5 = _mm256_madd_epi16(src_5, coeffs_45); + const __m256i res_6 = _mm256_madd_epi16(src_6, coeffs_67); + const __m256i res_7 = _mm256_madd_epi16(src_7, coeffs_67); + + // Calculate scalar product for even- and odd-indices separately, + // increasing to 32-bit precision + const __m256i res_even_sum = _mm256_add_epi32( + _mm256_add_epi32(res_0, res_4), _mm256_add_epi32(res_2, res_6)); + const __m256i res_even = _mm256_srai_epi32( + _mm256_add_epi32(res_even_sum, round_const), conv_params->round_0); + + const __m256i res_odd_sum = _mm256_add_epi32( + _mm256_add_epi32(res_1, res_5), _mm256_add_epi32(res_3, res_7)); + const __m256i res_odd = _mm256_srai_epi32( + _mm256_add_epi32(res_odd_sum, round_const), conv_params->round_0); + + // Reduce to 16-bit precision and pack even- and odd-index results + // back into one register. The _mm256_packs_epi32 intrinsic returns + // a register with the pixels ordered as follows: + // [ 15 13 11 9 14 12 10 8 ] [ 7 5 3 1 6 4 2 0 ] + const __m256i res = _mm256_packs_epi32(res_even, res_odd); + const __m256i res_clamped = + _mm256_min_epi16(_mm256_max_epi16(res, clamp_low), clamp_high_ep); + + // Store in a temporary array + yy_storeu_256(temp + i * MAX_SB_SIZE + j, res_clamped); + } + } + } + + /* Vertical filter */ + { + const __m256i clamp_high = _mm256_set1_epi16((1 << bd) - 1); + + // coeffs [ f7 f6 f5 f4 f3 f2 f1 f0 ] + const __m128i coeffs_y = _mm_add_epi16(xx_loadu_128(filter_y), offset); + + // coeffs [ f3 f2 f3 f2 f1 f0 f1 f0 ] + const __m128i coeffs_0123 = _mm_unpacklo_epi32(coeffs_y, coeffs_y); + // coeffs [ f7 f6 f7 f6 f5 f4 f5 f4 ] + const __m128i coeffs_4567 = _mm_unpackhi_epi32(coeffs_y, coeffs_y); + + // coeffs [ f1 f0 f1 f0 f1 f0 f1 f0 ] + const __m128i coeffs_01_128 = _mm_unpacklo_epi64(coeffs_0123, coeffs_0123); + // coeffs [ f3 f2 f3 f2 f3 f2 f3 f2 ] + const __m128i coeffs_23_128 = _mm_unpackhi_epi64(coeffs_0123, coeffs_0123); + // coeffs [ f5 f4 f5 f4 f5 f4 f5 f4 ] + const __m128i coeffs_45_128 = _mm_unpacklo_epi64(coeffs_4567, coeffs_4567); + // coeffs [ f7 f6 f7 f6 f7 f6 f7 f6 ] + const __m128i coeffs_67_128 = _mm_unpackhi_epi64(coeffs_4567, coeffs_4567); + + // coeffs [ f1 f0 f1 f0 f1 f0 f1 f0 ][ f1 f0 f1 f0 f1 f0 f1 f0 ] + const __m256i coeffs_01 = yy_set_m128i(coeffs_01_128, coeffs_01_128); + // coeffs [ f3 f2 f3 f2 f3 f2 f3 f2 ][ f3 f2 f3 f2 f3 f2 f3 f2 ] + const __m256i coeffs_23 = yy_set_m128i(coeffs_23_128, coeffs_23_128); + // coeffs [ f5 f4 f5 f4 f5 f4 f5 f4 ][ f5 f4 f5 f4 f5 f4 f5 f4 ] + const __m256i coeffs_45 = yy_set_m128i(coeffs_45_128, coeffs_45_128); + // coeffs [ f7 f6 f7 f6 f7 f6 f7 f6 ][ f7 f6 f7 f6 f7 f6 f7 f6 ] + const __m256i coeffs_67 = yy_set_m128i(coeffs_67_128, coeffs_67_128); + + const __m256i round_const = + _mm256_set1_epi32((1 << (conv_params->round_1 - 1)) - + (1 << (bd + conv_params->round_1 - 1))); + + for (int i = 0; i < h; ++i) { + for (int j = 0; j < w; j += 16) { + const uint16_t *temp_ij = temp + i * MAX_SB_SIZE + j; + + // Load 16-bit data from the output of the horizontal filter in + // which the pixels are ordered as follows: + // [ 15 13 11 9 14 12 10 8 ] [ 7 5 3 1 6 4 2 0 ] + const __m256i data_0 = yy_loadu_256(temp_ij + 0 * MAX_SB_SIZE); + const __m256i data_1 = yy_loadu_256(temp_ij + 1 * MAX_SB_SIZE); + const __m256i data_2 = yy_loadu_256(temp_ij + 2 * MAX_SB_SIZE); + const __m256i data_3 = yy_loadu_256(temp_ij + 3 * MAX_SB_SIZE); + const __m256i data_4 = yy_loadu_256(temp_ij + 4 * MAX_SB_SIZE); + const __m256i data_5 = yy_loadu_256(temp_ij + 5 * MAX_SB_SIZE); + const __m256i data_6 = yy_loadu_256(temp_ij + 6 * MAX_SB_SIZE); + const __m256i data_7 = yy_loadu_256(temp_ij + 7 * MAX_SB_SIZE); + + // Filter the even-indices, increasing to 32-bit precision + const __m256i src_0 = _mm256_unpacklo_epi16(data_0, data_1); + const __m256i src_2 = _mm256_unpacklo_epi16(data_2, data_3); + const __m256i src_4 = _mm256_unpacklo_epi16(data_4, data_5); + const __m256i src_6 = _mm256_unpacklo_epi16(data_6, data_7); + + const __m256i res_0 = _mm256_madd_epi16(src_0, coeffs_01); + const __m256i res_2 = _mm256_madd_epi16(src_2, coeffs_23); + const __m256i res_4 = _mm256_madd_epi16(src_4, coeffs_45); + const __m256i res_6 = _mm256_madd_epi16(src_6, coeffs_67); + + const __m256i res_even = _mm256_add_epi32( + _mm256_add_epi32(res_0, res_2), _mm256_add_epi32(res_4, res_6)); + + // Filter the odd-indices, increasing to 32-bit precision + const __m256i src_1 = _mm256_unpackhi_epi16(data_0, data_1); + const __m256i src_3 = _mm256_unpackhi_epi16(data_2, data_3); + const __m256i src_5 = _mm256_unpackhi_epi16(data_4, data_5); + const __m256i src_7 = _mm256_unpackhi_epi16(data_6, data_7); + + const __m256i res_1 = _mm256_madd_epi16(src_1, coeffs_01); + const __m256i res_3 = _mm256_madd_epi16(src_3, coeffs_23); + const __m256i res_5 = _mm256_madd_epi16(src_5, coeffs_45); + const __m256i res_7 = _mm256_madd_epi16(src_7, coeffs_67); + + const __m256i res_odd = _mm256_add_epi32( + _mm256_add_epi32(res_1, res_3), _mm256_add_epi32(res_5, res_7)); + + // Pixels are currently in the following order: + // res_even order: [ 14 12 10 8 ] [ 6 4 2 0 ] + // res_odd order: [ 15 13 11 9 ] [ 7 5 3 1 ] + // + // Rearrange the pixels into the following order: + // res_lo order: [ 11 10 9 8 ] [ 3 2 1 0 ] + // res_hi order: [ 15 14 13 12 ] [ 7 6 5 4 ] + const __m256i res_lo = _mm256_unpacklo_epi32(res_even, res_odd); + const __m256i res_hi = _mm256_unpackhi_epi32(res_even, res_odd); + + const __m256i res_lo_round = _mm256_srai_epi32( + _mm256_add_epi32(res_lo, round_const), conv_params->round_1); + const __m256i res_hi_round = _mm256_srai_epi32( + _mm256_add_epi32(res_hi, round_const), conv_params->round_1); + + // Reduce to 16-bit precision and pack into the correct order: + // [ 15 14 13 12 11 10 9 8 ][ 7 6 5 4 3 2 1 0 ] + const __m256i res_16bit = + _mm256_packs_epi32(res_lo_round, res_hi_round); + const __m256i res_16bit_clamped = _mm256_min_epi16( + _mm256_max_epi16(res_16bit, clamp_low), clamp_high); + + // Store in the dst array + yy_storeu_256(dst + i * dst_stride + j, res_16bit_clamped); + } + } + } +} diff --git a/third_party/aom/av1/common/x86/highbd_wiener_convolve_ssse3.c b/third_party/aom/av1/common/x86/highbd_wiener_convolve_ssse3.c new file mode 100644 index 0000000000..818b1099c9 --- /dev/null +++ b/third_party/aom/av1/common/x86/highbd_wiener_convolve_ssse3.c @@ -0,0 +1,202 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <tmmintrin.h> +#include <assert.h> + +#include "config/aom_dsp_rtcd.h" + +#include "av1/common/convolve.h" +#include "aom_dsp/aom_dsp_common.h" +#include "aom_dsp/aom_filter.h" + +void av1_highbd_wiener_convolve_add_src_ssse3( + const uint8_t *src8, ptrdiff_t src_stride, uint8_t *dst8, + ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, + const int16_t *filter_y, int y_step_q4, int w, int h, + const ConvolveParams *conv_params, int bd) { + assert(x_step_q4 == 16 && y_step_q4 == 16); + assert(!(w & 7)); + assert(bd + FILTER_BITS - conv_params->round_0 + 2 <= 16); + (void)x_step_q4; + (void)y_step_q4; + + const uint16_t *const src = CONVERT_TO_SHORTPTR(src8); + uint16_t *const dst = CONVERT_TO_SHORTPTR(dst8); + + DECLARE_ALIGNED(16, uint16_t, + temp[(MAX_SB_SIZE + SUBPEL_TAPS - 1) * MAX_SB_SIZE]); + int intermediate_height = h + SUBPEL_TAPS - 1; + int i, j; + const int center_tap = ((SUBPEL_TAPS - 1) / 2); + const uint16_t *const src_ptr = src - center_tap * src_stride - center_tap; + + const __m128i zero = _mm_setzero_si128(); + // Add an offset to account for the "add_src" part of the convolve function. + const __m128i offset = _mm_insert_epi16(zero, 1 << FILTER_BITS, 3); + + /* Horizontal filter */ + { + const __m128i coeffs_x = + _mm_add_epi16(_mm_loadu_si128((__m128i *)filter_x), offset); + + // coeffs 0 1 0 1 2 3 2 3 + const __m128i tmp_0 = _mm_unpacklo_epi32(coeffs_x, coeffs_x); + // coeffs 4 5 4 5 6 7 6 7 + const __m128i tmp_1 = _mm_unpackhi_epi32(coeffs_x, coeffs_x); + + // coeffs 0 1 0 1 0 1 0 1 + const __m128i coeff_01 = _mm_unpacklo_epi64(tmp_0, tmp_0); + // coeffs 2 3 2 3 2 3 2 3 + const __m128i coeff_23 = _mm_unpackhi_epi64(tmp_0, tmp_0); + // coeffs 4 5 4 5 4 5 4 5 + const __m128i coeff_45 = _mm_unpacklo_epi64(tmp_1, tmp_1); + // coeffs 6 7 6 7 6 7 6 7 + const __m128i coeff_67 = _mm_unpackhi_epi64(tmp_1, tmp_1); + + const __m128i round_const = _mm_set1_epi32( + (1 << (conv_params->round_0 - 1)) + (1 << (bd + FILTER_BITS - 1))); + + for (i = 0; i < intermediate_height; ++i) { + for (j = 0; j < w; j += 8) { + const __m128i data = + _mm_loadu_si128((__m128i *)&src_ptr[i * src_stride + j]); + const __m128i data2 = + _mm_loadu_si128((__m128i *)&src_ptr[i * src_stride + j + 8]); + + // Filter even-index pixels + const __m128i res_0 = _mm_madd_epi16(data, coeff_01); + const __m128i res_2 = + _mm_madd_epi16(_mm_alignr_epi8(data2, data, 4), coeff_23); + const __m128i res_4 = + _mm_madd_epi16(_mm_alignr_epi8(data2, data, 8), coeff_45); + const __m128i res_6 = + _mm_madd_epi16(_mm_alignr_epi8(data2, data, 12), coeff_67); + + __m128i res_even = _mm_add_epi32(_mm_add_epi32(res_0, res_4), + _mm_add_epi32(res_2, res_6)); + res_even = _mm_srai_epi32(_mm_add_epi32(res_even, round_const), + conv_params->round_0); + + // Filter odd-index pixels + const __m128i res_1 = + _mm_madd_epi16(_mm_alignr_epi8(data2, data, 2), coeff_01); + const __m128i res_3 = + _mm_madd_epi16(_mm_alignr_epi8(data2, data, 6), coeff_23); + const __m128i res_5 = + _mm_madd_epi16(_mm_alignr_epi8(data2, data, 10), coeff_45); + const __m128i res_7 = + _mm_madd_epi16(_mm_alignr_epi8(data2, data, 14), coeff_67); + + __m128i res_odd = _mm_add_epi32(_mm_add_epi32(res_1, res_5), + _mm_add_epi32(res_3, res_7)); + res_odd = _mm_srai_epi32(_mm_add_epi32(res_odd, round_const), + conv_params->round_0); + + // Pack in the column order 0, 2, 4, 6, 1, 3, 5, 7 + const __m128i maxval = + _mm_set1_epi16((WIENER_CLAMP_LIMIT(conv_params->round_0, bd)) - 1); + __m128i res = _mm_packs_epi32(res_even, res_odd); + res = _mm_min_epi16(_mm_max_epi16(res, zero), maxval); + _mm_storeu_si128((__m128i *)&temp[i * MAX_SB_SIZE + j], res); + } + } + } + + /* Vertical filter */ + { + const __m128i coeffs_y = + _mm_add_epi16(_mm_loadu_si128((__m128i *)filter_y), offset); + + // coeffs 0 1 0 1 2 3 2 3 + const __m128i tmp_0 = _mm_unpacklo_epi32(coeffs_y, coeffs_y); + // coeffs 4 5 4 5 6 7 6 7 + const __m128i tmp_1 = _mm_unpackhi_epi32(coeffs_y, coeffs_y); + + // coeffs 0 1 0 1 0 1 0 1 + const __m128i coeff_01 = _mm_unpacklo_epi64(tmp_0, tmp_0); + // coeffs 2 3 2 3 2 3 2 3 + const __m128i coeff_23 = _mm_unpackhi_epi64(tmp_0, tmp_0); + // coeffs 4 5 4 5 4 5 4 5 + const __m128i coeff_45 = _mm_unpacklo_epi64(tmp_1, tmp_1); + // coeffs 6 7 6 7 6 7 6 7 + const __m128i coeff_67 = _mm_unpackhi_epi64(tmp_1, tmp_1); + + const __m128i round_const = + _mm_set1_epi32((1 << (conv_params->round_1 - 1)) - + (1 << (bd + conv_params->round_1 - 1))); + + for (i = 0; i < h; ++i) { + for (j = 0; j < w; j += 8) { + // Filter even-index pixels + const uint16_t *data = &temp[i * MAX_SB_SIZE + j]; + const __m128i src_0 = + _mm_unpacklo_epi16(*(__m128i *)(data + 0 * MAX_SB_SIZE), + *(__m128i *)(data + 1 * MAX_SB_SIZE)); + const __m128i src_2 = + _mm_unpacklo_epi16(*(__m128i *)(data + 2 * MAX_SB_SIZE), + *(__m128i *)(data + 3 * MAX_SB_SIZE)); + const __m128i src_4 = + _mm_unpacklo_epi16(*(__m128i *)(data + 4 * MAX_SB_SIZE), + *(__m128i *)(data + 5 * MAX_SB_SIZE)); + const __m128i src_6 = + _mm_unpacklo_epi16(*(__m128i *)(data + 6 * MAX_SB_SIZE), + *(__m128i *)(data + 7 * MAX_SB_SIZE)); + + const __m128i res_0 = _mm_madd_epi16(src_0, coeff_01); + const __m128i res_2 = _mm_madd_epi16(src_2, coeff_23); + const __m128i res_4 = _mm_madd_epi16(src_4, coeff_45); + const __m128i res_6 = _mm_madd_epi16(src_6, coeff_67); + + const __m128i res_even = _mm_add_epi32(_mm_add_epi32(res_0, res_2), + _mm_add_epi32(res_4, res_6)); + + // Filter odd-index pixels + const __m128i src_1 = + _mm_unpackhi_epi16(*(__m128i *)(data + 0 * MAX_SB_SIZE), + *(__m128i *)(data + 1 * MAX_SB_SIZE)); + const __m128i src_3 = + _mm_unpackhi_epi16(*(__m128i *)(data + 2 * MAX_SB_SIZE), + *(__m128i *)(data + 3 * MAX_SB_SIZE)); + const __m128i src_5 = + _mm_unpackhi_epi16(*(__m128i *)(data + 4 * MAX_SB_SIZE), + *(__m128i *)(data + 5 * MAX_SB_SIZE)); + const __m128i src_7 = + _mm_unpackhi_epi16(*(__m128i *)(data + 6 * MAX_SB_SIZE), + *(__m128i *)(data + 7 * MAX_SB_SIZE)); + + const __m128i res_1 = _mm_madd_epi16(src_1, coeff_01); + const __m128i res_3 = _mm_madd_epi16(src_3, coeff_23); + const __m128i res_5 = _mm_madd_epi16(src_5, coeff_45); + const __m128i res_7 = _mm_madd_epi16(src_7, coeff_67); + + const __m128i res_odd = _mm_add_epi32(_mm_add_epi32(res_1, res_3), + _mm_add_epi32(res_5, res_7)); + + // Rearrange pixels back into the order 0 ... 7 + const __m128i res_lo = _mm_unpacklo_epi32(res_even, res_odd); + const __m128i res_hi = _mm_unpackhi_epi32(res_even, res_odd); + + const __m128i res_lo_round = _mm_srai_epi32( + _mm_add_epi32(res_lo, round_const), conv_params->round_1); + const __m128i res_hi_round = _mm_srai_epi32( + _mm_add_epi32(res_hi, round_const), conv_params->round_1); + + const __m128i maxval = _mm_set1_epi16((1 << bd) - 1); + __m128i res_16bit = _mm_packs_epi32(res_lo_round, res_hi_round); + res_16bit = _mm_min_epi16(_mm_max_epi16(res_16bit, zero), maxval); + + __m128i *const p = (__m128i *)&dst[i * dst_stride + j]; + _mm_storeu_si128(p, res_16bit); + } + } + } +} diff --git a/third_party/aom/av1/common/x86/intra_edge_sse4.c b/third_party/aom/av1/common/x86/intra_edge_sse4.c new file mode 100644 index 0000000000..0c857b5835 --- /dev/null +++ b/third_party/aom/av1/common/x86/intra_edge_sse4.c @@ -0,0 +1,318 @@ +/* + * Copyright (c) 2017, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <assert.h> +#include <smmintrin.h> + +#include "config/aom_config.h" +#include "config/av1_rtcd.h" + +void av1_filter_intra_edge_sse4_1(uint8_t *p, int sz, int strength) { + if (!strength) return; + + DECLARE_ALIGNED(16, static const int8_t, kern[3][16]) = { + { 4, 8, 4, 0, 4, 8, 4, 0, 4, 8, 4, 0, 4, 8, 4, 0 }, // strength 1: 4,8,4 + { 5, 6, 5, 0, 5, 6, 5, 0, 5, 6, 5, 0, 5, 6, 5, 0 }, // strength 2: 5,6,5 + { 2, 4, 4, 4, 2, 0, 0, 0, 2, 4, 4, 4, 2, 0, 0, 0 } // strength 3: 2,4,4,4,2 + }; + + DECLARE_ALIGNED(16, static const int8_t, v_const[5][16]) = { + { 0, 1, 2, 3, 1, 2, 3, 4, 2, 3, 4, 5, 3, 4, 5, 6 }, + { 4, 5, 6, 7, 5, 6, 7, 8, 6, 7, 8, 9, 7, 8, 9, 10 }, + { 0, 1, 2, 3, 4, 5, 6, 7, 1, 2, 3, 4, 5, 6, 7, 8 }, + { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 }, + }; + + // Extend the first and last samples to simplify the loop for the 5-tap case + p[-1] = p[0]; + __m128i last = _mm_set1_epi8(p[sz - 1]); + _mm_storeu_si128((__m128i *)&p[sz], last); + + // Adjust input pointer for filter support area + uint8_t *in = (strength == 3) ? p - 1 : p; + + // Avoid modifying first sample + uint8_t *out = p + 1; + int len = sz - 1; + + const int use_3tap_filter = (strength < 3); + + if (use_3tap_filter) { + __m128i coef0 = _mm_lddqu_si128((__m128i const *)kern[strength - 1]); + __m128i shuf0 = _mm_lddqu_si128((__m128i const *)v_const[0]); + __m128i shuf1 = _mm_lddqu_si128((__m128i const *)v_const[1]); + __m128i iden = _mm_lddqu_si128((__m128i *)v_const[3]); + __m128i in0 = _mm_lddqu_si128((__m128i *)in); + while (len > 0) { + int n_out = (len < 8) ? len : 8; + __m128i d0 = _mm_shuffle_epi8(in0, shuf0); + __m128i d1 = _mm_shuffle_epi8(in0, shuf1); + d0 = _mm_maddubs_epi16(d0, coef0); + d1 = _mm_maddubs_epi16(d1, coef0); + d0 = _mm_hadd_epi16(d0, d1); + __m128i eight = _mm_set1_epi16(8); + d0 = _mm_add_epi16(d0, eight); + d0 = _mm_srai_epi16(d0, 4); + d0 = _mm_packus_epi16(d0, d0); + __m128i out0 = _mm_lddqu_si128((__m128i *)out); + __m128i n0 = _mm_set1_epi8(n_out); + __m128i mask = _mm_cmpgt_epi8(n0, iden); + out0 = _mm_blendv_epi8(out0, d0, mask); + _mm_storel_epi64((__m128i *)out, out0); + __m128i in1 = _mm_lddqu_si128((__m128i *)(in + 16)); + in0 = _mm_alignr_epi8(in1, in0, 8); + in += 8; + out += 8; + len -= n_out; + } + } else { // 5-tap filter + __m128i coef0 = _mm_lddqu_si128((__m128i const *)kern[strength - 1]); + __m128i two = _mm_set1_epi8(2); + __m128i shuf_a = _mm_lddqu_si128((__m128i const *)v_const[2]); + __m128i shuf_b = _mm_add_epi8(shuf_a, two); + __m128i shuf_c = _mm_add_epi8(shuf_b, two); + __m128i shuf_d = _mm_add_epi8(shuf_c, two); + __m128i iden = _mm_lddqu_si128((__m128i *)v_const[3]); + __m128i in0 = _mm_lddqu_si128((__m128i *)in); + while (len > 0) { + int n_out = (len < 8) ? len : 8; + __m128i d0 = _mm_shuffle_epi8(in0, shuf_a); + __m128i d1 = _mm_shuffle_epi8(in0, shuf_b); + __m128i d2 = _mm_shuffle_epi8(in0, shuf_c); + __m128i d3 = _mm_shuffle_epi8(in0, shuf_d); + d0 = _mm_maddubs_epi16(d0, coef0); + d1 = _mm_maddubs_epi16(d1, coef0); + d2 = _mm_maddubs_epi16(d2, coef0); + d3 = _mm_maddubs_epi16(d3, coef0); + d0 = _mm_hadd_epi16(d0, d1); + d2 = _mm_hadd_epi16(d2, d3); + d0 = _mm_hadd_epi16(d0, d2); + __m128i eight = _mm_set1_epi16(8); + d0 = _mm_add_epi16(d0, eight); + d0 = _mm_srai_epi16(d0, 4); + d0 = _mm_packus_epi16(d0, d0); + __m128i out0 = _mm_lddqu_si128((__m128i *)out); + __m128i n0 = _mm_set1_epi8(n_out); + __m128i mask = _mm_cmpgt_epi8(n0, iden); + out0 = _mm_blendv_epi8(out0, d0, mask); + _mm_storel_epi64((__m128i *)out, out0); + __m128i in1 = _mm_lddqu_si128((__m128i *)(in + 16)); + in0 = _mm_alignr_epi8(in1, in0, 8); + in += 8; + out += 8; + len -= n_out; + } + } +} + +void av1_filter_intra_edge_high_sse4_1(uint16_t *p, int sz, int strength) { + if (!strength) return; + + DECLARE_ALIGNED(16, static const int16_t, kern[3][8]) = { + { 4, 8, 4, 8, 4, 8, 4, 8 }, // strength 1: 4,8,4 + { 5, 6, 5, 6, 5, 6, 5, 6 }, // strength 2: 5,6,5 + { 2, 4, 2, 4, 2, 4, 2, 4 } // strength 3: 2,4,4,4,2 + }; + + DECLARE_ALIGNED(16, static const int16_t, + v_const[1][8]) = { { 0, 1, 2, 3, 4, 5, 6, 7 } }; + + // Extend the first and last samples to simplify the loop for the 5-tap case + p[-1] = p[0]; + __m128i last = _mm_set1_epi16(p[sz - 1]); + _mm_storeu_si128((__m128i *)&p[sz], last); + + // Adjust input pointer for filter support area + uint16_t *in = (strength == 3) ? p - 1 : p; + + // Avoid modifying first sample + uint16_t *out = p + 1; + int len = sz - 1; + + const int use_3tap_filter = (strength < 3); + + if (use_3tap_filter) { + __m128i coef0 = _mm_lddqu_si128((__m128i const *)kern[strength - 1]); + __m128i iden = _mm_lddqu_si128((__m128i *)v_const[0]); + __m128i in0 = _mm_lddqu_si128((__m128i *)&in[0]); + __m128i in8 = _mm_lddqu_si128((__m128i *)&in[8]); + while (len > 0) { + int n_out = (len < 8) ? len : 8; + __m128i in1 = _mm_alignr_epi8(in8, in0, 2); + __m128i in2 = _mm_alignr_epi8(in8, in0, 4); + __m128i in02 = _mm_add_epi16(in0, in2); + __m128i d0 = _mm_unpacklo_epi16(in02, in1); + __m128i d1 = _mm_unpackhi_epi16(in02, in1); + d0 = _mm_mullo_epi16(d0, coef0); + d1 = _mm_mullo_epi16(d1, coef0); + d0 = _mm_hadd_epi16(d0, d1); + __m128i eight = _mm_set1_epi16(8); + d0 = _mm_add_epi16(d0, eight); + d0 = _mm_srli_epi16(d0, 4); + __m128i out0 = _mm_lddqu_si128((__m128i *)out); + __m128i n0 = _mm_set1_epi16(n_out); + __m128i mask = _mm_cmpgt_epi16(n0, iden); + out0 = _mm_blendv_epi8(out0, d0, mask); + _mm_storeu_si128((__m128i *)out, out0); + in += 8; + in0 = in8; + in8 = _mm_lddqu_si128((__m128i *)&in[8]); + out += 8; + len -= n_out; + } + } else { // 5-tap filter + __m128i coef0 = _mm_lddqu_si128((__m128i const *)kern[strength - 1]); + __m128i iden = _mm_lddqu_si128((__m128i *)v_const[0]); + __m128i in0 = _mm_lddqu_si128((__m128i *)&in[0]); + __m128i in8 = _mm_lddqu_si128((__m128i *)&in[8]); + while (len > 0) { + int n_out = (len < 8) ? len : 8; + __m128i in1 = _mm_alignr_epi8(in8, in0, 2); + __m128i in2 = _mm_alignr_epi8(in8, in0, 4); + __m128i in3 = _mm_alignr_epi8(in8, in0, 6); + __m128i in4 = _mm_alignr_epi8(in8, in0, 8); + __m128i in04 = _mm_add_epi16(in0, in4); + __m128i in123 = _mm_add_epi16(in1, in2); + in123 = _mm_add_epi16(in123, in3); + __m128i d0 = _mm_unpacklo_epi16(in04, in123); + __m128i d1 = _mm_unpackhi_epi16(in04, in123); + d0 = _mm_mullo_epi16(d0, coef0); + d1 = _mm_mullo_epi16(d1, coef0); + d0 = _mm_hadd_epi16(d0, d1); + __m128i eight = _mm_set1_epi16(8); + d0 = _mm_add_epi16(d0, eight); + d0 = _mm_srli_epi16(d0, 4); + __m128i out0 = _mm_lddqu_si128((__m128i *)out); + __m128i n0 = _mm_set1_epi16(n_out); + __m128i mask = _mm_cmpgt_epi16(n0, iden); + out0 = _mm_blendv_epi8(out0, d0, mask); + _mm_storeu_si128((__m128i *)out, out0); + in += 8; + in0 = in8; + in8 = _mm_lddqu_si128((__m128i *)&in[8]); + out += 8; + len -= n_out; + } + } +} + +void av1_upsample_intra_edge_sse4_1(uint8_t *p, int sz) { + // interpolate half-sample positions + assert(sz <= 24); + + DECLARE_ALIGNED(16, static const int8_t, kernel[1][16]) = { + { -1, 9, 9, -1, -1, 9, 9, -1, -1, 9, 9, -1, -1, 9, 9, -1 } + }; + + DECLARE_ALIGNED(16, static const int8_t, v_const[2][16]) = { + { 0, 1, 2, 3, 1, 2, 3, 4, 2, 3, 4, 5, 3, 4, 5, 6 }, + { 4, 5, 6, 7, 5, 6, 7, 8, 6, 7, 8, 9, 7, 8, 9, 10 } + }; + + // Extend first/last samples (upper-left p[-1], last p[sz-1]) + // to support 4-tap filter + p[-2] = p[-1]; + p[sz] = p[sz - 1]; + + uint8_t *in = &p[-2]; + uint8_t *out = &p[-2]; + + int n = sz + 1; // Input length including upper-left sample + + __m128i in0 = _mm_lddqu_si128((__m128i *)&in[0]); + __m128i in16 = _mm_lddqu_si128((__m128i *)&in[16]); + + __m128i coef0 = _mm_lddqu_si128((__m128i *)kernel[0]); + __m128i shuf0 = _mm_lddqu_si128((__m128i *)v_const[0]); + __m128i shuf1 = _mm_lddqu_si128((__m128i *)v_const[1]); + + while (n > 0) { + __m128i in8 = _mm_alignr_epi8(in16, in0, 8); + __m128i d0 = _mm_shuffle_epi8(in0, shuf0); + __m128i d1 = _mm_shuffle_epi8(in0, shuf1); + __m128i d2 = _mm_shuffle_epi8(in8, shuf0); + __m128i d3 = _mm_shuffle_epi8(in8, shuf1); + d0 = _mm_maddubs_epi16(d0, coef0); + d1 = _mm_maddubs_epi16(d1, coef0); + d2 = _mm_maddubs_epi16(d2, coef0); + d3 = _mm_maddubs_epi16(d3, coef0); + d0 = _mm_hadd_epi16(d0, d1); + d2 = _mm_hadd_epi16(d2, d3); + __m128i eight = _mm_set1_epi16(8); + d0 = _mm_add_epi16(d0, eight); + d2 = _mm_add_epi16(d2, eight); + d0 = _mm_srai_epi16(d0, 4); + d2 = _mm_srai_epi16(d2, 4); + d0 = _mm_packus_epi16(d0, d2); + __m128i in1 = _mm_alignr_epi8(in16, in0, 1); + __m128i out0 = _mm_unpacklo_epi8(in1, d0); + __m128i out1 = _mm_unpackhi_epi8(in1, d0); + _mm_storeu_si128((__m128i *)&out[0], out0); + _mm_storeu_si128((__m128i *)&out[16], out1); + in0 = in16; + in16 = _mm_setzero_si128(); + out += 32; + n -= 16; + } +} + +void av1_upsample_intra_edge_high_sse4_1(uint16_t *p, int sz, int bd) { + // interpolate half-sample positions + assert(sz <= 24); + + DECLARE_ALIGNED(16, static const int16_t, + kernel[1][8]) = { { -1, 9, -1, 9, -1, 9, -1, 9 } }; + + // Extend first/last samples (upper-left p[-1], last p[sz-1]) + // to support 4-tap filter + p[-2] = p[-1]; + p[sz] = p[sz - 1]; + + uint16_t *in = &p[-2]; + uint16_t *out = in; + int n = sz + 1; + + __m128i in0 = _mm_lddqu_si128((__m128i *)&in[0]); + __m128i in8 = _mm_lddqu_si128((__m128i *)&in[8]); + __m128i in16 = _mm_lddqu_si128((__m128i *)&in[16]); + __m128i in24 = _mm_lddqu_si128((__m128i *)&in[24]); + + while (n > 0) { + __m128i in1 = _mm_alignr_epi8(in8, in0, 2); + __m128i in2 = _mm_alignr_epi8(in8, in0, 4); + __m128i in3 = _mm_alignr_epi8(in8, in0, 6); + __m128i sum0 = _mm_add_epi16(in0, in3); + __m128i sum1 = _mm_add_epi16(in1, in2); + __m128i d0 = _mm_unpacklo_epi16(sum0, sum1); + __m128i d1 = _mm_unpackhi_epi16(sum0, sum1); + __m128i coef0 = _mm_lddqu_si128((__m128i *)kernel[0]); + d0 = _mm_madd_epi16(d0, coef0); + d1 = _mm_madd_epi16(d1, coef0); + __m128i eight = _mm_set1_epi32(8); + d0 = _mm_add_epi32(d0, eight); + d1 = _mm_add_epi32(d1, eight); + d0 = _mm_srai_epi32(d0, 4); + d1 = _mm_srai_epi32(d1, 4); + d0 = _mm_packus_epi32(d0, d1); + __m128i max0 = _mm_set1_epi16((1 << bd) - 1); + d0 = _mm_min_epi16(d0, max0); + __m128i out0 = _mm_unpacklo_epi16(in1, d0); + __m128i out1 = _mm_unpackhi_epi16(in1, d0); + _mm_storeu_si128((__m128i *)&out[0], out0); + _mm_storeu_si128((__m128i *)&out[8], out1); + in0 = in8; + in8 = in16; + in16 = in24; + in24 = _mm_setzero_si128(); + out += 16; + n -= 8; + } +} diff --git a/third_party/aom/av1/common/x86/jnt_convolve_avx2.c b/third_party/aom/av1/common/x86/jnt_convolve_avx2.c new file mode 100644 index 0000000000..9f2e2b4578 --- /dev/null +++ b/third_party/aom/av1/common/x86/jnt_convolve_avx2.c @@ -0,0 +1,633 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <immintrin.h> + +#include "config/aom_dsp_rtcd.h" + +#include "aom_dsp/x86/convolve_avx2.h" +#include "aom_dsp/x86/convolve_common_intrin.h" +#include "aom_dsp/x86/convolve_sse4_1.h" +#include "aom_dsp/aom_dsp_common.h" +#include "aom_dsp/aom_filter.h" +#include "av1/common/convolve.h" + +static INLINE __m256i unpack_weights_avx2(ConvolveParams *conv_params) { + const int w0 = conv_params->fwd_offset; + const int w1 = conv_params->bck_offset; + const __m256i wt0 = _mm256_set1_epi16(w0); + const __m256i wt1 = _mm256_set1_epi16(w1); + const __m256i wt = _mm256_unpacklo_epi16(wt0, wt1); + return wt; +} + +static INLINE __m256i load_line2_avx2(const void *a, const void *b) { + return _mm256_permute2x128_si256( + _mm256_castsi128_si256(_mm_loadu_si128((__m128i *)a)), + _mm256_castsi128_si256(_mm_loadu_si128((__m128i *)b)), 0x20); +} + +void av1_jnt_convolve_x_avx2(const uint8_t *src, int src_stride, uint8_t *dst0, + int dst_stride0, int w, int h, + const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, + const int subpel_x_q4, const int subpel_y_q4, + ConvolveParams *conv_params) { + CONV_BUF_TYPE *dst = conv_params->dst; + int dst_stride = conv_params->dst_stride; + const int bd = 8; + int i, j; + const int fo_horiz = filter_params_x->taps / 2 - 1; + const uint8_t *const src_ptr = src - fo_horiz; + const int bits = FILTER_BITS - conv_params->round_1; + const __m256i wt = unpack_weights_avx2(conv_params); + const int do_average = conv_params->do_average; + const int use_jnt_comp_avg = conv_params->use_jnt_comp_avg; + const int offset_0 = + bd + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const int offset = (1 << offset_0) + (1 << (offset_0 - 1)); + const __m256i offset_const = _mm256_set1_epi16(offset); + const int rounding_shift = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const __m256i rounding_const = _mm256_set1_epi16((1 << rounding_shift) >> 1); + __m256i filt[4], coeffs[4]; + + assert(bits >= 0); + assert(conv_params->round_0 > 0); + + filt[0] = _mm256_load_si256((__m256i const *)filt_global_avx2); + filt[1] = _mm256_load_si256((__m256i