diff options
Diffstat (limited to 'third_party/aom/av1/encoder/arm/neon/rdopt_neon.c')
-rw-r--r-- | third_party/aom/av1/encoder/arm/neon/rdopt_neon.c | 459 |
1 files changed, 459 insertions, 0 deletions
diff --git a/third_party/aom/av1/encoder/arm/neon/rdopt_neon.c b/third_party/aom/av1/encoder/arm/neon/rdopt_neon.c new file mode 100644 index 0000000000..7d3bd4c606 --- /dev/null +++ b/third_party/aom/av1/encoder/arm/neon/rdopt_neon.c @@ -0,0 +1,459 @@ +/* + * Copyright (c) 2020, 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 <arm_neon.h> + +#include "av1/encoder/rdopt.h" +#include "config/aom_config.h" +#include "config/av1_rtcd.h" + +// Process horizontal and vertical correlations in a 4x4 block of pixels. +// We actually use the 4x4 pixels to calculate correlations corresponding to +// the top-left 3x3 pixels, so this function must be called with 1x1 overlap, +// moving the window along/down by 3 pixels at a time. +INLINE static void horver_correlation_4x4(const int16_t *diff, int stride, + int32x4_t *xy_sum_32, + int32x4_t *xz_sum_32, + int32x4_t *x_sum_32, + int32x4_t *x2_sum_32) { + // Pixels in this 4x4 [ a b c d ] + // are referred to as: [ e f g h ] + // [ i j k l ] + // [ m n o p ] + + const int16x4_t pixelsa_2_lo = vld1_s16(diff + (0 * stride)); + const int16x4_t pixelsa_2_sli = + vreinterpret_s16_s64(vshl_n_s64(vreinterpret_s64_s16(pixelsa_2_lo), 16)); + const int16x4_t pixelsb_2_lo = vld1_s16(diff + (1 * stride)); + const int16x4_t pixelsb_2_sli = + vreinterpret_s16_s64(vshl_n_s64(vreinterpret_s64_s16(pixelsb_2_lo), 16)); + const int16x4_t pixelsa_1_lo = vld1_s16(diff + (2 * stride)); + const int16x4_t pixelsa_1_sli = + vreinterpret_s16_s64(vshl_n_s64(vreinterpret_s64_s16(pixelsa_1_lo), 16)); + const int16x4_t pixelsb_1_lo = vld1_s16(diff + (3 * stride)); + const int16x4_t pixelsb_1_sli = + vreinterpret_s16_s64(vshl_n_s64(vreinterpret_s64_s16(pixelsb_1_lo), 16)); + + const int16x8_t slli_a = vcombine_s16(pixelsa_1_sli, pixelsa_2_sli); + + *xy_sum_32 = vmlal_s16(*xy_sum_32, pixelsa_1_lo, pixelsa_1_sli); + *xy_sum_32 = vmlal_s16(*xy_sum_32, pixelsa_2_lo, pixelsa_2_sli); + *xy_sum_32 = vmlal_s16(*xy_sum_32, pixelsb_2_lo, pixelsb_2_sli); + + *xz_sum_32 = vmlal_s16(*xz_sum_32, pixelsa_1_sli, pixelsb_1_sli); + *xz_sum_32 = vmlal_s16(*xz_sum_32, pixelsa_2_sli, pixelsb_2_sli); + *xz_sum_32 = vmlal_s16(*xz_sum_32, pixelsa_1_sli, pixelsb_2_sli); + + // Now calculate the straight sums, x_sum += a+b+c+e+f+g+i+j+k + // (sum up every element in slli_a and swap_b) + *x_sum_32 = vpadalq_s16(*x_sum_32, slli_a); + *x_sum_32 = vaddw_s16(*x_sum_32, pixelsb_2_sli); + + // Also sum their squares + *x2_sum_32 = vmlal_s16(*x2_sum_32, pixelsa_1_sli, pixelsa_1_sli); + *x2_sum_32 = vmlal_s16(*x2_sum_32, pixelsa_2_sli, pixelsa_2_sli); + *x2_sum_32 = vmlal_s16(*x2_sum_32, pixelsb_2_sli, pixelsb_2_sli); +} + +void