generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Removing extra params from x_add_residual() functions

Browse Source 
Now that the predictor is the dest, we do not need the
extra parameters.

Change-Id: I31e2c3d2015f4a1cd12e7f04536d8db478582a0a
This commit is contained in:
Scott LaVarnway
2013-04-16 09:59:01 -04:00
parent 74610b1ae4
commit 6f95d53e37
3 changed files with 111 additions and 125 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
vp9/common/vp9_rtcd_defs.sh
View File

@@ -108,25 +108,25 @@ prototype void vp9_intra_uv4x4_predict "struct macroblockd *xd, struct blockd *x
specialize vp9_intra_uv4x4_predict; specialize vp9_intra_uv4x4_predict;
if [ "$CONFIG_VP9_DECODER" = "yes" ]; then if [ "$CONFIG_VP9_DECODER" = "yes" ]; then
prototype void vp9_add_residual_4x4 "const int16_t *diff, const uint8_t *pred, int pitch, uint8_t *dest, int stride" prototype void vp9_add_residual_4x4 "const int16_t *diff, uint8_t *dest, int stride"
specialize vp9_add_residual_4x4 sse2 specialize vp9_add_residual_4x4 sse2
prototype void vp9_add_residual_8x8 "const int16_t *diff, const uint8_t *pred, int pitch, uint8_t *dest, int stride" prototype void vp9_add_residual_8x8 "const int16_t *diff, uint8_t *dest, int stride"
specialize vp9_add_residual_8x8 sse2 specialize vp9_add_residual_8x8 sse2
prototype void vp9_add_residual_16x16 "const int16_t *diff, const uint8_t *pred, int pitch, uint8_t *dest, int stride" prototype void vp9_add_residual_16x16 "const int16_t *diff, uint8_t *dest, int stride"
specialize vp9_add_residual_16x16 sse2 specialize vp9_add_residual_16x16 sse2
prototype void vp9_add_residual_32x32 "const int16_t *diff, const uint8_t *pred, int pitch, uint8_t *dest, int stride" prototype void vp9_add_residual_32x32 "const int16_t *diff, uint8_t *dest, int stride"
specialize vp9_add_residual_32x32 sse2 specialize vp9_add_residual_32x32 sse2
prototype void vp9_add_constant_residual_8x8 "const int16_t diff, const uint8_t *pred, int pitch, uint8_t *dest, int stride" prototype void vp9_add_constant_residual_8x8 "const int16_t diff, uint8_t *dest, int stride"
specialize vp9_add_constant_residual_8x8 sse2 specialize vp9_add_constant_residual_8x8 sse2
prototype void vp9_add_constant_residual_16x16 "const int16_t diff, const uint8_t *pred, int pitch, uint8_t *dest, int stride" prototype void vp9_add_constant_residual_16x16 "const int16_t diff, uint8_t *dest, int stride"
specialize vp9_add_constant_residual_16x16 sse2 specialize vp9_add_constant_residual_16x16 sse2
prototype void vp9_add_constant_residual_32x32 "const int16_t diff, const uint8_t *pred, int pitch, uint8_t *dest, int stride" prototype void vp9_add_constant_residual_32x32 "const int16_t diff, uint8_t *dest, int stride"
specialize vp9_add_constant_residual_32x32 sse2 specialize vp9_add_constant_residual_32x32 sse2
fi fi

87
vp9/decoder/vp9_dequantize.c
View File

@@ -16,67 +16,60 @@
#include "vp9/common/vp9_common.h" #include "vp9/common/vp9_common.h"
static void add_residual(const int16_t *diff, const uint8_t *pred, int pitch, static void add_residual(const int16_t *diff, uint8_t *dest, int stride,
uint8_t *dest, int stride, int width, int height) { int width, int height) {
int r, c; int r, c;
for (r = 0; r < height; r++) { for (r = 0; r < height; r++) {
for (c = 0; c < width; c++) for (c = 0; c < width; c++)
dest[c] = clip_pixel(diff[c] + pred[c]); dest[c] = clip_pixel(diff[c] + dest[c]);
dest += stride; dest += stride;
diff += width; diff += width;
pred += pitch;
} }
} }