const *)(filt_global_avx2 + 32)); + filt[2] = _mm256_load_si256((__m256i const *)(filt_global_avx2 + 32 * 2)); + filt[3] = _mm256_load_si256((__m256i const *)(filt_global_avx2 + 32 * 3)); + + prepare_coeffs_lowbd(filter_params_x, subpel_x_q4, coeffs); + + const __m256i round_const = + _mm256_set1_epi16((1 << (conv_params->round_0 - 1)) >> 1); + const __m128i round_shift = _mm_cvtsi32_si128(conv_params->round_0 - 1); + + (void)filter_params_y; + (void)subpel_y_q4; + + for (i = 0; i < h; i += 2) { + const uint8_t *src_data = src_ptr + i * src_stride; + CONV_BUF_TYPE *dst_data = dst + i * dst_stride; + for (j = 0; j < w; j += 8) { + const __m256i data = + load_line2_avx2(&src_data[j], &src_data[j + src_stride]); + + __m256i res = convolve_lowbd_x(data, coeffs, filt); + + res = _mm256_sra_epi16(_mm256_add_epi16(res, round_const), round_shift); + + res = _mm256_slli_epi16(res, bits); + + const __m256i res_unsigned = _mm256_add_epi16(res, offset_const); + + // Accumulate values into the destination buffer + if (do_average) { + const __m256i data_ref_0 = + load_line2_avx2(&dst_data[j], &dst_data[j + dst_stride]); + const __m256i comp_avg_res = + comp_avg(&data_ref_0, &res_unsigned, &wt, use_jnt_comp_avg); + + const __m256i round_result = convolve_rounding( + &comp_avg_res, &offset_const, &rounding_const, rounding_shift); + + const __m256i res_8 = _mm256_packus_epi16(round_result, round_result); + const __m128i res_0 = _mm256_castsi256_si128(res_8); + const __m128i res_1 = _mm256_extracti128_si256(res_8, 1); + + if (w > 4) { + _mm_storel_epi64((__m128i *)(&dst0[i * dst_stride0 + j]), res_0); + _mm_storel_epi64( + (__m128i *)((&dst0[i * dst_stride0 + j + dst_stride0])), res_1); + } else { + *(uint32_t *)(&dst0[i * dst_stride0 + j]) = _mm_cvtsi128_si32(res_0); + *(uint32_t *)(&dst0[i * dst_stride0 + j + dst_stride0]) = + _mm_cvtsi128_si32(res_1); + } + } else { + const __m128i res_0 = _mm256_castsi256_si128(res_unsigned); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j]), res_0); + + const __m128i res_1 = _mm256_extracti128_si256(res_unsigned, 1); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j + dst_stride]), + res_1); + } + } + } +} + +void av1_jnt_convolve_y_avx2(const uint8_t *src, int src_stride, uint8_t *dst0, + int dst_stride0, int w, int h, + const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, + const int subpel_x_q4, const int subpel_y_q4, + ConvolveParams *conv_params) { + CONV_BUF_TYPE *dst = conv_params->dst; + int dst_stride = conv_params->dst_stride; + const int bd = 8; + int i, j; + const int fo_vert = filter_params_y->taps / 2 - 1; + const uint8_t *const src_ptr = src - fo_vert * src_stride; + // +1 to compensate for dividing the filter coeffs by 2 + const int left_shift = FILTER_BITS - conv_params->round_0 + 1; + const __m256i round_const = + _mm256_set1_epi32((1 << conv_params->round_1) >> 1); + const __m128i round_shift = _mm_cvtsi32_si128(conv_params->round_1); + const __m256i wt = unpack_weights_avx2(conv_params); + const int do_average = conv_params->do_average; + const int use_jnt_comp_avg = conv_params->use_jnt_comp_avg; + const int offset_0 = + bd + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const int offset = (1 << offset_0) + (1 << (offset_0 - 1)); + const __m256i offset_const = _mm256_set1_epi16(offset); + const int offset_1 = (1 << (bd + FILTER_BITS - 2)); + const __m256i offset_const_1 = _mm256_set1_epi16(offset_1); + const __m256i offset_const_2 = _mm256_set1_epi16((1 << offset_0)); + const int rounding_shift = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const __m256i rounding_const = _mm256_set1_epi16((1 << rounding_shift) >> 1); + const __m256i zero = _mm256_setzero_si256(); + __m256i coeffs[4], s[8]; + + assert((FILTER_BITS - conv_params->round_0) >= 0); + + prepare_coeffs_lowbd(filter_params_y, subpel_y_q4, coeffs); + + (void)conv_params; + (void)filter_params_x; + (void)subpel_x_q4; + + for (j = 0; j < w; j += 16) { + const uint8_t *data = &src_ptr[j]; + __m256i src6; + // Load lines a and b. Line a to lower 128, line b to upper 128 + { + __m256i src_ab[7]; + __m256i src_a[7]; + src_a[0] = _mm256_castsi128_si256(_mm_loadu_si128((__m128i *)data)); + for (int kk = 0; kk < 6; ++kk) { + data += src_stride; + src_a[kk + 1] = + _mm256_castsi128_si256(_mm_loadu_si128((__m128i *)data)); + src_ab[kk] = _mm256_permute2x128_si256(src_a[kk], src_a[kk + 1], 0x20); + } + src6 = src_a[6]; + s[0] = _mm256_unpacklo_epi8(src_ab[0], src_ab[1]); + s[1] = _mm256_unpacklo_epi8(src_ab[2], src_ab[3]); + s[2] = _mm256_unpacklo_epi8(src_ab[4], src_ab[5]); + s[4] = _mm256_unpackhi_epi8(src_ab[0], src_ab[1]); + s[5] = _mm256_unpackhi_epi8(src_ab[2], src_ab[3]); + s[6] = _mm256_unpackhi_epi8(src_ab[4], src_ab[5]); + } + + for (i = 0; i < h; i += 2) { + data = &src_ptr[(i + 7) * src_stride + j]; + const __m256i src7 = + _mm256_castsi128_si256(_mm_loadu_si128((__m128i *)data)); + const __m256i src_67a = _mm256_permute2x128_si256(src6, src7, 0x20); + + src6 = _mm256_castsi128_si256( + _mm_loadu_si128((__m128i *)(data + src_stride))); + const __m256i src_78a = _mm256_permute2x128_si256(src7, src6, 0x20); + + s[3] = _mm256_unpacklo_epi8(src_67a, src_78a); + s[7] = _mm256_unpackhi_epi8(src_67a, src_78a); + + __m256i res_lo = convolve_lowbd(s, coeffs); + + res_lo = _mm256_add_epi16(res_lo, offset_const_1); + + const __m256i res_lo_0_32b = _mm256_unpacklo_epi16(res_lo, zero); + const __m256i res_lo_0_shift = + _mm256_slli_epi32(res_lo_0_32b, left_shift); + const __m256i res_lo_0_round = _mm256_sra_epi32( + _mm256_add_epi32(res_lo_0_shift, round_const), round_shift); + + const __m256i res_lo_1_32b = _mm256_unpackhi_epi16(res_lo, zero); + const __m256i res_lo_1_shift = + _mm256_slli_epi32(res_lo_1_32b, left_shift); + const __m256i res_lo_1_round = _mm256_sra_epi32( + _mm256_add_epi32(res_lo_1_shift, round_const), round_shift); + + const __m256i res_lo_round = + _mm256_packs_epi32(res_lo_0_round, res_lo_1_round); + + const __m256i res_lo_unsigned = + _mm256_add_epi16(res_lo_round, offset_const_2); + + if (w - j < 16) { + if (do_average) { + const __m256i data_ref_0 = load_line2_avx2( + &dst[i * dst_stride + j], &dst[i * dst_stride + j + dst_stride]); + const __m256i comp_avg_res = + comp_avg(&data_ref_0, &res_lo_unsigned, &wt, use_jnt_comp_avg); + + const __m256i round_result = convolve_rounding( + &comp_avg_res, &offset_const, &rounding_const, rounding_shift); + + const __m256i res_8 = _mm256_packus_epi16(round_result, round_result); + const __m128i res_0 = _mm256_castsi256_si128(res_8); + const __m128i res_1 = _mm256_extracti128_si256(res_8, 1); + + if (w - j > 4) { + _mm_storel_epi64((__m128i *)(&dst0[i * dst_stride0 + j]), res_0); + _mm_storel_epi64( + (__m128i *)((&dst0[i * dst_stride0 + j + dst_stride0])), res_1); + } else { + *(uint32_t *)(&dst0[i * dst_stride0 + j]) = + _mm_cvtsi128_si32(res_0); + *(uint32_t *)(&dst0[i * dst_stride0 + j + dst_stride0]) = + _mm_cvtsi128_si32(res_1); + } + } else { + const __m128i res_0 = _mm256_castsi256_si128(res_lo_unsigned); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j]), res_0); + + const __m128i res_1 = _mm256_extracti128_si256(res_lo_unsigned, 1); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j + dst_stride]), + res_1); + } + } else { + __m256i res_hi = convolve_lowbd(s + 4, coeffs); + + res_hi = _mm256_add_epi16(res_hi, offset_const_1); + + const __m256i res_hi_0_32b = _mm256_unpacklo_epi16(res_hi, zero); + const __m256i res_hi_0_shift = + _mm256_slli_epi32(res_hi_0_32b, left_shift); + const __m256i res_hi_0_round = _mm256_sra_epi32( + _mm256_add_epi32(res_hi_0_shift, round_const), round_shift); + + const __m256i res_hi_1_32b = _mm256_unpackhi_epi16(res_hi, zero); + const __m256i res_hi_1_shift = + _mm256_slli_epi32(res_hi_1_32b, left_shift); + const __m256i res_hi_1_round = _mm256_sra_epi32( + _mm256_add_epi32(res_hi_1_shift, round_const), round_shift); + + const __m256i res_hi_round = + _mm256_packs_epi32(res_hi_0_round, res_hi_1_round); + + const __m256i res_hi_unsigned = + _mm256_add_epi16(res_hi_round, offset_const_2); + + if (do_average) { + const __m256i data_ref_0_lo = load_line2_avx2( + &dst[i * dst_stride + j], &dst[i * dst_stride + j + dst_stride]); + + const __m256i data_ref_0_hi = + load_line2_avx2(&dst[i * dst_stride + j + 8], + &dst[i * dst_stride + j + 8 + dst_stride]); + + const __m256i comp_avg_res_lo = + comp_avg(&data_ref_0_lo, &res_lo_unsigned, &wt, use_jnt_comp_avg); + + const __m256i comp_avg_res_hi = + comp_avg(&data_ref_0_hi, &res_hi_unsigned, &wt, use_jnt_comp_avg); + + const __m256i round_result_lo = convolve_rounding( + &comp_avg_res_lo, &offset_const, &rounding_const, rounding_shift); + + const __m256i round_result_hi = convolve_rounding( + &comp_avg_res_hi, &offset_const, &rounding_const, rounding_shift); + + const __m256i res_8 = + _mm256_packus_epi16(round_result_lo, round_result_hi); + const __m128i res_0 = _mm256_castsi256_si128(res_8); + const __m128i res_1 = _mm256_extracti128_si256(res_8, 1); + + _mm_store_si128((__m128i *)(&dst0[i * dst_stride0 + j]), res_0); + _mm_store_si128( + (__m128i *)((&dst0[i * dst_stride0 + j + dst_stride0])), res_1); + + } else { + const __m128i res_lo_0 = _mm256_castsi256_si128(res_lo_unsigned); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j]), res_lo_0); + + const __m128i res_lo_1 = _mm256_extracti128_si256(res_lo_unsigned, 1); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j + dst_stride]), + res_lo_1); + + const __m128i res_hi_0 = _mm256_castsi256_si128(res_hi_unsigned); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j + 8]), res_hi_0); + + const __m128i res_hi_1 = _mm256_extracti128_si256(res_hi_unsigned, 1); + _mm_store_si128( + (__m128i *)(&dst[i * dst_stride + j + 8 + dst_stride]), res_hi_1); + } + } + s[0] = s[1]; + s[1] = s[2]; + s[2] = s[3]; + + s[4] = s[5]; + s[5] = s[6]; + s[6] = s[7]; + } + } +} + +void av1_jnt_convolve_2d_avx2(const uint8_t *src, int src_stride, uint8_t *dst0, + int dst_stride0, int w, int h, + const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, + const int subpel_x_q4, const int subpel_y_q4, + ConvolveParams *conv_params) { + CONV_BUF_TYPE *dst = conv_params->dst; + int dst_stride = conv_params->dst_stride; + const int bd = 8; + + DECLARE_ALIGNED(32, int16_t, im_block[(MAX_SB_SIZE + MAX_FILTER_TAP) * 8]); + int im_h = h + filter_params_y->taps - 1; + int im_stride = 8; + int i, j; + const int fo_vert = filter_params_y->taps / 2 - 1; + const int fo_horiz = filter_params_x->taps / 2 - 1; + const uint8_t *const src_ptr = src - fo_vert * src_stride - fo_horiz; + const __m256i wt = unpack_weights_avx2(conv_params); + const int do_average = conv_params->do_average; + const int use_jnt_comp_avg = conv_params->use_jnt_comp_avg; + const int offset_0 = + bd + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const int offset = (1 << offset_0) + (1 << (offset_0 - 1)); + const __m256i offset_const = _mm256_set1_epi16(offset); + const int rounding_shift = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const __m256i rounding_const = _mm256_set1_epi16((1 << rounding_shift) >> 1); + __m256i filt[4], s[8], coeffs_x[4], coeffs_y[4]; + + assert(conv_params->round_0 > 0); + + filt[0] = _mm256_load_si256((__m256i const *)filt_global_avx2); + filt[1] = _mm256_load_si256((__m256i const *)(filt_global_avx2 + 32)); + filt[2] = _mm256_load_si256((__m256i const *)(filt_global_avx2 + 32 * 2)); + filt[3] = _mm256_load_si256((__m256i const *)(filt_global_avx2 + 32 * 3)); + + prepare_coeffs_lowbd(filter_params_x, subpel_x_q4, coeffs_x); + prepare_coeffs(filter_params_y, subpel_y_q4, coeffs_y); + + const __m256i round_const_h = _mm256_set1_epi16( + ((1 << (conv_params->round_0 - 1)) >> 1) + (1 << (bd + FILTER_BITS - 2))); + const __m128i round_shift_h = _mm_cvtsi32_si128(conv_params->round_0 - 1); + + const __m256i round_const_v = _mm256_set1_epi32( + ((1 << conv_params->round_1) >> 1) - + (1 << (bd + 2 * FILTER_BITS - conv_params->round_0 - 1))); + const __m128i round_shift_v = _mm_cvtsi32_si128(conv_params->round_1); + + for (j = 0; j < w; j += 8) { + /* Horizontal filter */ + { + const uint8_t *src_h = src_ptr + j; + for (i = 0; i < im_h; i += 2) { + __m256i data = + _mm256_castsi128_si256(_mm_loadu_si128((__m128i *)src_h)); + if (i + 1 < im_h) + data = _mm256_inserti128_si256( + data, _mm_loadu_si128((__m128i *)(src_h + src_stride)), 1); + src_h += (src_stride << 1); + __m256i res = convolve_lowbd_x(data, coeffs_x, filt); + + res = _mm256_sra_epi16(_mm256_add_epi16(res, round_const_h), + round_shift_h); + + _mm256_store_si256((__m256i *)&im_block[i * im_stride], res); + } + } + + /* Vertical filter */ + { + __m256i s0 = _mm256_loadu_si256((__m256i *)(im_block + 0 * im_stride)); + __m256i s1 = _mm256_loadu_si256((__m256i *)(im_block + 1 * im_stride)); + __m256i s2 = _mm256_loadu_si256((__m256i *)(im_block + 2 * im_stride)); + __m256i s3 = _mm256_loadu_si256((__m256i *)(im_block + 3 * im_stride)); + __m256i s4 = _mm256_loadu_si256((__m256i *)(im_block + 4 * im_stride)); + __m256i s5 = _mm256_loadu_si256((__m256i *)(im_block + 5 * im_stride)); + + s[0] = _mm256_unpacklo_epi16(s0, s1); + s[1] = _mm256_unpacklo_epi16(s2, s3); + s[2] = _mm256_unpacklo_epi16(s4, s5); + + s[4] = _mm256_unpackhi_epi16(s0, s1); + s[5] = _mm256_unpackhi_epi16(s2, s3); + s[6] = _mm256_unpackhi_epi16(s4, s5); + + for (i = 0; i < h; i += 2) { + const int16_t *data = &im_block[i * im_stride]; + + const __m256i s6 = + _mm256_loadu_si256((__m256i *)(data + 6 * im_stride)); + const __m256i s7 = + _mm256_loadu_si256((__m256i *)(data + 7 * im_stride)); + + s[3] = _mm256_unpacklo_epi16(s6, s7); + s[7] = _mm256_unpackhi_epi16(s6, s7); + + const __m256i res_a = convolve(s, coeffs_y); + const __m256i res_a_round = _mm256_sra_epi32( + _mm256_add_epi32(res_a, round_const_v), round_shift_v); + + if (w - j > 4) { + const __m256i res_b = convolve(s + 4, coeffs_y); + const __m256i res_b_round = _mm256_sra_epi32( + _mm256_add_epi32(res_b, round_const_v), round_shift_v); + const __m256i res_16b = _mm256_packs_epi32(res_a_round, res_b_round); + const __m256i res_unsigned = _mm256_add_epi16(res_16b, offset_const); + + if (do_average) { + const __m256i data_ref_0 = + load_line2_avx2(&dst[i * dst_stride + j], + &dst[i * dst_stride + j + dst_stride]); + const __m256i comp_avg_res = + comp_avg(&data_ref_0, &res_unsigned, &wt, use_jnt_comp_avg); + + const __m256i round_result = convolve_rounding( + &comp_avg_res, &offset_const, &rounding_const, rounding_shift); + + const __m256i res_8 = + _mm256_packus_epi16(round_result, round_result); + const __m128i res_0 = _mm256_castsi256_si128(res_8); + const __m128i res_1 = _mm256_extracti128_si256(res_8, 1); + + _mm_storel_epi64((__m128i *)(&dst0[i * dst_stride0 + j]), res_0); + _mm_storel_epi64( + (__m128i *)((&dst0[i * dst_stride0 + j + dst_stride0])), res_1); + } else { + const __m128i res_0 = _mm256_castsi256_si128(res_unsigned); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j]), res_0); + + const __m128i res_1 = _mm256_extracti128_si256(res_unsigned, 1); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j + dst_stride]), + res_1); + } + } else { + const __m256i res_16b = _mm256_packs_epi32(res_a_round, res_a_round); + const __m256i res_unsigned = _mm256_add_epi16(res_16b, offset_const); + + if (do_average) { + const __m256i data_ref_0 = + load_line2_avx2(&dst[i * dst_stride + j], + &dst[i * dst_stride + j + dst_stride]); + + const __m256i comp_avg_res = + comp_avg(&data_ref_0, &res_unsigned, &wt, use_jnt_comp_avg); + + const __m256i round_result = convolve_rounding( + &comp_avg_res, &offset_const, &rounding_const, rounding_shift); + + const __m256i res_8 = + _mm256_packus_epi16(round_result, round_result); + const __m128i res_0 = _mm256_castsi256_si128(res_8); + const __m128i res_1 = _mm256_extracti128_si256(res_8, 1); + + *(uint32_t *)(&dst0[i * dst_stride0 + j]) = + _mm_cvtsi128_si32(res_0); + *(uint32_t *)(&dst0[i * dst_stride0 + j + dst_stride0]) = + _mm_cvtsi128_si32(res_1); + + } else { + const __m128i res_0 = _mm256_castsi256_si128(res_unsigned); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j]), res_0); + + const __m128i res_1 = _mm256_extracti128_si256(res_unsigned, 1); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j + dst_stride]), + res_1); + } + } + + s[0] = s[1]; + s[1] = s[2]; + s[2] = s[3]; + + s[4] = s[5]; + s[5] = s[6]; + s[6] = s[7]; + } + } + } +} + +void av1_jnt_convolve_2d_copy_avx2(const uint8_t *src, int src_stride, + uint8_t *dst0, int dst_stride0, int w, int h, + const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, + const int subpel_x_q4, const int subpel_y_q4, + ConvolveParams *conv_params) { + const int bd = 8; + CONV_BUF_TYPE *dst = conv_params->dst; + int dst_stride = conv_params->dst_stride; + (void)filter_params_x; + (void)filter_params_y; + (void)subpel_x_q4; + (void)subpel_y_q4; + + const int bits = + FILTER_BITS * 2 - conv_params->round_1 - conv_params->round_0; + const __m128i left_shift = _mm_cvtsi32_si128(bits); + const int do_average = conv_params->do_average; + const int use_jnt_comp_avg = conv_params->use_jnt_comp_avg; + const __m256i wt = unpack_weights_avx2(conv_params); + const __m256i zero = _mm256_setzero_si256(); + + const int offset_0 = + bd + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const int offset = (1 << offset_0) + (1 << (offset_0 - 1)); + const __m256i offset_const = _mm256_set1_epi16(offset); + const int rounding_shift = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const __m256i rounding_const = _mm256_set1_epi16((1 << rounding_shift) >> 1); + int i, j; + + if (!(w % 16)) { + for (i = 0; i < h; i += 1) { + for (j = 0; j < w; j += 16) { + const __m256i src_16bit = _mm256_cvtepu8_epi16( + _mm_loadu_si128((__m128i *)(&src[i * src_stride + j]))); + + const __m256i res = _mm256_sll_epi16(src_16bit, left_shift); + const __m256i res_unsigned = _mm256_add_epi16(res, offset_const); + + if (do_average) { + const __m256i data_ref_0 = + _mm256_loadu_si256((__m256i *)(&dst[i * dst_stride + j])); + + const __m256i comp_avg_res = + comp_avg(&data_ref_0, &res_unsigned, &wt, use_jnt_comp_avg); + + const __m256i round_result = convolve_rounding( + &comp_avg_res, &offset_const, &rounding_const, rounding_shift); + + const __m256i res_8 = _mm256_packus_epi16(round_result, round_result); + const __m256i res_0 = _mm256_permute4x64_epi64(res_8, 0xD8); + + _mm_store_si128((__m128i *)(&dst0[i * dst_stride0 + j]), + _mm256_castsi256_si128(res_0)); + } else { + _mm256_store_si256((__m256i *)(&dst[i * dst_stride + j]), + res_unsigned); + } + } + } + } else if (!(w % 4)) { + for (i = 0; i < h; i += 2) { + for (j = 0; j < w; j += 8) { + const __m128i src_row_0 = + _mm_loadl_epi64((__m128i *)(&src[i * src_stride + j])); + const __m128i src_row_1 = + _mm_loadl_epi64((__m128i *)(&src[i * src_stride + j + src_stride])); + // since not all compilers yet support _mm256_set_m128i() + const __m256i src_10 = _mm256_insertf128_si256( + _mm256_castsi128_si256(src_row_0), src_row_1, 1); + + const __m256i src_16bit = _mm256_unpacklo_epi8(src_10, zero); + + const __m256i res = _mm256_sll_epi16(src_16bit, left_shift); + + const __m256i res_unsigned = _mm256_add_epi16(res, offset_const); + + // Accumulate values into the destination buffer + if (do_average) { + const __m256i data_ref_0 = load_line2_avx2( + &dst[i * dst_stride + j], &dst[i * dst_stride + j + dst_stride]); + const __m256i comp_avg_res = + comp_avg(&data_ref_0, &res_unsigned, &wt, use_jnt_comp_avg); + + const __m256i round_result = convolve_rounding( + &comp_avg_res, &offset_const, &rounding_const, rounding_shift); + + const __m256i res_8 = _mm256_packus_epi16(round_result, round_result); + const __m128i res_0 = _mm256_castsi256_si128(res_8); + const __m128i res_1 = _mm256_extracti128_si256(res_8, 1); + + if (w > 4) { + _mm_storel_epi64((__m128i *)(&dst0[i * dst_stride0 + j]), res_0); + _mm_storel_epi64( + (__m128i *)((&dst0[i * dst_stride0 + j + dst_stride0])), res_1); + } else { + *(uint32_t *)(&dst0[i * dst_stride0 + j]) = + _mm_cvtsi128_si32(res_0); + *(uint32_t *)(&dst0[i * dst_stride0 + j + dst_stride0]) = + _mm_cvtsi128_si32(res_1); + } + } else { + const __m128i res_0 = _mm256_castsi256_si128(res_unsigned); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j]), res_0); + + const __m128i res_1 = _mm256_extracti128_si256(res_unsigned, 1); + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j + dst_stride]), + res_1); + } + } + } + } +} diff --git a/third_party/aom/av1/common/x86/jnt_convolve_sse2.c b/third_party/aom/av1/common/x86/jnt_convolve_sse2.c new file mode 100644 index 0000000000..87dc3242e8 --- /dev/null +++ b/third_party/aom/av1/common/x86/jnt_convolve_sse2.c @@ -0,0 +1,385 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <emmintrin.h> + +#include "config/aom_dsp_rtcd.h" + +#include "aom_dsp/aom_filter.h" +#include "aom_dsp/x86/convolve_sse2.h" + +void av1_jnt_convolve_x_sse2(const uint8_t *src, int src_stride, uint8_t *dst0, + int dst_stride0, int w, int h, + const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, + const int subpel_x_q4, const int subpel_y_q4, + ConvolveParams *conv_params) { + const int bd = 8; + CONV_BUF_TYPE *dst = conv_params->dst; + const int dst_stride = conv_params->dst_stride; + const int fo_horiz = filter_params_x->taps / 2 - 1; + const uint8_t *src_ptr = src - fo_horiz; + const int bits = FILTER_BITS - conv_params->round_1; + const __m128i left_shift = _mm_cvtsi32_si128(bits); + const __m128i round_const = _mm_set1_epi32((1 << conv_params->round_0) >> 1); + const __m128i round_shift = _mm_cvtsi32_si128(conv_params->round_0); + const int w0 = conv_params->fwd_offset; + const int w1 = conv_params->bck_offset; + const __m128i wt0 = _mm_set1_epi16(w0); + const __m128i wt1 = _mm_set1_epi16(w1); + const __m128i wt = _mm_unpacklo_epi16(wt0, wt1); + const int do_average = conv_params->do_average; + const int use_jnt_comp_avg = conv_params->use_jnt_comp_avg; + const int offset_0 = + bd + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const int offset = (1 << offset_0) + (1 << (offset_0 - 1)); + const __m128i offset_const = _mm_set1_epi16(offset); + const int rounding_shift = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const __m128i rounding_const = _mm_set1_epi16((1 << rounding_shift) >> 1); + __m128i coeffs[4]; + + (void)filter_params_y; + (void)subpel_y_q4; + + prepare_coeffs(filter_params_x, subpel_x_q4, coeffs); + + if (w == 4) { + do { + const __m128i data = _mm_loadu_si128((__m128i *)src_ptr); + __m128i s[4]; + + s[0] = _mm_unpacklo_epi8(data, _mm_srli_si128(data, 1)); + s[1] = + _mm_unpacklo_epi8(_mm_srli_si128(data, 2), _mm_srli_si128(data, 3)); + s[2] = + _mm_unpacklo_epi8(_mm_srli_si128(data, 4), _mm_srli_si128(data, 5)); + s[3] = + _mm_unpacklo_epi8(_mm_srli_si128(data, 6), _mm_srli_si128(data, 7)); + const __m128i res_lo = convolve_lo_x(s, coeffs); + const __m128i res_lo_round = + _mm_sra_epi32(_mm_add_epi32(res_lo, round_const), round_shift); + const __m128i res_lo_shift = _mm_sll_epi32(res_lo_round, left_shift); + + const __m128i res_16b = _mm_packs_epi32(res_lo_shift, res_lo_shift); + const __m128i res_unsigned = _mm_add_epi16(res_16b, offset_const); + + // Accumulate values into the destination buffer + if (do_average) { + const __m128i data_ref_0 = _mm_loadu_si128((__m128i *)dst); + + const __m128i comp_avg_res = + comp_avg(&data_ref_0, &res_unsigned, &wt, use_jnt_comp_avg); + + const __m128i round_result = convolve_rounding( + &comp_avg_res, &offset_const, &rounding_const, rounding_shift); + + const __m128i res_8 = _mm_packus_epi16(round_result, round_result); + *(uint32_t *)(&dst0[0]) = _mm_cvtsi128_si32(res_8); + } else { + _mm_store_si128((__m128i *)(&dst[0]), res_unsigned); + } + src_ptr += src_stride; + dst += dst_stride; + dst0 += dst_stride0; + } while (--h); + } else { + assert(!(w % 8)); + int i = 0; + do { + int j = 0; + do { + const __m128i data = + _mm_loadu_si128((__m128i *)&src_ptr[i * src_stride + j]); + __m128i s[4]; + + // Filter even-index pixels + s[0] = data; + s[1] = _mm_srli_si128(data, 2); + s[2] = _mm_srli_si128(data, 4); + s[3] = _mm_srli_si128(data, 6); + const __m128i res_even = convolve_lo_x(s, coeffs); + + // Filter odd-index pixels + s[0] = _mm_srli_si128(data, 1); + s[1] = _mm_srli_si128(data, 3); + s[2] = _mm_srli_si128(data, 5); + s[3] = _mm_srli_si128(data, 7); + const __m128i res_odd = convolve_lo_x(s, coeffs); + + // Rearrange pixels back into the order 0 ... 