Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(2261)

Issues Search
    My Issues | Starred     Open | Closed | All

Unified Diff: source/libvpx/vp9/encoder/x86/vp9_variance_sse2.c

Issue 1162573005: libvpx: Pull from upstream (Closed) Base URL: https://chromium.googlesource.com/chromium/deps/libvpx.git@master
Patch Set: Created 5 years, 7 months ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View side-by-side diff with in-line comments
Download patch
« no previous file with comments | « source/libvpx/vp9/encoder/x86/vp9_variance_impl_intrin_avx2.c ('k') | source/libvpx/vp9/vp9_common.mk » ('j') | no next file with comments »
Expand Comments ('e') | Collapse Comments ('c') | Show Comments Hide Comments ('s')
Index: source/libvpx/vp9/encoder/x86/vp9_variance_sse2.c
diff --git a/source/libvpx/vp9/encoder/x86/vp9_variance_sse2.c b/source/libvpx/vp9/encoder/x86/vp9_variance_sse2.c
index 8490bbbdc2ebaedceee54b102a20fb22adc1952d..961efe34ee665b13c70a3d600fbc888127180e15 100644
--- a/source/libvpx/vp9/encoder/x86/vp9_variance_sse2.c
+++ b/source/libvpx/vp9/encoder/x86/vp9_variance_sse2.c
@@ -10,310 +10,12 @@
#include <emmintrin.h> // SSE2
+#include "./vp9_rtcd.h"
#include "./vpx_config.h"
#include "vp9/encoder/vp9_variance.h"
#include "vpx_ports/mem.h"
-typedef unsigned int (*variance_fn_t) (const unsigned char *src, int src_stride,
- const unsigned char *ref, int ref_stride,
- unsigned int *sse, int *sum);
-
-unsigned int vp9_get_mb_ss_sse2(const int16_t *src) {
- __m128i vsum = _mm_setzero_si128();
- int i;
-
- for (i = 0; i < 32; ++i) {
- const __m128i v = _mm_loadu_si128((const __m128i *)src);
- vsum = _mm_add_epi32(vsum, _mm_madd_epi16(v, v));
- src += 8;
- }
-
- vsum = _mm_add_epi32(vsum, _mm_srli_si128(vsum, 8));
- vsum = _mm_add_epi32(vsum, _mm_srli_si128(vsum, 4));
- return _mm_cvtsi128_si32(vsum);
-}
-
-#define READ64(p, stride, i) \
- _mm_unpacklo_epi8(_mm_cvtsi32_si128(*(const uint32_t *)(p + i * stride)), \
- _mm_cvtsi32_si128(*(const uint32_t *)(p + (i + 1) * stride)))
-
-unsigned int vp9_get4x4var_sse2(const uint8_t *src, int src_stride,
- const uint8_t *ref, int ref_stride,
- unsigned int *sse, int *sum) {
- const __m128i zero = _mm_setzero_si128();
- const __m128i src0 = _mm_unpacklo_epi8(READ64(src, src_stride, 0), zero);
- const __m128i src1 = _mm_unpacklo_epi8(READ64(src, src_stride, 2), zero);
- const __m128i ref0 = _mm_unpacklo_epi8(READ64(ref, ref_stride, 0), zero);
- const __m128i ref1 = _mm_unpacklo_epi8(READ64(ref, ref_stride, 2), zero);
- const __m128i diff0 = _mm_sub_epi16(src0, ref0);
- const __m128i diff1 = _mm_sub_epi16(src1, ref1);
-
- // sum
- __m128i vsum = _mm_add_epi16(diff0, diff1);
- vsum = _mm_add_epi16(vsum, _mm_srli_si128(vsum, 8));
- vsum = _mm_add_epi16(vsum, _mm_srli_si128(vsum, 4));
- vsum = _mm_add_epi16(vsum, _mm_srli_si128(vsum, 2));
- *sum = (int16_t)_mm_extract_epi16(vsum, 0);