av1_get_horver_correlation_full_neon(const int16_t *diff, int stride, + int width, int height, float *hcorr, + float *vcorr) { + // The following notation is used: + // x - current pixel + // y - right neighbour pixel + // z - below neighbour pixel + // w - down-right neighbour pixel + int64_t xy_sum = 0, xz_sum = 0; + int64_t x_sum = 0, x2_sum = 0; + int32x4_t zero = vdupq_n_s32(0); + int64x2_t v_x_sum = vreinterpretq_s64_s32(zero); + int64x2_t v_xy_sum = vreinterpretq_s64_s32(zero); + int64x2_t v_xz_sum = vreinterpretq_s64_s32(zero); + int64x2_t v_x2_sum = vreinterpretq_s64_s32(zero); + // Process horizontal and vertical correlations through the body in 4x4 + // blocks. This excludes the final row and column and possibly one extra + // column depending how 3 divides into width and height + + for (int i = 0; i <= height - 4; i += 3) { + int32x4_t xy_sum_32 = zero; + int32x4_t xz_sum_32 = zero; + int32x4_t x_sum_32 = zero; + int32x4_t x2_sum_32 = zero; + for (int j = 0; j <= width - 4; j += 3) { + horver_correlation_4x4(&diff[i * stride + j], stride, &xy_sum_32, + &xz_sum_32, &x_sum_32, &x2_sum_32); + } + v_xy_sum = vpadalq_s32(v_xy_sum, xy_sum_32); + v_xz_sum = vpadalq_s32(v_xz_sum, xz_sum_32); + v_x_sum = vpadalq_s32(v_x_sum, x_sum_32); + v_x2_sum = vpadalq_s32(v_x2_sum, x2_sum_32); + } +#if AOM_ARCH_AARCH64 + xy_sum = vaddvq_s64(v_xy_sum); + xz_sum = vaddvq_s64(v_xz_sum); + x2_sum = vaddvq_s64(v_x2_sum); + x_sum = vaddvq_s64(v_x_sum); +#else + xy_sum = vget_lane_s64( + vadd_s64(vget_low_s64(v_xy_sum), vget_high_s64(v_xy_sum)), 0); + xz_sum = vget_lane_s64( + vadd_s64(vget_low_s64(v_xz_sum), vget_high_s64(v_xz_sum)), 0); + x2_sum = vget_lane_s64( + vadd_s64(vget_low_s64(v_x2_sum), vget_high_s64(v_x2_sum)), 0); + x_sum = + vget_lane_s64(vadd_s64(vget_low_s64(v_x_sum), vget_high_s64(v_x_sum)), 0); +#endif + // x_sum now covers every pixel except the final 1-2 rows and 1-2 cols + int64_t x_finalrow = 0, x_finalcol = 0, x2_finalrow = 0, x2_finalcol = 0; + + // Do we have 2 rows remaining or just the one? Note that width and height + // are powers of 2, so each modulo 3 must be 1 or 2. + if (height % 3 == 1) { // Just horiz corrs on the final row + const int16_t x0 = diff[(height - 1) * stride]; + x_sum += x0; + x_finalrow += x0; + x2_sum += x0 * x0; + x2_finalrow += x0 * x0; + if (width >= 8) { + int32x4_t v_y_sum = zero; + int32x4_t v_y2_sum = zero; + int32x4_t v_xy_sum_a = zero; + int k = width - 1; + int j = 0; + while ((k - 8) > 0) { + const int16x8_t v_x = vld1q_s16(&diff[(height - 1) * stride + j]); + const int16x8_t v_y = vld1q_s16(&diff[(height - 1) * stride + j + 1]); + const int16x4_t v_x_lo = vget_low_s16(v_x); + const int16x4_t v_x_hi = vget_high_s16(v_x); + const int16x4_t v_y_lo = vget_low_s16(v_y); + const int16x4_t v_y_hi = vget_high_s16(v_y); + v_xy_sum_a = vmlal_s16(v_xy_sum_a, v_x_lo, v_y_lo); + v_xy_sum_a = vmlal_s16(v_xy_sum_a, v_x_hi, v_y_hi); + v_y2_sum = vmlal_s16(v_y2_sum, v_y_lo, v_y_lo); + v_y2_sum = vmlal_s16(v_y2_sum, v_y_hi, v_y_hi); + v_y_sum = vpadalq_s16(v_y_sum, v_y); + k -= 8; + j += 8; + } + + const int16x8_t v_l = vld1q_s16(&diff[(height - 1) * stride] + j); + const int16x8_t v_x = + vextq_s16(vextq_s16(vreinterpretq_s16_s32(zero), v_l, 7), + vreinterpretq_s16_s32(zero), 1); + const int16x8_t v_y = vextq_s16(v_l, vreinterpretq_s16_s32(zero), 1); + const int16x4_t v_x_lo = vget_low_s16(v_x); + const int16x4_t v_x_hi = vget_high_s16(v_x); + const int16x4_t v_y_lo = vget_low_s16(v_y); + const int16x4_t v_y_hi = vget_high_s16(v_y); + v_xy_sum_a = vmlal_s16(v_xy_sum_a, v_x_lo, v_y_lo); + v_xy_sum_a = vmlal_s16(v_xy_sum_a, v_x_hi, v_y_hi); + v_y2_sum = vmlal_s16(v_y2_sum, v_y_lo, v_y_lo); + v_y2_sum = vmlal_s16(v_y2_sum, v_y_hi, v_y_hi); + const int32x4_t v_y_sum_a = vpadalq_s16(v_y_sum, v_y); + const int64x2_t v_xy_sum2 = vpaddlq_s32(v_xy_sum_a); +#if AOM_ARCH_AARCH64 + const int64x2_t v_y2_sum_a = vpaddlq_s32(v_y2_sum); + xy_sum += vaddvq_s64(v_xy_sum2); + const int32_t y = vaddvq_s32(v_y_sum_a); + const int64_t y2 = vaddvq_s64(v_y2_sum_a); +#else + xy_sum += vget_lane_s64( + vadd_s64(vget_low_s64(v_xy_sum2), vget_high_s64(v_xy_sum2)), 0); + const int64x2_t v_y_a = vpaddlq_s32(v_y_sum_a); + const int64_t y = + vget_lane_s64(vadd_s64(vget_low_s64(v_y_a), vget_high_s64(v_y_a)), 0); + const int64x2_t v_y2_sum_b = vpaddlq_s32(v_y2_sum); + int64_t y2 = vget_lane_s64( + vadd_s64(vget_low_s64(v_y2_sum_b), vget_high_s64(v_y2_sum_b)), 0); +#endif + x_sum += y; + x2_sum += y2; + x_finalrow += y; + x2_finalrow += y2; + } else { + for (int j = 0; j < width - 1; ++j) { + const int16_t x = diff[(height - 1) * stride + j]; + const int16_t y = diff[(height - 1) * stride + j + 1]; + xy_sum += x * y; + x_sum += y; + x2_sum += y * y; + x_finalrow += y; + x2_finalrow += y * y; + } + } + } else { // Two rows remaining to do + const int16_t x0 = diff[(height - 2) * stride]; + const int16_t z0 = diff[(height - 1) * stride]; + x_sum += x0 + z0; + x2_sum += x0 * x0 + z0 * z0; + x_finalrow += z0; + x2_finalrow += z0 * z0; + if (width >= 8) { + int32x4_t v_y2_sum = zero; + int32x4_t v_w2_sum = zero; + int32x4_t v_xy_sum_a = zero; + int32x4_t v_xz_sum_a = zero; + int32x4_t v_x_sum_a = zero; + int32x4_t v_w_sum = zero; + int k = width - 1; + int j = 0; + while ((k - 8) > 0) { + const int16x8_t v_x = vld1q_s16(&diff[(height - 2) * stride + j]); + const int16x8_t v_y = vld1q_s16(&diff[(height - 2) * stride + j + 1]); + const int16x8_t v_z = vld1q_s16(&diff[(height - 1) * stride + j]); + const int16x8_t v_w = vld1q_s16(&diff[(height - 1) * stride + j + 1]); + + const int16x4_t v_x_lo = vget_low_s16(v_x); + const int16x4_t v_y_lo = vget_low_s16(v_y); + const int16x4_t v_z_lo = vget_low_s16(v_z); + const int16x4_t v_w_lo = vget_low_s16(v_w); + const int16x4_t v_x_hi = vget_high_s16(v_x); + const int16x4_t v_y_hi = vget_high_s16(v_y); + const int16x4_t v_z_hi = vget_high_s16(v_z); + const int16x4_t v_w_hi = vget_high_s16(v_w); + + v_xy_sum_a = vmlal_s16(v_xy_sum_a, v_x_lo, v_y_lo); + v_xy_sum_a = vmlal_s16(v_xy_sum_a, v_x_hi, v_y_hi); + v_xy_sum_a = vmlal_s16(v_xy_sum_a, v_z_lo, v_w_lo); + v_xy_sum_a = vmlal_s16(v_xy_sum_a, v_z_hi, v_w_hi); + + v_xz_sum_a = vmlal_s16(v_xz_sum_a, v_x_lo, v_z_lo); + v_xz_sum_a = vmlal_s16(v_xz_sum_a, v_x_hi, v_z_hi); + + v_w2_sum = vmlal_s16(v_w2_sum, v_w_lo, v_w_lo); + v_w2_sum = vmlal_s16(v_w2_sum, v_w_hi, v_w_hi); + v_y2_sum = vmlal_s16(v_y2_sum, v_y_lo, v_y_lo); + v_y2_sum = vmlal_s16(v_y2_sum, v_y_hi, v_y_hi); + + v_w_sum = vpadalq_s16(v_w_sum, v_w); + v_x_sum_a = vpadalq_s16(v_x_sum_a, v_y); + v_x_sum_a = vpadalq_s16(v_x_sum_a, v_w); + + k -= 8; + j += 8; + } + const int16x8_t v_l = vld1q_s16(&diff[(height - 2) * stride] + j); + const int16x8_t v_x = + vextq_s16(vextq_s16(vreinterpretq_s16_s32(zero), v_l, 7), + vreinterpretq_s16_s32(zero), 1); + const int16x8_t v_y = vextq_s16(v_l, vreinterpretq_s16_s32(zero), 1); + const int16x8_t v_l_2 = vld1q_s16(&diff[(height - 1) * stride] + j); + const int16x8_t v_z = + vextq_s16(vextq_s16(vreinterpretq_s16_s32(zero), v_l_2, 7), + vreinterpretq_s16_s32(zero), 1); + const int16x8_t v_w = vextq_s16(v_l_2, vreinterpretq_s16_s32(zero), 1); + + const int16x4_t v_x_lo = vget_low_s16(v_x); + const int16x4_t v_y_lo = vget_low_s16(v_y); + const int16x4_t v_z_lo = vget_low_s16(v_z); + const int16x4_t v_w_lo = vget_low_s16(v_w); + const int16x4_t v_x_hi = vget_high_s16(v_x); + const int16x4_t v_y_hi = vget_high_s16(v_y); + const int16x4_t v_z_hi = vget_high_s16(v_z); + const int16x4_t v_w_hi = vget_high_s16(v_w); + + v_xy_sum_a = vmlal_s16(v_xy_sum_a, v_x_lo, v_y_lo); + v_xy_sum_a = vmlal_s16(v_xy_sum_a, v_x_hi, v_y_hi); + v_xy_sum_a = vmlal_s16(v_xy_sum_a, v_z_lo, v_w_lo); + v_xy_sum_a = vmlal_s16(v_xy_sum_a, v_z_hi, v_w_hi); + + v_xz_sum_a = vmlal_s16(v_xz_sum_a, v_x_lo, v_z_lo); + v_xz_sum_a = vmlal_s16(v_xz_sum_a, v_x_hi, v_z_hi); + + v_w2_sum = vmlal_s16(v_w2_sum, v_w_lo, v_w_lo); + v_w2_sum = vmlal_s16(v_w2_sum, v_w_hi, v_w_hi); + v_y2_sum = vmlal_s16(v_y2_sum, v_y_lo, v_y_lo); + v_y2_sum = vmlal_s16(v_y2_sum, v_y_hi, v_y_hi); + + v_w_sum = vpadalq_s16(v_w_sum, v_w); + v_x_sum_a = vpadalq_s16(v_x_sum_a, v_y); + v_x_sum_a = vpadalq_s16(v_x_sum_a, v_w); + +#if AOM_ARCH_AARCH64 + xy_sum += vaddvq_s64(vpaddlq_s32(v_xy_sum_a)); + xz_sum += vaddvq_s64(vpaddlq_s32(v_xz_sum_a)); + x_sum += vaddvq_s32(v_x_sum_a); + x_finalrow += vaddvq_s32(v_w_sum); + int64_t y2 = vaddvq_s64(vpaddlq_s32(v_y2_sum)); + int64_t w2 = vaddvq_s64(vpaddlq_s32(v_w2_sum)); +#else + const int64x2_t v_xy_sum2 = vpaddlq_s32(v_xy_sum_a); + xy_sum += vget_lane_s64( + vadd_s64(vget_low_s64(v_xy_sum2), vget_high_s64(v_xy_sum2)), 0); + const int64x2_t v_xz_sum2 = vpaddlq_s32(v_xz_sum_a); + xz_sum += vget_lane_s64( + vadd_s64(vget_low_s64(v_xz_sum2), vget_high_s64(v_xz_sum2)), 0); + const int64x2_t v_x_sum2 = vpaddlq_s32(v_x_sum_a); + x_sum += vget_lane_s64( + vadd_s64(vget_low_s64(v_x_sum2), vget_high_s64(v_x_sum2)), 0); + const int64x2_t v_w_sum_a = vpaddlq_s32(v_w_sum); + x_finalrow += vget_lane_s64( + vadd_s64(vget_low_s64(v_w_sum_a), vget_high_s64(v_w_sum_a)), 0); + const int64x2_t v_y2_sum_a = vpaddlq_s32(v_y2_sum); + int64_t y2 = vget_lane_s64( + vadd_s64(vget_low_s64(v_y2_sum_a), vget_high_s64(v_y2_sum_a)), 0); + const int64x2_t v_w2_sum_a = vpaddlq_s32(v_w2_sum); + int64_t w2 = vget_lane_s64( + vadd_s64(vget_low_s64(v_w2_sum_a), vget_high_s64(v_w2_sum_a)), 0); +#endif + x2_sum += y2 + w2; + x2_finalrow += w2; + } else { + for (int j = 0; j < width - 1; ++j) { + const int16_t x = diff[(height - 2) * stride + j]; + const int16_t y = diff[(height - 2) * stride + j + 1]; + const int16_t z = diff[(height - 1) * stride + j]; + const int16_t w = diff[(height - 1) * stride + j + 1]; + + // Horizontal and vertical correlations for the penultimate row: + xy_sum += x * y; + xz_sum += x * z; + + // Now just horizontal correlations for the final row: + xy_sum += z * w; + + x_sum += y + w; + x2_sum += y * y + w * w; + x_finalrow += w; + x2_finalrow += w * w; + } + } + } + + // Do we have 2 columns remaining or just the one? + if (width % 3 == 1) { // Just vert corrs on the final col + const int16_t x0 = diff[width - 1]; + x_sum += x0; + x_finalcol += x0; + x2_sum += x0 * x0; + x2_finalcol += x0 * x0; + for (int i = 0; i < height - 1; ++i) { + const int16_t x = diff[i * stride + width - 1]; + const int16_t z = diff[(i + 1) * stride + width - 1]; + xz_sum += x * z; + x_finalcol += z; + x2_finalcol += z * z; + // So the bottom-right elements don't get counted twice: + if (i < height - (height % 3 == 1 ? 2 : 3)) { + x_sum += z; + x2_sum += z * z; + } + } + } else { // Two cols remaining + const int16_t x0 = diff[width - 2]; + const int16_t y0 = diff[width - 1]; + x_sum += x0 + y0; + x2_sum += x0 * x0 + y0 * y0; + x_finalcol += y0; + x2_finalcol += y0 * y0; + for (int i = 0; i < height - 1; ++i) { + const int16_t x = diff[i * stride + width - 2]; + const int16_t y = diff[i * stride + width - 1]; + const int16_t z = diff[(i + 1) * stride + width - 2]; + const int16_t w = diff[(i + 1) * stride + width - 1]; + + // Horizontal and vertical correlations for the penultimate col: + // Skip these on the last iteration of this loop if we also had two + // rows remaining, otherwise the final horizontal and vertical correlation + // get erroneously processed twice + if (i < height - 2 || height % 3 == 1) { + xy_sum += x * y; + xz_sum += x * z; + } + + x_finalcol += w; + x2_finalcol += w * w; + // So the bottom-right elements don't get counted twice: + if (i < height - (height % 3 == 1 ? 