7 + const __m128i res_lo = _mm_unpacklo_epi32(res_even, res_odd); + const __m128i res_hi = _mm_unpackhi_epi32(res_even, res_odd); + const __m128i res_lo_round = + _mm_sra_epi32(_mm_add_epi32(res_lo, round_const), round_shift); + const __m128i res_hi_round = + _mm_sra_epi32(_mm_add_epi32(res_hi, round_const), round_shift); + const __m128i res_lo_shift = _mm_sll_epi32(res_lo_round, left_shift); + const __m128i res_hi_shift = _mm_sll_epi32(res_hi_round, left_shift); + + const __m128i res_16b = _mm_packs_epi32(res_lo_shift, res_hi_shift); + const __m128i res_unsigned = _mm_add_epi16(res_16b, offset_const); + + // Accumulate values into the destination buffer + if (do_average) { + const __m128i data_ref_0 = + _mm_loadu_si128((__m128i *)(&dst[i * dst_stride + j])); + + const __m128i comp_avg_res = + comp_avg(&data_ref_0, &res_unsigned, &wt, use_jnt_comp_avg); + + const __m128i round_result = convolve_rounding( + &comp_avg_res, &offset_const, &rounding_const, rounding_shift); + + const __m128i res_8 = _mm_packus_epi16(round_result, round_result); + _mm_storel_epi64((__m128i *)(&dst0[i * dst_stride0 + j]), res_8); + } else { + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j]), res_unsigned); + } + j += 8; + } while (j < w); + } while (++i < h); + } +} + +void av1_jnt_convolve_y_sse2(const uint8_t *src, int src_stride, uint8_t *dst0, + int dst_stride0, int w, int h, + const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, + const int subpel_x_q4, const int subpel_y_q4, + ConvolveParams *conv_params) { + const int bd = 8; + CONV_BUF_TYPE *dst = conv_params->dst; + const int dst_stride = conv_params->dst_stride; + const int fo_vert = filter_params_y->taps / 2 - 1; + const uint8_t *src_ptr = src - fo_vert * src_stride; + const int bits = FILTER_BITS - conv_params->round_0; + const __m128i left_shift = _mm_cvtsi32_si128(bits); + const __m128i wt0 = _mm_set1_epi16(conv_params->fwd_offset); + const __m128i wt1 = _mm_set1_epi16(conv_params->bck_offset); + const __m128i wt = _mm_unpacklo_epi16(wt0, wt1); + const int do_average = conv_params->do_average; + const int use_jnt_comp_avg = conv_params->use_jnt_comp_avg; + const int offset_0 = + bd + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const int offset = (1 << offset_0) + (1 << (offset_0 - 1)); + const __m128i offset_const = _mm_set1_epi16(offset); + const int rounding_shift = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const __m128i rounding_const = _mm_set1_epi16((1 << rounding_shift) >> 1); + const __m128i round_const = _mm_set1_epi32((1 << conv_params->round_1) >> 1); + const __m128i round_shift = _mm_cvtsi32_si128(conv_params->round_1); + __m128i coeffs[4]; + + (void)filter_params_x; + (void)subpel_x_q4; + + prepare_coeffs(filter_params_y, subpel_y_q4, coeffs); + + if (w == 4) { + __m128i s[8], src6, res, res_shift; + src6 = _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 6 * src_stride)); + s[0] = _mm_unpacklo_epi8( + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 0 * src_stride)), + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 1 * src_stride))); + s[1] = _mm_unpacklo_epi8( + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 1 * src_stride)), + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 2 * src_stride))); + s[2] = _mm_unpacklo_epi8( + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 2 * src_stride)), + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 3 * src_stride))); + s[3] = _mm_unpacklo_epi8( + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 3 * src_stride)), + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 4 * src_stride))); + s[4] = _mm_unpacklo_epi8( + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 4 * src_stride)), + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 5 * src_stride))); + s[5] = _mm_unpacklo_epi8( + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 5 * src_stride)), src6); + + do { + s[6] = _mm_unpacklo_epi8( + src6, _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 7 * src_stride))); + src6 = _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 8 * src_stride)); + s[7] = _mm_unpacklo_epi8( + _mm_cvtsi32_si128(*(uint32_t *)(src_ptr + 7 * src_stride)), src6); + + res = convolve_lo_y(s + 0, coeffs); + res_shift = _mm_sll_epi32(res, left_shift); + res_shift = + _mm_sra_epi32(_mm_add_epi32(res_shift, round_const), round_shift); + + __m128i res_16b = _mm_packs_epi32(res_shift, res_shift); + __m128i res_unsigned = _mm_add_epi16(res_16b, offset_const); + + // Accumulate values into the destination buffer + if (do_average) { + const __m128i data_ref_0 = _mm_loadu_si128((__m128i *)dst); + + const __m128i comp_avg_res = + comp_avg(&data_ref_0, &res_unsigned, &wt, use_jnt_comp_avg); + + const __m128i round_result = convolve_rounding( + &comp_avg_res, &offset_const, &rounding_const, rounding_shift); + + const __m128i res_8 = _mm_packus_epi16(round_result, round_result); + *(uint32_t *)(&dst0[0]) = _mm_cvtsi128_si32(res_8); + + } else { + _mm_store_si128((__m128i *)dst, res_unsigned); + } + + src_ptr += src_stride; + dst += dst_stride; + dst0 += dst_stride0; + + res = convolve_lo_y(s + 1, coeffs); + res_shift = _mm_sll_epi32(res, left_shift); + res_shift = + _mm_sra_epi32(_mm_add_epi32(res_shift, round_const), round_shift); + + res_16b = _mm_packs_epi32(res_shift, res_shift); + res_unsigned = _mm_add_epi16(res_16b, offset_const); + + // Accumulate values into the destination buffer + if (do_average) { + const __m128i data_ref_0 = _mm_loadu_si128((__m128i *)dst); + + const __m128i comp_avg_res = + comp_avg(&data_ref_0, &res_unsigned, &wt, use_jnt_comp_avg); + + const __m128i round_result = convolve_rounding( + &comp_avg_res, &offset_const, &rounding_const, rounding_shift); + + const __m128i res_8 = _mm_packus_epi16(round_result, round_result); + *(uint32_t *)(&dst0[0]) = _mm_cvtsi128_si32(res_8); + + } else { + _mm_store_si128((__m128i *)dst, res_unsigned); + } + + src_ptr += src_stride; + dst += dst_stride; + dst0 += dst_stride0; + + s[0] = s[2]; + s[1] = s[3]; + s[2] = s[4]; + s[3] = s[5]; + s[4] = s[6]; + s[5] = s[7]; + h -= 2; + } while (h); + } else { + assert(!(w % 8)); + int j = 0; + do { + __m128i s[8], src6, res_lo, res_hi, res_lo_shift, res_hi_shift; + const uint8_t *data = &src_ptr[j]; + + src6 = _mm_loadl_epi64((__m128i *)(data + 6 * src_stride)); + s[0] = _mm_unpacklo_epi8( + _mm_loadl_epi64((__m128i *)(data + 0 * src_stride)), + _mm_loadl_epi64((__m128i *)(data + 1 * src_stride))); + s[1] = _mm_unpacklo_epi8( + _mm_loadl_epi64((__m128i *)(data + 1 * src_stride)), + _mm_loadl_epi64((__m128i *)(data + 2 * src_stride))); + s[2] = _mm_unpacklo_epi8( + _mm_loadl_epi64((__m128i *)(data + 2 * src_stride)), + _mm_loadl_epi64((__m128i *)(data + 3 * src_stride))); + s[3] = _mm_unpacklo_epi8( + _mm_loadl_epi64((__m128i *)(data + 3 * src_stride)), + _mm_loadl_epi64((__m128i *)(data + 4 * src_stride))); + s[4] = _mm_unpacklo_epi8( + _mm_loadl_epi64((__m128i *)(data + 4 * src_stride)), + _mm_loadl_epi64((__m128i *)(data + 5 * src_stride))); + s[5] = _mm_unpacklo_epi8( + _mm_loadl_epi64((__m128i *)(data + 5 * src_stride)), src6); + + int i = 0; + do { + data = &src_ptr[i * src_stride + j]; + s[6] = _mm_unpacklo_epi8( + src6, _mm_loadl_epi64((__m128i *)(data + 7 * src_stride))); + src6 = _mm_loadl_epi64((__m128i *)(data + 8 * src_stride)); + s[7] = _mm_unpacklo_epi8( + _mm_loadl_epi64((__m128i *)(data + 7 * src_stride)), src6); + + res_lo = convolve_lo_y(s, coeffs); // Filter low index pixels + res_hi = convolve_hi_y(s, coeffs); // Filter high index pixels + res_lo_shift = _mm_sll_epi32(res_lo, left_shift); + res_hi_shift = _mm_sll_epi32(res_hi, left_shift); + res_lo_shift = _mm_sra_epi32(_mm_add_epi32(res_lo_shift, round_const), + round_shift); + res_hi_shift = _mm_sra_epi32(_mm_add_epi32(res_hi_shift, round_const), + round_shift); + + __m128i res_16b = _mm_packs_epi32(res_lo_shift, res_hi_shift); + __m128i res_unsigned = _mm_add_epi16(res_16b, offset_const); + + // Accumulate values into the destination buffer + if (do_average) { + const __m128i data_ref_0 = + _mm_loadu_si128((__m128i *)(&dst[i * dst_stride + j])); + + const __m128i comp_avg_res = + comp_avg(&data_ref_0, &res_unsigned, &wt, use_jnt_comp_avg); + + const __m128i round_result = convolve_rounding( + &comp_avg_res, &offset_const, &rounding_const, rounding_shift); + + const __m128i res_8 = _mm_packus_epi16(round_result, round_result); + _mm_storel_epi64((__m128i *)(&dst0[i * dst_stride0 + j]), res_8); + } else { + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j]), res_unsigned); + } + i++; + + res_lo = convolve_lo_y(s + 1, coeffs); // Filter low index pixels + res_hi = convolve_hi_y(s + 1, coeffs); // Filter high index pixels + res_lo_shift = _mm_sll_epi32(res_lo, left_shift); + res_hi_shift = _mm_sll_epi32(res_hi, left_shift); + res_lo_shift = _mm_sra_epi32(_mm_add_epi32(res_lo_shift, round_const), + round_shift); + res_hi_shift = _mm_sra_epi32(_mm_add_epi32(res_hi_shift, round_const), + round_shift); + res_16b = _mm_packs_epi32(res_lo_shift, res_hi_shift); + res_unsigned = _mm_add_epi16(res_16b, offset_const); + + // Accumulate values into the destination buffer + if (do_average) { + __m128i data_ref_0 = + _mm_loadu_si128((__m128i *)(&dst[i * dst_stride + j])); + + const __m128i comp_avg_res = + comp_avg(&data_ref_0, &res_unsigned, &wt, use_jnt_comp_avg); + + const __m128i round_result = convolve_rounding( + &comp_avg_res, &offset_const, &rounding_const, rounding_shift); + + const __m128i res_8 = _mm_packus_epi16(round_result, round_result); + _mm_storel_epi64((__m128i *)(&dst0[i * dst_stride0 + j]), res_8); + } else { + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j]), res_unsigned); + } + i++; + + s[0] = s[2]; + s[1] = s[3]; + s[2] = s[4]; + s[3] = s[5]; + s[4] = s[6]; + s[5] = s[7]; + } while (i < h); + j += 8; + } while (j < w); + } +} diff --git a/third_party/aom/av1/common/x86/jnt_convolve_ssse3.c b/third_party/aom/av1/common/x86/jnt_convolve_ssse3.c new file mode 100644 index 0000000000..822772782b --- /dev/null +++ b/third_party/aom/av1/common/x86/jnt_convolve_ssse3.c @@ -0,0 +1,232 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <tmmintrin.h> + +#include "config/aom_dsp_rtcd.h" + +#include "aom_dsp/aom_filter.h" +#include "aom_dsp/x86/convolve_sse2.h" + +void av1_jnt_convolve_2d_ssse3(const uint8_t *src, int src_stride, + uint8_t *dst0, int dst_stride0, int w, int h, + const InterpFilterParams *filter_params_x, + const InterpFilterParams *filter_params_y, + const int subpel_x_q4, const int subpel_y_q4, + ConvolveParams *conv_params) { + CONV_BUF_TYPE *dst = conv_params->dst; + int dst_stride = conv_params->dst_stride; + const int bd = 8; + + DECLARE_ALIGNED(16, int16_t, + im_block[(MAX_SB_SIZE + MAX_FILTER_TAP - 1) * MAX_SB_SIZE]); + int im_h = h + filter_params_y->taps - 1; + int im_stride = MAX_SB_SIZE; + int i, j; + const int fo_vert = filter_params_y->taps / 2 - 1; + const int fo_horiz = filter_params_x->taps / 2 - 1; + const int do_average = conv_params->do_average; + const int use_jnt_comp_avg = conv_params->use_jnt_comp_avg; + const uint8_t *const src_ptr = src - fo_vert * src_stride - fo_horiz; + + const __m128i zero = _mm_setzero_si128(); + + const int w0 = conv_params->fwd_offset; + const int w1 = conv_params->bck_offset; + const __m128i wt0 = _mm_set1_epi16(w0); + const __m128i wt1 = _mm_set1_epi16(w1); + const __m128i wt = _mm_unpacklo_epi16(wt0, wt1); + + const int offset_0 = + bd + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const int offset = (1 << offset_0) + (1 << (offset_0 - 1)); + const __m128i offset_const = _mm_set1_epi16(offset); + const int rounding_shift = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const __m128i rounding_const = _mm_set1_epi16((1 << rounding_shift) >> 1); + + /* Horizontal filter */ + { + const int16_t *x_filter = av1_get_interp_filter_subpel_kernel( + filter_params_x, subpel_x_q4 & SUBPEL_MASK); + const __m128i coeffs_x = _mm_loadu_si128((__m128i *)x_filter); + + // coeffs 0 1 0 1 2 3 2 3 + const __m128i tmp_0 = _mm_unpacklo_epi32(coeffs_x, coeffs_x); + // coeffs 4 5 4 5 6 7 6 7 + const __m128i tmp_1 = _mm_unpackhi_epi32(coeffs_x, coeffs_x); + + // coeffs 0 1 0 1 0 1 0 1 + const __m128i coeff_01 = _mm_unpacklo_epi64(tmp_0, tmp_0); + // coeffs 2 3 2 3 2 3 2 3 + const __m128i coeff_23 = _mm_unpackhi_epi64(tmp_0, tmp_0); + // coeffs 4 5 4 5 4 5 4 5 + const __m128i coeff_45 = _mm_unpacklo_epi64(tmp_1, tmp_1); + // coeffs 6 7 6 7 6 7 6 7 + const __m128i coeff_67 = _mm_unpackhi_epi64(tmp_1, tmp_1); + + const __m128i round_const = _mm_set1_epi32( + ((1 << conv_params->round_0) >> 1) + (1 << (bd + FILTER_BITS - 1))); + const __m128i round_shift = _mm_cvtsi32_si128(conv_params->round_0); + + for (i = 0; i < im_h; ++i) { + for (j = 0; j < w; j += 8) { + const __m128i data = + _mm_loadu_si128((__m128i *)&src_ptr[i * src_stride + j]); + + const __m128i src_lo = _mm_unpacklo_epi8(data, zero); + const __m128i src_hi = _mm_unpackhi_epi8(data, zero); + + // Filter even-index pixels + const __m128i res_0 = _mm_madd_epi16(src_lo, coeff_01); + const __m128i src_2 = _mm_alignr_epi8(src_hi, src_lo, 4); + const __m128i res_2 = _mm_madd_epi16(src_2, coeff_23); + const __m128i src_4 = _mm_alignr_epi8(src_hi, src_lo, 8); + const __m128i res_4 = _mm_madd_epi16(src_4, coeff_45); + const __m128i src_6 = _mm_alignr_epi8(src_hi, src_lo, 12); + const __m128i res_6 = _mm_madd_epi16(src_6, coeff_67); + + __m128i res_even = _mm_add_epi32(_mm_add_epi32(res_0, res_4), + _mm_add_epi32(res_2, res_6)); + res_even = + _mm_sra_epi32(_mm_add_epi32(res_even, round_const), round_shift); + + // Filter odd-index pixels + const __m128i src_1 = _mm_alignr_epi8(src_hi, src_lo, 2); + const __m128i res_1 = _mm_madd_epi16(src_1, coeff_01); + const __m128i src_3 = _mm_alignr_epi8(src_hi, src_lo, 6); + const __m128i res_3 = _mm_madd_epi16(src_3, coeff_23); + const __m128i src_5 = _mm_alignr_epi8(src_hi, src_lo, 10); + const __m128i res_5 = _mm_madd_epi16(src_5, coeff_45); + const __m128i src_7 = _mm_alignr_epi8(src_hi, src_lo, 14); + const __m128i res_7 = _mm_madd_epi16(src_7, coeff_67); + + __m128i res_odd = _mm_add_epi32(_mm_add_epi32(res_1, res_5), + _mm_add_epi32(res_3, res_7)); + res_odd = + _mm_sra_epi32(_mm_add_epi32(res_odd, round_const), round_shift); + + // Pack in the column order 0, 2, 4, 6, 1, 3, 5, 7 + __m128i res = _mm_packs_epi32(res_even, res_odd); + _mm_store_si128((__m128i *)&im_block[i * im_stride + j], res); + } + } + } + + /* Vertical filter */ + { + const int16_t *y_filter = av1_get_interp_filter_subpel_kernel( + filter_params_y, subpel_y_q4 & SUBPEL_MASK); + const __m128i coeffs_y = _mm_loadu_si128((__m128i *)y_filter); + + // coeffs 0 1 0 1 2 3 2 3 + const __m128i tmp_0 = _mm_unpacklo_epi32(coeffs_y, coeffs_y); + // coeffs 4 5 4 5 6 7 6 7 + const __m128i tmp_1 = _mm_unpackhi_epi32(coeffs_y, coeffs_y); + + // coeffs 0 1 0 1 0 1 0 1 + const __m128i coeff_01 = _mm_unpacklo_epi64(tmp_0, tmp_0); + // coeffs 2 3 2 3 2 3 2 3 + const __m128i coeff_23 = _mm_unpackhi_epi64(tmp_0, tmp_0); + // coeffs 4 5 4 5 4 5 4 5 + const __m128i coeff_45 = _mm_unpacklo_epi64(tmp_1, tmp_1); + // coeffs 6 7 6 7 6 7 6 7 + const __m128i coeff_67 = _mm_unpackhi_epi64(tmp_1, tmp_1); + + const __m128i round_const = _mm_set1_epi32( + ((1 << conv_params->round_1) >> 1) - + (1 << (bd + 2 * FILTER_BITS - conv_params->round_0 - 1))); + const __m128i round_shift = _mm_cvtsi32_si128(conv_params->round_1); + + for (i = 0; i < h; ++i) { + for (j = 0; j < w; j += 8) { + // Filter even-index pixels + const int16_t *data = &im_block[i * im_stride + j]; + const __m128i src_0 = + _mm_unpacklo_epi16(*(__m128i *)(data + 0 * im_stride), + *(__m128i *)(data + 1 * im_stride)); + const __m128i src_2 = + _mm_unpacklo_epi16(*(__m128i *)(data + 2 * im_stride), + *(__m128i *)(data + 3 * im_stride)); + const __m128i src_4 = + _mm_unpacklo_epi16(*(__m128i *)(data + 4 * im_stride), + *(__m128i *)(data + 5 * im_stride)); + const __m128i src_6 = + _mm_unpacklo_epi16(*(__m128i *)(data + 6 * im_stride), + *(__m128i *)(data + 7 * im_stride)); + + const __m128i res_0 = _mm_madd_epi16(src_0, coeff_01); + const __m128i res_2 = _mm_madd_epi16(src_2, coeff_23); + const __m128i res_4 = _mm_madd_epi16(src_4, coeff_45); + const __m128i res_6 = _mm_madd_epi16(src_6, coeff_67); + + const __m128i res_even = _mm_add_epi32(_mm_add_epi32(res_0, res_2), + _mm_add_epi32(res_4, res_6)); + + // Filter odd-index pixels + const __m128i src_1 = + _mm_unpackhi_epi16(*(__m128i *)(data + 0 * im_stride), + *(__m128i *)(data + 1 * im_stride)); + const __m128i src_3 = + _mm_unpackhi_epi16(*(__m128i *)(data + 2 * im_stride), + *(__m128i *)(data + 3 * im_stride)); + const __m128i src_5 = + _mm_unpackhi_epi16(*(__m128i *)(data + 4 * im_stride), + *(__m128i *)(data + 5 * im_stride)); + const __m128i src_7 = + _mm_unpackhi_epi16(*(__m128i *)(data + 6 * im_stride), + *(__m128i *)(data + 7 * im_stride)); + + const __m128i res_1 = _mm_madd_epi16(src_1, coeff_01); + const __m128i res_3 = _mm_madd_epi16(src_3, coeff_23); + const __m128i res_5 = _mm_madd_epi16(src_5, coeff_45); + const __m128i res_7 = _mm_madd_epi16(src_7, coeff_67); + + const __m128i res_odd = _mm_add_epi32(_mm_add_epi32(res_1, res_3), + _mm_add_epi32(res_5, res_7)); + + // Rearrange pixels back into the order 0 ... 7 + const __m128i res_lo = _mm_unpacklo_epi32(res_even, res_odd); + const __m128i res_hi = _mm_unpackhi_epi32(res_even, res_odd); + + const __m128i res_lo_round = + _mm_sra_epi32(_mm_add_epi32(res_lo, round_const), round_shift); + const __m128i res_hi_round = + _mm_sra_epi32(_mm_add_epi32(res_hi, round_const), round_shift); + + const __m128i res_16b = _mm_packs_epi32(res_lo_round, res_hi_round); + const __m128i res_unsigned = _mm_add_epi16(res_16b, offset_const); + + // Accumulate values into the destination buffer + if (do_average) { + const __m128i data_ref_0 = + _mm_loadu_si128((__m128i *)(&dst[i * dst_stride + j])); + + const __m128i comp_avg_res = + comp_avg(&data_ref_0, &res_unsigned, &wt, use_jnt_comp_avg); + + const __m128i round_result = convolve_rounding( + &comp_avg_res, &offset_const, &rounding_const, rounding_shift); + + const __m128i res_8 = _mm_packus_epi16(round_result, round_result); + + if (w > 4) + _mm_storel_epi64((__m128i *)(&dst0[i * dst_stride0 + j]), res_8); + else + *(uint32_t *)(&dst0[i * dst_stride0 + j]) = + _mm_cvtsi128_si32(res_8); + } else { + _mm_store_si128((__m128i *)(&dst[i * dst_stride + j]), res_unsigned); + } + } + } + } +} diff --git a/third_party/aom/av1/common/x86/reconinter_avx2.c b/third_party/aom/av1/common/x86/reconinter_avx2.c new file mode 100644 index 0000000000..f645e04541 --- /dev/null +++ b/third_party/aom/av1/common/x86/reconinter_avx2.c @@ -0,0 +1,620 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <immintrin.h> + +#include "config/av1_rtcd.h" + +#include "aom/aom_integer.h" +#include "aom_dsp/blend.h" +#include "aom_dsp/x86/synonyms.h" +#include "aom_dsp/x86/synonyms_avx2.h" +#include "av1/common/blockd.h" + +static INLINE __m256i calc_mask_avx2(const __m256i mask_base, const __m256i s0, + const __m256i s1) { + const __m256i diff = _mm256_abs_epi16(_mm256_sub_epi16(s0, s1)); + return _mm256_abs_epi16( + _mm256_add_epi16(mask_base, _mm256_srli_epi16(diff, 4))); + // clamp(diff, 0, 64) can be skiped for diff is always in the range ( 38, 54) +} +void av1_build_compound_diffwtd_mask_avx2(uint8_t *mask, + DIFFWTD_MASK_TYPE mask_type, + const uint8_t *src0, int stride0, + const uint8_t *src1, int stride1, + int h, int w) { + const int mb = (mask_type == DIFFWTD_38_INV) ? AOM_BLEND_A64_MAX_ALPHA : 0; + const __m256i y_mask_base = _mm256_set1_epi16(38 - mb); + int i = 0; + if (4 == w) { + do { + const __m128i s0A = xx_loadl_32(src0); + const __m128i s0B = xx_loadl_32(src0 + stride0); + const __m128i s0C = xx_loadl_32(src0 + stride0 * 2); + const __m128i s0D = xx_loadl_32(src0 + stride0 * 3); + const __m128i s0AB = _mm_unpacklo_epi32(s0A, s0B); + const __m128i s0CD = _mm_unpacklo_epi32(s0C, s0D); + const __m128i s0ABCD = _mm_unpacklo_epi64(s0AB, s0CD); + const __m256i s0ABCD_w = _mm256_cvtepu8_epi16(s0ABCD); + + const __m128i s1A = xx_loadl_32(src1); + const __m128i s1B = xx_loadl_32(src1 + stride1); + const __m128i s1C = xx_loadl_32(src1 + stride1 * 2); + const __m128i s1D = xx_loadl_32(src1 + stride1 * 3); + const __m128i s1AB = _mm_unpacklo_epi32(s1A, s1B); + const __m128i s1CD = _mm_unpacklo_epi32(s1C, s1D); + const __m128i s1ABCD = _mm_unpacklo_epi64(s1AB, s1CD); + const __m256i s1ABCD_w = _mm256_cvtepu8_epi16(s1ABCD); + const __m256i m16 = calc_mask_avx2(y_mask_base, s0ABCD_w, s1ABCD_w); + const __m256i m8 = _mm256_packus_epi16(m16, _mm256_setzero_si256()); + const __m128i x_m8 = + _mm256_castsi256_si128(_mm256_permute4x64_epi64(m8, 0xd8)); + xx_storeu_128(mask, x_m8); + src0 += (stride0 << 2); + src1 += (stride1 << 2); + mask += 16; + i += 4; + } while (i < h); + } else if (8 == w) { + do { + const __m128i s0A = xx_loadl_64(src0); + const __m128i s0B = xx_loadl_64(src0 + stride0); + const __m128i s0C = xx_loadl_64(src0 + stride0 * 2); + const __m128i s0D = xx_loadl_64(src0 + stride0 * 3); + const __m256i s0AC_w = _mm256_cvtepu8_epi16(_mm_unpacklo_epi64(s0A, s0C)); + const __m256i s0BD_w = _mm256_cvtepu8_epi16(_mm_unpacklo_epi64(s0B, s0D)); + const __m128i s1A = xx_loadl_64(src1); + const __m128i s1B = xx_loadl_64(src1 + stride1); + const __m128i s1C = xx_loadl_64(src1 + stride1 * 2); + const __m128i s1D = xx_loadl_64(src1 + stride1 * 3); + const __m256i s1AB_w = _mm256_cvtepu8_epi16(_mm_unpacklo_epi64(s1A, s1C)); + const __m256i s1CD_w = _mm256_cvtepu8_epi16(_mm_unpacklo_epi64(s1B, s1D)); + const __m256i m16AC = calc_mask_avx2(y_mask_base, s0AC_w, s1AB_w); + const __m256i m16BD = calc_mask_avx2(y_mask_base, s0BD_w, s1CD_w); + const __m256i m8 = _mm256_packus_epi16(m16AC, m16BD); + yy_storeu_256(mask, m8); + src0 += stride0 << 2; + src1 += stride1 << 2; + mask += 32; + i += 4; + } while (i < h); + } else if (16 == w) { + do { + const __m128i s0A = xx_load_128(src0); + const __m128i s0B = xx_load_128(src0 + stride0); + const __m128i s1A = xx_load_128(src1); + const __m128i s1B = xx_load_128(src1 + stride1); + const __m256i s0AL = _mm256_cvtepu8_epi16(s0A); + const __m256i s0BL = _mm256_cvtepu8_epi16(s0B); + const __m256i s1AL = _mm256_cvtepu8_epi16(s1A); + const __m256i s1BL = _mm256_cvtepu8_epi16(s1B); + + const __m256i m16AL = calc_mask_avx2(y_mask_base, s0AL, s1AL); + const __m256i m16BL = calc_mask_avx2(y_mask_base, s0BL, s1BL); + + const __m256i m8 = + _mm256_permute4x64_epi64(_mm256_packus_epi16(m16AL, m16BL), 0xd8); + yy_storeu_256(mask, m8); + src0 += stride0 << 1; + src1 += stride1 << 1; + mask += 32; + i += 2; + } while (i < h); + } else { + do { + int j = 0; + do { + const __m256i s0 = yy_loadu_256(src0 + j); + const __m256i s1 = yy_loadu_256(src1 + j); + const __m256i s0L = _mm256_cvtepu8_epi16(_mm256_castsi256_si128(s0)); + const __m256i s1L = _mm256_cvtepu8_epi16(_mm256_castsi256_si128(s1)); + const __m256i s0H = + _mm256_cvtepu8_epi16(_mm256_extracti128_si256(s0, 1)); + const __m256i s1H = + _mm256_cvtepu8_epi16(_mm256_extracti128_si256(s1, 1)); + const __m256i m16L = calc_mask_avx2(y_mask_base, s0L, s1L); + const __m256i m16H = calc_mask_avx2(y_mask_base, s0H, s1H); + const __m256i m8 = + _mm256_permute4x64_epi64(_mm256_packus_epi16(m16L, m16H), 0xd8); + yy_storeu_256(mask + j, m8); + j += 32; + } while (j < w); + src0 += stride0; + src1 += stride1; + mask += w; + i += 1; + } while (i < h); + } +} + +static INLINE __m256i calc_mask_d16_avx2(const __m256i *data_src0, + const __m256i *data_src1, + const __m256i *round_const, + const __m256i *mask_base_16, + const __m256i *clip_diff, int round) { + const __m256i diffa = _mm256_subs_epu16(*data_src0, *data_src1); + const __m256i diffb = _mm256_subs_epu16(*data_src1, *data_src0); + const __m256i diff = _mm256_max_epu16(diffa, diffb); + const __m256i diff_round = + _mm256_srli_epi16(_mm256_adds_epu16(diff, *round_const), round); + const __m256i diff_factor = _mm256_srli_epi16(diff_round, DIFF_FACTOR_LOG2); + const __m256i diff_mask = _mm256_adds_epi16(diff_factor, *mask_base_16); + const __m256i diff_clamp = _mm256_min_epi16(diff_mask, *clip_diff); + return diff_clamp; +} + +static INLINE __m256i calc_mask_d16_inv_avx2(const __m256i *data_src0, + const __m256i *data_src1, + const __m256i *round_const, + const __m256i *mask_base_16, + const __m256i *clip_diff, + int round) { + const __m256i diffa = _mm256_subs_epu16(*data_src0, *data_src1); + const __m256i diffb = _mm256_subs_epu16(*data_src1, *data_src0); + const __m256i diff = _mm256_max_epu16(diffa, diffb); + const __m256i diff_round = + _mm256_srli_epi16(_mm256_adds_epu16(diff, *round_const), round); + const __m256i diff_factor = _mm256_srli_epi16(diff_round, DIFF_FACTOR_LOG2); + const __m256i diff_mask = _mm256_adds_epi16(diff_factor, *mask_base_16); + const __m256i diff_clamp = _mm256_min_epi16(diff_mask, *clip_diff); + const __m256i diff_const_16 = _mm256_sub_epi16(*clip_diff, diff_clamp); + return diff_const_16; +} + +static INLINE void build_compound_diffwtd_mask_d16_avx2( + uint8_t *mask, const CONV_BUF_TYPE *src0, int src0_stride, + const CONV_BUF_TYPE *src1, int src1_stride, int h, int w, int shift) { + const int mask_base = 38; + const __m256i _r = _mm256_set1_epi16((1 << shift) >> 1); + const __m256i y38 = _mm256_set1_epi16(mask_base); + const __m256i y64 = _mm256_set1_epi16(AOM_BLEND_A64_MAX_ALPHA); + int i = 0; + if (w == 4) { + do { + const __m128i s0A = xx_loadl_64(src0); + const __m128i s0B = xx_loadl_64(src0 + src0_stride); + const __m128i s0C = xx_loadl_64(src0 + src0_stride * 2); + const __m128i s0D = xx_loadl_64(src0 + src0_stride * 3); + const __m128i s1A = xx_loadl_64(src1); + const __m128i s1B = xx_loadl_64(src1 + src1_stride); + const __m128i s1C = xx_loadl_64(src1 + src1_stride * 2); + const __m128i s1D = xx_loadl_64(src1 + src1_stride * 3); + const __m256i s0 = yy_set_m128i(_mm_unpacklo_epi64(s0C, s0D), + _mm_unpacklo_epi64(s0A, s0B)); + const __m256i s1 = yy_set_m128i(_mm_unpacklo_epi64(s1C, s1D), + _mm_unpacklo_epi64(s1A, s1B)); + const __m256i m16 = calc_mask_d16_avx2(&s0, &s1, &_r, &y38, &y64, shift); + const __m256i m8 = _mm256_packus_epi16(m16, _mm256_setzero_si256()); + xx_storeu_128(mask, + _mm256_castsi256_si128(_mm256_permute4x64_epi64(m8, 0xd8))); + src0 += src0_stride << 2; + src1 += src1_stride << 2; + mask += 16; + i += 4; + } while (i < h); + } else if (w == 8) { + do { + const __m256i s0AB = yy_loadu2_128(src0 + src0_stride, src0); + const __m256i s0CD = + yy_loadu2_128(src0 + src0_stride * 3, src0 + src0_stride * 2); + const __m256i s1AB = yy_loadu2_128(src1 + src1_stride, src1); + const __m256i s1CD = + yy_loadu2_128(src1 + src1_stride * 3, src1 + src1_stride * 2); + const __m256i m16AB = + calc_mask_d16_avx2(&s0AB, &s1AB, &_r, &y38, &y64, shift); + const __m256i m16CD = + calc_mask_d16_avx2(&s0CD, &s1CD, &_r, &y38, &y64, shift); + const __m256i m8 = _mm256_packus_epi16(m16AB, m16CD); + yy_storeu_256(mask, _mm256_permute4x64_epi64(m8, 0xd8)); + src0 += src0_stride << 2; + src1 += src1_stride << 2; + mask += 32; + i += 4; + } while (i < h); + } else if (w == 16) { + do { + const __m256i s0A = yy_loadu_256(src0); + const __m256i s0B = yy_loadu_256(src0 + src0_stride); + const __m256i s1A = yy_loadu_256(src1); + const __m256i s1B = yy_loadu_256(src1 + src1_stride); + const __m256i m16A = + calc_mask_d16_avx2(&s0A, &s1A, &_r, &y38, &y64, shift); + const __m256i m16B = + calc_mask_d16_avx2(&s0B, &s1B, &_r, &y38, &y64, shift); + const __m256i m8 = _mm256_packus_epi16(m16A, m16B); + yy_storeu_256(mask, _mm256_permute4x64_epi64(m8, 0xd8)); + src0 += src0_stride << 1; + src1 += src1_stride << 1; + mask += 32; + i += 2; + } while (i < h); + } else if (w == 32) { + do { + const __m256i s0A = yy_loadu_256(src0); + const __m256i s0B = yy_loadu_256(src0 + 16); + const __m256i s1A = yy_loadu_256(src1); + const __m256i s1B = yy_loadu_256(src1 + 16); + const __m256i m16A = + calc_mask_d16_avx2(&s0A, &s1A, &_r, &y38, &y64, shift); + const __m256i m16B = + calc_mask_d16_avx2(&s0B, &s1B, &_r, &y38, &y64, shift); + const __m256i m8 = _mm256_packus_epi16(m16A, m16B); + yy_storeu_256(mask, _mm256_permute4x64_epi64(m8, 0xd8)); + src0 += src0_stride; + src1 += src1_stride; + mask += 32; + i += 1; + } while (i < h); + } else if (w == 64) { + do { + const __m256i s0A = yy_loadu_256(src0); + const __m256i s0B = yy_loadu_256(src0 + 16); + const __m256i s0C = yy_loadu_256(src0 + 32); + const __m256i s0D = yy_loadu_256(src0 + 48); + const __m256i