void vp9_add_residual_4x4_c(const int16_t *diff, const uint8_t *pred, int pitch, void vp9_add_residual_4x4_c(const int16_t *diff, uint8_t *dest, int stride) {
uint8_t *dest, int stride) { add_residual(diff, dest, stride, 4, 4);
add_residual(diff, pred, pitch, dest, stride, 4, 4);
} }
void vp9_add_residual_8x8_c(const int16_t *diff, const uint8_t *pred, int pitch, void vp9_add_residual_8x8_c(const int16_t *diff, uint8_t *dest, int stride) {
uint8_t *dest, int stride) { add_residual(diff, dest, stride, 8, 8);
add_residual(diff, pred, pitch, dest, stride, 8, 8);
} }
void vp9_add_residual_16x16_c(const int16_t *diff, const uint8_t *pred, void vp9_add_residual_16x16_c(const int16_t *diff, uint8_t *dest, int stride) {
int pitch, uint8_t *dest, int stride) { add_residual(diff, dest, stride, 16, 16);
add_residual(diff, pred, pitch, dest, stride, 16, 16);
} }
void vp9_add_residual_32x32_c(const int16_t *diff, const uint8_t *pred, void vp9_add_residual_32x32_c(const int16_t *diff,uint8_t *dest, int stride) {
int pitch, uint8_t *dest, int stride) { add_residual(diff, dest, stride, 32, 32);
add_residual(diff, pred, pitch, dest, stride, 32, 32);
} }
static void add_constant_residual(const int16_t diff, const uint8_t *pred, static void add_constant_residual(const int16_t diff, uint8_t *dest, int stride,
int pitch, uint8_t *dest, int stride,
int width, int height) { int width, int height) {
int r, c; int r, c;
for (r = 0; r < height; r++) { for (r = 0; r < height; r++) {
for (c = 0; c < width; c++) for (c = 0; c < width; c++)
dest[c] = clip_pixel(diff + pred[c]); dest[c] = clip_pixel(diff + dest[c]);
dest += stride; dest += stride;
pred += pitch;
} }
} }
void vp9_add_constant_residual_8x8_c(const int16_t diff, const uint8_t *pred, void vp9_add_constant_residual_8x8_c(const int16_t diff, uint8_t *dest,
int pitch, uint8_t *dest, int stride) { int stride) {
add_constant_residual(diff, pred, pitch, dest, stride, 8, 8); add_constant_residual(diff, dest, stride, 8, 8);
} }
void vp9_add_constant_residual_16x16_c(const int16_t diff, const uint8_t *pred, void vp9_add_constant_residual_16x16_c(const int16_t diff, uint8_t *dest,
int pitch, uint8_t *dest, int stride) { int stride) {
add_constant_residual(diff, pred, pitch, dest, stride, 16, 16); add_constant_residual(diff, dest, stride, 16, 16);
} }
void vp9_add_constant_residual_32x32_c(const int16_t diff, const uint8_t *pred, void vp9_add_constant_residual_32x32_c(const int16_t diff, uint8_t *dest,
int pitch, uint8_t *dest, int stride) { int stride) {
add_constant_residual(diff, pred, pitch, dest, stride, 32, 32); add_constant_residual(diff, dest, stride, 32, 32);
} }
void vp9_dequant_iht_add_c(TX_TYPE tx_type, int16_t *input, void vp9_dequant_iht_add_c(TX_TYPE tx_type, int16_t *input,
@@ -90,7 +83,7 @@ void vp9_dequant_iht_add_c(TX_TYPE tx_type, int16_t *input,
vp9_short_iht4x4(input, output, 4, tx_type); vp9_short_iht4x4(input, output, 4, tx_type);
vpx_memset(input, 0, 32); vpx_memset(input, 0, 32);
vp9_add_residual_4x4(output, dest, stride, dest, stride); vp9_add_residual_4x4(output, dest, stride);
} }
void vp9_dequant_iht_add_8x8_c(TX_TYPE tx_type, int16_t *input, void vp9_dequant_iht_add_8x8_c(TX_TYPE tx_type, int16_t *input,
@@ -107,7 +100,7 @@ void vp9_dequant_iht_add_8x8_c(TX_TYPE tx_type, int16_t *input,
vp9_short_iht8x8(input, output, 8, tx_type); vp9_short_iht8x8(input, output, 8, tx_type);
vpx_memset(input, 0, 128); vpx_memset(input, 0, 128);
vp9_add_residual_8x8(output, dest, stride, dest, stride); vp9_add_residual_8x8(output, dest, stride);
} }
} }
