summaryrefslogtreecommitdiffstats
path: root/third_party/aom/aom_dsp/x86/sad_avx2.c
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-07 09:22:09 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-07 09:22:09 +0000
commit43a97878ce14b72f0981164f87f2e35e14151312 (patch)
tree620249daf56c0258faa40cbdcf9cfba06de2a846 /third_party/aom/aom_dsp/x86/sad_avx2.c
parentInitial commit. (diff)
downloadfirefox-43a97878ce14b72f0981164f87f2e35e14151312.tar.xz
firefox-43a97878ce14b72f0981164f87f2e35e14151312.zip
Adding upstream version 110.0.1.upstream/110.0.1upstream
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'third_party/aom/aom_dsp/x86/sad_avx2.c')
-rw-r--r--third_party/aom/aom_dsp/x86/sad_avx2.c189
1 files changed, 189 insertions, 0 deletions
diff --git a/third_party/aom/aom_dsp/x86/sad_avx2.c b/third_party/aom/aom_dsp/x86/sad_avx2.c
new file mode 100644
index 0000000000..a50dba64a1
--- /dev/null
+++ b/third_party/aom/aom_dsp/x86/sad_avx2.c
@@ -0,0 +1,189 @@
+/*
+ * 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 <immintrin.h>
+
+#include "config/aom_dsp_rtcd.h"
+
+#include "aom_ports/mem.h"
+
+#define FSAD64_H(h) \
+ unsigned int aom_sad64x##h##_avx2(const uint8_t *src_ptr, int src_stride, \
+ const uint8_t *ref_ptr, int ref_stride) { \
+ int i, res; \
+ __m256i sad1_reg, sad2_reg, ref1_reg, ref2_reg; \
+ __m256i sum_sad = _mm256_setzero_si256(); \
+ __m256i sum_sad_h; \
+ __m128i sum_sad128; \
+ for (i = 0; i < h; i++) { \
+ ref1_reg = _mm256_loadu_si256((__m256i const *)ref_ptr); \
+ ref2_reg = _mm256_loadu_si256((__m256i const *)(ref_ptr + 32)); \
+ sad1_reg = _mm256_sad_epu8( \
+ ref1_reg, _mm256_loadu_si256((__m256i const *)src_ptr)); \
+ sad2_reg = _mm256_sad_epu8( \
+ ref2_reg, _mm256_loadu_si256((__m256i const *)(src_ptr + 32))); \
+ sum_sad = \
+ _mm256_add_epi32(sum_sad, _mm256_add_epi32(sad1_reg, sad2_reg)); \
+ ref_ptr += ref_stride; \
+ src_ptr += src_stride; \
+ } \
+ sum_sad_h = _mm256_srli_si256(sum_sad, 8); \
+ sum_sad = _mm256_add_epi32(sum_sad, sum_sad_h); \
+ sum_sad128 = _mm256_extracti128_si256(sum_sad, 1); \
+ sum_sad128 = _mm_add_epi32(_mm256_castsi256_si128(sum_sad), sum_sad128); \
+ res = _mm_cvtsi128_si32(sum_sad128); \
+ _mm256_zeroupper(); \
+ return res; \
+ }
+
+#define FSAD32_H(h) \
+ unsigned int aom_sad32x##h##_avx2(const uint8_t *src_ptr, int src_stride, \
+ const uint8_t *ref_ptr, int ref_stride) { \
+ int i, res; \
+ __m256i sad1_reg, sad2_reg, ref1_reg, ref2_reg; \
+ __m256i sum_sad = _mm256_setzero_si256(); \
+ __m256i sum_sad_h; \
+ __m128i sum_sad128; \
+ int ref2_stride = ref_stride << 1; \
+ int src2_stride = src_stride << 1; \
+ int max = h >> 1; \
+ for (i = 0; i < max; i++) { \
+ ref1_reg = _mm256_loadu_si256((__m256i const *)ref_ptr); \
+ ref2_reg = _mm256_loadu_si256((__m256i const *)(ref_ptr + ref_stride)); \
+ sad1_reg = _mm256_sad_epu8( \
+ ref1_reg, _mm256_loadu_si256((__m256i const *)src_ptr)); \
+ sad2_reg = _mm256_sad_epu8( \
+ ref2_reg, \
+ _mm256_loadu_si256((__m256i const *)(src_ptr + src_stride))); \
+ sum_sad = \
+ _mm256_add_epi32(sum_sad, _mm256_add_epi32(sad1_reg, sad2_reg)); \
+ ref_ptr += ref2_stride; \
+ src_ptr += src2_stride; \
+ } \
+ sum_sad_h = _mm256_srli_si256(sum_sad, 8); \
+ sum_sad = _mm256_add_epi32(sum_sad, sum_sad_h); \
+ sum_sad128 = _mm256_extracti128_si256(sum_sad, 1); \
+ sum_sad128 = _mm_add_epi32(_mm256_castsi256_si128(sum_sad), sum_sad128); \
+ res = _mm_cvtsi128_si32(sum_sad128); \
+ _mm256_zeroupper(); \
+ return res; \
+ }
+