s1A = yy_loadu_256(src1); + const __m256i s1B = yy_loadu_256(src1 + 16); + const __m256i s1C = yy_loadu_256(src1 + 32); + const __m256i s1D = yy_loadu_256(src1 + 48); + const __m256i m16A = + calc_mask_d16_avx2(&s0A, &s1A, &_r, &y38, &y64, shift); + const __m256i m16B = + calc_mask_d16_avx2(&s0B, &s1B, &_r, &y38, &y64, shift); + const __m256i m16C = + calc_mask_d16_avx2(&s0C, &s1C, &_r, &y38, &y64, shift); + const __m256i m16D = + calc_mask_d16_avx2(&s0D, &s1D, &_r, &y38, &y64, shift); + const __m256i m8AB = _mm256_packus_epi16(m16A, m16B); + const __m256i m8CD = _mm256_packus_epi16(m16C, m16D); + yy_storeu_256(mask, _mm256_permute4x64_epi64(m8AB, 0xd8)); + yy_storeu_256(mask + 32, _mm256_permute4x64_epi64(m8CD, 0xd8)); + src0 += src0_stride; + src1 += src1_stride; + mask += 64; + i += 1; + } while (i < h); + } else { + do { + const __m256i s0A = yy_loadu_256(src0); + const __m256i s0B = yy_loadu_256(src0 + 16); + const __m256i s0C = yy_loadu_256(src0 + 32); + const __m256i s0D = yy_loadu_256(src0 + 48); + const __m256i s0E = yy_loadu_256(src0 + 64); + const __m256i s0F = yy_loadu_256(src0 + 80); + const __m256i s0G = yy_loadu_256(src0 + 96); + const __m256i s0H = yy_loadu_256(src0 + 112); + const __m256i s1A = yy_loadu_256(src1); + const __m256i s1B = yy_loadu_256(src1 + 16); + const __m256i s1C = yy_loadu_256(src1 + 32); + const __m256i s1D = yy_loadu_256(src1 + 48); + const __m256i s1E = yy_loadu_256(src1 + 64); + const __m256i s1F = yy_loadu_256(src1 + 80); + const __m256i s1G = yy_loadu_256(src1 + 96); + const __m256i s1H = yy_loadu_256(src1 + 112); + const __m256i m16A = + calc_mask_d16_avx2(&s0A, &s1A, &_r, &y38, &y64, shift); + const __m256i m16B = + calc_mask_d16_avx2(&s0B, &s1B, &_r, &y38, &y64, shift); + const __m256i m16C = + calc_mask_d16_avx2(&s0C, &s1C, &_r, &y38, &y64, shift); + const __m256i m16D = + calc_mask_d16_avx2(&s0D, &s1D, &_r, &y38, &y64, shift); + const __m256i m16E = + calc_mask_d16_avx2(&s0E, &s1E, &_r, &y38, &y64, shift); + const __m256i m16F = + calc_mask_d16_avx2(&s0F, &s1F, &_r, &y38, &y64, shift); + const __m256i m16G = + calc_mask_d16_avx2(&s0G, &s1G, &_r, &y38, &y64, shift); + const __m256i m16H = + calc_mask_d16_avx2(&s0H, &s1H, &_r, &y38, &y64, shift); + const __m256i m8AB = _mm256_packus_epi16(m16A, m16B); + const __m256i m8CD = _mm256_packus_epi16(m16C, m16D); + const __m256i m8EF = _mm256_packus_epi16(m16E, m16F); + const __m256i m8GH = _mm256_packus_epi16(m16G, m16H); + yy_storeu_256(mask, _mm256_permute4x64_epi64(m8AB, 0xd8)); + yy_storeu_256(mask + 32, _mm256_permute4x64_epi64(m8CD, 0xd8)); + yy_storeu_256(mask + 64, _mm256_permute4x64_epi64(m8EF, 0xd8)); + yy_storeu_256(mask + 96, _mm256_permute4x64_epi64(m8GH, 0xd8)); + src0 += src0_stride; + src1 += src1_stride; + mask += 128; + i += 1; + } while (i < h); + } +} + +static INLINE void build_compound_diffwtd_mask_d16_inv_avx2( + uint8_t *mask, const CONV_BUF_TYPE *src0, int src0_stride, + const CONV_BUF_TYPE *src1, int src1_stride, int h, int w, int shift) { + const int mask_base = 38; + const __m256i _r = _mm256_set1_epi16((1 << shift) >> 1); + const __m256i y38 = _mm256_set1_epi16(mask_base); + const __m256i y64 = _mm256_set1_epi16(AOM_BLEND_A64_MAX_ALPHA); + int i = 0; + if (w == 4) { + do { + const __m128i s0A = xx_loadl_64(src0); + const __m128i s0B = xx_loadl_64(src0 + src0_stride); + const __m128i s0C = xx_loadl_64(src0 + src0_stride * 2); + const __m128i s0D = xx_loadl_64(src0 + src0_stride * 3); + const __m128i s1A = xx_loadl_64(src1); + const __m128i s1B = xx_loadl_64(src1 + src1_stride); + const __m128i s1C = xx_loadl_64(src1 + src1_stride * 2); + const __m128i s1D = xx_loadl_64(src1 + src1_stride * 3); + const __m256i s0 = yy_set_m128i(_mm_unpacklo_epi64(s0C, s0D), + _mm_unpacklo_epi64(s0A, s0B)); + const __m256i s1 = yy_set_m128i(_mm_unpacklo_epi64(s1C, s1D), + _mm_unpacklo_epi64(s1A, s1B)); + const __m256i m16 = + calc_mask_d16_inv_avx2(&s0, &s1, &_r, &y38, &y64, shift); + const __m256i m8 = _mm256_packus_epi16(m16, _mm256_setzero_si256()); + xx_storeu_128(mask, + _mm256_castsi256_si128(_mm256_permute4x64_epi64(m8, 0xd8))); + src0 += src0_stride << 2; + src1 += src1_stride << 2; + mask += 16; + i += 4; + } while (i < h); + } else if (w == 8) { + do { + const __m256i s0AB = yy_loadu2_128(src0 + src0_stride, src0); + const __m256i s0CD = + yy_loadu2_128(src0 + src0_stride * 3, src0 + src0_stride * 2); + const __m256i s1AB = yy_loadu2_128(src1 + src1_stride, src1); + const __m256i s1CD = + yy_loadu2_128(src1 + src1_stride * 3, src1 + src1_stride * 2); + const __m256i m16AB = + calc_mask_d16_inv_avx2(&s0AB, &s1AB, &_r, &y38, &y64, shift); + const __m256i m16CD = + calc_mask_d16_inv_avx2(&s0CD, &s1CD, &_r, &y38, &y64, shift); + const __m256i m8 = _mm256_packus_epi16(m16AB, m16CD); + yy_storeu_256(mask, _mm256_permute4x64_epi64(m8, 0xd8)); + src0 += src0_stride << 2; + src1 += src1_stride << 2; + mask += 32; + i += 4; + } while (i < h); + } else if (w == 16) { + do { + const __m256i s0A = yy_loadu_256(src0); + const __m256i s0B = yy_loadu_256(src0 + src0_stride); + const __m256i s1A = yy_loadu_256(src1); + const __m256i s1B = yy_loadu_256(src1 + src1_stride); + const __m256i m16A = + calc_mask_d16_inv_avx2(&s0A, &s1A, &_r, &y38, &y64, shift); + const __m256i m16B = + calc_mask_d16_inv_avx2(&s0B, &s1B, &_r, &y38, &y64, shift); + const __m256i m8 = _mm256_packus_epi16(m16A, m16B); + yy_storeu_256(mask, _mm256_permute4x64_epi64(m8, 0xd8)); + src0 += src0_stride << 1; + src1 += src1_stride << 1; + mask += 32; + i += 2; + } while (i < h); + } else if (w == 32) { + do { + const __m256i s0A = yy_loadu_256(src0); + const __m256i s0B = yy_loadu_256(src0 + 16); + const __m256i s1A = yy_loadu_256(src1); + const __m256i s1B = yy_loadu_256(src1 + 16); + const __m256i m16A = + calc_mask_d16_inv_avx2(&s0A, &s1A, &_r, &y38, &y64, shift); + const __m256i m16B = + calc_mask_d16_inv_avx2(&s0B, &s1B, &_r, &y38, &y64, shift); + const __m256i m8 = _mm256_packus_epi16(m16A, m16B); + yy_storeu_256(mask, _mm256_permute4x64_epi64(m8, 0xd8)); + src0 += src0_stride; + src1 += src1_stride; + mask += 32; + i += 1; + } while (i < h); + } else if (w == 64) { + do { + const __m256i s0A = yy_loadu_256(src0); + const __m256i s0B = yy_loadu_256(src0 + 16); + const __m256i s0C = yy_loadu_256(src0 + 32); + const __m256i s0D = yy_loadu_256(src0 + 48); + const __m256i s1A = yy_loadu_256(src1); + const __m256i s1B = yy_loadu_256(src1 + 16); + const __m256i s1C = yy_loadu_256(src1 + 32); + const __m256i s1D = yy_loadu_256(src1 + 48); + const __m256i m16A = + calc_mask_d16_inv_avx2(&s0A, &s1A, &_r, &y38, &y64, shift); + const __m256i m16B = + calc_mask_d16_inv_avx2(&s0B, &s1B, &_r, &y38, &y64, shift); + const __m256i m16C = + calc_mask_d16_inv_avx2(&s0C, &s1C, &_r, &y38, &y64, shift); + const __m256i m16D = + calc_mask_d16_inv_avx2(&s0D, &s1D, &_r, &y38, &y64, shift); + const __m256i m8AB = _mm256_packus_epi16(m16A, m16B); + const __m256i m8CD = _mm256_packus_epi16(m16C, m16D); + yy_storeu_256(mask, _mm256_permute4x64_epi64(m8AB, 0xd8)); + yy_storeu_256(mask + 32, _mm256_permute4x64_epi64(m8CD, 0xd8)); + src0 += src0_stride; + src1 += src1_stride; + mask += 64; + i += 1; + } while (i < h); + } else { + do { + const __m256i s0A = yy_loadu_256(src0); + const __m256i s0B = yy_loadu_256(src0 + 16); + const __m256i s0C = yy_loadu_256(src0 + 32); + const __m256i s0D = yy_loadu_256(src0 + 48); + const __m256i s0E = yy_loadu_256(src0 + 64); + const __m256i s0F = yy_loadu_256(src0 + 80); + const __m256i s0G = yy_loadu_256(src0 + 96); + const __m256i s0H = yy_loadu_256(src0 + 112); + const __m256i s1A = yy_loadu_256(src1); + const __m256i s1B = yy_loadu_256(src1 + 16); + const __m256i s1C = yy_loadu_256(src1 + 32); + const __m256i s1D = yy_loadu_256(src1 + 48); + const __m256i s1E = yy_loadu_256(src1 + 64); + const __m256i s1F = yy_loadu_256(src1 + 80); + const __m256i s1G = yy_loadu_256(src1 + 96); + const __m256i s1H = yy_loadu_256(src1 + 112); + const __m256i m16A = + calc_mask_d16_inv_avx2(&s0A, &s1A, &_r, &y38, &y64, shift); + const __m256i m16B = + calc_mask_d16_inv_avx2(&s0B, &s1B, &_r, &y38, &y64, shift); + const __m256i m16C = + calc_mask_d16_inv_avx2(&s0C, &s1C, &_r, &y38, &y64, shift); + const __m256i m16D = + calc_mask_d16_inv_avx2(&s0D, &s1D, &_r, &y38, &y64, shift); + const __m256i m16E = + calc_mask_d16_inv_avx2(&s0E, &s1E, &_r, &y38, &y64, shift); + const __m256i m16F = + calc_mask_d16_inv_avx2(&s0F, &s1F, &_r, &y38, &y64, shift); + const __m256i m16G = + calc_mask_d16_inv_avx2(&s0G, &s1G, &_r, &y38, &y64, shift); + const __m256i m16H = + calc_mask_d16_inv_avx2(&s0H, &s1H, &_r, &y38, &y64, shift); + const __m256i m8AB = _mm256_packus_epi16(m16A, m16B); + const __m256i m8CD = _mm256_packus_epi16(m16C, m16D); + const __m256i m8EF = _mm256_packus_epi16(m16E, m16F); + const __m256i m8GH = _mm256_packus_epi16(m16G, m16H); + yy_storeu_256(mask, _mm256_permute4x64_epi64(m8AB, 0xd8)); + yy_storeu_256(mask + 32, _mm256_permute4x64_epi64(m8CD, 0xd8)); + yy_storeu_256(mask + 64, _mm256_permute4x64_epi64(m8EF, 0xd8)); + yy_storeu_256(mask + 96, _mm256_permute4x64_epi64(m8GH, 0xd8)); + src0 += src0_stride; + src1 += src1_stride; + mask += 128; + i += 1; + } while (i < h); + } +} + +void av1_build_compound_diffwtd_mask_d16_avx2( + uint8_t *mask, DIFFWTD_MASK_TYPE mask_type, const CONV_BUF_TYPE *src0, + int src0_stride, const CONV_BUF_TYPE *src1, int src1_stride, int h, int w, + ConvolveParams *conv_params, int bd) { + const int shift = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1 + (bd - 8); + // When rounding constant is added, there is a possibility of overflow. + // However that much precision is not required. Code should very well work for + // other values of DIFF_FACTOR_LOG2 and AOM_BLEND_A64_MAX_ALPHA as well. But + // there is a possibility of corner case bugs. + assert(DIFF_FACTOR_LOG2 == 4); + assert(AOM_BLEND_A64_MAX_ALPHA == 64); + + if (mask_type == DIFFWTD_38) { + build_compound_diffwtd_mask_d16_avx2(mask, src0, src0_stride, src1, + src1_stride, h, w, shift); + } else { + build_compound_diffwtd_mask_d16_inv_avx2(mask, src0, src0_stride, src1, + src1_stride, h, w, shift); + } +} + +void av1_build_compound_diffwtd_mask_highbd_avx2( + uint8_t *mask, DIFFWTD_MASK_TYPE mask_type, const uint8_t *src0, + int src0_stride, const uint8_t *src1, int src1_stride, int h, int w, + int bd) { + if (w < 16) { + av1_build_compound_diffwtd_mask_highbd_ssse3( + mask, mask_type, src0, src0_stride, src1, src1_stride, h, w, bd); + } else { + assert(mask_type == DIFFWTD_38 || mask_type == DIFFWTD_38_INV); + assert(bd >= 8); + assert((w % 16) == 0); + const __m256i y0 = _mm256_setzero_si256(); + const __m256i yAOM_BLEND_A64_MAX_ALPHA = + _mm256_set1_epi16(AOM_BLEND_A64_MAX_ALPHA); + const int mask_base = 38; + const __m256i ymask_base = _mm256_set1_epi16(mask_base); + const uint16_t *ssrc0 = CONVERT_TO_SHORTPTR(src0); + const uint16_t *ssrc1 = CONVERT_TO_SHORTPTR(src1); + if (bd == 8) { + if (mask_type == DIFFWTD_38_INV) { + for (int i = 0; i < h; ++i) { + for (int j = 0; j < w; j += 16) { + __m256i s0 = _mm256_loadu_si256((const __m256i *)&ssrc0[j]); + __m256i s1 = _mm256_loadu_si256((const __m256i *)&ssrc1[j]); + __m256i diff = _mm256_srai_epi16( + _mm256_abs_epi16(_mm256_sub_epi16(s0, s1)), DIFF_FACTOR_LOG2); + __m256i m = _mm256_min_epi16( + _mm256_max_epi16(y0, _mm256_add_epi16(diff, ymask_base)), + yAOM_BLEND_A64_MAX_ALPHA); + m = _mm256_sub_epi16(yAOM_BLEND_A64_MAX_ALPHA, m); + m = _mm256_packus_epi16(m, m); + m = _mm256_permute4x64_epi64(m, _MM_SHUFFLE(0, 0, 2, 0)); + __m128i m0 = _mm256_castsi256_si128(m); + _mm_storeu_si128((__m128i *)&mask[j], m0); + } + ssrc0 += src0_stride; + ssrc1 += src1_stride; + mask += w; + } + } else { + for (int i = 0; i < h; ++i) { + for (int j = 0; j < w; j += 16) { + __m256i s0 = _mm256_loadu_si256((const __m256i *)&ssrc0[j]); + __m256i s1 = _mm256_loadu_si256((const __m256i *)&ssrc1[j]); + __m256i diff = _mm256_srai_epi16( + _mm256_abs_epi16(_mm256_sub_epi16(s0, s1)), DIFF_FACTOR_LOG2); + __m256i m = _mm256_min_epi16( + _mm256_max_epi16(y0, _mm256_add_epi16(diff, ymask_base)), + yAOM_BLEND_A64_MAX_ALPHA); + m = _mm256_packus_epi16(m, m); + m = _mm256_permute4x64_epi64(m, _MM_SHUFFLE(0, 0, 2, 0)); + __m128i m0 = _mm256_castsi256_si128(m); + _mm_storeu_si128((__m128i *)&mask[j], m0); + } + ssrc0 += src0_stride; + ssrc1 += src1_stride; + mask += w; + } + } + } else { + const __m128i xshift = xx_set1_64_from_32i(bd - 8 + DIFF_FACTOR_LOG2); + if (mask_type == DIFFWTD_38_INV) { + for (int i = 0; i < h; ++i) { + for (int j = 0; j < w; j += 16) { + __m256i s0 = _mm256_loadu_si256((const __m256i *)&ssrc0[j]); + __m256i s1 = _mm256_loadu_si256((const __m256i *)&ssrc1[j]); + __m256i diff = _mm256_sra_epi16( + _mm256_abs_epi16(_mm256_sub_epi16(s0, s1)), xshift); + __m256i m = _mm256_min_epi16( + _mm256_max_epi16(y0, _mm256_add_epi16(diff, ymask_base)), + yAOM_BLEND_A64_MAX_ALPHA); + m = _mm256_sub_epi16(yAOM_BLEND_A64_MAX_ALPHA, m); + m = _mm256_packus_epi16(m, m); + m = _mm256_permute4x64_epi64(m, _MM_SHUFFLE(0, 0, 2, 0)); + __m128i m0 = _mm256_castsi256_si128(m); + _mm_storeu_si128((__m128i *)&mask[j], m0); + } + ssrc0 += src0_stride; + ssrc1 += src1_stride; + mask += w; + } + } else { + for (int i = 0; i < h; ++i) { + for (int j = 0; j < w; j += 16) { + __m256i s0 = _mm256_loadu_si256((const __m256i *)&ssrc0[j]); + __m256i s1 = _mm256_loadu_si256((const __m256i *)&ssrc1[j]); + __m256i diff = _mm256_sra_epi16( + _mm256_abs_epi16(_mm256_sub_epi16(s0, s1)), xshift); + __m256i m = _mm256_min_epi16( + _mm256_max_epi16(y0, _mm256_add_epi16(diff, ymask_base)), + yAOM_BLEND_A64_MAX_ALPHA); + m = _mm256_packus_epi16(m, m); + m = _mm256_permute4x64_epi64(m, _MM_SHUFFLE(0, 0, 2, 0)); + __m128i m0 = _mm256_castsi256_si128(m); + _mm_storeu_si128((__m128i *)&mask[j], m0); + } + ssrc0 += src0_stride; + ssrc1 += src1_stride; + mask += w; + } + } + } + } +} diff --git a/third_party/aom/av1/common/x86/reconinter_sse4.c b/third_party/aom/av1/common/x86/reconinter_sse4.c new file mode 100644 index 0000000000..5171ca4934 --- /dev/null +++ b/third_party/aom/av1/common/x86/reconinter_sse4.c @@ -0,0 +1,153 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <emmintrin.h> // SSE2 +#include <smmintrin.h> /* SSE4.1 */ + +#include "aom/aom_integer.h" +#include "aom_dsp/blend.h" +#include "av1/common/blockd.h" + +static INLINE __m128i calc_mask(const __m128i mask_base, const __m128i s0, + const __m128i s1) { + const __m128i diff = _mm_abs_epi16(_mm_sub_epi16(s0, s1)); + return _mm_abs_epi16(_mm_add_epi16(mask_base, _mm_srli_epi16(diff, 4))); + // clamp(diff, 0, 64) can be skiped for diff is always in the range ( 38, 54) +} + +void av1_build_compound_diffwtd_mask_sse4_1(uint8_t *mask, + DIFFWTD_MASK_TYPE mask_type, + const uint8_t *src0, int stride0, + const uint8_t *src1, int stride1, + int h, int w) { + const int mb = (mask_type == DIFFWTD_38_INV) ? AOM_BLEND_A64_MAX_ALPHA : 0; + const __m128i mask_base = _mm_set1_epi16(38 - mb); + int i = 0; + if (4 == w) { + do { + const __m128i s0A = _mm_cvtsi32_si128(*(uint32_t *)src0); + const __m128i s0B = _mm_cvtsi32_si128(*(uint32_t *)(src0 + stride0)); + const __m128i s0AB = _mm_unpacklo_epi32(s0A, s0B); + const __m128i s0 = _mm_cvtepu8_epi16(s0AB); + + const __m128i s1A = _mm_cvtsi32_si128(*(uint32_t *)src1); + const __m128i s1B = _mm_cvtsi32_si128(*(uint32_t *)(src1 + stride1)); + const __m128i s1AB = _mm_unpacklo_epi32(s1A, s1B); + const __m128i s1 = _mm_cvtepu8_epi16(s1AB); + + const __m128i m16 = calc_mask(mask_base, s0, s1); + const __m128i m8 = _mm_packus_epi16(m16, m16); + + *(uint32_t *)mask = _mm_cvtsi128_si32(m8); + *(uint32_t *)(mask + w) = _mm_extract_epi32(m8, 1); + src0 += (stride0 << 1); + src1 += (stride1 << 1); + mask += 8; + i += 2; + } while (i < h); + } else if (8 == w) { + do { + __m128i s0 = _mm_loadl_epi64((__m128i const *)src0); + __m128i s1 = _mm_loadl_epi64((__m128i const *)src1); + s0 = _mm_cvtepu8_epi16(s0); + s1 = _mm_cvtepu8_epi16(s1); + const __m128i m16 = calc_mask(mask_base, s0, s1); + const __m128i m8 = _mm_packus_epi16(m16, m16); + _mm_storel_epi64((__m128i *)mask, m8); + src0 += stride0; + src1 += stride1; + mask += 8; + i += 1; + } while (i < h); + } else { + const __m128i zero = _mm_setzero_si128(); + do { + int j = 0; + do { + const __m128i s0 = _mm_load_si128((__m128i const *)(src0 + j)); + const __m128i s1 = _mm_load_si128((__m128i const *)(src1 + j)); + const __m128i s0L = _mm_cvtepu8_epi16(s0); + const __m128i s1L = _mm_cvtepu8_epi16(s1); + const __m128i s0H = _mm_unpackhi_epi8(s0, zero); + const __m128i s1H = _mm_unpackhi_epi8(s1, zero); + + const __m128i m16L = calc_mask(mask_base, s0L, s1L); + const __m128i m16H = calc_mask(mask_base, s0H, s1H); + + const __m128i m8 = _mm_packus_epi16(m16L, m16H); + _mm_store_si128((__m128i *)(mask + j), m8); + j += 16; + } while (j < w); + src0 += stride0; + src1 += stride1; + mask += w; + i += 1; + } while (i < h); + } +} + +void av1_build_compound_diffwtd_mask_d16_sse4_1( + uint8_t *mask, DIFFWTD_MASK_TYPE mask_type, const CONV_BUF_TYPE *src0, + int src0_stride, const CONV_BUF_TYPE *src1, int src1_stride, int h, int w, + ConvolveParams *conv_params, int bd) { + const int which_inverse = (mask_type == DIFFWTD_38) ? 0 : 1; + const int mask_base = 38; + int round = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1 + (bd - 8); + const __m128i round_const = _mm_set1_epi16((1 << round) >> 1); + const __m128i mask_base_16 = _mm_set1_epi16(mask_base); + const __m128i clip_diff = _mm_set1_epi16(AOM_BLEND_A64_MAX_ALPHA); + const __m128i add_const = + _mm_set1_epi16((which_inverse ? AOM_BLEND_A64_MAX_ALPHA : 0)); + const __m128i add_sign = _mm_set1_epi16((which_inverse ? -1 : 1)); + + int i, j; + // When rounding constant is added, there is a possibility of overflow. + // However that much precision is not required. Code should very well work for + // other values of DIFF_FACTOR_LOG2 and AOM_BLEND_A64_MAX_ALPHA as well. But + // there is a possibility of corner case bugs. + assert(DIFF_FACTOR_LOG2 == 4); + assert(AOM_BLEND_A64_MAX_ALPHA == 64); + for (i = 0; i < h; ++i) { + for (j = 0; j < w; j += 8) { + const __m128i data_src0 = + _mm_loadu_si128((__m128i *)&src0[(i * src0_stride) + j]); + const __m128i data_src1 = + _mm_loadu_si128((__m128i *)&src1[(i * src1_stride) + j]); + + const __m128i diffa = _mm_subs_epu16(data_src0, data_src1); + const __m128i diffb = _mm_subs_epu16(data_src1, data_src0); + const __m128i diff = _mm_max_epu16(diffa, diffb); + const __m128i diff_round = + _mm_srli_epi16(_mm_adds_epu16(diff, round_const), round); + const __m128i diff_factor = _mm_srli_epi16(diff_round, DIFF_FACTOR_LOG2); + const __m128i diff_mask = _mm_adds_epi16(diff_factor, mask_base_16); + __m128i diff_clamp = _mm_min_epi16(diff_mask, clip_diff); + // clamp to 0 can be skipped since we are using add and saturate + // instruction + + const __m128i diff_sign = _mm_sign_epi16(diff_clamp, add_sign); + const __m128i diff_const_16 = _mm_add_epi16(diff_sign, add_const); + + // 8 bit conversion and saturation to uint8 + const __m128i res_8 = _mm_packus_epi16(diff_const_16, diff_const_16); + + // Store values into the destination buffer + __m128i *const dst = (__m128i *)&mask[i * w + j]; + + if ((w - j) > 4) { + _mm_storel_epi64(dst, res_8); + } else { // w==4 + *(uint32_t *)dst = _mm_cvtsi128_si32(res_8); + } + } + } +} diff --git a/third_party/aom/av1/common/x86/reconinter_ssse3.c b/third_party/aom/av1/common/x86/reconinter_ssse3.c new file mode 100644 index 0000000000..cf684447c5 --- /dev/null +++ b/third_party/aom/av1/common/x86/reconinter_ssse3.c @@ -0,0 +1,116 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <tmmintrin.h> + +#include "config/av1_rtcd.h" + +#include "aom/aom_integer.h" +#include "aom_dsp/blend.h" +#include "aom_dsp/x86/synonyms.h" +#include "av1/common/blockd.h" + +void av1_build_compound_diffwtd_mask_highbd_ssse3( + uint8_t *mask, DIFFWTD_MASK_TYPE mask_type, const uint8_t *src0, + int src0_stride, const uint8_t *src1, int src1_stride, int h, int w, + int bd) { + if (w < 8) { + av1_build_compound_diffwtd_mask_highbd_c(mask, mask_type, src0, src0_stride, + src1, src1_stride, h, w, bd); + } else { + assert(bd >= 8); + assert((w % 8) == 0); + assert(mask_type == DIFFWTD_38 || mask_type == DIFFWTD_38_INV); + const __m128i x0 = _mm_setzero_si128(); + const __m128i xAOM_BLEND_A64_MAX_ALPHA = + _mm_set1_epi16(AOM_BLEND_A64_MAX_ALPHA); + const int mask_base = 38; + const __m128i xmask_base = _mm_set1_epi16(mask_base); + const uint16_t *ssrc0 = CONVERT_TO_SHORTPTR(src0); + const uint16_t *ssrc1 = CONVERT_TO_SHORTPTR(src1); + if (bd == 8) { + if (mask_type == DIFFWTD_38_INV) { + for (int i = 0; i < h; ++i) { + for (int j = 0; j < w; j += 8) { + __m128i s0 = _mm_loadu_si128((const __m128i *)&ssrc0[j]); + __m128i s1 = _mm_loadu_si128((const __m128i *)&ssrc1[j]); + __m128i diff = _mm_srai_epi16(_mm_abs_epi16(_mm_sub_epi16(s0, s1)), + DIFF_FACTOR_LOG2); + __m128i m = _mm_min_epi16( + _mm_max_epi16(x0, _mm_add_epi16(diff, xmask_base)), + xAOM_BLEND_A64_MAX_ALPHA); + m = _mm_sub_epi16(xAOM_BLEND_A64_MAX_ALPHA, m); + m = _mm_packus_epi16(m, m); + _mm_storel_epi64((__m128i *)&mask[j], m); + } + ssrc0 += src0_stride; + ssrc1 += src1_stride; + mask += w; + } + } else { + for (int i = 0; i < h; ++i) { + for (int j = 0; j < w; j += 8) { + __m128i s0 = _mm_loadu_si128((const __m128i *)&ssrc0[j]); + __m128i s1 = _mm_loadu_si128((const __m128i *)&ssrc1[j]); + __m128i diff = _mm_srai_epi16(_mm_abs_epi16(_mm_sub_epi16(s0, s1)), + DIFF_FACTOR_LOG2); + __m128i m = _mm_min_epi16( + _mm_max_epi16(x0, _mm_add_epi16(diff, xmask_base)), + xAOM_BLEND_A64_MAX_ALPHA); + m = _mm_packus_epi16(m, m); + _mm_storel_epi64((__m128i *)&mask[j], m); + } + ssrc0 += src0_stride; + ssrc1 += src1_stride; + mask += w; + } + } + } else { + const __m128i xshift = xx_set1_64_from_32i(bd - 8 + DIFF_FACTOR_LOG2); + if (mask_type == DIFFWTD_38_INV) { + for (int i = 0; i < h; ++i) { + for (int j = 0; j < w; j += 8) { + __m128i s0 = _mm_loadu_si128((const __m128i *)&ssrc0[j]); + __m128i s1 = _mm_loadu_si128((const __m128i *)&ssrc1[j]); + __m128i diff = + _mm_sra_epi16(_mm_abs_epi16(_mm_sub_epi16(s0, s1)), xshift); + __m128i m = _mm_min_epi16( + _mm_max_epi16(x0, _mm_add_epi16(diff, xmask_base)), + xAOM_BLEND_A64_MAX_ALPHA); + m = _mm_sub_epi16(xAOM_BLEND_A64_MAX_ALPHA, m); + m = _mm_packus_epi16(m, m); + _mm_storel_epi64((__m128i *)&mask[j], m); + } + ssrc0 += src0_stride; + ssrc1 += src1_stride; + mask += w; + } + } else { + for (int i = 0; i < h; ++i) { + for (int j = 0; j < w; j += 8) { + __m128i s0 = _mm_loadu_si128((const __m128i *)&ssrc0[j]); + __m128i s1 = _mm_loadu_si128((const __m128i *)&ssrc1[j]); + __m128i diff = + _mm_sra_epi16(_mm_abs_epi16(_mm_sub_epi16(s0, s1)), xshift); + __m128i m = _mm_min_epi16( + _mm_max_epi16(x0, _mm_add_epi16(diff, xmask_base)), + xAOM_BLEND_A64_MAX_ALPHA); + m = _mm_packus_epi16(m, m); + _mm_storel_epi64((__m128i *)&mask[j], m); + } + ssrc0 += src0_stride; + ssrc1 += src1_stride; + mask += w; + } + } + } + } +} diff --git a/third_party/aom/av1/common/x86/selfguided_avx2.c b/third_party/aom/av1/common/x86/selfguided_avx2.c new file mode 100644 index 0000000000..0aaf1f4547 --- /dev/null +++ b/third_party/aom/av1/common/x86/selfguided_avx2.c @@ -0,0 +1,724 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <immintrin.h> + +#include "config/aom_config.h" +#include "config/av1_rtcd.h" + +#include "av1/common/restoration.h" +#include "aom_dsp/x86/synonyms.h" +#include "aom_dsp/x86/synonyms_avx2.h" + +// Load 8 bytes from the possibly-misaligned pointer p, extend each byte to +// 32-bit precision and return them in an AVX2 register. +static __m256i yy256_load_extend_8_32(const void *p) { + return _mm256_cvtepu8_epi32(xx_loadl_64(p)); +} + +// Load 8 halfwords from the possibly-misaligned pointer p, extend each +// halfword to 32-bit precision and return them in an AVX2 register. +static __m256i yy256_load_extend_16_32(const void *p) { + return _mm256_cvtepu16_epi32(xx_loadu_128(p)); +} + +// Compute the scan of an AVX2 register holding 8 32-bit integers. If the +// register holds x0..x7 then the scan will hold x0, x0+x1, x0+x1+x2, ..., +// x0+x1+...