-
- // sse
- vsum = _mm_add_epi32(_mm_madd_epi16(diff0, diff0),
- _mm_madd_epi16(diff1, diff1));
- vsum = _mm_add_epi32(vsum, _mm_srli_si128(vsum, 8));
- vsum = _mm_add_epi32(vsum, _mm_srli_si128(vsum, 4));
- *sse = _mm_cvtsi128_si32(vsum);
-
- return 0;
-}
-
-unsigned int vp9_get8x8var_sse2(const uint8_t *src, int src_stride,
- const uint8_t *ref, int ref_stride,
- unsigned int *sse, int *sum) {
- const __m128i zero = _mm_setzero_si128();
- __m128i vsum = _mm_setzero_si128();
- __m128i vsse = _mm_setzero_si128();
- int i;
-
- for (i = 0; i < 8; i += 2) {
- const __m128i src0 = _mm_unpacklo_epi8(_mm_loadl_epi64(
- (const __m128i *)(src + i * src_stride)), zero);
- const __m128i ref0 = _mm_unpacklo_epi8(_mm_loadl_epi64(
- (const __m128i *)(ref + i * ref_stride)), zero);
- const __m128i diff0 = _mm_sub_epi16(src0, ref0);
-
- const __m128i src1 = _mm_unpacklo_epi8(_mm_loadl_epi64(
- (const __m128i *)(src + (i + 1) * src_stride)), zero);
- const __m128i ref1 = _mm_unpacklo_epi8(_mm_loadl_epi64(
- (const __m128i *)(ref + (i + 1) * ref_stride)), zero);
- const __m128i diff1 = _mm_sub_epi16(src1, ref1);
-
- vsum = _mm_add_epi16(vsum, diff0);
- vsum = _mm_add_epi16(vsum, diff1);
- vsse = _mm_add_epi32(vsse, _mm_madd_epi16(diff0, diff0));
- vsse = _mm_add_epi32(vsse, _mm_madd_epi16(diff1, diff1));
- }
-
- // sum
- vsum = _mm_add_epi16(vsum, _mm_srli_si128(vsum, 8));
- vsum = _mm_add_epi16(vsum, _mm_srli_si128(vsum, 4));
- vsum = _mm_add_epi16(vsum, _mm_srli_si128(vsum, 2));
- *sum = (int16_t)_mm_extract_epi16(vsum, 0);
-
- // sse
- vsse = _mm_add_epi32(vsse, _mm_srli_si128(vsse, 8));
- vsse = _mm_add_epi32(vsse, _mm_srli_si128(vsse, 4));
- *sse = _mm_cvtsi128_si32(vsse);
-
- return 0;
-}
-
-unsigned int vp9_get16x16var_sse2(const uint8_t *src, int src_stride,
- const uint8_t *ref, int ref_stride,
- unsigned int *sse, int *sum) {
- const __m128i zero = _mm_setzero_si128();
- __m128i vsum = _mm_setzero_si128();
- __m128i vsse = _mm_setzero_si128();
- int i;
-
- for (i = 0; i < 16; ++i) {
- const __m128i s = _mm_loadu_si128((const __m128i *)src);
- const __m128i r = _mm_loadu_si128((const __m128i *)ref);
-
- const __m128i src0 = _mm_unpacklo_epi8(s, zero);
- const __m128i ref0 = _mm_unpacklo_epi8(r, zero);
- const __m128i diff0 = _mm_sub_epi16(src0, ref0);
-
- const __m128i src1 = _mm_unpackhi_epi8(s, zero);
- const __m128i ref1 = _mm_unpackhi_epi8(r, zero);
- const __m128i diff1 = _mm_sub_epi16(src1, ref1);
-
- vsum = _mm_add_epi16(vsum, diff0);
- vsum = _mm_add_epi16(vsum, diff1);
- vsse = _mm_add_epi32(vsse, _mm_madd_epi16(diff0, diff0));
- vsse = _mm_add_epi32(vsse, _mm_madd_epi16(diff1, diff1));
-
- src += src_stride;
- ref += ref_stride;
- }
-
- // sum
- vsum = _mm_add_epi16(vsum, _mm_srli_si128(vsum, 8));
- vsum = _mm_add_epi16(vsum, _mm_srli_si128(vsum, 4));
- *sum = (int16_t)_mm_extract_epi16(vsum, 0) +