@@ -123,15 +116,15 @@ void vp9_dequant_idct_add_c(int16_t *input, const int16_t *dq, uint8_t *dest,
// the idct halves ( >> 1) the pitch // the idct halves ( >> 1) the pitch
vp9_short_idct4x4(input, output, 4 << 1); vp9_short_idct4x4(input, output, 4 << 1);
vpx_memset(input, 0, 32); vpx_memset(input, 0, 32);
vp9_add_residual_4x4(output, dest, stride, dest, stride); vp9_add_residual_4x4(output, dest, stride);
} else { } else {
vp9_dc_only_idct_add(input[0]*dq[0], dest, dest, stride, stride); vp9_dc_only_idct_add(input[0]*dq[0], dest, dest, stride, stride);
((int *)input)[0] = 0; ((int *)input)[0] = 0;
} }
} }
void vp9_dequant_dc_idct_add_c(int16_t *input, const int16_t *dq, uint8_t *pred, void vp9_dequant_dc_idct_add_c(int16_t *input, const int16_t *dq, uint8_t *dest,
uint8_t *dest, int pitch, int stride, int dc) { int stride, int dc) {
int i; int i;
DECLARE_ALIGNED_ARRAY(16, int16_t, output, 16); DECLARE_ALIGNED_ARRAY(16, int16_t, output, 16);
@@ -143,7 +136,7 @@ void vp9_dequant_dc_idct_add_c(int16_t *input, const int16_t *dq, uint8_t *pred,
// the idct halves ( >> 1) the pitch // the idct halves ( >> 1) the pitch
vp9_short_idct4x4(input, output, 4 << 1); vp9_short_idct4x4(input, output, 4 << 1);
vpx_memset(input, 0, 32); vpx_memset(input, 0, 32);
vp9_add_residual_4x4(output, pred, pitch, dest, stride); vp9_add_residual_4x4(output, dest, stride);
} }
void vp9_dequant_idct_add_lossless_c(int16_t *input, const int16_t *dq, void vp9_dequant_idct_add_lossless_c(int16_t *input, const int16_t *dq,
@@ -157,7 +150,7 @@ void vp9_dequant_idct_add_lossless_c(int16_t *input, const int16_t *dq,
vp9_short_iwalsh4x4_c(input, output, 4 << 1); vp9_short_iwalsh4x4_c(input, output, 4 << 1);
vpx_memset(input, 0, 32); vpx_memset(input, 0, 32);
vp9_add_residual_4x4(output, dest, stride, dest, stride); vp9_add_residual_4x4(output, dest, stride);
} else { } else {
vp9_dc_only_inv_walsh_add(input[0]*dq[0], dest, dest, stride, stride); vp9_dc_only_inv_walsh_add(input[0]*dq[0], dest, dest, stride, stride);
((int *)input)[0] = 0; ((int *)input)[0] = 0;
@@ -176,7 +169,7 @@ void vp9_dequant_dc_idct_add_lossless_c(int16_t *input, const int16_t *dq,
vp9_short_iwalsh4x4_c(input, output, 4 << 1); vp9_short_iwalsh4x4_c(input, output, 4 << 1);
vpx_memset(input, 0, 32); vpx_memset(input, 0, 32);
vp9_add_residual_4x4(output, dest, stride, dest, stride); vp9_add_residual_4x4(output, dest, stride);
} }
void vp9_dequant_idct_add_8x8_c(int16_t *input, const int16_t *dq, void vp9_dequant_idct_add_8x8_c(int16_t *input, const int16_t *dq,
@@ -202,7 +195,7 @@ void vp9_dequant_idct_add_8x8_c(int16_t *input, const int16_t *dq,
vp9_short_idct1_8x8_c(&in, &out); vp9_short_idct1_8x8_c(&in, &out);
input[0] = 0; input[0] = 0;
vp9_add_constant_residual_8x8(out, dest, stride, dest, stride); vp9_add_constant_residual_8x8(out, dest, stride);
#if !CONFIG_SCATTERSCAN #if !CONFIG_SCATTERSCAN
} else if (eob <= 10) { } else if (eob <= 10) {
input[1] *= dq[1]; input[1] *= dq[1];
@@ -222,7 +215,7 @@ void vp9_dequant_idct_add_8x8_c(int16_t *input, const int16_t *dq,
input[16] = input[17] = 0; input[16] = input[17] = 0;
input[24] = 0; input[24] = 0;
vp9_add_residual_8x8(output, dest, stride, dest, stride); vp9_add_residual_8x8(output, dest, stride);
#endif #endif
} else { } else {
int i; int i;
@@ -234,7 +227,7 @@ void vp9_dequant_idct_add_8x8_c(int16_t *input, const int16_t *dq,
// the idct halves ( >> 1) the pitch // the idct halves ( >> 1) the pitch
vp9_short_idct8x8(input, output, 8 << 1); vp9_short_idct8x8(input, output, 8 << 1);
