From 36d22d82aa202bb199967e9512281e9a53db42c9 Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Sun, 7 Apr 2024 21:33:14 +0200 Subject: Adding upstream version 115.7.0esr. Signed-off-by: Daniel Baumann --- third_party/aom/av1/encoder/x86/encodetxb_sse4.c | 92 ++++++++++++++++++++++++ 1 file changed, 92 insertions(+) create mode 100644 third_party/aom/av1/encoder/x86/encodetxb_sse4.c (limited to 'third_party/aom/av1/encoder/x86/encodetxb_sse4.c') diff --git a/third_party/aom/av1/encoder/x86/encodetxb_sse4.c b/third_party/aom/av1/encoder/x86/encodetxb_sse4.c new file mode 100644 index 0000000000..5e0687cd38 --- /dev/null +++ b/third_party/aom/av1/encoder/x86/encodetxb_sse4.c @@ -0,0 +1,92 @@ +/* + * 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 +#include // SSE2 +#include /* SSE4.1 */ + +#include "aom/aom_integer.h" +#include "av1/common/onyxc_int.h" +#include "av1/common/txb_common.h" +#include "aom_dsp/x86/synonyms.h" + +void av1_txb_init_levels_sse4_1(const tran_low_t *const coeff, const int width, + const int height, uint8_t *const levels) { + const int stride = width + TX_PAD_HOR; + const __m128i zeros = _mm_setzero_si128(); + + const int32_t pre_len = sizeof(*levels) * TX_PAD_TOP * stride; + uint8_t *pre_buf = levels - TX_PAD_TOP * stride; + uint8_t *pre_buf_end = pre_buf + pre_len; + do { + _mm_storeu_si128((__m128i *)(pre_buf), zeros); + pre_buf += 16; + } while (pre_buf < pre_buf_end); + + const int32_t bottom_len = sizeof(*levels) * (TX_PAD_BOTTOM * stride); + uint8_t *bottom_buf = levels + stride * height; + uint8_t *bottom_buf_end = bottom_buf + bottom_len; + do { + _mm_storeu_si128((__m128i *)(bottom_buf), zeros); + bottom_buf += 16; + } while (bottom_buf < bottom_buf_end); + + int i = 0; + uint8_t *ls = levels; + const tran_low_t *cf = coeff; + if (width == 4) { + do { + const __m128i coeffA = xx_loadu_128(cf); + const __m128i coeffB = xx_loadu_128(cf + 4); + const __m128i coeffAB = _mm_packs_epi32(coeffA, coeffB); + const __m128i absAB = _mm_abs_epi16(coeffAB); + const __m128i absAB8 = _mm_packs_epi16(absAB, zeros); + const __m128i lsAB = _mm_unpacklo_epi32(absAB8, zeros); + xx_storeu_128(ls, lsAB); + ls += (stride << 1); + cf += (width << 1); + i += 2; + } while (i < height); + } else if (width == 8) { + do { + const __m128i coeffA = xx_loadu_128(cf); + const __m128i coeffB = xx_loadu_128(cf + 4); + const __m128i coeffAB = _mm_packs_epi32(coeffA, coeffB); + const __m128i absAB = _mm_abs_epi16(coeffAB); + const __m128i absAB8 = _mm_packs_epi16(absAB, zeros); + xx_storeu_128(ls, absAB8); + ls += stride; + cf += width; + i += 1; + } while (i < height); + } else { + do { + int j = 0; + do { + const __m128i coeffA = xx_loadu_128(cf); + const __m128i coeffB = xx_loadu_128(cf + 4); + const __m128i coeffC = xx_loadu_128(cf + 8); + const __m128i coeffD = xx_loadu_128(cf + 12); + const __m128i coeffAB = _mm_packs_epi32(coeffA, coeffB); + const __m128i coeffCD = _mm_packs_epi32(coeffC, coeffD); + const __m128i absAB = _mm_abs_epi16(coeffAB); + const __m128i absCD = _mm_abs_epi16(coeffCD); + const __m128i absABCD = _mm_packs_epi16(absAB, absCD); + xx_storeu_128(ls + j, absABCD); + j += 16; + cf += 16; + } while (j < width); + *(int32_t *)(ls + width) = 0; + ls += stride; + i += 1; + } while (i < height); + } +} -- cgit v1.2.3