- (int16_t)_mm_extract_epi16(vsum, 1);
-
- // sse
- vsse = _mm_add_epi32(vsse, _mm_srli_si128(vsse, 8));
- vsse = _mm_add_epi32(vsse, _mm_srli_si128(vsse, 4));
- *sse = _mm_cvtsi128_si32(vsse);
-
- return 0;
-}
-
-
-static void variance_sse2(const unsigned char *src, int src_stride,
- const unsigned char *ref, int ref_stride,
- int w, int h, unsigned int *sse, int *sum,
- variance_fn_t var_fn, int block_size) {
- int i, j;
-
- *sse = 0;
- *sum = 0;
-
- for (i = 0; i < h; i += block_size) {
- for (j = 0; j < w; j += block_size) {
- unsigned int sse0;
- int sum0;
- var_fn(src + src_stride * i + j, src_stride,
- ref + ref_stride * i + j, ref_stride, &sse0, &sum0);
- *sse += sse0;
- *sum += sum0;
- }
- }
-}
-
-unsigned int vp9_variance4x4_sse2(const unsigned char *src, int src_stride,
- const unsigned char *ref, int ref_stride,
- unsigned int *sse) {
- int sum;
- vp9_get4x4var_sse2(src, src_stride, ref, ref_stride, sse, &sum);
- return *sse - (((unsigned int)sum * sum) >> 4);
-}
-
-unsigned int vp9_variance8x4_sse2(const uint8_t *src, int src_stride,
- const uint8_t *ref, int ref_stride,
- unsigned int *sse) {
- int sum;
- variance_sse2(src, src_stride, ref, ref_stride, 8, 4,
- sse, &sum, vp9_get4x4var_sse2, 4);
- return *sse - (((unsigned int)sum * sum) >> 5);
-}
-
-unsigned int vp9_variance4x8_sse2(const uint8_t *src, int src_stride,
- const uint8_t *ref, int ref_stride,
- unsigned int *sse) {
- int sum;
- variance_sse2(src, src_stride, ref, ref_stride, 4, 8,
- sse, &sum, vp9_get4x4var_sse2, 4);
- return *sse - (((unsigned int)sum * sum) >> 5);
-}
-
-unsigned int vp9_variance8x8_sse2(const unsigned char *src, int src_stride,
- const unsigned char *ref, int ref_stride,
- unsigned int *sse) {
- int sum;
- vp9_get8x8var_sse2(src, src_stride, ref, ref_stride, sse, &sum);
- return *sse - (((unsigned int)sum * sum) >> 6);
-}
-
-unsigned int vp9_variance16x8_sse2(const unsigned char *src, int src_stride,
- const unsigned char *ref, int ref_stride,
- unsigned int *sse) {
- int sum;
- variance_sse2(src, src_stride, ref, ref_stride, 16, 8,
- sse, &sum, vp9_get8x8var_sse2, 8);
- return *sse - (((unsigned int)sum * sum) >> 7);
-}
-
-unsigned int vp9_variance8x16_sse2(const unsigned char *src, int src_stride,
- const unsigned char *ref, int ref_stride,
- unsigned int *sse) {
- int sum;
- variance_sse2(src, src_stride, ref, ref_stride, 8, 16,
- sse, &sum, vp9_get8x8var_sse2, 8);
- return *sse - (((unsigned int)sum * sum) >> 7);
-}
-
-unsigned int vp9_variance16x16_sse2(const unsigned char *src, int src_stride,
- const unsigned char *ref, int ref_stride,
- unsigned int *sse) {
- int sum;
- vp9_get16x16var_sse2(src, src_stride, ref, ref_stride, sse, &sum);
- return *sse - (((unsigned int)sum * sum) >> 8);
-}
-
-unsigned int vp9_variance32x32_sse2(const uint8_t *src, int src_stride,
- const uint8_t *ref, int ref_stride,
- unsigned int *sse) {