vpx_memset(input, 0, 128); vpx_memset(input, 0, 128);
vp9_add_residual_8x8(output, dest, stride, dest, stride); vp9_add_residual_8x8(output, dest, stride);
} }
} }
} }
@@ -262,7 +255,7 @@ void vp9_dequant_iht_add_16x16_c(TX_TYPE tx_type, int16_t *input,
vpx_memset(input, 0, 512); vpx_memset(input, 0, 512);
vp9_add_residual_16x16(output, dest, stride, dest, stride); vp9_add_residual_16x16(output, dest, stride);
} }
} }
@@ -282,7 +275,7 @@ void vp9_dequant_idct_add_16x16_c(int16_t *input, const int16_t *dq,
vp9_short_idct1_16x16_c(&in, &out); vp9_short_idct1_16x16_c(&in, &out);
input[0] = 0; input[0] = 0;
vp9_add_constant_residual_16x16(out, dest, stride, dest, stride); vp9_add_constant_residual_16x16(out, dest, stride);
#if !CONFIG_SCATTERSCAN #if !CONFIG_SCATTERSCAN
} else if (eob <= 10) { } else if (eob <= 10) {
input[0] *= dq[0]; input[0] *= dq[0];
@@ -305,7 +298,7 @@ void vp9_dequant_idct_add_16x16_c(int16_t *input, const int16_t *dq,
input[32] = input[33] = 0; input[32] = input[33] = 0;
input[48] = 0; input[48] = 0;
vp9_add_residual_16x16(output, dest, stride, dest, stride); vp9_add_residual_16x16(output, dest, stride);
#endif #endif
} else { } else {
int i; int i;
@@ -319,7 +312,7 @@ void vp9_dequant_idct_add_16x16_c(int16_t *input, const int16_t *dq,
// the idct halves ( >> 1) the pitch // the idct halves ( >> 1) the pitch
vp9_short_idct16x16(input, output, 16 << 1); vp9_short_idct16x16(input, output, 16 << 1);
vpx_memset(input, 0, 512); vpx_memset(input, 0, 512);
vp9_add_residual_16x16(output, dest, stride, dest, stride); vp9_add_residual_16x16(output, dest, stride);
} }
} }
} }
@@ -332,7 +325,7 @@ void vp9_dequant_idct_add_32x32_c(int16_t *input, const int16_t *dq,
input[0] = input[0] * dq[0] / 2; input[0] = input[0] * dq[0] / 2;
if (eob == 1) { if (eob == 1) {
vp9_short_idct1_32x32(input, output); vp9_short_idct1_32x32(input, output);
vp9_add_constant_residual_32x32(output[0], dest, stride, dest, stride); vp9_add_constant_residual_32x32(output[0], dest, stride);
input[0] = 0; input[0] = 0;
#if !CONFIG_SCATTERSCAN #if !CONFIG_SCATTERSCAN
} else if (eob <= 10) { } else if (eob <= 10) {
@@ -354,7 +347,7 @@ void vp9_dequant_idct_add_32x32_c(int16_t *input, const int16_t *dq,
input[64] = input[65] = 0; input[64] = input[65] = 0;
input[96] = 0; input[96] = 0;
vp9_add_residual_32x32(output, dest, stride, dest, stride); vp9_add_residual_32x32(output, dest, stride);
#endif #endif
} else { } else {
int i; int i;
@@ -362,7 +355,7 @@ void vp9_dequant_idct_add_32x32_c(int16_t *input, const int16_t *dq,
input[i] = input[i] * dq[1] / 2; input[i] = input[i] * dq[1] / 2;
vp9_short_idct32x32(input, output, 64); vp9_short_idct32x32(input, output, 64);
vpx_memset(input, 0, 2048); vpx_memset(input, 0, 2048);
vp9_add_residual_32x32(output, dest, stride, dest, stride); vp9_add_residual_32x32(output, dest, stride);
} }
} }
} }

135
vp9/decoder/x86/vp9_dequantize_x86.c
View File

@@ -17,8 +17,7 @@
#if HAVE_SSE2 #if HAVE_SSE2
void vp9_add_residual_4x4_sse2(const int16_t *diff, const uint8_t *pred, void vp9_add_residual_4x4_sse2(const int16_t *diff, uint8_t *dest, int stride) {
int pitch, uint8_t *dest, int stride) {
const int width = 4; const int width = 4;
const __m128i zero = _mm_setzero_si128(); const __m128i zero = _mm_setzero_si128();
@@ -29,10 +28,10 @@ void vp9_add_residual_4x4_sse2(const int16_t *diff, const uint8_t *pred,
const __m128i d3 = _mm_loadl_epi64((const __m128i *)(diff + 3 * width)); const __m128i d3 = _mm_loadl_epi64((const __m128i *)(diff + 3 * width));
// Prediction data. // Prediction data.