+x7 +// +// Let [...] represent a 128-bit block, and let a, ..., h be 32-bit integers +// (assumed small enough to be able to add them without overflow). +// +// Use -> as shorthand for summing, i.e. h->a = h + g + f + e + d + c + b + a. +// +// x = [h g f e][d c b a] +// x01 = [g f e 0][c b a 0] +// x02 = [g+h f+g e+f e][c+d b+c a+b a] +// x03 = [e+f e 0 0][a+b a 0 0] +// x04 = [e->h e->g e->f e][a->d a->c a->b a] +// s = a->d +// s01 = [a->d a->d a->d a->d] +// s02 = [a->d a->d a->d a->d][0 0 0 0] +// ret = [a->h a->g a->f a->e][a->d a->c a->b a] +static __m256i scan_32(__m256i x) { + const __m256i x01 = _mm256_slli_si256(x, 4); + const __m256i x02 = _mm256_add_epi32(x, x01); + const __m256i x03 = _mm256_slli_si256(x02, 8); + const __m256i x04 = _mm256_add_epi32(x02, x03); + const int32_t s = _mm256_extract_epi32(x04, 3); + const __m128i s01 = _mm_set1_epi32(s); + const __m256i s02 = _mm256_insertf128_si256(_mm256_setzero_si256(), s01, 1); + return _mm256_add_epi32(x04, s02); +} + +// Compute two integral images from src. B sums elements; A sums their +// squares. The images are offset by one pixel, so will have width and height +// equal to width + 1, height + 1 and the first row and column will be zero. +// +// A+1 and B+1 should be aligned to 32 bytes. buf_stride should be a multiple +// of 8. + +static void *memset_zero_avx(int32_t *dest, const __m256i *zero, size_t count) { + unsigned int i = 0; + for (i = 0; i < (count & 0xffffffe0); i += 32) { + _mm256_storeu_si256((__m256i *)(dest + i), *zero); + _mm256_storeu_si256((__m256i *)(dest + i + 8), *zero); + _mm256_storeu_si256((__m256i *)(dest + i + 16), *zero); + _mm256_storeu_si256((__m256i *)(dest + i + 24), *zero); + } + for (; i < (count & 0xfffffff8); i += 8) { + _mm256_storeu_si256((__m256i *)(dest + i), *zero); + } + for (; i < count; i++) { + dest[i] = 0; + } + return dest; +} + +static void integral_images(const uint8_t *src, int src_stride, int width, + int height, int32_t *A, int32_t *B, + int buf_stride) { + const __m256i zero = _mm256_setzero_si256(); + // Write out the zero top row + memset_zero_avx(A, &zero, (width + 8)); + memset_zero_avx(B, &zero, (width + 8)); + for (int i = 0; i < height; ++i) { + // Zero the left column. + A[(i + 1) * buf_stride] = B[(i + 1) * buf_stride] = 0; + + // ldiff is the difference H - D where H is the output sample immediately + // to the left and D is the output sample above it. These are scalars, + // replicated across the eight lanes. + __m256i ldiff1 = zero, ldiff2 = zero; + for (int j = 0; j < width; j += 8) { + const int ABj = 1 + j; + + const __m256i above1 = yy_load_256(B + ABj + i * buf_stride); + const __m256i above2 = yy_load_256(A + ABj + i * buf_stride); + + const __m256i x1 = yy256_load_extend_8_32(src + j + i * src_stride); + const __m256i x2 = _mm256_madd_epi16(x1, x1); + + const __m256i sc1 = scan_32(x1); + const __m256i sc2 = scan_32(x2); + + const __m256i row1 = + _mm256_add_epi32(_mm256_add_epi32(sc1, above1), ldiff1); + const __m256i row2 = + _mm256_add_epi32(_mm256_add_epi32(sc2, above2), ldiff2); + + yy_store_256(B + ABj + (i + 1) * buf_stride, row1); + yy_store_256(A + ABj + (i + 1) * buf_stride, row2); + + // Calculate the new H - D. + ldiff1 = _mm256_set1_epi32( + _mm256_extract_epi32(_mm256_sub_epi32(row1, above1), 7)); + ldiff2 = _mm256_set1_epi32( + _mm256_extract_epi32(_mm256_sub_epi32(row2, above2), 7)); + } + } +} + +// Compute two integral images from src. B sums elements; A sums their squares +// +// A and B should be aligned to 32 bytes. buf_stride should be a multiple of 8. +static void integral_images_highbd(const uint16_t *src, int src_stride, + int width, int height, int32_t *A, + int32_t *B, int buf_stride) { + const __m256i zero = _mm256_setzero_si256(); + // Write out the zero top row + memset_zero_avx(A, &zero, (width + 8)); + memset_zero_avx(B, &zero, (width + 8)); + + for (int i = 0; i < height; ++i) { + // Zero the left column. + A[(i + 1) * buf_stride] = B[(i + 1) * buf_stride] = 0; + + // ldiff is the difference H - D where H is the output sample immediately + // to the left and D is the output sample above it. These are scalars, + // replicated across the eight lanes. + __m256i ldiff1 = zero, ldiff2 = zero; + for (int j = 0; j < width; j += 8) { + const int ABj = 1 + j; + + const __m256i above1 = yy_load_256(B + ABj + i * buf_stride); + const __m256i above2 = yy_load_256(A + ABj + i * buf_stride); + + const __m256i x1 = yy256_load_extend_16_32(src + j + i * src_stride); + const __m256i x2 = _mm256_madd_epi16(x1, x1); + + const __m256i sc1 = scan_32(x1); + const __m256i sc2 = scan_32(x2); + + const __m256i row1 = + _mm256_add_epi32(_mm256_add_epi32(sc1, above1), ldiff1); + const __m256i row2 = + _mm256_add_epi32(_mm256_add_epi32(sc2, above2), ldiff2); + + yy_store_256(B + ABj + (i + 1) * buf_stride, row1); + yy_store_256(A + ABj + (i + 1) * buf_stride, row2); + + // Calculate the new H - D. + ldiff1 = _mm256_set1_epi32( + _mm256_extract_epi32(_mm256_sub_epi32(row1, above1), 7)); + ldiff2 = _mm256_set1_epi32( + _mm256_extract_epi32(_mm256_sub_epi32(row2, above2), 7)); + } + } +} + +// Compute 8 values of boxsum from the given integral image. ii should point +// at the middle of the box (for the first value). r is the box radius. +static INLINE __m256i boxsum_from_ii(const int32_t *ii, int stride, int r) { + const __m256i tl = yy_loadu_256(ii - (r + 1) - (r + 1) * stride); + const __m256i tr = yy_loadu_256(ii + (r + 0) - (r + 1) * stride); + const __m256i bl = yy_loadu_256(ii - (r + 1) + r * stride); + const __m256i br = yy_loadu_256(ii + (r + 0) + r * stride); + const __m256i u = _mm256_sub_epi32(tr, tl); + const __m256i v = _mm256_sub_epi32(br, bl); + return _mm256_sub_epi32(v, u); +} + +static __m256i round_for_shift(unsigned shift) { + return _mm256_set1_epi32((1 << shift) >> 1); +} + +static __m256i compute_p(__m256i sum1, __m256i sum2, int bit_depth, int n) { + __m256i an, bb; + if (bit_depth > 8) { + const __m256i rounding_a = round_for_shift(2 * (bit_depth - 8)); + const __m256i rounding_b = round_for_shift(bit_depth - 8); + const __m128i shift_a = _mm_cvtsi32_si128(2 * (bit_depth - 8)); + const __m128i shift_b = _mm_cvtsi32_si128(bit_depth - 8); + const __m256i a = + _mm256_srl_epi32(_mm256_add_epi32(sum2, rounding_a), shift_a); + const __m256i b = + _mm256_srl_epi32(_mm256_add_epi32(sum1, rounding_b), shift_b); + // b < 2^14, so we can use a 16-bit madd rather than a 32-bit + // mullo to square it + bb = _mm256_madd_epi16(b, b); + an = _mm256_max_epi32(_mm256_mullo_epi32(a, _mm256_set1_epi32(n)), bb); + } else { + bb = _mm256_madd_epi16(sum1, sum1); + an = _mm256_mullo_epi32(sum2, _mm256_set1_epi32(n)); + } + return _mm256_sub_epi32(an, bb); +} + +// Assumes that C, D are integral images for the original buffer which has been +// extended to have a padding of SGRPROJ_BORDER_VERT/SGRPROJ_BORDER_HORZ pixels +// on the sides. A, B, C, D point at logical position (0, 0). +static void calc_ab(int32_t *A, int32_t *B, const int32_t *C, const int32_t *D, + int width, int height, int buf_stride, int bit_depth, + int sgr_params_idx, int radius_idx) { + const sgr_params_type *const params = &sgr_params[sgr_params_idx]; + const int r = params->r[radius_idx]; + const int n = (2 * r + 1) * (2 * r + 1); + const __m256i s = _mm256_set1_epi32(params->s[radius_idx]); + // one_over_n[n-1] is 2^12/n, so easily fits in an int16 + const __m256i one_over_n = _mm256_set1_epi32(one_by_x[n - 1]); + + const __m256i rnd_z = round_for_shift(SGRPROJ_MTABLE_BITS); + const __m256i rnd_res = round_for_shift(SGRPROJ_RECIP_BITS); + + // Set up masks + const __m128i ones32 = _mm_set_epi32(0, 0, 0xffffffff, 0xffffffff); + __m256i mask[8]; + for (int idx = 0; idx < 8; idx++) { + const __m128i shift = _mm_cvtsi32_si128(8 * (8 - idx)); + mask[idx] = _mm256_cvtepi8_epi32(_mm_srl_epi64(ones32, shift)); + } + + for (int i = -1; i < height + 1; ++i) { + for (int j = -1; j < width + 1; j += 8) { + const int32_t *Cij = C + i * buf_stride + j; + const int32_t *Dij = D + i * buf_stride + j; + + __m256i sum1 = boxsum_from_ii(Dij, buf_stride, r); + __m256i sum2 = boxsum_from_ii(Cij, buf_stride, r); + + // When width + 2 isn't a multiple of 8, sum1 and sum2 will contain + // some uninitialised data in their upper words. We use a mask to + // ensure that these bits are set to 0. + int idx = AOMMIN(8, width + 1 - j); + assert(idx >= 1); + + if (idx < 8) { + sum1 = _mm256_and_si256(mask[idx], sum1); + sum2 = _mm256_and_si256(mask[idx], sum2); + } + + const __m256i p = compute_p(sum1, sum2, bit_depth, n); + + const __m256i z = _mm256_min_epi32( + _mm256_srli_epi32(_mm256_add_epi32(_mm256_mullo_epi32(p, s), rnd_z), + SGRPROJ_MTABLE_BITS), + _mm256_set1_epi32(255)); + + const __m256i a_res = _mm256_i32gather_epi32(x_by_xplus1, z, 4); + + yy_storeu_256(A + i * buf_stride + j, a_res); + + const __m256i a_complement = + _mm256_sub_epi32(_mm256_set1_epi32(SGRPROJ_SGR), a_res); + + // sum1 might have lanes greater than 2^15, so we can't use madd to do + // multiplication involving sum1. However, a_complement and one_over_n + // are both less than 256, so we can multiply them first. + const __m256i a_comp_over_n = _mm256_madd_epi16(a_complement, one_over_n); + const __m256i b_int = _mm256_mullo_epi32(a_comp_over_n, sum1); + const __m256i b_res = _mm256_srli_epi32(_mm256_add_epi32(b_int, rnd_res), + SGRPROJ_RECIP_BITS); + + yy_storeu_256(B + i * buf_stride + j, b_res); + } + } +} + +// Calculate 8 values of the "cross sum" starting at buf. This is a 3x3 filter +// where the outer four corners have weight 3 and all other pixels have weight +// 4. +// +// Pixels are indexed as follows: +// xtl xt xtr +// xl x xr +// xbl xb xbr +// +// buf points to x +// +// fours = xl + xt + xr + xb + x +// threes = xtl + xtr + xbr + xbl +// cross_sum = 4 * fours + 3 * threes +// = 4 * (fours + threes) - threes +// = (fours + threes) << 2 - threes +static INLINE __m256i cross_sum(const int32_t *buf, int stride) { + const __m256i xtl = yy_loadu_256(buf - 1 - stride); + const __m256i xt = yy_loadu_256(buf - stride); + const __m256i xtr = yy_loadu_256(buf + 1 - stride); + const __m256i xl = yy_loadu_256(buf - 1); + const __m256i x = yy_loadu_256(buf); + const __m256i xr = yy_loadu_256(buf + 1); + const __m256i xbl = yy_loadu_256(buf - 1 + stride); + const __m256i xb = yy_loadu_256(buf + stride); + const __m256i xbr = yy_loadu_256(buf + 1 + stride); + + const __m256i fours = _mm256_add_epi32( + xl, _mm256_add_epi32(xt, _mm256_add_epi32(xr, _mm256_add_epi32(xb, x)))); + const __m256i threes = + _mm256_add_epi32(xtl, _mm256_add_epi32(xtr, _mm256_add_epi32(xbr, xbl))); + + return _mm256_sub_epi32(_mm256_slli_epi32(_mm256_add_epi32(fours, threes), 2), + threes); +} + +// The final filter for self-guided restoration. Computes a weighted average +// across A, B with "cross sums" (see cross_sum implementation above). +static void final_filter(int32_t *dst, int dst_stride, const int32_t *A, + const int32_t *B, int buf_stride, const void *dgd8, + int dgd_stride, int width, int height, int highbd) { + const int nb = 5; + const __m256i rounding = + round_for_shift(SGRPROJ_SGR_BITS + nb - SGRPROJ_RST_BITS); + const uint8_t *dgd_real = + highbd ? (const uint8_t *)CONVERT_TO_SHORTPTR(dgd8) : dgd8; + + for (int i = 0; i < height; ++i) { + for (int j = 0; j < width; j += 8) { + const __m256i a = cross_sum(A + i * buf_stride + j, buf_stride); + const __m256i b = cross_sum(B + i * buf_stride + j, buf_stride); + + const __m128i raw = + xx_loadu_128(dgd_real + ((i * dgd_stride + j) << highbd)); + const __m256i src = + highbd ? _mm256_cvtepu16_epi32(raw) : _mm256_cvtepu8_epi32(raw); + + __m256i v = _mm256_add_epi32(_mm256_madd_epi16(a, src), b); + __m256i w = _mm256_srai_epi32(_mm256_add_epi32(v, rounding), + SGRPROJ_SGR_BITS + nb - SGRPROJ_RST_BITS); + + yy_storeu_256(dst + i * dst_stride + j, w); + } + } +} + +// Assumes that C, D are integral images for the original buffer which has been +// extended to have a padding of SGRPROJ_BORDER_VERT/SGRPROJ_BORDER_HORZ pixels +// on the sides. A, B, C, D point at logical position (0, 0). +static void calc_ab_fast(int32_t *A, int32_t *B, const int32_t *C, + const int32_t *D, int width, int height, + int buf_stride, int bit_depth, int sgr_params_idx, + int radius_idx) { + const sgr_params_type *const params = &sgr_params[sgr_params_idx]; + const int r = params->r[radius_idx]; + const int n = (2 * r + 1) * (2 * r + 1); + const __m256i s = _mm256_set1_epi32(params->s[radius_idx]); + // one_over_n[n-1] is 2^12/n, so easily fits in an int16 + const __m256i one_over_n = _mm256_set1_epi32(one_by_x[n - 1]); + + const __m256i rnd_z = round_for_shift(SGRPROJ_MTABLE_BITS); + const __m256i rnd_res = round_for_shift(SGRPROJ_RECIP_BITS); + + // Set up masks + const __m128i ones32 = _mm_set_epi32(0, 0, 0xffffffff, 0xffffffff); + __m256i mask[8]; + for (int idx = 0; idx < 8; idx++) { + const __m128i shift = _mm_cvtsi32_si128(8 * (8 - idx)); + mask[idx] = _mm256_cvtepi8_epi32(_mm_srl_epi64(ones32, shift)); + } + + for (int i = -1; i < height + 1; i += 2) { + for (int j = -1; j < width + 1; j += 8) { + const int32_t *Cij = C + i * buf_stride + j; + const int32_t *Dij = D + i * buf_stride + j; + + __m256i sum1 = boxsum_from_ii(Dij, buf_stride, r); + __m256i sum2 = boxsum_from_ii(Cij, buf_stride, r); + + // When width + 2 isn't a multiple of 8, sum1 and sum2 will contain + // some uninitialised data in their upper words. We use a mask to + // ensure that these bits are set to 0. + int idx = AOMMIN(8, width + 1 - j); + assert(idx >= 1); + + if (idx < 8) { + sum1 = _mm256_and_si256(mask[idx], sum1); + sum2 = _mm256_and_si256(mask[idx], sum2); + } + + const __m256i p = compute_p(sum1, sum2, bit_depth, n); + + const __m256i z = _mm256_min_epi32( + _mm256_srli_epi32(_mm256_add_epi32(_mm256_mullo_epi32(p, s), rnd_z), + SGRPROJ_MTABLE_BITS), + _mm256_set1_epi32(255)); + + const __m256i a_res = _mm256_i32gather_epi32(x_by_xplus1, z, 4); + + yy_storeu_256(A + i * buf_stride + j, a_res); + + const __m256i a_complement = + _mm256_sub_epi32(_mm256_set1_epi32(SGRPROJ_SGR), a_res); + + // sum1 might have lanes greater than 2^15, so we can't use madd to do + // multiplication involving sum1. However, a_complement and one_over_n + // are both less than 256, so we can multiply them first. + const __m256i a_comp_over_n = _mm256_madd_epi16(a_complement, one_over_n); + const __m256i b_int = _mm256_mullo_epi32(a_comp_over_n, sum1); + const __m256i b_res = _mm256_srli_epi32(_mm256_add_epi32(b_int, rnd_res), + SGRPROJ_RECIP_BITS); + + yy_storeu_256(B + i * buf_stride + j, b_res); + } + } +} + +// Calculate 8 values of the "cross sum" starting at buf. +// +// Pixels are indexed like this: +// xtl xt xtr +// - buf - +// xbl xb xbr +// +// Pixels are weighted like this: +// 5 6 5 +// 0 0 0 +// 5 6 5 +// +// fives = xtl + xtr + xbl + xbr +// sixes = xt + xb +// cross_sum = 6 * sixes + 5 * fives +// = 5 * (fives + sixes) - sixes +// = (fives + sixes) << 2 + (fives + sixes) + sixes +static INLINE __m256i cross_sum_fast_even_row(const int32_t *buf, int stride) { + const __m256i xtl = yy_loadu_256(buf - 1 - stride); + const __m256i xt = yy_loadu_256(buf - stride); + const __m256i xtr = yy_loadu_256(buf + 1 - stride); + const __m256i xbl = yy_loadu_256(buf - 1 + stride); + const __m256i xb = yy_loadu_256(buf + stride); + const __m256i xbr = yy_loadu_256(buf + 1 + stride); + + const __m256i fives = + _mm256_add_epi32(xtl, _mm256_add_epi32(xtr, _mm256_add_epi32(xbr, xbl))); + const __m256i sixes = _mm256_add_epi32(xt, xb); + const __m256i fives_plus_sixes = _mm256_add_epi32(fives, sixes); + + return _mm256_add_epi32( + _mm256_add_epi32(_mm256_slli_epi32(fives_plus_sixes, 2), + fives_plus_sixes), + sixes); +} + +// Calculate 8 values of the "cross sum" starting at buf. +// +// Pixels are indexed like this: +// xl x xr +// +// Pixels are weighted like this: +// 5 6 5 +// +// buf points to x +// +// fives = xl + xr +// sixes = x +// cross_sum = 5 * fives + 6 * sixes +// = 4 * (fives + sixes) + (fives + sixes) + sixes +// = (fives + sixes) << 2 + (fives + sixes) + sixes +static INLINE __m256i cross_sum_fast_odd_row(const int32_t *buf) { + const __m256i xl = yy_loadu_256(buf - 1); + const __m256i x = yy_loadu_256(buf); + const __m256i xr = yy_loadu_256(buf + 1); + + const __m256i fives = _mm256_add_epi32(xl, xr); + const __m256i sixes = x; + + const __m256i fives_plus_sixes = _mm256_add_epi32(fives, sixes); + + return _mm256_add_epi32( + _mm256_add_epi32(_mm256_slli_epi32(fives_plus_sixes, 2), + fives_plus_sixes), + sixes); +} + +// The final filter for the self-guided restoration. Computes a +// weighted average across A, B with "cross sums" (see cross_sum_... +// implementations above). +static void final_filter_fast(int32_t *dst, int dst_stride, const int32_t *A, + const int32_t *B, int buf_stride, + const void *dgd8, int dgd_stride, int width, + int height, int highbd) { + const int nb0 = 5; + const int nb1 = 4; + + const __m256i rounding0 = + round_for_shift(SGRPROJ_SGR_BITS + nb0 - SGRPROJ_RST_BITS); + const __m256i rounding1 = + round_for_shift(SGRPROJ_SGR_BITS + nb1 - SGRPROJ_RST_BITS); + + const uint8_t *dgd_real = + highbd ? (const uint8_t *)CONVERT_TO_SHORTPTR(dgd8) : dgd8; + + for (int i = 0; i < height; ++i) { + if (!(i & 1)) { // even row + for (int j = 0; j < width; j += 8) { + const __m256i a = + cross_sum_fast_even_row(A + i * buf_stride + j, buf_stride); + const __m256i b = + cross_sum_fast_even_row(B + i * buf_stride + j, buf_stride); + + const __m128i raw = + xx_loadu_128(dgd_real + ((i * dgd_stride + j) << highbd)); + const __m256i src = + highbd ? _mm256_cvtepu16_epi32(raw) : _mm256_cvtepu8_epi32(raw); + + __m256i v = _mm256_add_epi32(_mm256_madd_epi16(a, src), b); + __m256i w = + _mm256_srai_epi32(_mm256_add_epi32(v, rounding0), + SGRPROJ_SGR_BITS + nb0 - SGRPROJ_RST_BITS); + + yy_storeu_256(dst + i * dst_stride + j, w); + } + } else { // odd row + for (int j = 0; j < width; j += 8) { + const __m256i a = cross_sum_fast_odd_row(A + i * buf_stride + j); + const __m256i b = cross_sum_fast_odd_row(B + i * buf_stride + j); + + const __m128i raw = + xx_loadu_128(dgd_real + ((i * dgd_stride + j) << highbd)); + const __m256i src = + highbd ? _mm256_cvtepu16_epi32(raw) : _mm256_cvtepu8_epi32(raw); + + __m256i v = _mm256_add_epi32(_mm256_madd_epi16(a, src), b); + __m256i w = + _mm256_srai_epi32(_mm256_add_epi32(v, rounding1), + SGRPROJ_SGR_BITS + nb1 - SGRPROJ_RST_BITS); + + yy_storeu_256(dst + i * dst_stride + j, w); + } + } + } +} + +int av1_selfguided_restoration_avx2(const uint8_t *dgd8, int width, int height, + int dgd_stride, int32_t *flt0, + int32_t *flt1, int flt_stride, + int sgr_params_idx, int bit_depth, + int highbd) { + // The ALIGN_POWER_OF_TWO macro here ensures that column 1 of Atl, Btl, + // Ctl and Dtl is 32-byte aligned. + const int buf_elts = ALIGN_POWER_OF_TWO(RESTORATION_PROC_UNIT_PELS, 3); + + int32_t *buf = aom_memalign( + 32, 4 * sizeof(*buf) * ALIGN_POWER_OF_TWO(RESTORATION_PROC_UNIT_PELS, 3)); + if (!buf) return -1; + + const int width_ext = width + 2 * SGRPROJ_BORDER_HORZ; + const int height_ext = height + 2 * SGRPROJ_BORDER_VERT; + + // Adjusting the stride of A and B here appears to avoid bad cache effects, + // leading to a significant speed improvement. + // We also align the stride to a multiple of 32 bytes for efficiency. + int buf_stride = ALIGN_POWER_OF_TWO(width_ext + 16, 3); + + // The "tl" pointers point at the top-left of the initialised data for the + // array. + int32_t *Atl = buf + 0 * buf_elts + 7; + int32_t *Btl = buf + 1 * buf_elts + 7; + int32_t *Ctl = buf + 2 * buf_elts + 7; + int32_t *Dtl = buf + 3 * buf_elts + 7; + + // The "0" pointers are (- SGRPROJ_BORDER_VERT, -SGRPROJ_BORDER_HORZ). Note + // there's a zero row and column in A, B (integral images), so we move down + // and right one for them. + const int buf_diag_border = + SGRPROJ_BORDER_HORZ + buf_stride * SGRPROJ_BORDER_VERT; + + int32_t *A0 = Atl + 1 + buf_stride; + int32_t *B0 = Btl + 1 + buf_stride; + int32_t *C0 = Ctl + 1 + buf_stride; + int32_t *D0 = Dtl + 1 + buf_stride; + + // Finally, A, B, C, D point at position (0, 0). + int32_t *A = A0 + buf_diag_border; + int32_t *B = B0 + buf_diag_border; + int32_t *C = C0 + buf_diag_border; + int32_t *D = D0 + buf_diag_border; + + const int dgd_diag_border = + SGRPROJ_BORDER_HORZ + dgd_stride * SGRPROJ_BORDER_VERT; + const uint8_t *dgd0 = dgd8 - dgd_diag_border; + + // Generate integral images from the input. C will contain sums of squares; D + // will contain just sums + if (highbd) + integral_images_highbd(CONVERT_TO_SHORTPTR(dgd0), dgd_stride, width_ext, + height_ext, Ctl, Dtl, buf_stride); + else + integral_images(dgd0, dgd_stride, width_ext, height_ext, Ctl, Dtl, + buf_stride); + + const sgr_params_type *const params = &sgr_params[sgr_params_idx]; + // Write to flt0 and flt1 + // If params->r == 0 we skip the corresponding filter. We only allow one of + // the radii to be 0, as having both equal to 0 would be equivalent to + // skipping SGR entirely. + assert(!(params->r[0] == 0 && params->r[1] == 0)); + assert(params->r[0] < AOMMIN(SGRPROJ_BORDER_VERT, SGRPROJ_BORDER_HORZ)); + assert(params->r[1] < AOMMIN(SGRPROJ_BORDER_VERT, SGRPROJ_BORDER_HORZ)); + + if (params->r[0] > 0) { + calc_ab_fast(A, B, C, D, width, height, buf_stride, bit_depth, + sgr_params_idx, 0); + final_filter_fast(flt0, flt_stride, A, B, buf_stride, dgd8, dgd_stride, + width, height, highbd); + } + + if (params->r[1] > 0) { + calc_ab(A, B, C, D, width, height, buf_stride, bit_depth, sgr_params_idx, + 1); + final_filter(flt1, flt_stride, A, B, buf_stride, dgd8, dgd_stride, width, + height, highbd); + } + aom_free(buf); + return 0; +} + +void apply_selfguided_restoration_avx2(const uint8_t *dat8, int width, + int height, int stride, int eps, + const int *xqd, uint8_t *dst8, + int dst_stride, int32_t *tmpbuf, + int bit_depth, int highbd) { + int32_t *flt0 = tmpbuf; + int32_t *flt1 = flt0 + RESTORATION_UNITPELS_MAX; + assert(width * height <= RESTORATION_UNITPELS_MAX); + const int ret = av1_selfguided_restoration_avx2( + dat8, width, height, stride, flt0, flt1, width, eps, bit_depth, highbd); + (void)ret; + assert(!ret); + const sgr_params_type *const params = &sgr_params[eps]; + int xq[2]; + decode_xq(xqd, xq, params); + + __m256i xq0 = _mm256_set1_epi32(xq[0]); + __m256i xq1 = _mm256_set1_epi32(xq[1]); + + for (int i = 0; i < height; ++i) { + // Calculate output in batches of 16 pixels + for (int j = 0; j < width; j += 16) { + const int k = i * width + j; + const int m = i * dst_stride + j; + + const uint8_t *dat8ij = dat8 + i * stride + j; + __m256i ep_0, ep_1; + __m128i src_0, src_1; + if (highbd) { + src_0 = xx_loadu_128(CONVERT_TO_SHORTPTR(dat8ij)); + src_1 = xx_loadu_128(CONVERT_TO_SHORTPTR(dat8ij + 8)); + ep_0 = _mm256_cvtepu16_epi32(src_0); + ep_1 = _mm256_cvtepu16_epi32(src_1); + } else { + src_0 = xx_loadu_128(dat8ij); + ep_0 = _mm256_cvtepu8_epi32(src_0); + ep_1 = _mm256_cvtepu8_epi32(_mm_srli_si128(src_0, 8)); + } + + const __m256i u_0 = _mm256_slli_epi32(ep_0, SGRPROJ_RST_BITS); + const __m256i u_1 = _mm256_slli_epi32(ep_1, SGRPROJ_RST_BITS); + + __m256i v_0 = _mm256_slli_epi32(u_0, SGRPROJ_PRJ_BITS); + __m256i v_1 = _mm256_slli_epi32(u_1, SGRPROJ_PRJ_BITS); + + if (params->r[0] > 0) { + const __m256i f1_0 = _mm256_sub_epi32(yy_loadu_256(&flt0[k]), u_0); + v_0 = _mm256_add_epi32(v_0, _mm256_mullo_epi32(xq0, f1_0)); + + const __m256i f1_1 = _mm256_sub_epi32(yy_loadu_256(&flt0[k + 8]), u_1); + v_1 = _mm256_add_epi32(v_1, _mm256_mullo_epi32(xq0, f1_1)); + } + + if (params->r[1] > 0) { + const __m256i f2_0 = _mm256_sub_epi32(yy_loadu_256(&flt1[k]), u_0); + v_0 = _mm256_add_epi32(v_0, _mm256_mullo_epi32(xq1, f2_0)); + + const __m256i f2_1 = _mm256_sub_epi32(yy_loadu_256(&flt1[k + 8]), u_1); + v_1 = _mm256_add_epi32(v_1, _mm256_mullo_epi32(xq1, f2_1)); + } + + const __m256i rounding = + round_for_shift(SGRPROJ_PRJ_BITS + SGRPROJ_RST_BITS); + const __m256i w_0 = _mm256_srai_epi32( + _mm256_add_epi32(v_0, rounding), SGRPROJ_PRJ_BITS + SGRPROJ_RST_BITS); + const __m256i w_1 = _mm256_srai_epi32( + _mm256_add_epi32(v_1, rounding), SGRPROJ_PRJ_BITS + SGRPROJ_RST_BITS); + + if (highbd) { + // Pack into 16 bits and clamp to [0, 2^bit_depth) + // Note that packing into 16 bits messes up the order of the bits, + // so we use a permute function to correct this + const __m256i tmp = _mm256_packus_epi32(w_0, w_1); + const __m256i tmp2 = _mm256_permute4x64_epi64(tmp, 0xd8); + const __m256i max = _mm256_set1_epi16((1 << bit_depth) - 1); + const __m256i res = _mm256_min_epi16(tmp2, max); + yy_storeu_256(CONVERT_TO_SHORTPTR(dst8 + m), res); + } else { + // Pack into 8 bits and clamp to [0, 256) + // Note that each pack messes up the order of the bits, + // so we use a permute function to correct this + const __m256i tmp = _mm256_packs_epi32(w_0, w_1); + const __m256i tmp2 = _mm256_permute4x64_epi64(tmp, 0xd8); + const __m256i res = + _mm256_packus_epi16(tmp2, tmp2 /* "don't care" value */); + const __m128i res2 = + _mm256_castsi256_si128(_mm256_permute4x64_epi64(res, 0xd8)); + xx_storeu_128(dst8 + m, res2); + } + } + } +} diff --git a/third_party/aom/av1/common/x86/selfguided_sse4.c b/third_party/aom/av1/common/x86/selfguided_sse4.c new file mode 100644 index 0000000000..ea3f6d9422 --- /dev/null +++ b/third_party/aom/av1/common/x86/selfguided_sse4.c @@ -0,0 +1,660 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <smmintrin.h> + +#include "config/aom_config.h" +#include "config/av1_rtcd.h" + +#include "av1/common/restoration.h" +#include "aom_dsp/x86/synonyms.h" + +// Load 4 bytes from the possibly-misaligned pointer p, extend each byte to +// 32-bit precision and return them in an SSE register. +static __m128i xx_load_extend_8_32(const void *p) { + return _mm_cvtepu8_epi32(xx_loadl_32(p)); +} + +// Load 4 halfwords from the possibly-misaligned pointer p, extend each +// halfword to 32-bit precision and return them in an SSE register. +static __m128i xx_load_extend_16_32(const void *p) { + return _mm_cvtepu16_epi32(xx_loadl_64(p)); +} + +// Compute the scan of an SSE register holding 4 32-bit integers. If the +// register holds x0..x3 then the scan will hold x0, x0+x1, x0+x1+x2, +// x0+x1+x2+x3 +static __m128i scan_32(__m128i x) { + const __m128i x01 = _mm_add_epi32(x, _mm_slli_si128(x, 4)); + return _mm_add_epi32(x01, _mm_slli_si128(x01, 8)); +} + +// Compute two integral images from src. B sums elements; A sums their +// squares. The images are offset by one pixel, so will have width and height +// equal to width + 1, height + 1 and the first row and column will be zero. +// +// A+1 and B+1 should be aligned to 16 bytes. buf_stride should be a multiple +// of 4. +static void integral_images(const uint8_t *src, int src_stride, int width, + int height, int32_t *A, int32_t *B, + int buf_stride) { + // Write out the zero top row + memset(A, 0, sizeof(*A) * (width + 1)); + memset(B, 0, sizeof(*B) * (width + 1)); + + const __m128i zero = _mm_setzero_si128(); + for (int i = 0; i < height; ++i) { + // Zero the left column. + A[(i + 1) * buf_stride] = B[(i + 1) * buf_stride] = 0; + + // ldiff is the difference H - D where H is the output sample immediately + // to the left and D is the output sample above it. These are scalars, + // replicated across the four lanes. + __m128i ldiff1 = zero, ldiff2 = zero; + for (int j = 0; j < width; j += 4) { + const int ABj = 1 + j; + + const __m128i above1 = xx_load_128(B + ABj + i * buf_stride); + const __m128i above2 = xx_load_128(A + ABj + i * buf_stride); + + const __m128i x1 = xx_load_extend_8_32(src + j + i * src_stride); + const __m128i x2 = _mm_madd_epi16(x1, x1); + + const __m128i sc1 = scan_32(x1); + const __m128i sc2 = scan_32(x2); + + const __m128i row1 = _mm_add_epi32(_mm_add_epi32(sc1, above1), ldiff1); + const __m128i row2 = _mm_add_epi32(_mm_add_epi32(sc2, above2), ldiff2); + + xx_store_128(B + ABj + (i + 1) * buf_stride, row1); + xx_store_128(A + ABj + (i + 1) * buf_stride, row2); + + // Calculate the new H - D. + ldiff1 = _mm_shuffle_epi32(_mm_sub_epi32(row1, above1), 0xff); + ldiff2 = _mm_shuffle_epi32(_mm_sub_epi32(row2, above2), 0xff); + } + } +} + +// Compute two integral images from src. B sums elements; A sums their squares +// +// A and B should be aligned to 16 bytes. buf_stride should be a multiple of 4. +static void integral_images_highbd(const uint16_t *src, int src_stride, + int width, int height, int32_t *A, + int32_t *B, int buf_stride) { + // Write out the zero top row + memset(A, 0, sizeof(*A) * (width + 1)); + memset(B, 0, sizeof(*B) * (width + 1)); + + const __m128i zero = _mm_setzero_si128(); + for (int i = 0; i < height; ++i) { + // Zero the left column. + A[(i + 1) * buf_stride] = B[(i + 1) * buf_stride] = 0; + + // ldiff