2 : 3)) { + x_sum += z + w; + x2_sum += z * z + w * w; + } + + // Now just vertical correlations for the final column: + xz_sum += y * w; + } + } + + // Calculate the simple sums and squared-sums + int64_t x_firstrow = 0, x_firstcol = 0; + int64_t x2_firstrow = 0, x2_firstcol = 0; + + if (width >= 8) { + int32x4_t v_x_firstrow = zero; + int32x4_t v_x2_firstrow = zero; + for (int j = 0; j < width; j += 8) { + const int16x8_t v_diff = vld1q_s16(diff + j); + const int16x4_t v_diff_lo = vget_low_s16(v_diff); + const int16x4_t v_diff_hi = vget_high_s16(v_diff); + v_x_firstrow = vpadalq_s16(v_x_firstrow, v_diff); + v_x2_firstrow = vmlal_s16(v_x2_firstrow, v_diff_lo, v_diff_lo); + v_x2_firstrow = vmlal_s16(v_x2_firstrow, v_diff_hi, v_diff_hi); + } +#if AOM_ARCH_AARCH64 + x_firstrow += vaddvq_s32(v_x_firstrow); + x2_firstrow += vaddvq_s32(v_x2_firstrow); +#else + const int64x2_t v_x_firstrow_64 = vpaddlq_s32(v_x_firstrow); + x_firstrow += vget_lane_s64( + vadd_s64(vget_low_s64(v_x_firstrow_64), vget_high_s64(v_x_firstrow_64)), + 0); + const int64x2_t v_x2_firstrow_64 = vpaddlq_s32(v_x2_firstrow); + x2_firstrow += vget_lane_s64(vadd_s64(vget_low_s64(v_x2_firstrow_64), + vget_high_s64(v_x2_firstrow_64)), + 0); +#endif + } else { + for (int j = 0; j < width; ++j) { + x_firstrow += diff[j]; + x2_firstrow += diff[j] * diff[j]; + } + } + for (int i = 0; i < height; ++i) { + x_firstcol += diff[i * stride]; + x2_firstcol += diff[i * stride] * diff[i * stride]; + } + + int64_t xhor_sum = x_sum - x_finalcol; + int64_t xver_sum = x_sum - x_finalrow; + int64_t y_sum = x_sum - x_firstcol; + int64_t z_sum = x_sum - x_firstrow; + int64_t x2hor_sum = x2_sum - x2_finalcol; + int64_t x2ver_sum = x2_sum - x2_finalrow; + int64_t y2_sum = x2_sum - x2_firstcol; + int64_t z2_sum = x2_sum - x2_firstrow; + + const float num_hor = (float)(height * (width - 1)); + const float num_ver = (float)((height - 1) * width); + + const float xhor_var_n = x2hor_sum - (xhor_sum * xhor_sum) / num_hor; + const float xver_var_n = x2ver_sum - (xver_sum * xver_sum) / num_ver; + + const float y_var_n = y2_sum - (y_sum * y_sum) / num_hor; + const float z_var_n = z2_sum - (z_sum * z_sum) / num_ver; + + const float xy_var_n = xy_sum - (xhor_sum * y_sum) / num_hor; + const float xz_var_n = xz_sum - (xver_sum * z_sum) / num_ver; + + if (xhor_var_n > 0 && y_var_n > 0) { + *hcorr = xy_var_n / sqrtf(xhor_var_n * y_var_n); + *hcorr = *hcorr < 0 ? 0 : *hcorr; + } else { + *hcorr = 1.0; + } + if (xver_var_n > 0 && z_var_n > 0) { + *vcorr = xz_var_n / sqrtf(xver_var_n * z_var_n); + *vcorr = *vcorr < 0 ? 0 : *vcorr; + } else { + *vcorr = 1.0; + } +} |