__m128i p0 = _mm_cvtsi32_si128(*(const int *)(pred + 0 * pitch)); __m128i p0 = _mm_cvtsi32_si128(*(const int *)(dest + 0 * stride));
__m128i p1 = _mm_cvtsi32_si128(*(const int *)(pred + 1 * pitch)); __m128i p1 = _mm_cvtsi32_si128(*(const int *)(dest + 1 * stride));
__m128i p2 = _mm_cvtsi32_si128(*(const int *)(pred + 2 * pitch)); __m128i p2 = _mm_cvtsi32_si128(*(const int *)(dest + 2 * stride));
__m128i p3 = _mm_cvtsi32_si128(*(const int *)(pred + 3 * pitch)); __m128i p3 = _mm_cvtsi32_si128(*(const int *)(dest + 3 * stride));
p0 = _mm_unpacklo_epi8(p0, zero); p0 = _mm_unpacklo_epi8(p0, zero);
p1 = _mm_unpacklo_epi8(p1, zero); p1 = _mm_unpacklo_epi8(p1, zero);
@@ -61,8 +60,7 @@ void vp9_add_residual_4x4_sse2(const int16_t *diff, const uint8_t *pred,
*(int *)dest = _mm_cvtsi128_si32(p2); *(int *)dest = _mm_cvtsi128_si32(p2);
} }
void vp9_add_residual_8x8_sse2(const int16_t *diff, const uint8_t *pred, void vp9_add_residual_8x8_sse2(const int16_t *diff, uint8_t *dest, int stride) {
int pitch, uint8_t *dest, int stride) {
const int width = 8; const int width = 8;
const __m128i zero = _mm_setzero_si128(); const __m128i zero = _mm_setzero_si128();
@@ -77,14 +75,14 @@ void vp9_add_residual_8x8_sse2(const int16_t *diff, const uint8_t *pred,
const __m128i d7 = _mm_load_si128((const __m128i *)(diff + 7 * width)); const __m128i d7 = _mm_load_si128((const __m128i *)(diff + 7 * width));
// Prediction data. // Prediction data.
__m128i p0 = _mm_loadl_epi64((const __m128i *)(pred + 0 * pitch)); __m128i p0 = _mm_loadl_epi64((const __m128i *)(dest + 0 * stride));
__m128i p1 = _mm_loadl_epi64((const __m128i *)(pred + 1 * pitch)); __m128i p1 = _mm_loadl_epi64((const __m128i *)(dest + 1 * stride));
__m128i p2 = _mm_loadl_epi64((const __m128i *)(pred + 2 * pitch)); __m128i p2 = _mm_loadl_epi64((const __m128i *)(dest + 2 * stride));
__m128i p3 = _mm_loadl_epi64((const __m128i *)(pred + 3 * pitch)); __m128i p3 = _mm_loadl_epi64((const __m128i *)(dest + 3 * stride));
__m128i p4 = _mm_loadl_epi64((const __m128i *)(pred + 4 * pitch)); __m128i p4 = _mm_loadl_epi64((const __m128i *)(dest + 4 * stride));
__m128i p5 = _mm_loadl_epi64((const __m128i *)(pred + 5 * pitch)); __m128i p5 = _mm_loadl_epi64((const __m128i *)(dest + 5 * stride));
__m128i p6 = _mm_loadl_epi64((const __m128i *)(pred + 6 * pitch)); __m128i p6 = _mm_loadl_epi64((const __m128i *)(dest + 6 * stride));
__m128i p7 = _mm_loadl_epi64((const __m128i *)(pred + 7 * pitch)); __m128i p7 = _mm_loadl_epi64((const __m128i *)(dest + 7 * stride));
p0 = _mm_unpacklo_epi8(p0, zero); p0 = _mm_unpacklo_epi8(p0, zero);
p1 = _mm_unpacklo_epi8(p1, zero); p1 = _mm_unpacklo_epi8(p1, zero);
@@ -126,8 +124,8 @@ void vp9_add_residual_8x8_sse2(const int16_t *diff, const uint8_t *pred,
_mm_storel_epi64((__m128i *)(dest + 7 * stride), p6); _mm_storel_epi64((__m128i *)(dest + 7 * stride), p6);
} }
void vp9_add_residual_16x16_sse2(const int16_t *diff, const uint8_t *pred, void vp9_add_residual_16x16_sse2(const int16_t *diff, uint8_t *dest,
int pitch, uint8_t *dest, int stride) { int stride) {
const int width = 16; const int width = 16;
int i = 4; int i = 4;
const __m128i zero = _mm_setzero_si128(); const __m128i zero = _mm_setzero_si128();
@@ -147,10 +145,10 @@ void vp9_add_residual_16x16_sse2(const int16_t *diff, const uint8_t *pred,
d7 = _mm_load_si128((const __m128i *)(diff + 3 * width + 8)); d7 = _mm_load_si128((const __m128i *)(diff + 3 * width + 8));
// Prediction data. // Prediction data.