is the difference H - D where H is the output sample immediately + // to the left and D is the output sample above it. These are scalars, + // replicated across the four lanes. + __m128i ldiff1 = zero, ldiff2 = zero; + for (int j = 0; j < width; j += 4) { + const int ABj = 1 + j; + + const __m128i above1 = xx_load_128(B + ABj + i * buf_stride); + const __m128i above2 = xx_load_128(A + ABj + i * buf_stride); + + const __m128i x1 = xx_load_extend_16_32(src + j + i * src_stride); + const __m128i x2 = _mm_madd_epi16(x1, x1); + + const __m128i sc1 = scan_32(x1); + const __m128i sc2 = scan_32(x2); + + const __m128i row1 = _mm_add_epi32(_mm_add_epi32(sc1, above1), ldiff1); + const __m128i row2 = _mm_add_epi32(_mm_add_epi32(sc2, above2), ldiff2); + + xx_store_128(B + ABj + (i + 1) * buf_stride, row1); + xx_store_128(A + ABj + (i + 1) * buf_stride, row2); + + // Calculate the new H - D. + ldiff1 = _mm_shuffle_epi32(_mm_sub_epi32(row1, above1), 0xff); + ldiff2 = _mm_shuffle_epi32(_mm_sub_epi32(row2, above2), 0xff); + } + } +} + +// Compute 4 values of boxsum from the given integral image. ii should point +// at the middle of the box (for the first value). r is the box radius. +static INLINE __m128i boxsum_from_ii(const int32_t *ii, int stride, int r) { + const __m128i tl = xx_loadu_128(ii - (r + 1) - (r + 1) * stride); + const __m128i tr = xx_loadu_128(ii + (r + 0) - (r + 1) * stride); + const __m128i bl = xx_loadu_128(ii - (r + 1) + r * stride); + const __m128i br = xx_loadu_128(ii + (r + 0) + r * stride); + const __m128i u = _mm_sub_epi32(tr, tl); + const __m128i v = _mm_sub_epi32(br, bl); + return _mm_sub_epi32(v, u); +} + +static __m128i round_for_shift(unsigned shift) { + return _mm_set1_epi32((1 << shift) >> 1); +} + +static __m128i compute_p(__m128i sum1, __m128i sum2, int bit_depth, int n) { + __m128i an, bb; + if (bit_depth > 8) { + const __m128i rounding_a = round_for_shift(2 * (bit_depth - 8)); + const __m128i rounding_b = round_for_shift(bit_depth - 8); + const __m128i shift_a = _mm_cvtsi32_si128(2 * (bit_depth - 8)); + const __m128i shift_b = _mm_cvtsi32_si128(bit_depth - 8); + const __m128i a = _mm_srl_epi32(_mm_add_epi32(sum2, rounding_a), shift_a); + const __m128i b = _mm_srl_epi32(_mm_add_epi32(sum1, rounding_b), shift_b); + // b < 2^14, so we can use a 16-bit madd rather than a 32-bit + // mullo to square it + bb = _mm_madd_epi16(b, b); + an = _mm_max_epi32(_mm_mullo_epi32(a, _mm_set1_epi32(n)), bb); + } else { + bb = _mm_madd_epi16(sum1, sum1); + an = _mm_mullo_epi32(sum2, _mm_set1_epi32(n)); + } + return _mm_sub_epi32(an, bb); +} + +// Assumes that C, D are integral images for the original buffer which has been +// extended to have a padding of SGRPROJ_BORDER_VERT/SGRPROJ_BORDER_HORZ pixels +// on the sides. A, B, C, D point at logical position (0, 0). +static void calc_ab(int32_t *A, int32_t *B, const int32_t *C, const int32_t *D, + int width, int height, int buf_stride, int bit_depth, + int sgr_params_idx, int radius_idx) { + const sgr_params_type *const params = &sgr_params[sgr_params_idx]; + const int r = params->r[radius_idx]; + const int n = (2 * r + 1) * (2 * r + 1); + const __m128i s = _mm_set1_epi32(params->s[radius_idx]); + // one_over_n[n-1] is 2^12/n, so easily fits in an int16 + const __m128i one_over_n = _mm_set1_epi32(one_by_x[n - 1]); + + const __m128i rnd_z = round_for_shift(SGRPROJ_MTABLE_BITS); + const __m128i rnd_res = round_for_shift(SGRPROJ_RECIP_BITS); + + // Set up masks + const __m128i ones32 = _mm_set_epi32(0, 0, 0xffffffff, 0xffffffff); + __m128i mask[4]; + for (int idx = 0; idx < 4; idx++) { + const __m128i shift = _mm_cvtsi32_si128(8 * (4 - idx)); + mask[idx] = _mm_cvtepi8_epi32(_mm_srl_epi64(ones32, shift)); + } + + for (int i = -1; i < height + 1; ++i) { + for (int j = -1; j < width + 1; j += 4) { + const int32_t *Cij = C + i * buf_stride + j; + const int32_t *Dij = D + i * buf_stride + j; + + __m128i sum1 = boxsum_from_ii(Dij, buf_stride, r); + __m128i sum2 = boxsum_from_ii(Cij, buf_stride, r); + + // When width + 2 isn't a multiple of 4, sum1 and sum2 will contain + // some uninitialised data in their upper words. We use a mask to + // ensure that these bits are set to 0. + int idx = AOMMIN(4, width + 1 - j); + assert(idx >= 1); + + if (idx < 4) { + sum1 = _mm_and_si128(mask[idx], sum1); + sum2 = _mm_and_si128(mask[idx], sum2); + } + + const __m128i p = compute_p(sum1, sum2, bit_depth, n); + + const __m128i z = _mm_min_epi32( + _mm_srli_epi32(_mm_add_epi32(_mm_mullo_epi32(p, s), rnd_z), + SGRPROJ_MTABLE_BITS), + _mm_set1_epi32(255)); + + // 'Gather' type instructions are not available pre-AVX2, so synthesize a + // gather using scalar loads. + const __m128i a_res = _mm_set_epi32(x_by_xplus1[_mm_extract_epi32(z, 3)], + x_by_xplus1[_mm_extract_epi32(z, 2)], + x_by_xplus1[_mm_extract_epi32(z, 1)], + x_by_xplus1[_mm_extract_epi32(z, 0)]); + + xx_storeu_128(A + i * buf_stride + j, a_res); + + const __m128i a_complement = + _mm_sub_epi32(_mm_set1_epi32(SGRPROJ_SGR), a_res); + + // sum1 might have lanes greater than 2^15, so we can't use madd to do + // multiplication involving sum1. However, a_complement and one_over_n + // are both less than 256, so we can multiply them first. + const __m128i a_comp_over_n = _mm_madd_epi16(a_complement, one_over_n); + const __m128i b_int = _mm_mullo_epi32(a_comp_over_n, sum1); + const __m128i b_res = + _mm_srli_epi32(_mm_add_epi32(b_int, rnd_res), SGRPROJ_RECIP_BITS); + + xx_storeu_128(B + i * buf_stride + j, b_res); + } + } +} + +// Calculate 4 values of the "cross sum" starting at buf. This is a 3x3 filter +// where the outer four corners have weight 3 and all other pixels have weight +// 4. +// +// Pixels are indexed like this: +// xtl xt xtr +// xl x xr +// xbl xb xbr +// +// buf points to x +// +// fours = xl + xt + xr + xb + x +// threes = xtl + xtr + xbr + xbl +// cross_sum = 4 * fours + 3 * threes +// = 4 * (fours + threes) - threes +// = (fours + threes) << 2 - threes +static INLINE __m128i cross_sum(const int32_t *buf, int stride) { + const __m128i xtl = xx_loadu_128(buf - 1 - stride); + const __m128i xt = xx_loadu_128(buf - stride); + const __m128i xtr = xx_loadu_128(buf + 1 - stride); + const __m128i xl = xx_loadu_128(buf - 1); + const __m128i x = xx_loadu_128(buf); + const __m128i xr = xx_loadu_128(buf + 1); + const __m128i xbl = xx_loadu_128(buf - 1 + stride); + const __m128i xb = xx_loadu_128(buf + stride); + const __m128i xbr = xx_loadu_128(buf + 1 + stride); + + const __m128i fours = _mm_add_epi32( + xl, _mm_add_epi32(xt, _mm_add_epi32(xr, _mm_add_epi32(xb, x)))); + const __m128i threes = + _mm_add_epi32(xtl, _mm_add_epi32(xtr, _mm_add_epi32(xbr, xbl))); + + return _mm_sub_epi32(_mm_slli_epi32(_mm_add_epi32(fours, threes), 2), threes); +} + +// The final filter for self-guided restoration. Computes a weighted average +// across A, B with "cross sums" (see cross_sum implementation above). +static void final_filter(int32_t *dst, int dst_stride, const int32_t *A, + const int32_t *B, int buf_stride, const void *dgd8, + int dgd_stride, int width, int height, int highbd) { + const int nb = 5; + const __m128i rounding = + round_for_shift(SGRPROJ_SGR_BITS + nb - SGRPROJ_RST_BITS); + const uint8_t *dgd_real = + highbd ? (const uint8_t *)CONVERT_TO_SHORTPTR(dgd8) : dgd8; + + for (int i = 0; i < height; ++i) { + for (int j = 0; j < width; j += 4) { + const __m128i a = cross_sum(A + i * buf_stride + j, buf_stride); + const __m128i b = cross_sum(B + i * buf_stride + j, buf_stride); + const __m128i raw = + xx_loadl_64(dgd_real + ((i * dgd_stride + j) << highbd)); + const __m128i src = + highbd ? _mm_cvtepu16_epi32(raw) : _mm_cvtepu8_epi32(raw); + + __m128i v = _mm_add_epi32(_mm_madd_epi16(a, src), b); + __m128i w = _mm_srai_epi32(_mm_add_epi32(v, rounding), + SGRPROJ_SGR_BITS + nb - SGRPROJ_RST_BITS); + + xx_storeu_128(dst + i * dst_stride + j, w); + } + } +} + +// Assumes that C, D are integral images for the original buffer which has been +// extended to have a padding of SGRPROJ_BORDER_VERT/SGRPROJ_BORDER_HORZ pixels +// on the sides. A, B, C, D point at logical position (0, 0). +static void calc_ab_fast(int32_t *A, int32_t *B, const int32_t *C, + const int32_t *D, int width, int height, + int buf_stride, int bit_depth, int sgr_params_idx, + int radius_idx) { + const sgr_params_type *const params = &sgr_params[sgr_params_idx]; + const int r = params->r[radius_idx]; + const int n = (2 * r + 1) * (2 * r + 1); + const __m128i s = _mm_set1_epi32(params->s[radius_idx]); + // one_over_n[n-1] is 2^12/n, so easily fits in an int16 + const __m128i one_over_n = _mm_set1_epi32(one_by_x[n - 1]); + + const __m128i rnd_z = round_for_shift(SGRPROJ_MTABLE_BITS); + const __m128i rnd_res = round_for_shift(SGRPROJ_RECIP_BITS); + + // Set up masks + const __m128i ones32 = _mm_set_epi32(0, 0, 0xffffffff, 0xffffffff); + __m128i mask[4]; + for (int idx = 0; idx < 4; idx++) { + const __m128i shift = _mm_cvtsi32_si128(8 * (4 - idx)); + mask[idx] = _mm_cvtepi8_epi32(_mm_srl_epi64(ones32, shift)); + } + + for (int i = -1; i < height + 1; i += 2) { + for (int j = -1; j < width + 1; j += 4) { + const int32_t *Cij = C + i * buf_stride + j; + const int32_t *Dij = D + i * buf_stride + j; + + __m128i sum1 = boxsum_from_ii(Dij, buf_stride, r); + __m128i sum2 = boxsum_from_ii(Cij, buf_stride, r); + + // When width + 2 isn't a multiple of 4, sum1 and sum2 will contain + // some uninitialised data in their upper words. We use a mask to + // ensure that these bits are set to 0. + int idx = AOMMIN(4, width + 1 - j); + assert(idx >= 1); + + if (idx < 4) { + sum1 = _mm_and_si128(mask[idx], sum1); + sum2 = _mm_and_si128(mask[idx], sum2); + } + + const __m128i p = compute_p(sum1, sum2, bit_depth, n); + + const __m128i z = _mm_min_epi32( + _mm_srli_epi32(_mm_add_epi32(_mm_mullo_epi32(p, s), rnd_z), + SGRPROJ_MTABLE_BITS), + _mm_set1_epi32(255)); + + // 'Gather' type instructions are not available pre-AVX2, so synthesize a + // gather using scalar loads. + const __m128i a_res = _mm_set_epi32(x_by_xplus1[_mm_extract_epi32(z, 3)], + x_by_xplus1[_mm_extract_epi32(z, 2)], + x_by_xplus1[_mm_extract_epi32(z, 1)], + x_by_xplus1[_mm_extract_epi32(z, 0)]); + + xx_storeu_128(A + i * buf_stride + j, a_res); + + const __m128i a_complement = + _mm_sub_epi32(_mm_set1_epi32(SGRPROJ_SGR), a_res); + + // sum1 might have lanes greater than 2^15, so we can't use madd to do + // multiplication involving sum1. However, a_complement and one_over_n + // are both less than 256, so we can multiply them first. + const __m128i a_comp_over_n = _mm_madd_epi16(a_complement, one_over_n); + const __m128i b_int = _mm_mullo_epi32(a_comp_over_n, sum1); + const __m128i b_res = + _mm_srli_epi32(_mm_add_epi32(b_int, rnd_res), SGRPROJ_RECIP_BITS); + + xx_storeu_128(B + i * buf_stride + j, b_res); + } + } +} + +// Calculate 4 values of the "cross sum" starting at buf. +// +// Pixels are indexed like this: +// xtl xt xtr +// - buf - +// xbl xb xbr +// +// Pixels are weighted like this: +// 5 6 5 +// 0 0 0 +// 5 6 5 +// +// fives = xtl + xtr + xbl + xbr +// sixes = xt + xb +// cross_sum = 6 * sixes + 5 * fives +// = 5 * (fives + sixes) - sixes +// = (fives + sixes) << 2 + (fives + sixes) + sixes +static INLINE __m128i cross_sum_fast_even_row(const int32_t *buf, int stride) { + const __m128i xtl = xx_loadu_128(buf - 1 - stride); + const __m128i xt = xx_loadu_128(buf - stride); + const __m128i xtr = xx_loadu_128(buf + 1 - stride); + const __m128i xbl = xx_loadu_128(buf - 1 + stride); + const __m128i xb = xx_loadu_128(buf + stride); + const __m128i xbr = xx_loadu_128(buf + 1 + stride); + + const __m128i fives = + _mm_add_epi32(xtl, _mm_add_epi32(xtr, _mm_add_epi32(xbr, xbl))); + const __m128i sixes = _mm_add_epi32(xt, xb); + const __m128i fives_plus_sixes = _mm_add_epi32(fives, sixes); + + return _mm_add_epi32( + _mm_add_epi32(_mm_slli_epi32(fives_plus_sixes, 2), fives_plus_sixes), + sixes); +} + +// Calculate 4 values of the "cross sum" starting at buf. +// +// Pixels are indexed like this: +// xl x xr +// +// Pixels are weighted like this: +// 5 6 5 +// +// buf points to x +// +// fives = xl + xr +// sixes = x +// cross_sum = 5 * fives + 6 * sixes +// = 4 * (fives + sixes) + (fives + sixes) + sixes +// = (fives + sixes) << 2 + (fives + sixes) + sixes +static INLINE __m128i cross_sum_fast_odd_row(const int32_t *buf) { + const __m128i xl = xx_loadu_128(buf - 1); + const __m128i x = xx_loadu_128(buf); + const __m128i xr = xx_loadu_128(buf + 1); + + const __m128i fives = _mm_add_epi32(xl, xr); + const __m128i sixes = x; + + const __m128i fives_plus_sixes = _mm_add_epi32(fives, sixes); + + return _mm_add_epi32( + _mm_add_epi32(_mm_slli_epi32(fives_plus_sixes, 2), fives_plus_sixes), + sixes); +} + +// The final filter for the self-guided restoration. Computes a +// weighted average across A, B with "cross sums" (see cross_sum_... +// implementations above). +static void final_filter_fast(int32_t *dst, int dst_stride, const int32_t *A, + const int32_t *B, int buf_stride, + const void *dgd8, int dgd_stride, int width, + int height, int highbd) { + const int nb0 = 5; + const int nb1 = 4; + + const __m128i rounding0 = + round_for_shift(SGRPROJ_SGR_BITS + nb0 - SGRPROJ_RST_BITS); + const __m128i rounding1 = + round_for_shift(SGRPROJ_SGR_BITS + nb1 - SGRPROJ_RST_BITS); + + const uint8_t *dgd_real = + highbd ? (const uint8_t *)CONVERT_TO_SHORTPTR(dgd8) : dgd8; + + for (int i = 0; i < height; ++i) { + if (!(i & 1)) { // even row + for (int j = 0; j < width; j += 4) { + const __m128i a = + cross_sum_fast_even_row(A + i * buf_stride + j, buf_stride); + const __m128i b = + cross_sum_fast_even_row(B + i * buf_stride + j, buf_stride); + const __m128i raw = + xx_loadl_64(dgd_real + ((i * dgd_stride + j) << highbd)); + const __m128i src = + highbd ? _mm_cvtepu16_epi32(raw) : _mm_cvtepu8_epi32(raw); + + __m128i v = _mm_add_epi32(_mm_madd_epi16(a, src), b); + __m128i w = _mm_srai_epi32(_mm_add_epi32(v, rounding0), + SGRPROJ_SGR_BITS + nb0 - SGRPROJ_RST_BITS); + + xx_storeu_128(dst + i * dst_stride + j, w); + } + } else { // odd row + for (int j = 0; j < width; j += 4) { + const __m128i a = cross_sum_fast_odd_row(A + i * buf_stride + j); + const __m128i b = cross_sum_fast_odd_row(B + i * buf_stride + j); + const __m128i raw = + xx_loadl_64(dgd_real + ((i * dgd_stride + j) << highbd)); + const __m128i src = + highbd ? _mm_cvtepu16_epi32(raw) : _mm_cvtepu8_epi32(raw); + + __m128i v = _mm_add_epi32(_mm_madd_epi16(a, src), b); + __m128i w = _mm_srai_epi32(_mm_add_epi32(v, rounding1), + SGRPROJ_SGR_BITS + nb1 - SGRPROJ_RST_BITS); + + xx_storeu_128(dst + i * dst_stride + j, w); + } + } + } +} + +int av1_selfguided_restoration_sse4_1(const uint8_t *dgd8, int width, + int height, int dgd_stride, int32_t *flt0, + int32_t *flt1, int flt_stride, + int sgr_params_idx, int bit_depth, + int highbd) { + int32_t *buf = (int32_t *)aom_memalign( + 16, 4 * sizeof(*buf) * RESTORATION_PROC_UNIT_PELS); + if (!buf) return -1; + memset(buf, 0, 4 * sizeof(*buf) * RESTORATION_PROC_UNIT_PELS); + + const int width_ext = width + 2 * SGRPROJ_BORDER_HORZ; + const int height_ext = height + 2 * SGRPROJ_BORDER_VERT; + + // Adjusting the stride of A and B here appears to avoid bad cache effects, + // leading to a significant speed improvement. + // We also align the stride to a multiple of 16 bytes for efficiency. + int buf_stride = ((width_ext + 3) & ~3) + 16; + + // The "tl" pointers point at the top-left of the initialised data for the + // array. Adding 3 here ensures that column 1 is 16-byte aligned. + int32_t *Atl = buf + 0 * RESTORATION_PROC_UNIT_PELS + 3; + int32_t *Btl = buf + 1 * RESTORATION_PROC_UNIT_PELS + 3; + int32_t *Ctl = buf + 2 * RESTORATION_PROC_UNIT_PELS + 3; + int32_t *Dtl = buf + 3 * RESTORATION_PROC_UNIT_PELS + 3; + + // The "0" pointers are (- SGRPROJ_BORDER_VERT, -SGRPROJ_BORDER_HORZ). Note + // there's a zero row and column in A, B (integral images), so we move down + // and right one for them. + const int buf_diag_border = + SGRPROJ_BORDER_HORZ + buf_stride * SGRPROJ_BORDER_VERT; + + int32_t *A0 = Atl + 1 + buf_stride; + int32_t *B0 = Btl + 1 + buf_stride; + int32_t *C0 = Ctl + 1 + buf_stride; + int32_t *D0 = Dtl + 1 + buf_stride; + + // Finally, A, B, C, D point at position (0, 0). + int32_t *A = A0 + buf_diag_border; + int32_t *B = B0 + buf_diag_border; + int32_t *C = C0 + buf_diag_border; + int32_t *D = D0 + buf_diag_border; + + const int dgd_diag_border = + SGRPROJ_BORDER_HORZ + dgd_stride * SGRPROJ_BORDER_VERT; + const uint8_t *dgd0 = dgd8 - dgd_diag_border; + + // Generate integral images from the input. C will contain sums of squares; D + // will contain just sums + if (highbd) + integral_images_highbd(CONVERT_TO_SHORTPTR(dgd0), dgd_stride, width_ext, + height_ext, Ctl, Dtl, buf_stride); + else + integral_images(dgd0, dgd_stride, width_ext, height_ext, Ctl, Dtl, + buf_stride); + + const sgr_params_type *const params = &sgr_params[sgr_params_idx]; + // Write to flt0 and flt1 + // If params->r == 0 we skip the corresponding filter. We only allow one of + // the radii to be 0, as having both equal to 0 would be equivalent to + // skipping SGR entirely. + assert(!(params->r[0] == 0 && params->r[1] == 0)); + assert(params->r[0] < AOMMIN(SGRPROJ_BORDER_VERT, SGRPROJ_BORDER_HORZ)); + assert(params->r[1] < AOMMIN(SGRPROJ_BORDER_VERT, SGRPROJ_BORDER_HORZ)); + + if (params->r[0] > 0) { + calc_ab_fast(A, B, C, D, width, height, buf_stride, bit_depth, + sgr_params_idx, 0); + final_filter_fast(flt0, flt_stride, A, B, buf_stride, dgd8, dgd_stride, + width, height, highbd); + } + + if (params->r[1] > 0) { + calc_ab(A, B, C, D, width, height, buf_stride, bit_depth, sgr_params_idx, + 1); + final_filter(flt1, flt_stride, A, B, buf_stride, dgd8, dgd_stride, width, + height, highbd); + } + aom_free(buf); + return 0; +} + +void apply_selfguided_restoration_sse4_1(const uint8_t *dat8, int width, + int height, int stride, int eps, + const int *xqd, uint8_t *dst8, + int dst_stride, int32_t *tmpbuf, + int bit_depth, int highbd) { + int32_t *flt0 = tmpbuf; + int32_t *flt1 = flt0 + RESTORATION_UNITPELS_MAX; + assert(width * height <= RESTORATION_UNITPELS_MAX); + const int ret = av1_selfguided_restoration_sse4_1( + dat8, width, height, stride, flt0, flt1, width, eps, bit_depth, highbd); + (void)ret; + assert(!ret); + const sgr_params_type *const params = &sgr_params[eps]; + int xq[2]; + decode_xq(xqd, xq, params); + + __m128i xq0 = _mm_set1_epi32(xq[0]); + __m128i xq1 = _mm_set1_epi32(xq[1]); + + for (int i = 0; i < height; ++i) { + // Calculate output in batches of 8 pixels + for (int j = 0; j < width; j += 8) { + const int k = i * width + j; + const int m = i * dst_stride + j; + + const uint8_t *dat8ij = dat8 + i * stride + j; + __m128i src; + if (highbd) { + src = xx_loadu_128(CONVERT_TO_SHORTPTR(dat8ij)); + } else { + src = _mm_cvtepu8_epi16(xx_loadl_64(dat8ij)); + } + + const __m128i u = _mm_slli_epi16(src, SGRPROJ_RST_BITS); + const __m128i u_0 = _mm_cvtepu16_epi32(u); + const __m128i u_1 = _mm_cvtepu16_epi32(_mm_srli_si128(u, 8)); + + __m128i v_0 = _mm_slli_epi32(u_0, SGRPROJ_PRJ_BITS); + __m128i v_1 = _mm_slli_epi32(u_1, SGRPROJ_PRJ_BITS); + + if (params->r[0] > 0) { + const __m128i f1_0 = _mm_sub_epi32(xx_loadu_128(&flt0[k]), u_0); + v_0 = _mm_add_epi32(v_0, _mm_mullo_epi32(xq0, f1_0)); + + const __m128i f1_1 = _mm_sub_epi32(xx_loadu_128(&flt0[k + 4]), u_1); + v_1 = _mm_add_epi32(v_1, _mm_mullo_epi32(xq0, f1_1)); + } + + if (params->r[1] > 0) { + const __m128i f2_0 = _mm_sub_epi32(xx_loadu_128(&flt1[k]), u_0); + v_0 = _mm_add_epi32(v_0, _mm_mullo_epi32(xq1, f2_0)); + + const __m128i f2_1 = _mm_sub_epi32(xx_loadu_128(&flt1[k + 4]), u_1); + v_1 = _mm_add_epi32(v_1, _mm_mullo_epi32(xq1, f2_1)); + } + + const __m128i rounding = + round_for_shift(SGRPROJ_PRJ_BITS + SGRPROJ_RST_BITS); + const __m128i w_0 = _mm_srai_epi32(_mm_add_epi32(v_0, rounding), + SGRPROJ_PRJ_BITS + SGRPROJ_RST_BITS); + const __m128i w_1 = _mm_srai_epi32(_mm_add_epi32(v_1, rounding), + SGRPROJ_PRJ_BITS + SGRPROJ_RST_BITS); + + if (highbd) { + // Pack into 16 bits and clamp to [0, 2^bit_depth) + const __m128i tmp = _mm_packus_epi32(w_0, w_1); + const __m128i max = _mm_set1_epi16((1 << bit_depth) - 1); + const __m128i res = _mm_min_epi16(tmp, max); + xx_storeu_128(CONVERT_TO_SHORTPTR(dst8 + m), res); + } else { + // Pack into 8 bits and clamp to [0, 256) + const __m128i tmp = _mm_packs_epi32(w_0, w_1); + const __m128i res = _mm_packus_epi16(tmp, tmp /* "don't care" value */); + xx_storel_64(dst8 + m, res); + } + } + } +} diff --git a/third_party/aom/av1/common/x86/warp_plane_sse4.c b/third_party/aom/av1/common/x86/warp_plane_sse4.c new file mode 100644 index 0000000000..b810cea2e5 --- /dev/null +++ b/third_party/aom/av1/common/x86/warp_plane_sse4.c @@ -0,0 +1,942 @@ +/* + * Copyright (c) 2017, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <emmintrin.h> +#include <smmintrin.h> + +#include "config/av1_rtcd.h" + +#include "av1/common/warped_motion.h" + +/* This is a modified version of 'warped_filter' from warped_motion.c: + * Each coefficient is stored in 8 bits instead of 16 bits + * The coefficients are rearranged in the column order 0, 2, 4, 6, 1, 3, 5, 7 + + This is done in order to avoid overflow: Since the tap with the largest + coefficient could be any of taps 2, 3, 4 or 5, we can't use the summation + order ((0 + 1) + (4 + 5)) + ((2 + 3) + (6 + 7)) used in the regular + convolve functions. + + Instead, we use the summation order + ((0 + 2) + (4 + 6)) + ((1 + 3) + (5 + 7)). + The rearrangement of coefficients in this table is so that we can get the + coefficients into the correct order more quickly. +*/ +/* clang-format off */ +DECLARE_ALIGNED(8, static const int8_t, + filter_8bit[WARPEDPIXEL_PREC_SHIFTS * 3 + 1][8]) = { +#if WARPEDPIXEL_PREC_BITS == 6 + // [-1, 0) + { 0, 127, 0, 0, 0, 1, 0, 0}, { 0, 127, 0, 0, -1, 2, 0, 0}, + { 1, 127, -1, 0, -3, 4, 0, 0}, { 1, 126, -2, 0, -4, 6, 1, 0}, + { 1, 126, -3, 0, -5, 8, 1, 0}, { 1, 125, -4, 0, -6, 11, 1, 0}, + { 1, 124, -4, 0, -7, 13, 1, 0}, { 2, 123, -5, 0, -8, 15, 1, 0}, + { 2, 122, -6, 0, -9, 18, 1, 0}, { 2, 121, -6, 0, -10, 20, 1, 0}, + { 2, 120, -7, 0, -11, 22, 2, 0}, { 2, 119, -8, 0, -12, 25, 2, 0}, + { 3, 117, -8, 0, -13, 27, 2, 0}, { 3, 116, -9, 0, -13, 29, 2, 0}, + { 3, 114, -10, 0, -14, 32, 3, 0}, { 3, 113, -10, 0, -15, 35, 2, 0}, + { 3, 111, -11, 0, -15, 37, 3, 0}, { 3, 109, -11, 0, -16, 40, 3, 0}, + { 3, 108, -12, 0, -16, 42, 3, 0}, { 4, 106, -13, 0, -17, 45, 3, 0}, + { 4, 104, -13, 0, -17, 47, 3, 0}, { 4, 102, -14, 0, -17, 50, 3, 0}, + { 4, 100, -14, 0, -17, 52, 3, 0}, { 4, 98, -15, 0, -18, 55, 4, 0}, + { 4, 96, -15, 0, -18, 58, 3, 0}, { 4, 94, -16, 0, -18, 60, 4, 0}, + { 4, 91, -16, 0, -18, 63, 4, 0}, { 4, 89, -16, 0, -18, 65, 4, 0}, + { 4, 87, -17, 0, -18, 68, 4, 0}, { 4, 85, -17, 0, -18, 70, 4, 0}, + { 4, 82, -17, 0, -18, 73, 4, 0}, { 4, 80, -17, 0, -18, 75, 4, 0}, + { 4, 78, -18, 0, -18, 78, 4, 0}, { 4, 75, -18, 0, -17, 80, 4, 0}, + { 4, 73, -18, 0, -17, 82, 4, 0}, { 4, 70, -18, 0, -17, 85, 4, 0}, + { 4, 68, -18, 0, -17, 87, 4, 0}, { 4, 65, -18, 0, -16, 89, 4, 0}, + { 4, 63, -18, 0, -16, 91, 4, 0}, { 4, 60, -18, 0, -16, 94, 4, 0}, + { 3, 58, -18, 0, -15, 96, 4, 0}, { 4, 55, -18, 0, -15, 98, 4, 0}, + { 3, 52, -17, 0, -14, 100, 4, 0}, { 3, 50, -17, 0, -14, 102, 4, 0}, + { 3, 47, -17, 0, -13, 104, 4, 0}, { 3, 45, -17, 0, -13, 106, 4, 0}, + { 3, 42, -16, 0, -12, 108, 3, 0}, { 3, 40, -16, 0, -11, 109, 3, 0}, + { 3, 37, -15, 0, -11, 111, 3, 0}, { 2, 35, -15, 0, -10, 113, 3, 0}, + { 3, 32, -14, 0, -10, 114, 3, 0}, { 2, 29, -13, 0, -9, 116, 3, 0}, + { 2, 27, -13, 0, -8, 117, 3, 0}, { 2, 25, -12, 0, -8, 119, 2, 0}, + { 2, 22, -11, 0, -7, 120, 2, 0}, { 1, 20, -10, 0, -6, 121, 2, 0}, + { 1, 18, -9, 0, -6, 122, 2, 0}, { 1, 15, -8, 0, -5, 123, 2, 0}, + { 1, 13, -7, 0, -4, 124, 1, 0}, { 1, 11, -6, 0, -4, 125, 1, 0}, + { 1, 8, -5, 0, -3, 126, 1, 0}, { 1, 6, -4, 0, -2, 126, 1, 0}, + { 0, 4, -3, 0, -1, 127, 1, 0}, { 0, 2, -1, 0, 0, 127, 0, 0}, + // [0, 1) + { 0, 0, 1, 0, 0, 127, 0, 0}, { 0, -1, 2, 0, 0, 127, 0, 0}, + { 0, -3, 4, 1, 1, 127, -2, 0}, { 0, -5, 6, 1, 1, 127, -2, 0}, + { 0, -6, 8, 1, 2, 126, -3, 0}, {-1, -7, 11, 2, 2, 126, -4, -1}, + {-1, -8, 13, 2, 3, 125, -5, -1}, {-1, -10, 16, 3, 3, 124, -6, -1}, + {-1, -11, 18, 3, 4, 123, -7, -1}, {-1, -12, 20, 3, 4, 122, -7, -1}, + {-1, -13, 23, 3, 4, 121, -8, -1}, {-2, -14, 25, 4, 5, 120, -9, -1}, + {-1, -15, 27, 4, 5, 119, -10, -1}, {-1, -16, 30, 4, 5, 118, -11, -1}, + {-2, -17, 33, 5, 6, 116, -12, -1}, {-2, -17, 35, 5, 6, 114, -12, -1}, + {-2, -18, 38, 5, 6, 113, -13, -1}, {-2, -19, 41, 6, 7, 111, -14, -2}, + {-2, -19, 43, 6, 7, 110, -15, -2}, {-2, -20, 46, 6, 7, 108, -15, -2}, + {-2, -20, 49, 6, 7, 106, -16, -2}, {-2, -21, 51, 7, 7, 104, -16, -2}, + {-2, -21, 54, 7, 7, 102, -17, -2}, {-2, -21, 56, 7, 8, 100, -18, -2}, + {-2, -22, 59, 7, 8, 98, -18, -2}, {-2, -22, 62, 7, 8, 96, -19, -2}, + {-2, -22, 64, 7, 8, 94, -19, -2}, {-2, -22, 67, 8, 8, 91, -20, -2}, + {-2, -22, 69, 8, 8, 89, -20, -2}, {-2, -22, 72, 8, 8, 87, -21, -2}, + {-2, -21, 74, 8, 8, 84, -21, -2}, {-2, -22, 77, 8, 8, 82, -21, -2}, + {-2, -21, 79, 8, 8, 79, -21, -2}, {-2, -21, 82, 8, 8, 77, -22, -2}, + {-2, -21, 84, 8, 8, 74, -21, -2}, {-2, -21, 87, 8, 8, 72, -22, -2}, + {-2, -20, 89, 8, 8, 69, -22, -2}, {-2, -20, 91, 8, 8, 67, -22, -2}, + {-2, -19, 94, 8, 7, 64, -22, -2}, {-2, -19, 96, 8, 7, 62, -22, -2}, + {-2, -18, 98, 8, 7, 59, -22, -2}, {-2, -18, 100, 8, 7, 56, -21, -2}, + {-2, -17, 102, 7, 7, 54, -21, -2}, {-2, -16, 104, 7, 7, 51, -21, -2}, + {-2, -16, 106, 7, 6, 49, -20, -2}, {-2, -15, 108, 7, 6, 46, -20, -2}, + {-2, -15, 110, 7, 6, 43, -19, -2}, {-2, -14, 111, 7, 6, 41, -19, -2}, + {-1, -13, 113, 6, 5, 38, -18, -2}, {-1, -12, 114, 6, 5, 35, -17, -2}, + {-1, -12, 116, 6, 5, 33, -17, -2}, {-1, -11, 118, 5, 4, 30, -16, -1}, + {-1, -10, 119, 5, 4, 27, -15, -1}, {-1, -9, 120, 5, 4, 25, -14, -2}, + {-1, -8, 121, 4, 3, 23, -13, -1}, {-1, -7, 122, 4, 3, 20, -12, -1}, + {-1, -7, 123, 4, 3, 18, -11, -1}, {-1, -6, 124, 3, 3, 16, -10, -1}, + {-1, -5, 125, 3, 2, 13, -8, -1}, {-1, -4, 126, 2, 2, 11, -7, -1}, + { 0, -3, 126, 2, 1, 8, -6, 0}, { 0, -2, 127, 1, 1, 6, -5, 0}, + { 0, -2, 127, 1, 1, 4, -3, 0}, { 0, 0, 127, 0, 0, 2, -1, 0}, + // [1, 2) + { 0, 0, 127, 0, 0, 1, 0, 0}, { 0, 0, 127, 0, 0, -1, 2, 0}, + { 0, 1, 127, -1, 0, -3, 4, 0}, { 0, 1, 126, -2, 0, -4, 6, 1}, + { 0, 1, 126, -3, 0, -5, 8, 1}, { 0, 1, 125, -4, 0, -6, 11, 1}, + { 0, 1, 124, -4, 0, -7, 13, 1}, { 0, 2, 123, -5, 0, -8, 15, 1}, + { 0, 2, 