+#define FSAD64 \
+ FSAD64_H(64); \
+ FSAD64_H(32);
+
+#define FSAD32 \
+ FSAD32_H(64); \
+ FSAD32_H(32); \
+ FSAD32_H(16);
+
+/* clang-format off */
+FSAD64
+FSAD32
+/* clang-format on */
+
+#undef FSAD64
+#undef FSAD32
+#undef FSAD64_H
+#undef FSAD32_H
+
+#define FSADAVG64_H(h) \
+ unsigned int aom_sad64x##h##_avg_avx2( \
+ const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, \
+ int ref_stride, const uint8_t *second_pred) { \
+ int i, res; \
+ __m256i sad1_reg, sad2_reg, ref1_reg, ref2_reg; \
+ __m256i sum_sad = _mm256_setzero_si256(); \
+ __m256i sum_sad_h; \
+ __m128i sum_sad128; \
+ for (i = 0; i < h; i++) { \
+ ref1_reg = _mm256_loadu_si256((__m256i const *)ref_ptr); \
+ ref2_reg = _mm256_loadu_si256((__m256i const *)(ref_ptr + 32)); \
+ ref1_reg = _mm256_avg_epu8( \
+ ref1_reg, _mm256_loadu_si256((__m256i const *)second_pred)); \
+ ref2_reg = _mm256_avg_epu8( \
+ ref2_reg, _mm256_loadu_si256((__m256i const *)(second_pred + 32))); \
+ sad1_reg = _mm256_sad_epu8( \
+ ref1_reg, _mm256_loadu_si256((__m256i const *)src_ptr)); \
+ sad2_reg = _mm256_sad_epu8( \
+ ref2_reg, _mm256_loadu_si256((__m256i const *)(src_ptr + 32))); \
+ sum_sad = \
+ _mm256_add_epi32(sum_sad, _mm256_add_epi32(sad1_reg, sad2_reg)); \
+ ref_ptr += ref_stride; \
+ src_ptr += src_stride; \
+ second_pred += 64; \
+ } \
+ sum_sad_h = _mm256_srli_si256(sum_sad, 8); \
+ sum_sad = _mm256_add_epi32(sum_sad, sum_sad_h); \
+ sum_sad128 = _mm256_extracti128_si256(sum_sad, 1); \
+ sum_sad128 = _mm_add_epi32(_mm256_castsi256_si128(sum_sad), sum_sad128); \
+ res = _mm_cvtsi128_si32(sum_sad128); \
+ _mm256_zeroupper(); \
+ return res; \
+ }
+
+#define FSADAVG32_H(h) \
+ unsigned int aom_sad32x##h##_avg_avx2( \
+ const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, \
+ int ref_stride, const uint8_t *second_pred) { \
+ int i, res; \
+ __m256i sad1_reg, sad2_reg, ref1_reg, ref2_reg; \
+ __m256i sum_sad = _mm256_setzero_si256(); \
+ __m256i sum_sad_h; \
+ __m128i sum_sad128; \
+ int ref2_stride = ref_stride << 1; \
+ int src2_stride = src_stride << 1; \
+ int max = h >> 1; \
+ for (i = 0; i < max; i++) { \
+ ref1_reg = _mm256_loadu_si256((__m256i const *)ref_ptr); \
+ ref2_reg = _mm256_loadu_si256((__m256i const *)(ref_ptr + ref_stride)); \
+ ref1_reg = _mm256_avg_epu8( \
+ ref1_reg, _mm256_loadu_si256((__m256i const *)second_pred)); \
+ ref2_reg = _mm256_avg_epu8( \
+ ref2_reg, _mm256_loadu_si256((__m256i const *)(second_pred + 32))); \
+ sad1_reg = _mm256_sad_epu8( \
+ ref1_reg, _mm256_loadu_si256((__m256i const *)src_ptr)); \
+ sad2_reg = _mm256_sad_epu8( \
+ ref2_reg, \
+ _mm256_loadu_si256((__m256i const *)(src_ptr + src_stride))); \
+ sum_sad = \
+ _mm256_add_epi32(sum_sad, _mm256_add_epi32(sad1_reg, sad2_reg)); \
+ ref_ptr += ref2_stride; \
+ src_ptr += src2_stride; \
+ second_pred += 64; \
+ } \
+ sum_sad_h = _mm256_srli_si256(sum_sad, 8); \
+ sum_sad = _mm256_add_epi32(sum_sad, sum_sad_h); \
+ sum_sad128 = _mm256_extracti128_si256(sum_sad, 1); \
+ sum_sad128 = _mm_add_epi32(_mm256_castsi256_si128(sum_sad), sum_sad128); \
+ res = _mm_cvtsi128_si32(sum_sad128); \
+ _mm256_zeroupper(); \
+ return res; \
+ }
+
+#define FSADAVG64 \
+ FSADAVG64_H(64); \
+ FSADAVG64_H(32);
+
+#define FSADAVG32 \
+ FSADAVG32_H(64); \
+ FSADAVG32_H(32); \
+ FSADAVG32_H(16);
+
+/* clang-format off */
+FSADAVG64
+FSADAVG32
+/* clang-format on */
+
+#undef FSADAVG64
+#undef FSADAVG32
+#undef FSADAVG64_H
+#undef FSADAVG32_H
generated by cgit v1.2.3 (git 2.39.1) at 2024-12-17 06:30:56 +0000