p1 = _mm_load_si128((const __m128i *)(pred + 0 * pitch)); p1 = _mm_load_si128((const __m128i *)(dest + 0 * stride));
p3 = _mm_load_si128((const __m128i *)(pred + 1 * pitch)); p3 = _mm_load_si128((const __m128i *)(dest + 1 * stride));
p5 = _mm_load_si128((const __m128i *)(pred + 2 * pitch)); p5 = _mm_load_si128((const __m128i *)(dest + 2 * stride));
p7 = _mm_load_si128((const __m128i *)(pred + 3 * pitch)); p7 = _mm_load_si128((const __m128i *)(dest + 3 * stride));
p0 = _mm_unpacklo_epi8(p1, zero); p0 = _mm_unpacklo_epi8(p1, zero);
p1 = _mm_unpackhi_epi8(p1, zero); p1 = _mm_unpackhi_epi8(p1, zero);
@@ -181,13 +179,12 @@ void vp9_add_residual_16x16_sse2(const int16_t *diff, const uint8_t *pred,
_mm_store_si128((__m128i *)(dest + 3 * stride), p3); _mm_store_si128((__m128i *)(dest + 3 * stride), p3);
diff += 4 * width; diff += 4 * width;
pred += 4 * pitch;
dest += 4 * stride; dest += 4 * stride;
} while (--i); } while (--i);
} }
void vp9_add_residual_32x32_sse2(const int16_t *diff, const uint8_t *pred, void vp9_add_residual_32x32_sse2(const int16_t *diff, uint8_t *dest,
int pitch, uint8_t *dest, int stride) { int stride) {
const int width = 32; const int width = 32;
int i = 16; int i = 16;
const __m128i zero = _mm_setzero_si128(); const __m128i zero = _mm_setzero_si128();
@@ -207,10 +204,10 @@ void vp9_add_residual_32x32_sse2(const int16_t *diff, const uint8_t *pred,
d7 = _mm_load_si128((const __m128i *)(diff + 1 * width + 24)); d7 = _mm_load_si128((const __m128i *)(diff + 1 * width + 24));
// Prediction data. // Prediction data.
p1 = _mm_load_si128((const __m128i *)(pred + 0 * pitch)); p1 = _mm_load_si128((const __m128i *)(dest + 0 * stride));
p3 = _mm_load_si128((const __m128i *)(pred + 0 * pitch + 16)); p3 = _mm_load_si128((const __m128i *)(dest + 0 * stride + 16));
p5 = _mm_load_si128((const __m128i *)(pred + 1 * pitch)); p5 = _mm_load_si128((const __m128i *)(dest + 1 * stride));
p7 = _mm_load_si128((const __m128i *)(pred + 1 * pitch + 16)); p7 = _mm_load_si128((const __m128i *)(dest + 1 * stride + 16));
p0 = _mm_unpacklo_epi8(p1, zero); p0 = _mm_unpacklo_epi8(p1, zero);
p1 = _mm_unpackhi_epi8(p1, zero); p1 = _mm_unpackhi_epi8(p1, zero);
@@ -241,25 +238,24 @@ void vp9_add_residual_32x32_sse2(const int16_t *diff, const uint8_t *pred,
_mm_store_si128((__m128i *)(dest + 1 * stride + 16), p3); _mm_store_si128((__m128i *)(dest + 1 * stride + 16), p3);
diff += 2 * width; diff += 2 * width;
pred += 2 * pitch;
dest += 2 * stride; dest += 2 * stride;
} while (--i); } while (--i);
} }
void vp9_add_constant_residual_8x8_sse2(const int16_t diff, const uint8_t *pred, void vp9_add_constant_residual_8x8_sse2(const int16_t diff, uint8_t *dest,
int pitch, uint8_t *dest, int stride) { int stride) {
uint8_t abs_diff; uint8_t abs_diff;
__m128i d; __m128i d;
// Prediction data. // Prediction data.