122, -6, 0, -9, 18, 1}, { 0, 2, 121, -6, 0, -10, 20, 1}, + { 0, 2, 120, -7, 0, -11, 22, 2}, { 0, 2, 119, -8, 0, -12, 25, 2}, + { 0, 3, 117, -8, 0, -13, 27, 2}, { 0, 3, 116, -9, 0, -13, 29, 2}, + { 0, 3, 114, -10, 0, -14, 32, 3}, { 0, 3, 113, -10, 0, -15, 35, 2}, + { 0, 3, 111, -11, 0, -15, 37, 3}, { 0, 3, 109, -11, 0, -16, 40, 3}, + { 0, 3, 108, -12, 0, -16, 42, 3}, { 0, 4, 106, -13, 0, -17, 45, 3}, + { 0, 4, 104, -13, 0, -17, 47, 3}, { 0, 4, 102, -14, 0, -17, 50, 3}, + { 0, 4, 100, -14, 0, -17, 52, 3}, { 0, 4, 98, -15, 0, -18, 55, 4}, + { 0, 4, 96, -15, 0, -18, 58, 3}, { 0, 4, 94, -16, 0, -18, 60, 4}, + { 0, 4, 91, -16, 0, -18, 63, 4}, { 0, 4, 89, -16, 0, -18, 65, 4}, + { 0, 4, 87, -17, 0, -18, 68, 4}, { 0, 4, 85, -17, 0, -18, 70, 4}, + { 0, 4, 82, -17, 0, -18, 73, 4}, { 0, 4, 80, -17, 0, -18, 75, 4}, + { 0, 4, 78, -18, 0, -18, 78, 4}, { 0, 4, 75, -18, 0, -17, 80, 4}, + { 0, 4, 73, -18, 0, -17, 82, 4}, { 0, 4, 70, -18, 0, -17, 85, 4}, + { 0, 4, 68, -18, 0, -17, 87, 4}, { 0, 4, 65, -18, 0, -16, 89, 4}, + { 0, 4, 63, -18, 0, -16, 91, 4}, { 0, 4, 60, -18, 0, -16, 94, 4}, + { 0, 3, 58, -18, 0, -15, 96, 4}, { 0, 4, 55, -18, 0, -15, 98, 4}, + { 0, 3, 52, -17, 0, -14, 100, 4}, { 0, 3, 50, -17, 0, -14, 102, 4}, + { 0, 3, 47, -17, 0, -13, 104, 4}, { 0, 3, 45, -17, 0, -13, 106, 4}, + { 0, 3, 42, -16, 0, -12, 108, 3}, { 0, 3, 40, -16, 0, -11, 109, 3}, + { 0, 3, 37, -15, 0, -11, 111, 3}, { 0, 2, 35, -15, 0, -10, 113, 3}, + { 0, 3, 32, -14, 0, -10, 114, 3}, { 0, 2, 29, -13, 0, -9, 116, 3}, + { 0, 2, 27, -13, 0, -8, 117, 3}, { 0, 2, 25, -12, 0, -8, 119, 2}, + { 0, 2, 22, -11, 0, -7, 120, 2}, { 0, 1, 20, -10, 0, -6, 121, 2}, + { 0, 1, 18, -9, 0, -6, 122, 2}, { 0, 1, 15, -8, 0, -5, 123, 2}, + { 0, 1, 13, -7, 0, -4, 124, 1}, { 0, 1, 11, -6, 0, -4, 125, 1}, + { 0, 1, 8, -5, 0, -3, 126, 1}, { 0, 1, 6, -4, 0, -2, 126, 1}, + { 0, 0, 4, -3, 0, -1, 127, 1}, { 0, 0, 2, -1, 0, 0, 127, 0}, + // dummy (replicate row index 191) + { 0, 0, 2, -1, 0, 0, 127, 0}, + +#else + // [-1, 0) + { 0, 127, 0, 0, 0, 1, 0, 0}, { 1, 127, -1, 0, -3, 4, 0, 0}, + { 1, 126, -3, 0, -5, 8, 1, 0}, { 1, 124, -4, 0, -7, 13, 1, 0}, + { 2, 122, -6, 0, -9, 18, 1, 0}, { 2, 120, -7, 0, -11, 22, 2, 0}, + { 3, 117, -8, 0, -13, 27, 2, 0}, { 3, 114, -10, 0, -14, 32, 3, 0}, + { 3, 111, -11, 0, -15, 37, 3, 0}, { 3, 108, -12, 0, -16, 42, 3, 0}, + { 4, 104, -13, 0, -17, 47, 3, 0}, { 4, 100, -14, 0, -17, 52, 3, 0}, + { 4, 96, -15, 0, -18, 58, 3, 0}, { 4, 91, -16, 0, -18, 63, 4, 0}, + { 4, 87, -17, 0, -18, 68, 4, 0}, { 4, 82, -17, 0, -18, 73, 4, 0}, + { 4, 78, -18, 0, -18, 78, 4, 0}, { 4, 73, -18, 0, -17, 82, 4, 0}, + { 4, 68, -18, 0, -17, 87, 4, 0}, { 4, 63, -18, 0, -16, 91, 4, 0}, + { 3, 58, -18, 0, -15, 96, 4, 0}, { 3, 52, -17, 0, -14, 100, 4, 0}, + { 3, 47, -17, 0, -13, 104, 4, 0}, { 3, 42, -16, 0, -12, 108, 3, 0}, + { 3, 37, -15, 0, -11, 111, 3, 0}, { 3, 32, -14, 0, -10, 114, 3, 0}, + { 2, 27, -13, 0, -8, 117, 3, 0}, { 2, 22, -11, 0, -7, 120, 2, 0}, + { 1, 18, -9, 0, -6, 122, 2, 0}, { 1, 13, -7, 0, -4, 124, 1, 0}, + { 1, 8, -5, 0, -3, 126, 1, 0}, { 0, 4, -3, 0, -1, 127, 1, 0}, + // [0, 1) + { 0, 0, 1, 0, 0, 127, 0, 0}, { 0, -3, 4, 1, 1, 127, -2, 0}, + { 0, -6, 8, 1, 2, 126, -3, 0}, {-1, -8, 13, 2, 3, 125, -5, -1}, + {-1, -11, 18, 3, 4, 123, -7, -1}, {-1, -13, 23, 3, 4, 121, -8, -1}, + {-1, -15, 27, 4, 5, 119, -10, -1}, {-2, -17, 33, 5, 6, 116, -12, -1}, + {-2, -18, 38, 5, 6, 113, -13, -1}, {-2, -19, 43, 6, 7, 110, -15, -2}, + {-2, -20, 49, 6, 7, 106, -16, -2}, {-2, -21, 54, 7, 7, 102, -17, -2}, + {-2, -22, 59, 7, 8, 98, -18, -2}, {-2, -22, 64, 7, 8, 94, -19, -2}, + {-2, -22, 69, 8, 8, 89, -20, -2}, {-2, -21, 74, 8, 8, 84, -21, -2}, + {-2, -21, 79, 8, 8, 79, -21, -2}, {-2, -21, 84, 8, 8, 74, -21, -2}, + {-2, -20, 89, 8, 8, 69, -22, -2}, {-2, -19, 94, 8, 7, 64, -22, -2}, + {-2, -18, 98, 8, 7, 59, -22, -2}, {-2, -17, 102, 7, 7, 54, -21, -2}, + {-2, -16, 106, 7, 6, 49, -20, -2}, {-2, -15, 110, 7, 6, 43, -19, -2}, + {-1, -13, 113, 6, 5, 38, -18, -2}, {-1, -12, 116, 6, 5, 33, -17, -2}, + {-1, -10, 119, 5, 4, 27, -15, -1}, {-1, -8, 121, 4, 3, 23, -13, -1}, + {-1, -7, 123, 4, 3, 18, -11, -1}, {-1, -5, 125, 3, 2, 13, -8, -1}, + { 0, -3, 126, 2, 1, 8, -6, 0}, { 0, -2, 127, 1, 1, 4, -3, 0}, + // [1, 2) + { 0, 0, 127, 0, 0, 1, 0, 0}, { 0, 1, 127, -1, 0, -3, 4, 0}, + { 0, 1, 126, -3, 0, -5, 8, 1}, { 0, 1, 124, -4, 0, -7, 13, 1}, + { 0, 2, 122, -6, 0, -9, 18, 1}, { 0, 2, 120, -7, 0, -11, 22, 2}, + { 0, 3, 117, -8, 0, -13, 27, 2}, { 0, 3, 114, -10, 0, -14, 32, 3}, + { 0, 3, 111, -11, 0, -15, 37, 3}, { 0, 3, 108, -12, 0, -16, 42, 3}, + { 0, 4, 104, -13, 0, -17, 47, 3}, { 0, 4, 100, -14, 0, -17, 52, 3}, + { 0, 4, 96, -15, 0, -18, 58, 3}, { 0, 4, 91, -16, 0, -18, 63, 4}, + { 0, 4, 87, -17, 0, -18, 68, 4}, { 0, 4, 82, -17, 0, -18, 73, 4}, + { 0, 4, 78, -18, 0, -18, 78, 4}, { 0, 4, 73, -18, 0, -17, 82, 4}, + { 0, 4, 68, -18, 0, -17, 87, 4}, { 0, 4, 63, -18, 0, -16, 91, 4}, + { 0, 3, 58, -18, 0, -15, 96, 4}, { 0, 3, 52, -17, 0, -14, 100, 4}, + { 0, 3, 47, -17, 0, -13, 104, 4}, { 0, 3, 42, -16, 0, -12, 108, 3}, + { 0, 3, 37, -15, 0, -11, 111, 3}, { 0, 3, 32, -14, 0, -10, 114, 3}, + { 0, 2, 27, -13, 0, -8, 117, 3}, { 0, 2, 22, -11, 0, -7, 120, 2}, + { 0, 1, 18, -9, 0, -6, 122, 2}, { 0, 1, 13, -7, 0, -4, 124, 1}, + { 0, 1, 8, -5, 0, -3, 126, 1}, { 0, 0, 4, -3, 0, -1, 127, 1}, + // dummy (replicate row index 95) + { 0, 0, 4, -3, 0, -1, 127, 1}, +#endif // WARPEDPIXEL_PREC_BITS == 6 +}; +/* clang-format on */ + +// Shuffle masks: we want to convert a sequence of bytes 0, 1, 2, ..., 15 +// in an SSE register into two sequences: +// 0, 2, 2, 4, ..., 12, 12, 14, <don't care> +// 1, 3, 3, 5, ..., 13, 13, 15, <don't care> +static const uint8_t even_mask[16] = { 0, 2, 2, 4, 4, 6, 6, 8, + 8, 10, 10, 12, 12, 14, 14, 0 }; +static const uint8_t odd_mask[16] = { 1, 3, 3, 5, 5, 7, 7, 9, + 9, 11, 11, 13, 13, 15, 15, 0 }; + +static const uint8_t shuffle_alpha0_mask01[16] = { 0, 1, 0, 1, 0, 1, 0, 1, + 0, 1, 0, 1, 0, 1, 0, 1 }; + +static const uint8_t shuffle_alpha0_mask23[16] = { 2, 3, 2, 3, 2, 3, 2, 3, + 2, 3, 2, 3, 2, 3, 2, 3 }; + +static const uint8_t shuffle_alpha0_mask45[16] = { 4, 5, 4, 5, 4, 5, 4, 5, + 4, 5, 4, 5, 4, 5, 4, 5 }; + +static const uint8_t shuffle_alpha0_mask67[16] = { 6, 7, 6, 7, 6, 7, 6, 7, + 6, 7, 6, 7, 6, 7, 6, 7 }; + +static const uint8_t shuffle_gamma0_mask0[16] = { 0, 1, 2, 3, 0, 1, 2, 3, + 0, 1, 2, 3, 0, 1, 2, 3 }; +static const uint8_t shuffle_gamma0_mask1[16] = { 4, 5, 6, 7, 4, 5, 6, 7, + 4, 5, 6, 7, 4, 5, 6, 7 }; +static const uint8_t shuffle_gamma0_mask2[16] = { 8, 9, 10, 11, 8, 9, 10, 11, + 8, 9, 10, 11, 8, 9, 10, 11 }; +static const uint8_t shuffle_gamma0_mask3[16] = { + 12, 13, 14, 15, 12, 13, 14, 15, 12, 13, 14, 15, 12, 13, 14, 15 +}; + +static INLINE void filter_src_pixels(__m128i src, __m128i *tmp, __m128i *coeff, + const int offset_bits_horiz, + const int reduce_bits_horiz, int k) { + const __m128i src_even = + _mm_shuffle_epi8(src, _mm_loadu_si128((__m128i *)even_mask)); + const __m128i src_odd = + _mm_shuffle_epi8(src, _mm_loadu_si128((__m128i *)odd_mask)); + // The pixel order we need for 'src' is: + // 0 2 2 4 4 6 6 8 1 3 3 5 5 7 7 9 + const __m128i src_02 = _mm_unpacklo_epi64(src_even, src_odd); + const __m128i res_02 = _mm_maddubs_epi16(src_02, coeff[0]); + // 4 6 6 8 8 10 10 12 5 7 7 9 9 11 11 13 + const __m128i src_46 = _mm_unpacklo_epi64(_mm_srli_si128(src_even, 4), + _mm_srli_si128(src_odd, 4)); + const __m128i res_46 = _mm_maddubs_epi16(src_46, coeff[1]); + // 1 3 3 5 5 7 7 9 2 4 4 6 6 8 8 10 + const __m128i src_13 = + _mm_unpacklo_epi64(src_odd, _mm_srli_si128(src_even, 2)); + const __m128i res_13 = _mm_maddubs_epi16(src_13, coeff[2]); + // 5 7 7 9 9 11 11 13 6 8 8 10 10 12 12 14 + const __m128i src_57 = _mm_unpacklo_epi64(_mm_srli_si128(src_odd, 4), + _mm_srli_si128(src_even, 6)); + const __m128i res_57 = _mm_maddubs_epi16(src_57, coeff[3]); + + const __m128i round_const = _mm_set1_epi16((1 << offset_bits_horiz) + + ((1 << reduce_bits_horiz) >> 1)); + + // Note: The values res_02 + res_46 and res_13 + res_57 both + // fit into int16s at this point, but their sum may be too wide to fit + // into an int16. However, once we also add round_const, the sum of + // all of these fits into a uint16. + // + // The wrapping behaviour of _mm_add_* is used here to make sure we + // get the correct result despite converting between different + // (implicit) types. + const __m128i res_even = _mm_add_epi16(res_02, res_46); + const __m128i res_odd = _mm_add_epi16(res_13, res_57); + const __m128i res = + _mm_add_epi16(_mm_add_epi16(res_even, res_odd), round_const); + tmp[k + 7] = _mm_srl_epi16(res, _mm_cvtsi32_si128(reduce_bits_horiz)); +} + +static INLINE void prepare_horizontal_filter_coeff(int alpha, int sx, + __m128i *coeff) { + // Filter even-index pixels + const __m128i tmp_0 = _mm_loadl_epi64( + (__m128i *)&filter_8bit[(sx + 0 * alpha) >> WARPEDDIFF_PREC_BITS]); + const __m128i tmp_1 = _mm_loadl_epi64( + (__m128i *)&filter_8bit[(sx + 1 * alpha) >> WARPEDDIFF_PREC_BITS]); + const __m128i tmp_2 = _mm_loadl_epi64( + (__m128i *)&filter_8bit[(sx + 2 * alpha) >> WARPEDDIFF_PREC_BITS]); + const __m128i tmp_3 = _mm_loadl_epi64( + (__m128i *)&filter_8bit[(sx + 3 * alpha) >> WARPEDDIFF_PREC_BITS]); + const __m128i tmp_4 = _mm_loadl_epi64( + (__m128i *)&filter_8bit[(sx + 4 * alpha) >> WARPEDDIFF_PREC_BITS]); + const __m128i tmp_5 = _mm_loadl_epi64( + (__m128i *)&filter_8bit[(sx + 5 * alpha) >> WARPEDDIFF_PREC_BITS]); + const __m128i tmp_6 = _mm_loadl_epi64( + (__m128i *)&filter_8bit[(sx + 6 * alpha) >> WARPEDDIFF_PREC_BITS]); + const __m128i tmp_7 = _mm_loadl_epi64( + (__m128i *)&filter_8bit[(sx + 7 * alpha) >> WARPEDDIFF_PREC_BITS]); + + // Coeffs 0 2 0 2 4 6 4 6 1 3 1 3 5 7 5 7 for pixels 0 2 + const __m128i tmp_8 = _mm_unpacklo_epi16(tmp_0, tmp_2); + // Coeffs 0 2 0 2 4 6 4 6 1 3 1 3 5 7 5 7 for pixels 1 3 + const __m128i tmp_9 = _mm_unpacklo_epi16(tmp_1, tmp_3); + // Coeffs 0 2 0 2 4 6 4 6 1 3 1 3 5 7 5 7 for pixels 4 6 + const __m128i tmp_10 = _mm_unpacklo_epi16(tmp_4, tmp_6); + // Coeffs 0 2 0 2 4 6 4 6 1 3 1 3 5 7 5 7 for pixels 5 7 + const __m128i tmp_11 = _mm_unpacklo_epi16(tmp_5, tmp_7); + + // Coeffs 0 2 0 2 0 2 0 2 4 6 4 6 4 6 4 6 for pixels 0 2 4 6 + const __m128i tmp_12 = _mm_unpacklo_epi32(tmp_8, tmp_10); + // Coeffs 1 3 1 3 1 3 1 3 5 7 5 7 5 7 5 7 for pixels 0 2 4 6 + const __m128i tmp_13 = _mm_unpackhi_epi32(tmp_8, tmp_10); + // Coeffs 0 2 0 2 0 2 0 2 4 6 4 6 4 6 4 6 for pixels 1 3 5 7 + const __m128i tmp_14 = _mm_unpacklo_epi32(tmp_9, tmp_11); + // Coeffs 1 3 1 3 1 3 1 3 5 7 5 7 5 7 5 7 for pixels 1 3 5 7 + const __m128i tmp_15 = _mm_unpackhi_epi32(tmp_9, tmp_11); + + // Coeffs 0 2 for pixels 0 2 4 6 1 3 5 7 + coeff[0] = _mm_unpacklo_epi64(tmp_12, tmp_14); + // Coeffs 4 6 for pixels 0 2 4 6 1 3 5 7 + coeff[1] = _mm_unpackhi_epi64(tmp_12, tmp_14); + // Coeffs 1 3 for pixels 0 2 4 6 1 3 5 7 + coeff[2] = _mm_unpacklo_epi64(tmp_13, tmp_15); + // Coeffs 5 7 for pixels 0 2 4 6 1 3 5 7 + coeff[3] = _mm_unpackhi_epi64(tmp_13, tmp_15); +} + +static INLINE void prepare_horizontal_filter_coeff_alpha0(int sx, + __m128i *coeff) { + // Filter even-index pixels + const __m128i tmp_0 = + _mm_loadl_epi64((__m128i *)&filter_8bit[sx >> WARPEDDIFF_PREC_BITS]); + + // Coeffs 0 2 for pixels 0 2 4 6 1 3 5 7 + coeff[0] = _mm_shuffle_epi8( + tmp_0, _mm_loadu_si128((__m128i *)shuffle_alpha0_mask01)); + // Coeffs 4 6 for pixels 0 2 4 6 1 3 5 7 + coeff[1] = _mm_shuffle_epi8( + tmp_0, _mm_loadu_si128((__m128i *)shuffle_alpha0_mask23)); + // Coeffs 1 3 for pixels 0 2 4 6 1 3 5 7 + coeff[2] = _mm_shuffle_epi8( + tmp_0, _mm_loadu_si128((__m128i *)shuffle_alpha0_mask45)); + // Coeffs 5 7 for pixels 0 2 4 6 1 3 5 7 + coeff[3] = _mm_shuffle_epi8( + tmp_0, _mm_loadu_si128((__m128i *)shuffle_alpha0_mask67)); +} + +static INLINE void horizontal_filter(__m128i src, __m128i *tmp, int sx, + int alpha, int k, + const int offset_bits_horiz, + const int reduce_bits_horiz) { + __m128i coeff[4]; + prepare_horizontal_filter_coeff(alpha, sx, coeff); + filter_src_pixels(src, tmp, coeff, offset_bits_horiz, reduce_bits_horiz, k); +} + +static INLINE void warp_horizontal_filter(const uint8_t *ref, __m128i *tmp, + int stride, int32_t ix4, int32_t iy4, + int32_t sx4, int alpha, int beta, + int p_height, int height, int i, + const int offset_bits_horiz, + const int reduce_bits_horiz) { + int k; + for (k = -7; k < AOMMIN(8, p_height - i); ++k) { + int iy = iy4 + k; + if (iy < 0) + iy = 0; + else if (iy > height - 1) + iy = height - 1; + int sx = sx4 + beta * (k + 4); + + // Load source pixels + const __m128i src = + _mm_loadu_si128((__m128i *)(ref + iy * stride + ix4 - 7)); + horizontal_filter(src, tmp, sx, alpha, k, offset_bits_horiz, + reduce_bits_horiz); + } +} + +static INLINE void warp_horizontal_filter_alpha0( + const uint8_t *ref, __m128i *tmp, int stride, int32_t ix4, int32_t iy4, + int32_t sx4, int alpha, int beta, int p_height, int height, int i, + const int offset_bits_horiz, const int reduce_bits_horiz) { + (void)alpha; + int k; + for (k = -7; k < AOMMIN(8, p_height - i); ++k) { + int iy = iy4 + k; + if (iy < 0) + iy = 0; + else if (iy > height - 1) + iy = height - 1; + int sx = sx4 + beta * (k + 4); + + // Load source pixels + const __m128i src = + _mm_loadu_si128((__m128i *)(ref + iy * stride + ix4 - 7)); + + __m128i coeff[4]; + prepare_horizontal_filter_coeff_alpha0(sx, coeff); + filter_src_pixels(src, tmp, coeff, offset_bits_horiz, reduce_bits_horiz, k); + } +} + +static INLINE void warp_horizontal_filter_beta0( + const uint8_t *ref, __m128i *tmp, int stride, int32_t ix4, int32_t iy4, + int32_t sx4, int alpha, int beta, int p_height, int height, int i, + const int offset_bits_horiz, const int reduce_bits_horiz) { + (void)beta; + int k; + __m128i coeff[4]; + prepare_horizontal_filter_coeff(alpha, sx4, coeff); + + for (k = -7; k < AOMMIN(8, p_height - i); ++k) { + int iy = iy4 + k; + if (iy < 0) + iy = 0; + else if (iy > height - 1) + iy = height - 1; + + // Load source pixels + const __m128i src = + _mm_loadu_si128((__m128i *)(ref + iy * stride + ix4 - 7)); + filter_src_pixels(src, tmp, coeff, offset_bits_horiz, reduce_bits_horiz, k); + } +} + +static INLINE void warp_horizontal_filter_alpha0_beta0( + const uint8_t *ref, __m128i *tmp, int stride, int32_t ix4, int32_t iy4, + int32_t sx4, int alpha, int beta, int p_height, int height, int i, + const int offset_bits_horiz, const int reduce_bits_horiz) { + (void)beta; + (void)alpha; + int k; + + __m128i coeff[4]; + prepare_horizontal_filter_coeff_alpha0(sx4, coeff); + + for (k = -7; k < AOMMIN(8, p_height - i); ++k) { + int iy = iy4 + k; + if (iy < 0) + iy = 0; + else if (iy > height - 1) + iy = height - 1; + + // Load source pixels + const __m128i src = + _mm_loadu_si128((__m128i *)(ref + iy * stride + ix4 - 7)); + filter_src_pixels(src, tmp, coeff, offset_bits_horiz, reduce_bits_horiz, k); + } +} + +static INLINE void unpack_weights_and_set_round_const( + ConvolveParams *conv_params, const int round_bits, const int offset_bits, + __m128i *res_sub_const, __m128i *round_bits_const, __m128i *wt) { + *res_sub_const = + _mm_set1_epi16(-(1 << (offset_bits - conv_params->round_1)) - + (1 << (offset_bits - conv_params->round_1 - 1))); + *round_bits_const = _mm_set1_epi16(((1 << round_bits) >> 1)); + + const int w0 = conv_params->fwd_offset; + const int w1 = conv_params->bck_offset; + const __m128i wt0 = _mm_set1_epi16(w0); + const __m128i wt1 = _mm_set1_epi16(w1); + *wt = _mm_unpacklo_epi16(wt0, wt1); +} + +static INLINE void prepare_vertical_filter_coeffs(int gamma, int sy, + __m128i *coeffs) { + const __m128i tmp_0 = _mm_loadu_si128( + (__m128i *)(warped_filter + ((sy + 0 * gamma) >> WARPEDDIFF_PREC_BITS))); + const __m128i tmp_2 = _mm_loadu_si128( + (__m128i *)(warped_filter + ((sy + 2 * gamma) >> WARPEDDIFF_PREC_BITS))); + const __m128i tmp_4 = _mm_loadu_si128( + (__m128i *)(warped_filter + ((sy + 4 * gamma) >> WARPEDDIFF_PREC_BITS))); + const __m128i tmp_6 = _mm_loadu_si128( + (__m128i *)(warped_filter + ((sy + 6 * gamma) >> WARPEDDIFF_PREC_BITS))); + + const __m128i tmp_8 = _mm_unpacklo_epi32(tmp_0, tmp_2); + const __m128i tmp_10 = _mm_unpacklo_epi32(tmp_4, tmp_6); + const __m128i tmp_12 = _mm_unpackhi_epi32(tmp_0, tmp_2); + const __m128i tmp_14 = _mm_unpackhi_epi32(tmp_4, tmp_6); + + // even coeffs + coeffs[0] = _mm_unpacklo_epi64(tmp_8, tmp_10); + coeffs[1] = _mm_unpackhi_epi64(tmp_8, tmp_10); + coeffs[2] = _mm_unpacklo_epi64(tmp_12, tmp_14); + coeffs[3] = _mm_unpackhi_epi64(tmp_12, tmp_14); + + const __m128i tmp_1 = _mm_loadu_si128( + (__m128i *)(warped_filter + ((sy + 1 * gamma) >> WARPEDDIFF_PREC_BITS))); + const __m128i tmp_3 = _mm_loadu_si128( + (__m128i *)(warped_filter + ((sy + 3 * gamma) >> WARPEDDIFF_PREC_BITS))); + const __m128i tmp_5 = _mm_loadu_si128( + (__m128i *)(warped_filter + ((sy + 5 * gamma) >> WARPEDDIFF_PREC_BITS))); + const __m128i tmp_7 = _mm_loadu_si128( + (__m128i *)(warped_filter + ((sy + 7 * gamma) >> WARPEDDIFF_PREC_BITS))); + + const __m128i tmp_9 = _mm_unpacklo_epi32(tmp_1, tmp_3); + const __m128i tmp_11 = _mm_unpacklo_epi32(tmp_5, tmp_7); + const __m128i tmp_13 = _mm_unpackhi_epi32(tmp_1, tmp_3); + const __m128i tmp_15 = _mm_unpackhi_epi32(tmp_5, tmp_7); + + // odd coeffs + coeffs[4] = _mm_unpacklo_epi64(tmp_9, tmp_11); + coeffs[5] = _mm_unpackhi_epi64(tmp_9, tmp_11); + coeffs[6] = _mm_unpacklo_epi64(tmp_13, tmp_15); + coeffs[7] = _mm_unpackhi_epi64(tmp_13, tmp_15); +} + +static INLINE void prepare_vertical_filter_coeffs_gamma0(int sy, + __m128i *coeffs) { + const __m128i tmp_0 = _mm_loadu_si128( + (__m128i *)(warped_filter + (sy >> WARPEDDIFF_PREC_BITS))); + + // even coeffs + coeffs[0] = + _mm_shuffle_epi8(tmp_0, _mm_loadu_si128((__m128i *)shuffle_gamma0_mask0)); + coeffs[1] = + _mm_shuffle_epi8(tmp_0, _mm_loadu_si128((__m128i *)shuffle_gamma0_mask1)); + coeffs[2] = + _mm_shuffle_epi8(tmp_0, _mm_loadu_si128((__m128i *)shuffle_gamma0_mask2)); + coeffs[3] = + _mm_shuffle_epi8(tmp_0, _mm_loadu_si128((__m128i *)shuffle_gamma0_mask3)); + + // odd coeffs + coeffs[4] = coeffs[0]; + coeffs[5] = coeffs[1]; + coeffs[6] = coeffs[2]; + coeffs[7] = coeffs[3]; +} + +static INLINE void filter_src_pixels_vertical(__m128i *tmp, __m128i *coeffs, + __m128i *res_lo, __m128i *res_hi, + int k) { + // Load from tmp and rearrange pairs of consecutive rows into the + // column order 0 0 2 2 4 4 6 6; 1 1 3 3 5 5 7 7 + const __m128i *src = tmp + (k + 4); + const __m128i src_0 = _mm_unpacklo_epi16(src[0], src[1]); + const __m128i src_2 = _mm_unpacklo_epi16(src[2], src[3]); + const __m128i src_4 = _mm_unpacklo_epi16(src[4], src[5]); + const __m128i src_6 = _mm_unpacklo_epi16(src[6], src[7]); + + const __m128i res_0 = _mm_madd_epi16(src_0, coeffs[0]); + const __m128i res_2 = _mm_madd_epi16(src_2, coeffs[1]); + const __m128i res_4 = _mm_madd_epi16(src_4, coeffs[2]); + const __m128i res_6 = _mm_madd_epi16(src_6, coeffs[3]); + + const __m128i res_even = + _mm_add_epi32(_mm_add_epi32(res_0, res_2), _mm_add_epi32(res_4, res_6)); + + // Filter odd-index pixels + const __m128i src_1 = _mm_unpackhi_epi16(src[0], src[1]); + const __m128i src_3 = _mm_unpackhi_epi16(src[2], src[3]); + const __m128i src_5 = _mm_unpackhi_epi16(src[4], src[5]); + const __m128i src_7 = _mm_unpackhi_epi16(src[6], src[7]); + + const __m128i res_1 = _mm_madd_epi16(src_1, coeffs[4]); + const __m128i res_3 = _mm_madd_epi16(src_3, coeffs[5]); + const __m128i res_5 = _mm_madd_epi16(src_5, coeffs[6]); + const __m128i res_7 = _mm_madd_epi16(src_7, coeffs[7]); + + const __m128i res_odd = + _mm_add_epi32(_mm_add_epi32(res_1, res_3), _mm_add_epi32(res_5, res_7)); + + // Rearrange pixels back into the order 0 ... 7 + *res_lo = _mm_unpacklo_epi32(res_even, res_odd); + *res_hi = _mm_unpackhi_epi32(res_even, res_odd); +} + +static INLINE void store_vertical_filter_output( + __m128i *res_lo, __m128i *res_hi, const __m128i *res_add_const, + const __m128i *wt, const __m128i *res_sub_const, __m128i *round_bits_const, + uint8_t *pred, ConvolveParams *conv_params, int i, int j, int k, + const int reduce_bits_vert, int p_stride, int p_width, + const int round_bits) { + __m128i res_lo_1 = *res_lo; + __m128i res_hi_1 = *res_hi; + + if (conv_params->is_compound) { + __m128i *const p = + (__m128i *)&conv_params->dst[(i + k + 4) * conv_params->dst_stride + j]; + res_lo_1 = _mm_srai_epi32(_mm_add_epi32(res_lo_1, *res_add_const), + reduce_bits_vert); + const __m128i temp_lo_16 = _mm_packus_epi32(res_lo_1, res_lo_1); + __m128i res_lo_16; + if (conv_params->do_average) { + __m128i *const dst8 = (__m128i *)&pred[(i + k + 4) * p_stride + j]; + const __m128i p_16 = _mm_loadl_epi64(p); + + if (conv_params->use_jnt_comp_avg) { + const __m128i p_16_lo = _mm_unpacklo_epi16(p_16, temp_lo_16); + const __m128i wt_res_lo = _mm_madd_epi16(p_16_lo, *wt); + const __m128i shifted_32 = + _mm_srai_epi32(wt_res_lo, DIST_PRECISION_BITS); + res_lo_16 = _mm_packus_epi32(shifted_32, shifted_32); + } else { + res_lo_16 = _mm_srai_epi16(_mm_add_epi16(p_16, temp_lo_16), 1); + } + + res_lo_16 = _mm_add_epi16(res_lo_16, *res_sub_const); + + res_lo_16 = _mm_srai_epi16(_mm_add_epi16(res_lo_16, *round_bits_const), + round_bits); + __m128i res_8_lo = _mm_packus_epi16(res_lo_16, res_lo_16); + *(uint32_t *)dst8 = _mm_cvtsi128_si32(res_8_lo); + } else { + _mm_storel_epi64(p, temp_lo_16); + } + if (p_width > 4) { + __m128i *const p4 = + (__m128i *)&conv_params + ->dst[(i + k + 4) * conv_params->dst_stride + j + 4]; + res_hi_1 = _mm_srai_epi32(_mm_add_epi32(res_hi_1, *res_add_const), + reduce_bits_vert); + const __m128i temp_hi_16 = _mm_packus_epi32(res_hi_1, res_hi_1); + __m128i res_hi_16; + + if (conv_params->do_average) { + __m128i *const dst8_4 = + (__m128i *)&pred[(i + k + 4) * p_stride + j + 4]; + const __m128i p4_16 = _mm_loadl_epi64(p4); + + if (conv_params->use_jnt_comp_avg) { + const __m128i p_16_hi = _mm_unpacklo_epi16(p4_16, temp_hi_16); + const __m128i wt_res_hi = _mm_madd_epi16(p_16_hi, *wt); + const __m128i shifted_32 = + _mm_srai_epi32(wt_res_hi, DIST_PRECISION_BITS); + res_hi_16 = _mm_packus_epi32(shifted_32, shifted_32); + } else { + res_hi_16 = _mm_srai_epi16(_mm_add_epi16(p4_16, temp_hi_16), 1); + } + res_hi_16 = _mm_add_epi16(res_hi_16, *res_sub_const); + + res_hi_16 = _mm_srai_epi16(_mm_add_epi16(res_hi_16, *round_bits_const), + round_bits); + __m128i res_8_hi = _mm_packus_epi16(res_hi_16, res_hi_16); + *(uint32_t *)dst8_4 = _mm_cvtsi128_si32(res_8_hi); + + } else { + _mm_storel_epi64(p4, temp_hi_16); + } + } + } else { + const __m128i res_lo_round = _mm_srai_epi32( + _mm_add_epi32(res_lo_1, *res_add_const), reduce_bits_vert); + const __m128i res_hi_round = _mm_srai_epi32( + _mm_add_epi32(res_hi_1, *res_add_const), reduce_bits_vert); + + const __m128i res_16bit = _mm_packs_epi32(res_lo_round, res_hi_round); + __m128i res_8bit = _mm_packus_epi16(res_16bit, res_16bit); + + // Store, blending with 'pred' if needed + __m128i *const p = (__m128i *)&pred[(i + k + 4) * p_stride + j]; + + // Note: If we're outputting a 4x4 block, we need to be very careful + // to only output 4 pixels at this point, to avoid encode/decode + // mismatches when encoding with multiple threads. + if (p_width == 4) { + *(uint32_t *)p = _mm_cvtsi128_si32(res_8bit); + } else { + _mm_storel_epi64(p, res_8bit); + } + } +} + +static INLINE void warp_vertical_filter( + uint8_t *pred, __m128i *tmp, ConvolveParams *conv_params, int16_t gamma, + int16_t delta, int p_height, int p_stride, int p_width, int i, int j, + int sy4, const int reduce_bits_vert, const __m128i *res_add_const, + const int round_bits, const int offset_bits) { + int k; + __m128i res_sub_const, round_bits_const, wt; + unpack_weights_and_set_round_const(conv_params, round_bits, offset_bits, + &res_sub_const, &round_bits_const, &wt); + // Vertical filter + for (k = -4; k < AOMMIN(4, p_height - i - 4); ++k) { + int sy = sy4 + delta * (k + 4); + + __m128i coeffs[8]; + prepare_vertical_filter_coeffs(gamma, sy, coeffs); + + __m128i res_lo; + __m128i res_hi; + filter_src_pixels_vertical(tmp, coeffs, &res_lo, &res_hi, k); + + store_vertical_filter_output(&res_lo, &res_hi, res_add_const, &wt, + &res_sub_const, &round_bits_const, pred, + conv_params, i, j, k, reduce_bits_vert, + p_stride, p_width, round_bits); + } +} + +static INLINE void warp_vertical_filter_gamma0( + uint8_t *pred, __m128i *tmp, ConvolveParams *conv_params, int16_t gamma, + int16_t delta, int p_height, int p_stride, int p_width, int i, int j, + int sy4, const int reduce_bits_vert, const __m128i *res_add_const, + const int round_bits, const int offset_bits) { + int k; + (void)gamma; + __m128i res_sub_const, round_bits_const, wt; + unpack_weights_and_set_round_const(conv_params, round_bits, offset_bits, + &res_sub_const, &round_bits_const, &wt); + // Vertical filter + for (k = -4; k < AOMMIN(4, p_height - i - 4); ++k) { + int sy = sy4 + delta * (k + 4); + + __m128i coeffs[8]; + prepare_vertical_filter_coeffs_gamma0(sy, coeffs); + + __m128i res_lo; + __m128i res_hi; + filter_src_pixels_vertical(tmp, coeffs, &res_lo, &res_hi, k); + + store_vertical_filter_output(&res_lo, &res_hi, res_add_const, &wt, + &res_sub_const, &round_bits_const, pred, + conv_params, i, j, k, reduce_bits_vert, + p_stride, p_width, round_bits); + } +} + +static INLINE void warp_vertical_filter_delta0( + uint8_t *pred, __m128i *tmp, ConvolveParams *conv_params, int16_t gamma, + int16_t delta, int p_height, int p_stride, int p_width, int i, int j, + int sy4, const int reduce_bits_vert, const __m128i *res_add_const, + const int round_bits, const int offset_bits) { + (void)delta; + int k; + __m128i res_sub_const, round_bits_const, wt; + unpack_weights_and_set_round_const(conv_params, round_bits, offset_bits, + &res_sub_const, &round_bits_const, &wt); + + __m128i coeffs[8]; + prepare_vertical_filter_coeffs(gamma, sy4, coeffs); + // Vertical filter + for (k = -4; k < AOMMIN(4, p_height - i - 4); ++k) { + __m128i res_lo; + __m128i res_hi; + filter_src_pixels_vertical(tmp, coeffs, &res_lo, &res_hi, k); + + store_vertical_filter_output(&res_lo, &res_hi, res_add_const, &wt, + &res_sub_const, &round_bits_const, pred, + conv_params, i, j, k, reduce_bits_vert, + p_stride, p_width, round_bits); + } +} + +static INLINE void warp_vertical_filter_gamma0_delta0( + uint8_t *pred, __m128i *tmp, ConvolveParams *conv_params, int16_t gamma, + int16_t delta, int p_height, int p_stride, int p_width, int i, int j, + int sy4, const int reduce_bits_vert, const __m128i *res_add_const, + const int round_bits, const int offset_bits) { + (void)delta; + (void)gamma; + int