- int sum;
- variance_sse2(src, src_stride, ref, ref_stride, 32, 32,
- sse, &sum, vp9_get16x16var_sse2, 16);
- return *sse - (((int64_t)sum * sum) >> 10);
-}
-
-unsigned int vp9_variance32x16_sse2(const uint8_t *src, int src_stride,
- const uint8_t *ref, int ref_stride,
- unsigned int *sse) {
- int sum;
- variance_sse2(src, src_stride, ref, ref_stride, 32, 16,
- sse, &sum, vp9_get16x16var_sse2, 16);
- return *sse - (((int64_t)sum * sum) >> 9);
-}
-
-unsigned int vp9_variance16x32_sse2(const uint8_t *src, int src_stride,
- const uint8_t *ref, int ref_stride,
- unsigned int *sse) {
- int sum;
- variance_sse2(src, src_stride, ref, ref_stride, 16, 32,
- sse, &sum, vp9_get16x16var_sse2, 16);
- return *sse - (((int64_t)sum * sum) >> 9);
-}
-
-unsigned int vp9_variance64x64_sse2(const uint8_t *src, int src_stride,
- const uint8_t *ref, int ref_stride,
- unsigned int *sse) {
- int sum;
- variance_sse2(src, src_stride, ref, ref_stride, 64, 64,
- sse, &sum, vp9_get16x16var_sse2, 16);
- return *sse - (((int64_t)sum * sum) >> 12);
-}
-
-unsigned int vp9_variance64x32_sse2(const uint8_t *src, int src_stride,
- const uint8_t *ref, int ref_stride,
- unsigned int *sse) {
- int sum;
- variance_sse2(src, src_stride, ref, ref_stride, 64, 32,
- sse, &sum, vp9_get16x16var_sse2, 16);
- return *sse - (((int64_t)sum * sum) >> 11);
-}
-
-unsigned int vp9_variance32x64_sse2(const uint8_t *src, int src_stride,
- const uint8_t *ref, int ref_stride,
- unsigned int *sse) {
- int sum;
- variance_sse2(src, src_stride, ref, ref_stride, 32, 64,
- sse, &sum, vp9_get16x16var_sse2, 16);
- return *sse - (((int64_t)sum * sum) >> 11);
-}
-
-unsigned int vp9_mse8x8_sse2(const uint8_t *src, int src_stride,
- const uint8_t *ref, int ref_stride,
- unsigned int *sse) {
- vp9_variance8x8_sse2(src, src_stride, ref, ref_stride, sse);
- return *sse;
-}
-
-unsigned int vp9_mse8x16_sse2(const uint8_t *src, int src_stride,
- const uint8_t *ref, int ref_stride,
- unsigned int *sse) {
- vp9_variance8x16_sse2(src, src_stride, ref, ref_stride, sse);
- return *sse;
-}
-
-unsigned int vp9_mse16x8_sse2(const uint8_t *src, int src_stride,
- const uint8_t *ref, int ref_stride,
- unsigned int *sse) {
- vp9_variance16x8_sse2(src, src_stride, ref, ref_stride, sse);
- return *sse;
-}
-
-unsigned int vp9_mse16x16_sse2(const uint8_t *src, int src_stride,
- const uint8_t *ref, int ref_stride,
- unsigned int *sse) {
- vp9_variance16x16_sse2(src, src_stride, ref, ref_stride, sse);
- return *sse;
-}
-
// The 2 unused parameters are place holders for PIC enabled build.
#define DECL(w, opt) \
int vp9_sub_pixel_variance##w##xh_##opt(const uint8_t *src, \
« no previous file with comments | « source/libvpx/vp9/encoder/x86/vp9_variance_impl_intrin_avx2.c ('k') | source/libvpx/vp9/vp9_common.mk » ('j') | no next file with comments »

Powered by Google App Engine
This is Rietveld 408576698