__m128i p0 = _mm_loadl_epi64((const __m128i *)(pred + 0 * pitch)); __m128i p0 = _mm_loadl_epi64((const __m128i *)(dest + 0 * stride));
__m128i p1 = _mm_loadl_epi64((const __m128i *)(pred + 1 * pitch)); __m128i p1 = _mm_loadl_epi64((const __m128i *)(dest + 1 * stride));
__m128i p2 = _mm_loadl_epi64((const __m128i *)(pred + 2 * pitch)); __m128i p2 = _mm_loadl_epi64((const __m128i *)(dest + 2 * stride));
__m128i p3 = _mm_loadl_epi64((const __m128i *)(pred + 3 * pitch)); __m128i p3 = _mm_loadl_epi64((const __m128i *)(dest + 3 * stride));
__m128i p4 = _mm_loadl_epi64((const __m128i *)(pred + 4 * pitch)); __m128i p4 = _mm_loadl_epi64((const __m128i *)(dest + 4 * stride));
__m128i p5 = _mm_loadl_epi64((const __m128i *)(pred + 5 * pitch)); __m128i p5 = _mm_loadl_epi64((const __m128i *)(dest + 5 * stride));
__m128i p6 = _mm_loadl_epi64((const __m128i *)(pred + 6 * pitch)); __m128i p6 = _mm_loadl_epi64((const __m128i *)(dest + 6 * stride));
__m128i p7 = _mm_loadl_epi64((const __m128i *)(pred + 7 * pitch)); __m128i p7 = _mm_loadl_epi64((const __m128i *)(dest + 7 * stride));
p0 = _mm_unpacklo_epi64(p0, p1); p0 = _mm_unpacklo_epi64(p0, p1);
p2 = _mm_unpacklo_epi64(p2, p3); p2 = _mm_unpacklo_epi64(p2, p3);
@@ -303,29 +299,28 @@ void vp9_add_constant_residual_8x8_sse2(const int16_t diff, const uint8_t *pred,
_mm_storel_epi64((__m128i *)(dest + 7 * stride), p6); _mm_storel_epi64((__m128i *)(dest + 7 * stride), p6);
} }
void vp9_add_constant_residual_16x16_sse2(const int16_t diff, void vp9_add_constant_residual_16x16_sse2(const int16_t diff, uint8_t *dest,
const uint8_t *pred, int pitch, int stride) {
uint8_t *dest, int stride) {
uint8_t abs_diff; uint8_t abs_diff;
__m128i d; __m128i d;
// Prediction data. // Prediction data.
__m128i p0 = _mm_load_si128((const __m128i *)(pred + 0 * pitch)); __m128i p0 = _mm_load_si128((const __m128i *)(dest + 0 * stride));
__m128i p1 = _mm_load_si128((const __m128i *)(pred + 1 * pitch)); __m128i p1 = _mm_load_si128((const __m128i *)(dest + 1 * stride));
__m128i p2 = _mm_load_si128((const __m128i *)(pred + 2 * pitch)); __m128i p2 = _mm_load_si128((const __m128i *)(dest + 2 * stride));
__m128i p3 = _mm_load_si128((const __m128i *)(pred + 3 * pitch)); __m128i p3 = _mm_load_si128((const __m128i *)(dest + 3 * stride));
__m128i p4 = _mm_load_si128((const __m128i *)(pred + 4 * pitch)); __m128i p4 = _mm_load_si128((const __m128i *)(dest + 4 * stride));
__m128i p5 = _mm_load_si128((const __m128i *)(pred + 5 * pitch)); __m128i p5 = _mm_load_si128((const __m128i *)(dest + 5 * stride));
__m128i p6 = _mm_load_si128((const __m128i *)(pred + 6 * pitch)); __m128i p6 = _mm_load_si128((const __m128i *)(dest + 6 * stride));
__m128i p7 = _mm_load_si128((const __m128i *)(pred + 7 * pitch)); __m128i p7 = _mm_load_si128((const __m128i *)(dest + 7 * stride));
__m128i p8 = _mm_load_si128((const __m128i *)(pred + 8 * pitch)); __m128i p8 = _mm_load_si128((const __m128i *)(dest + 8 * stride));
__m128i p9 = _mm_load_si128((const __m128i *)(pred + 9 * pitch)); __m128i p9 = _mm_load_si128((const __m128i *)(dest + 9 * stride));
__m128i p10 = _mm_load_si128((const __m128i *)(pred + 10 * pitch)); __m128i p10 = _mm_load_si128((const __m128i *)(dest + 10 * stride));