k; + __m128i res_sub_const, round_bits_const, wt; + unpack_weights_and_set_round_const(conv_params, round_bits, offset_bits, + &res_sub_const, &round_bits_const, &wt); + + __m128i coeffs[8]; + prepare_vertical_filter_coeffs_gamma0(sy4, coeffs); + // Vertical filter + for (k = -4; k < AOMMIN(4, p_height - i - 4); ++k) { + __m128i res_lo; + __m128i res_hi; + filter_src_pixels_vertical(tmp, coeffs, &res_lo, &res_hi, k); + + store_vertical_filter_output(&res_lo, &res_hi, res_add_const, &wt, + &res_sub_const, &round_bits_const, pred, + conv_params, i, j, k, reduce_bits_vert, + p_stride, p_width, round_bits); + } +} + +static INLINE void prepare_warp_vertical_filter( + uint8_t *pred, __m128i *tmp, ConvolveParams *conv_params, int16_t gamma, + int16_t delta, int p_height, int p_stride, int p_width, int i, int j, + int sy4, const int reduce_bits_vert, const __m128i *res_add_const, + const int round_bits, const int offset_bits) { + if (gamma == 0 && delta == 0) + warp_vertical_filter_gamma0_delta0( + pred, tmp, conv_params, gamma, delta, p_height, p_stride, p_width, i, j, + sy4, reduce_bits_vert, res_add_const, round_bits, offset_bits); + else if (gamma == 0 && delta != 0) + warp_vertical_filter_gamma0(pred, tmp, conv_params, gamma, delta, p_height, + p_stride, p_width, i, j, sy4, reduce_bits_vert, + res_add_const, round_bits, offset_bits); + else if (gamma != 0 && delta == 0) + warp_vertical_filter_delta0(pred, tmp, conv_params, gamma, delta, p_height, + p_stride, p_width, i, j, sy4, reduce_bits_vert, + res_add_const, round_bits, offset_bits); + else + warp_vertical_filter(pred, tmp, conv_params, gamma, delta, p_height, + p_stride, p_width, i, j, sy4, reduce_bits_vert, + res_add_const, round_bits, offset_bits); +} + +static INLINE void prepare_warp_horizontal_filter( + const uint8_t *ref, __m128i *tmp, int stride, int32_t ix4, int32_t iy4, + int32_t sx4, int alpha, int beta, int p_height, int height, int i, + const int offset_bits_horiz, const int reduce_bits_horiz) { + if (alpha == 0 && beta == 0) + warp_horizontal_filter_alpha0_beta0(ref, tmp, stride, ix4, iy4, sx4, alpha, + beta, p_height, height, i, + offset_bits_horiz, reduce_bits_horiz); + else if (alpha == 0 && beta != 0) + warp_horizontal_filter_alpha0(ref, tmp, stride, ix4, iy4, sx4, alpha, beta, + p_height, height, i, offset_bits_horiz, + reduce_bits_horiz); + else if (alpha != 0 && beta == 0) + warp_horizontal_filter_beta0(ref, tmp, stride, ix4, iy4, sx4, alpha, beta, + p_height, height, i, offset_bits_horiz, + reduce_bits_horiz); + else + warp_horizontal_filter(ref, tmp, stride, ix4, iy4, sx4, alpha, beta, + p_height, height, i, offset_bits_horiz, + reduce_bits_horiz); +} + +void av1_warp_affine_sse4_1(const int32_t *mat, const uint8_t *ref, int width, + int height, int stride, uint8_t *pred, int p_col, + int p_row, int p_width, int p_height, int p_stride, + int subsampling_x, int subsampling_y, + ConvolveParams *conv_params, int16_t alpha, + int16_t beta, int16_t gamma, int16_t delta) { + __m128i tmp[15]; + int i, j, k; + const int bd = 8; + const int reduce_bits_horiz = conv_params->round_0; + const int reduce_bits_vert = conv_params->is_compound + ? conv_params->round_1 + : 2 * FILTER_BITS - reduce_bits_horiz; + const int offset_bits_horiz = bd + FILTER_BITS - 1; + assert(IMPLIES(conv_params->is_compound, conv_params->dst != NULL)); + + const int offset_bits_vert = bd + 2 * FILTER_BITS - reduce_bits_horiz; + const __m128i reduce_bits_vert_const = + _mm_set1_epi32(((1 << reduce_bits_vert) >> 1)); + const __m128i res_add_const = _mm_set1_epi32(1 << offset_bits_vert); + const int round_bits = + 2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1; + const int offset_bits = bd + 2 * FILTER_BITS - conv_params->round_0; + assert(IMPLIES(conv_params->do_average, conv_params->is_compound)); + + /* Note: For this code to work, the left/right frame borders need to be + extended by at least 13 pixels each. By the time we get here, other + code will have set up this border, but we allow an explicit check + for debugging purposes. + */ + /*for (i = 0; i < height; ++i) { + for (j = 0; j < 13; ++j) { + assert(ref[i * stride - 13 + j] == ref[i * stride]); + assert(ref[i * stride + width + j] == ref[i * stride + (width - 1)]); + } + }*/ + __m128i res_add_const_1; + if (conv_params->is_compound == 1) { + res_add_const_1 = _mm_add_epi32(reduce_bits_vert_const, res_add_const); + } else { + res_add_const_1 = _mm_set1_epi32(-(1 << (bd + reduce_bits_vert - 1)) + + ((1 << reduce_bits_vert) >> 1)); + } + + for (i = 0; i < p_height; i += 8) { + for (j = 0; j < p_width; j += 8) { + const int32_t src_x = (p_col + j + 4) << subsampling_x; + const int32_t src_y = (p_row + i + 4) << subsampling_y; + const int32_t dst_x = mat[2] * src_x + mat[3] * src_y + mat[0]; + const int32_t dst_y = mat[4] * src_x + mat[5] * src_y + mat[1]; + const int32_t x4 = dst_x >> subsampling_x; + const int32_t y4 = dst_y >> subsampling_y; + + int32_t ix4 = x4 >> WARPEDMODEL_PREC_BITS; + int32_t sx4 = x4 & ((1 << WARPEDMODEL_PREC_BITS) - 1); + int32_t iy4 = y4 >> WARPEDMODEL_PREC_BITS; + int32_t sy4 = y4 & ((1 << WARPEDMODEL_PREC_BITS) - 1); + + // Add in all the constant terms, including rounding and offset + sx4 += alpha * (-4) + beta * (-4) + (1 << (WARPEDDIFF_PREC_BITS - 1)) + + (WARPEDPIXEL_PREC_SHIFTS << WARPEDDIFF_PREC_BITS); + sy4 += gamma * (-4) + delta * (-4) + (1 << (WARPEDDIFF_PREC_BITS - 1)) + + (WARPEDPIXEL_PREC_SHIFTS << WARPEDDIFF_PREC_BITS); + + sx4 &= ~((1 << WARP_PARAM_REDUCE_BITS) - 1); + sy4 &= ~((1 << WARP_PARAM_REDUCE_BITS) - 1); + + // Horizontal filter + // If the block is aligned such that, after clamping, every sample + // would be taken from the leftmost/rightmost column, then we can + // skip the expensive horizontal filter. + if (ix4 <= -7) { + for (k = -7; k < AOMMIN(8, p_height - i); ++k) { + int iy = iy4 + k; + if (iy < 0) + iy = 0; + else if (iy > height - 1) + iy = height - 1; + tmp[k + 7] = _mm_set1_epi16( + (1 << (bd + FILTER_BITS - reduce_bits_horiz - 1)) + + ref[iy * stride] * (1 << (FILTER_BITS - reduce_bits_horiz))); + } + } else if (ix4 >= width + 6) { + for (k = -7; k < AOMMIN(8, p_height - i); ++k) { + int iy = iy4 + k; + if (iy < 0) + iy = 0; + else if (iy > height - 1) + iy = height - 1; + tmp[k + 7] = + _mm_set1_epi16((1 << (bd + FILTER_BITS - reduce_bits_horiz - 1)) + + ref[iy * stride + (width - 1)] * + (1 << (FILTER_BITS - reduce_bits_horiz))); + } + } else if (((ix4 - 7) < 0) || ((ix4 + 9) > width)) { + const int out_of_boundary_left = -(ix4 - 6); + const int out_of_boundary_right = (ix4 + 8) - width; + for (k = -7; k < AOMMIN(8, p_height - i); ++k) { + int iy = iy4 + k; + if (iy < 0) + iy = 0; + else if (iy > height - 1) + iy = height - 1; + int sx = sx4 + beta * (k + 4); + + // Load source pixels + __m128i src = + _mm_loadu_si128((__m128i *)(ref + iy * stride + ix4 - 7)); + if (out_of_boundary_left >= 0) { + const __m128i shuffle_reg_left = + _mm_loadu_si128((__m128i *)warp_pad_left[out_of_boundary_left]); + src = _mm_shuffle_epi8(src, shuffle_reg_left); + } + if (out_of_boundary_right >= 0) { + const __m128i shuffle_reg_right = _mm_loadu_si128( + (__m128i *)warp_pad_right[out_of_boundary_right]); + src = _mm_shuffle_epi8(src, shuffle_reg_right); + } + horizontal_filter(src, tmp, sx, alpha, k, offset_bits_horiz, + reduce_bits_horiz); + } + } else { + prepare_warp_horizontal_filter(ref, tmp, stride, ix4, iy4, sx4, alpha, + beta, p_height, height, i, + offset_bits_horiz, reduce_bits_horiz); + } + + // Vertical filter + prepare_warp_vertical_filter( + pred, tmp, conv_params, gamma, delta, p_height, p_stride, p_width, i, + j, sy4, reduce_bits_vert, &res_add_const_1, round_bits, offset_bits); + } + } +} diff --git a/third_party/aom/av1/common/x86/wiener_convolve_avx2.c b/third_party/aom/av1/common/x86/wiener_convolve_avx2.c new file mode 100644 index 0000000000..87a6e12396 --- /dev/null +++ b/third_party/aom/av1/common/x86/wiener_convolve_avx2.c @@ -0,0 +1,261 @@ +/* + * Copyright (c) 2018, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <immintrin.h> +#include <assert.h> + +#include "config/av1_rtcd.h" + +#include "av1/common/convolve.h" +#include "aom_dsp/aom_dsp_common.h" +#include "aom_dsp/aom_filter.h" +#include "aom_dsp/x86/synonyms.h" +#include "aom_dsp/x86/synonyms_avx2.h" + +// 128-bit xmmwords are written as [ ... ] with the MSB on the left. +// 256-bit ymmwords are written as two xmmwords, [ ... ][ ... ] with the MSB +// on the left. +// A row of, say, 8-bit pixels with values p0, p1, p2, ..., p30, p31 will be +// loaded and stored as [ p31 ... p17 p16 ][ p15 ... p1 p0 ]. +void av1_wiener_convolve_add_src_avx2(const uint8_t *src, ptrdiff_t src_stride, + uint8_t *dst, ptrdiff_t dst_stride, + const int16_t *filter_x, int x_step_q4, + const int16_t *filter_y, int y_step_q4, + int w, int h, + const ConvolveParams *conv_params) { + const int bd = 8; + assert(x_step_q4 == 16 && y_step_q4 == 16); + assert(!(w & 7)); + (void)x_step_q4; + (void)y_step_q4; + + DECLARE_ALIGNED(32, uint16_t, + temp[(MAX_SB_SIZE + SUBPEL_TAPS - 1) * MAX_SB_SIZE]); + int intermediate_height = h + SUBPEL_TAPS - 2; + memset(temp + (intermediate_height * MAX_SB_SIZE), 0, MAX_SB_SIZE); + const int center_tap = ((SUBPEL_TAPS - 1) / 2); + const uint8_t *const src_ptr = src - center_tap * src_stride - center_tap; + + const __m128i zero_128 = _mm_setzero_si128(); + const __m256i zero_256 = _mm256_setzero_si256(); + + // Add an offset to account for the "add_src" part of the convolve function. + const __m128i offset = _mm_insert_epi16(zero_128, 1 << FILTER_BITS, 3); + + const __m256i clamp_low = zero_256; + const __m256i clamp_high = + _mm256_set1_epi16(WIENER_CLAMP_LIMIT(conv_params->round_0, bd) - 1); + + /* Horizontal filter */ + { + // coeffs [ f7 f6 f5 f4 f3 f2 f1 f0 ] + const __m128i coeffs_x = _mm_add_epi16(xx_loadu_128(filter_x), offset); + + // coeffs [ f3 f2 f3 f2 f1 f0 f1 f0 ] + const __m128i coeffs_0123 = _mm_unpacklo_epi32(coeffs_x, coeffs_x); + // coeffs [ f7 f6 f7 f6 f5 f4 f5 f4 ] + const __m128i coeffs_4567 = _mm_unpackhi_epi32(coeffs_x, coeffs_x); + + // coeffs [ f1 f0 f1 f0 f1 f0 f1 f0 ] + const __m128i coeffs_01_128 = _mm_unpacklo_epi64(coeffs_0123, coeffs_0123); + // coeffs [ f3 f2 f3 f2 f3 f2 f3 f2 ] + const __m128i coeffs_23_128 = _mm_unpackhi_epi64(coeffs_0123, coeffs_0123); + // coeffs [ f5 f4 f5 f4 f5 f4 f5 f4 ] + const __m128i coeffs_45_128 = _mm_unpacklo_epi64(coeffs_4567, coeffs_4567); + // coeffs [ f7 f6 f7 f6 f7 f6 f7 f6 ] + const __m128i coeffs_67_128 = _mm_unpackhi_epi64(coeffs_4567, coeffs_4567); + + // coeffs [ f1 f0 f1 f0 f1 f0 f1 f0 ][ f1 f0 f1 f0 f1 f0 f1 f0 ] + const __m256i coeffs_01 = yy_set_m128i(coeffs_01_128, coeffs_01_128); + // coeffs [ f3 f2 f3 f2 f3 f2 f3 f2 ][ f3 f2 f3 f2 f3 f2 f3 f2 ] + const __m256i coeffs_23 = yy_set_m128i(coeffs_23_128, coeffs_23_128); + // coeffs [ f5 f4 f5 f4 f5 f4 f5 f4 ][ f5 f4 f5 f4 f5 f4 f5 f4 ] + const __m256i coeffs_45 = yy_set_m128i(coeffs_45_128, coeffs_45_128); + // coeffs [ f7 f6 f7 f6 f7 f6 f7 f6 ][ f7 f6 f7 f6 f7 f6 f7 f6 ] + const __m256i coeffs_67 = yy_set_m128i(coeffs_67_128, coeffs_67_128); + + const __m256i round_const = _mm256_set1_epi32( + (1 << (conv_params->round_0 - 1)) + (1 << (bd + FILTER_BITS - 1))); + + for (int i = 0; i < intermediate_height; ++i) { + for (int j = 0; j < w; j += 16) { + const uint8_t *data_ij = src_ptr + i * src_stride + j; + + // Load 8-bit src data + const __m128i data_0 = xx_loadu_128(data_ij + 0); + const __m128i data_1 = xx_loadu_128(data_ij + 1); + const __m128i data_2 = xx_loadu_128(data_ij + 2); + const __m128i data_3 = xx_loadu_128(data_ij + 3); + const __m128i data_4 = xx_loadu_128(data_ij + 4); + const __m128i data_5 = xx_loadu_128(data_ij + 5); + const __m128i data_6 = xx_loadu_128(data_ij + 6); + const __m128i data_7 = xx_loadu_128(data_ij + 7); + + // (Zero-)Extend 8-bit data to 16-bit data + const __m256i src_0 = _mm256_cvtepu8_epi16(data_0); + const __m256i src_1 = _mm256_cvtepu8_epi16(data_1); + const __m256i src_2 = _mm256_cvtepu8_epi16(data_2); + const __m256i src_3 = _mm256_cvtepu8_epi16(data_3); + const __m256i src_4 = _mm256_cvtepu8_epi16(data_4); + const __m256i src_5 = _mm256_cvtepu8_epi16(data_5); + const __m256i src_6 = _mm256_cvtepu8_epi16(data_6); + const __m256i src_7 = _mm256_cvtepu8_epi16(data_7); + + // Multiply src data by filter coeffs and sum pairs + const __m256i res_0 = _mm256_madd_epi16(src_0, coeffs_01); + const __m256i res_1 = _mm256_madd_epi16(src_1, coeffs_01); + const __m256i res_2 = _mm256_madd_epi16(src_2, coeffs_23); + const __m256i res_3 = _mm256_madd_epi16(src_3, coeffs_23); + const __m256i res_4 = _mm256_madd_epi16(src_4, coeffs_45); + const __m256i res_5 = _mm256_madd_epi16(src_5, coeffs_45); + const __m256i res_6 = _mm256_madd_epi16(src_6, coeffs_67); + const __m256i res_7 = _mm256_madd_epi16(src_7, coeffs_67); + + // Calculate scalar product for even- and odd-indices separately, + // increasing to 32-bit precision + const __m256i res_even_sum = _mm256_add_epi32( + _mm256_add_epi32(res_0, res_4), _mm256_add_epi32(res_2, res_6)); + const __m256i res_odd_sum = _mm256_add_epi32( + _mm256_add_epi32(res_1, res_5), _mm256_add_epi32(res_3, res_7)); + + const __m256i res_even = _mm256_srai_epi32( + _mm256_add_epi32(res_even_sum, round_const), conv_params->round_0); + const __m256i res_odd = _mm256_srai_epi32( + _mm256_add_epi32(res_odd_sum, round_const), conv_params->round_0); + + // Reduce to 16-bit precision and pack even- and odd-index results + // back into one register. The _mm256_packs_epi32 intrinsic returns + // a register with the pixels ordered as follows: + // [ 15 13 11 9 14 12 10 8 ] [ 7 5 3 1 6 4 2 0 ] + const __m256i res = _mm256_packs_epi32(res_even, res_odd); + const __m256i res_clamped = + _mm256_min_epi16(_mm256_max_epi16(res, clamp_low), clamp_high); + + // Store in a temporary array + yy_storeu_256(temp + i * MAX_SB_SIZE + j, res_clamped); + } + } + } + + /* Vertical filter */ + { + // coeffs [ g7 g6 g5 g4 g3 g2 g1 g0 ] + const __m128i coeffs_y = _mm_add_epi16(xx_loadu_128(filter_y), offset); + + // coeffs [ g3 g2 g3 g2 g1 g0 g1 g0 ] + const __m128i coeffs_0123 = _mm_unpacklo_epi32(coeffs_y, coeffs_y); + // coeffs [ g7 g6 g7 g6 g5 g4 g5 g4 ] + const __m128i coeffs_4567 = _mm_unpackhi_epi32(coeffs_y, coeffs_y); + + // coeffs [ g1 g0 g1 g0 g1 g0 g1 g0 ] + const __m128i coeffs_01_128 = _mm_unpacklo_epi64(coeffs_0123, coeffs_0123); + // coeffs [ g3 g2 g3 g2 g3 g2 g3 g2 ] + const __m128i coeffs_23_128 = _mm_unpackhi_epi64(coeffs_0123, coeffs_0123); + // coeffs [ g5 g4 g5 g4 g5 g4 g5 g4 ] + const __m128i coeffs_45_128 = _mm_unpacklo_epi64(coeffs_4567, coeffs_4567); + // coeffs [ g7 g6 g7 g6 g7 g6 g7 g6 ] + const __m128i coeffs_67_128 = _mm_unpackhi_epi64(coeffs_4567, coeffs_4567); + + // coeffs [ g1 g0 g1 g0 g1 g0 g1 g0 ][ g1 g0 g1 g0 g1 g0 g1 g0 ] + const __m256i coeffs_01 = yy_set_m128i(coeffs_01_128, coeffs_01_128); + // coeffs [ g3 g2 g3 g2 g3 g2 g3 g2 ][ g3 g2 g3 g2 g3 g2 g3 g2 ] + const __m256i coeffs_23 = yy_set_m128i(coeffs_23_128, coeffs_23_128); + // coeffs [ g5 g4 g5 g4 g5 g4 g5 g4 ][ g5 g4 g5 g4 g5 g4 g5 g4 ] + const __m256i coeffs_45 = yy_set_m128i(coeffs_45_128, coeffs_45_128); + // coeffs [ g7 g6 g7 g6 g7 g6 g7 g6 ][ g7 g6 g7 g6 g7 g6 g7 g6 ] + const __m256i coeffs_67 = yy_set_m128i(coeffs_67_128, coeffs_67_128); + + const __m256i round_const = + _mm256_set1_epi32((1 << (conv_params->round_1 - 1)) - + (1 << (bd + conv_params->round_1 - 1))); + + for (int i = 0; i < h; ++i) { + for (int j = 0; j < w; j += 16) { + const uint16_t *data_ij = temp + i * MAX_SB_SIZE + j; + + // Load 16-bit data from the output of the horizontal filter in + // which the pixels are ordered as follows: + // [ 15 13 11 9 14 12 10 8 ] [ 7 5 3 1 6 4 2 0 ] + const __m256i data_0 = yy_loadu_256(data_ij + 0 * MAX_SB_SIZE); + const __m256i data_1 = yy_loadu_256(data_ij + 1 * MAX_SB_SIZE); + const __m256i data_2 = yy_loadu_256(data_ij + 2 * MAX_SB_SIZE); + const __m256i data_3 = yy_loadu_256(data_ij + 3 * MAX_SB_SIZE); + const __m256i data_4 = yy_loadu_256(data_ij + 4 * MAX_SB_SIZE); + const __m256i data_5 = yy_loadu_256(data_ij + 5 * MAX_SB_SIZE); + const __m256i data_6 = yy_loadu_256(data_ij + 6 * MAX_SB_SIZE); + const __m256i data_7 = yy_loadu_256(data_ij + 7 * MAX_SB_SIZE); + + // Filter the even-indices, increasing to 32-bit precision + const __m256i src_0 = _mm256_unpacklo_epi16(data_0, data_1); + const __m256i src_2 = _mm256_unpacklo_epi16(data_2, data_3); + const __m256i src_4 = _mm256_unpacklo_epi16(data_4, data_5); + const __m256i src_6 = _mm256_unpacklo_epi16(data_6, data_7); + + const __m256i res_0 = _mm256_madd_epi16(src_0, coeffs_01); + const __m256i res_2 = _mm256_madd_epi16(src_2, coeffs_23); + const __m256i res_4 = _mm256_madd_epi16(src_4, coeffs_45); + const __m256i res_6 = _mm256_madd_epi16(src_6, coeffs_67); + + const __m256i res_even = _mm256_add_epi32( + _mm256_add_epi32(res_0, res_2), _mm256_add_epi32(res_4, res_6)); + + // Filter the odd-indices, increasing to 32-bit precision + const __m256i src_1 = _mm256_unpackhi_epi16(data_0, data_1); + const __m256i src_3 = _mm256_unpackhi_epi16(data_2, data_3); + const __m256i src_5 = _mm256_unpackhi_epi16(data_4, data_5); + const __m256i src_7 = _mm256_unpackhi_epi16(data_6, data_7); + + const __m256i res_1 = _mm256_madd_epi16(src_1, coeffs_01); + const __m256i res_3 = _mm256_madd_epi16(src_3, coeffs_23); + const __m256i res_5 = _mm256_madd_epi16(src_5, coeffs_45); + const __m256i res_7 = _mm256_madd_epi16(src_7, coeffs_67); + + const __m256i res_odd = _mm256_add_epi32( + _mm256_add_epi32(res_1, res_3), _mm256_add_epi32(res_5, res_7)); + + // Pixels are currently in the following order: + // res_even order: [ 14 12 10 8 ] [ 6 4 2 0 ] + // res_odd order: [ 15 13 11 9 ] [ 7 5 3 1 ] + // + // Rearrange the pixels into the following order: + // res_lo order: [ 11 10 9 8 ] [ 3 2 1 0 ] + // res_hi order: [ 15 14 13 12 ] [ 7 6 5 4 ] + const __m256i res_lo = _mm256_unpacklo_epi32(res_even, res_odd); + const __m256i res_hi = _mm256_unpackhi_epi32(res_even, res_odd); + + const __m256i res_lo_round = _mm256_srai_epi32( + _mm256_add_epi32(res_lo, round_const), conv_params->round_1); + const __m256i res_hi_round = _mm256_srai_epi32( + _mm256_add_epi32(res_hi, round_const), conv_params->round_1); + + // Reduce to 16-bit precision and pack into the correct order: + // [ 15 14 13 12 11 10 9 8 ][ 7 6 5 4 3 2 1 0 ] + const __m256i res_16bit = + _mm256_packs_epi32(res_lo_round, res_hi_round); + + // Reduce to 8-bit precision. This messes up the order: + // [ - - - - - - - - 15 14 13 12 11 10 9 8 ] + // [ - - - - - - - - 7 6 5 4 3 2 1 0 ] + const __m256i res_8bit = + _mm256_packus_epi16(res_16bit, zero_256 /* don't care value */); + + // Swap the two central 32-bit values to get the order: + // [ - - - - - - - - - - - - - - - - ] + // [ 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0 ] + const __m256i res_8bit2 = _mm256_permute4x64_epi64(res_8bit, 0xd8); + + // Store the lower 128-bit lane in the dst array + xx_storeu_128(dst + i * dst_stride + j, + _mm256_castsi256_si128(res_8bit2)); + } + } + } +} diff --git a/third_party/aom/av1/common/x86/wiener_convolve_sse2.c b/third_party/aom/av1/common/x86/wiener_convolve_sse2.c new file mode 100644 index 0000000000..f9d00b7330 --- /dev/null +++ b/third_party/aom/av1/common/x86/wiener_convolve_sse2.c @@ -0,0 +1,199 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <emmintrin.h> +#include <assert.h> + +#include "config/av1_rtcd.h" + +#include "av1/common/convolve.h" +#include "aom_dsp/aom_dsp_common.h" +#include "aom_dsp/aom_filter.h" + +void av1_wiener_convolve_add_src_sse2(const uint8_t *src, ptrdiff_t src_stride, + uint8_t *dst, ptrdiff_t dst_stride, + const int16_t *filter_x, int x_step_q4, + const int16_t *filter_y, int y_step_q4, + int w, int h, + const ConvolveParams *conv_params) { + const int bd = 8; + assert(x_step_q4 == 16 && y_step_q4 == 16); + assert(!(w & 7)); + (void)x_step_q4; + (void)y_step_q4; + + DECLARE_ALIGNED(16, uint16_t, + temp[(MAX_SB_SIZE + SUBPEL_TAPS - 1) * MAX_SB_SIZE]); + int intermediate_height = h + SUBPEL_TAPS - 2; + memset(temp + (intermediate_height * MAX_SB_SIZE), 0, MAX_SB_SIZE); + int i, j; + const int center_tap = ((SUBPEL_TAPS - 1) / 2); + const uint8_t *const src_ptr = src - center_tap * src_stride - center_tap; + + const __m128i zero = _mm_setzero_si128(); + // Add an offset to account for the "add_src" part of the convolve function. + const __m128i offset = _mm_insert_epi16(zero, 1 << FILTER_BITS, 3); + + /* Horizontal filter */ + { + const __m128i coeffs_x = + _mm_add_epi16(_mm_loadu_si128((__m128i *)filter_x), offset); + + // coeffs 0 1 0 1 2 3 2 3 + const __m128i tmp_0 = _mm_unpacklo_epi32(coeffs_x, coeffs_x); + // coeffs 4 5 4 5 6 7 6 7 + const __m128i tmp_1 = _mm_unpackhi_epi32(coeffs_x, coeffs_x); + + // coeffs 0 1 0 1 0 1 0 1 + const __m128i coeff_01 = _mm_unpacklo_epi64(tmp_0, tmp_0); + // coeffs 2 3 2 3 2 3 2 3 + const __m128i coeff_23 = _mm_unpackhi_epi64(tmp_0, tmp_0); + // coeffs 4 5 4 5 4 5 4 5 + const __m128i coeff_45 = _mm_unpacklo_epi64(tmp_1, tmp_1); + // coeffs 6 7 6 7 6 7 6 7 + const __m128i coeff_67 = _mm_unpackhi_epi64(tmp_1, tmp_1); + + const __m128i round_const = _mm_set1_epi32( + (1 << (conv_params->round_0 - 1)) + (1 << (bd + FILTER_BITS - 1))); + + for (i = 0; i < intermediate_height; ++i) { + for (j = 0; j < w; j += 8) { + const __m128i data = + _mm_loadu_si128((__m128i *)&src_ptr[i * src_stride + j]); + + // Filter even-index pixels + const __m128i src_0 = _mm_unpacklo_epi8(data, zero); + const __m128i res_0 = _mm_madd_epi16(src_0, coeff_01); + const __m128i src_2 = _mm_unpacklo_epi8(_mm_srli_si128(data, 2), zero); + const __m128i res_2 = _mm_madd_epi16(src_2, coeff_23); + const __m128i src_4 = _mm_unpacklo_epi8(_mm_srli_si128(data, 4), zero); + const __m128i res_4 = _mm_madd_epi16(src_4, coeff_45); + const __m128i src_6 = _mm_unpacklo_epi8(_mm_srli_si128(data, 6), zero); + const __m128i res_6 = _mm_madd_epi16(src_6, coeff_67); + + __m128i res_even = _mm_add_epi32(_mm_add_epi32(res_0, res_4), + _mm_add_epi32(res_2, res_6)); + res_even = _mm_srai_epi32(_mm_add_epi32(res_even, round_const), + conv_params->round_0); + + // Filter odd-index pixels + const __m128i src_1 = _mm_unpacklo_epi8(_mm_srli_si128(data, 1), zero); + const __m128i res_1 = _mm_madd_epi16(src_1, coeff_01); + const __m128i src_3 = _mm_unpacklo_epi8(_mm_srli_si128(data, 3), zero); + const __m128i res_3 = _mm_madd_epi16(src_3, coeff_23); + const __m128i src_5 = _mm_unpacklo_epi8(_mm_srli_si128(data, 5), zero); + const __m128i res_5 = _mm_madd_epi16(src_5, coeff_45); + const __m128i src_7 = _mm_unpacklo_epi8(_mm_srli_si128(data, 7), zero); + const __m128i res_7 = _mm_madd_epi16(src_7, coeff_67); + + __m128i res_odd = _mm_add_epi32(_mm_add_epi32(res_1, res_5), + _mm_add_epi32(res_3, res_7)); + res_odd = _mm_srai_epi32(_mm_add_epi32(res_odd, round_const), + conv_params->round_0); + + // Pack in the column order 0, 2, 4, 6, 1, 3, 5, 7 + __m128i res = _mm_packs_epi32(res_even, res_odd); + res = _mm_min_epi16( + _mm_max_epi16(res, zero), + _mm_set1_epi16(WIENER_CLAMP_LIMIT(conv_params->round_0, bd) - 1)); + _mm_storeu_si128((__m128i *)&temp[i * MAX_SB_SIZE + j], res); + } + } + } + + /* Vertical filter */ + { + const __m128i coeffs_y = + _mm_add_epi16(_mm_loadu_si128((__m128i *)filter_y), offset); + + // coeffs 0 1 0 1 2 3 2 3 + const __m128i tmp_0 = _mm_unpacklo_epi32(coeffs_y, coeffs_y); + // coeffs 4 5 4 5 6 7 6 7 + const __m128i tmp_1 = _mm_unpackhi_epi32(coeffs_y, coeffs_y); + + // coeffs 0 1 0 1 0 1 0 1 + const __m128i coeff_01 = _mm_unpacklo_epi64(tmp_0, tmp_0); + // coeffs 2 3 2 3 2 3 2 3 + const __m128i coeff_23 = _mm_unpackhi_epi64(tmp_0, tmp_0); + // coeffs 4 5 4 5 4 5 4 5 + const __m128i coeff_45 = _mm_unpacklo_epi64(tmp_1, tmp_1); + // coeffs 6 7 6 7 6 7 6 7 + const __m128i coeff_67 = _mm_unpackhi_epi64(tmp_1, tmp_1); + + const __m128i round_const = + _mm_set1_epi32((1 << (conv_params->round_1 - 1)) - + (1 << (bd + conv_params->round_1 - 1))); + + for (i = 0; i < h; ++i) { + for (j = 0; j < w; j += 8) { + // Filter even-index pixels + const uint16_t *data = &temp[i * MAX_SB_SIZE + j]; + const __m128i src_0 = + _mm_unpacklo_epi16(*(__m128i *)(data + 0 * MAX_SB_SIZE), + *(__m128i *)(data + 1 * MAX_SB_SIZE)); + const __m128i src_2 = + _mm_unpacklo_epi16(*(__m128i *)(data + 2 * MAX_SB_SIZE), + *(__m128i *)(data + 3 * MAX_SB_SIZE)); + const __m128i src_4 = + _mm_unpacklo_epi16(*(__m128i *)(data + 4 * MAX_SB_SIZE), + *(__m128i *)(data + 5 * MAX_SB_SIZE)); + const __m128i src_6 = + _mm_unpacklo_epi16(*(__m128i *)(data + 6 * MAX_SB_SIZE), + *(__m128i *)(data + 7 * MAX_SB_SIZE)); + + const __m128i res_0 = _mm_madd_epi16(src_0, coeff_01); + const __m128i res_2 = _mm_madd_epi16(src_2, coeff_23); + const __m128i res_4 = _mm_madd_epi16(src_4, coeff_45); + const __m128i res_6 = _mm_madd_epi16(src_6, coeff_67); + + const __m128i res_even = _mm_add_epi32(_mm_add_epi32(res_0, res_2), + _mm_add_epi32(res_4, res_6)); + + // Filter odd-index pixels + const __m128i src_1 = + _mm_unpackhi_epi16(*(__m128i *)(data + 0 * MAX_SB_SIZE), + *(__m128i *)(data + 1 * MAX_SB_SIZE)); + const __m128i src_3 = + _mm_unpackhi_epi16(*(__m128i *)(data + 2 * MAX_SB_SIZE), + *(__m128i *)(data + 3 * MAX_SB_SIZE)); + const __m128i src_5 = + _mm_unpackhi_epi16(*(__m128i *)(data + 4 * MAX_SB_SIZE), + *(__m128i *)(data + 5 * MAX_SB_SIZE)); + const __m128i src_7 = + _mm_unpackhi_epi16(*(__m128i *)(data + 6 * MAX_SB_SIZE), + *(__m128i *)(data + 7 * MAX_SB_SIZE)); + + const __m128i res_1 = _mm_madd_epi16(src_1, coeff_01); + const __m128i res_3 = _mm_madd_epi16(src_3, coeff_23); + const __m128i res_5 = _mm_madd_epi16(src_5, coeff_45); + const __m128i res_7 = _mm_madd_epi16(src_7, coeff_67); + + const __m128i res_odd = _mm_add_epi32(_mm_add_epi32(res_1, res_3), + _mm_add_epi32(res_5, res_7)); + + // Rearrange pixels back into the order 0 ... 7 + const __m128i res_lo = _mm_unpacklo_epi32(res_even, res_odd); + const __m128i res_hi = _mm_unpackhi_epi32(res_even, res_odd); + + const __m128i res_lo_round = _mm_srai_epi32( + _mm_add_epi32(res_lo, round_const), conv_params->round_1); + const __m128i res_hi_round = _mm_srai_epi32( + _mm_add_epi32(res_hi, round_const), conv_params->round_1); + + const __m128i res_16bit = _mm_packs_epi32(res_lo_round, res_hi_round); + __m128i res_8bit = _mm_packus_epi16(res_16bit, res_16bit); + + __m128i *const p = (__m128i *)&dst[i * dst_stride + j]; + _mm_storel_epi64(p, res_8bit); + } + } + } +} |