__m128i p11 = _mm_load_si128((const __m128i *)(pred + 11 * pitch)); __m128i p11 = _mm_load_si128((const __m128i *)(dest + 11 * stride));
__m128i p12 = _mm_load_si128((const __m128i *)(pred + 12 * pitch)); __m128i p12 = _mm_load_si128((const __m128i *)(dest + 12 * stride));
__m128i p13 = _mm_load_si128((const __m128i *)(pred + 13 * pitch)); __m128i p13 = _mm_load_si128((const __m128i *)(dest + 13 * stride));
__m128i p14 = _mm_load_si128((const __m128i *)(pred + 14 * pitch)); __m128i p14 = _mm_load_si128((const __m128i *)(dest + 14 * stride));
__m128i p15 = _mm_load_si128((const __m128i *)(pred + 15 * pitch)); __m128i p15 = _mm_load_si128((const __m128i *)(dest + 15 * stride));
// Clip diff value to [0, 255] range. Then, do addition or subtraction // Clip diff value to [0, 255] range. Then, do addition or subtraction
// according to its sign. // according to its sign.
@@ -390,9 +385,8 @@ void vp9_add_constant_residual_16x16_sse2(const int16_t diff,
_mm_store_si128((__m128i *)(dest + 15 * stride), p15); _mm_store_si128((__m128i *)(dest + 15 * stride), p15);
} }
void vp9_add_constant_residual_32x32_sse2(const int16_t diff, void vp9_add_constant_residual_32x32_sse2(const int16_t diff, uint8_t *dest,
const uint8_t *pred, int pitch, int stride) {
uint8_t *dest, int stride) {
uint8_t abs_diff; uint8_t abs_diff;
__m128i d; __m128i d;
int i = 8; int i = 8;
@@ -407,14 +401,14 @@ void vp9_add_constant_residual_32x32_sse2(const int16_t diff,
do { do {
// Prediction data. // Prediction data.
__m128i p0 = _mm_load_si128((const __m128i *)(pred + 0 * pitch)); __m128i p0 = _mm_load_si128((const __m128i *)(dest + 0 * stride));
__m128i p1 = _mm_load_si128((const __m128i *)(pred + 0 * pitch + 16)); __m128i p1 = _mm_load_si128((const __m128i *)(dest + 0 * stride + 16));
__m128i p2 = _mm_load_si128((const __m128i *)(pred + 1 * pitch)); __m128i p2 = _mm_load_si128((const __m128i *)(dest + 1 * stride));
__m128i p3 = _mm_load_si128((const __m128i *)(pred + 1 * pitch + 16)); __m128i p3 = _mm_load_si128((const __m128i *)(dest + 1 * stride + 16));
__m128i p4 = _mm_load_si128((const __m128i *)(pred + 2 * pitch)); __m128i p4 = _mm_load_si128((const __m128i *)(dest + 2 * stride));
__m128i p5 = _mm_load_si128((const __m128i *)(pred + 2 * pitch + 16)); __m128i p5 = _mm_load_si128((const __m128i *)(dest + 2 * stride + 16));
__m128i p6 = _mm_load_si128((const __m128i *)(pred + 3 * pitch)); __m128i p6 = _mm_load_si128((const __m128i *)(dest + 3 * stride));
__m128i p7 = _mm_load_si128((const __m128i *)(pred + 3 * pitch + 16)); __m128i p7 = _mm_load_si128((const __m128i *)(dest + 3 * stride + 16));
// Clip diff value to [0, 255] range. Then, do addition or subtraction // Clip diff value to [0, 255] range. Then, do addition or subtraction
// according to its sign. // according to its sign.
@@ -448,7 +442,6 @@ void vp9_add_constant_residual_32x32_sse2(const int16_t diff,
_mm_store_si128((__m128i *)(dest + 3 * stride), p6); _mm_store_si128((__m128i *)(dest + 3 * stride), p6);
_mm_store_si128((__m128i *)(dest + 3 * stride + 16), p7); _mm_store_si128((__m128i *)(dest + 3 * stride + 16), p7);
pred += 4 * pitch;
dest += 4 * stride; dest += 4 * stride;
} while (--i); } while (--i);
} }