Commit d5d92898 authored by Johann's avatar Johann

Move shared SAD code to vpx_dsp

Create a new component, vpx_dsp, for code that can be shared
between codecs. Move the SAD code into the component.

This reduces the size of vpxenc/dec by 36k on x86_64 builds.

Change-Id: I73f837ddaecac6b350bf757af0cfe19c4ab9327a
parent c77b1f5a
......@@ -54,6 +54,9 @@ CODEC_SRCS-yes += $(addprefix vpx_scale/,$(call enabled,SCALE_SRCS))
include $(SRC_PATH_BARE)/vpx_ports/vpx_ports.mk
CODEC_SRCS-yes += $(addprefix vpx_ports/,$(call enabled,PORTS_SRCS))
include $(SRC_PATH_BARE)/vpx_dsp/vpx_dsp.mk
CODEC_SRCS-yes += $(addprefix vpx_dsp/,$(call enabled,DSP_SRCS))
ifneq ($(CONFIG_VP8_ENCODER)$(CONFIG_VP8_DECODER),)
VP8_PREFIX=vp8/
include $(SRC_PATH_BARE)/$(VP8_PREFIX)vp8_common.mk
......
This diff is collapsed.
......@@ -19,6 +19,7 @@ extern void vp8_rtcd();
#if CONFIG_VP9
extern void vp9_rtcd();
#endif // CONFIG_VP9
extern void vpx_dsp_rtcd();
extern void vpx_scale_rtcd();
}
#include "third_party/googletest/src/include/gtest/gtest.h"
......@@ -64,6 +65,7 @@ int main(int argc, char **argv) {
#if CONFIG_VP9
vp9_rtcd();
#endif // CONFIG_VP9
vpx_dsp_rtcd();
vpx_scale_rtcd();
#endif // !CONFIG_SHARED
......
/*
* Copyright (c) 2014 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include <arm_neon.h>
unsigned int vp8_sad8x8_neon(
unsigned char *src_ptr,
int src_stride,
unsigned char *ref_ptr,
int ref_stride) {
uint8x8_t d0, d8;
uint16x8_t q12;
uint32x4_t q1;
uint64x2_t q3;
uint32x2_t d5;
int i;
d0 = vld1_u8(src_ptr);
src_ptr += src_stride;
d8 = vld1_u8(ref_ptr);
ref_ptr += ref_stride;
q12 = vabdl_u8(d0, d8);
for (i = 0; i < 7; i++) {
d0 = vld1_u8(src_ptr);
src_ptr += src_stride;
d8 = vld1_u8(ref_ptr);
ref_ptr += ref_stride;
q12 = vabal_u8(q12, d0, d8);
}
q1 = vpaddlq_u16(q12);
q3 = vpaddlq_u32(q1);
d5 = vadd_u32(vreinterpret_u32_u64(vget_low_u64(q3)),
vreinterpret_u32_u64(vget_high_u64(q3)));
return vget_lane_u32(d5, 0);
}
unsigned int vp8_sad8x16_neon(
unsigned char *src_ptr,
int src_stride,
unsigned char *ref_ptr,
int ref_stride) {
uint8x8_t d0, d8;
uint16x8_t q12;
uint32x4_t q1;
uint64x2_t q3;
uint32x2_t d5;
int i;
d0 = vld1_u8(src_ptr);
src_ptr += src_stride;
d8 = vld1_u8(ref_ptr);
ref_ptr += ref_stride;
q12 = vabdl_u8(d0, d8);
for (i = 0; i < 15; i++) {
d0 = vld1_u8(src_ptr);
src_ptr += src_stride;
d8 = vld1_u8(ref_ptr);
ref_ptr += ref_stride;
q12 = vabal_u8(q12, d0, d8);
}
q1 = vpaddlq_u16(q12);
q3 = vpaddlq_u32(q1);
d5 = vadd_u32(vreinterpret_u32_u64(vget_low_u64(q3)),
vreinterpret_u32_u64(vget_high_u64(q3)));
return vget_lane_u32(d5, 0);
}
unsigned int vp8_sad4x4_neon(
unsigned char *src_ptr,
int src_stride,
unsigned char *ref_ptr,
int ref_stride) {
uint8x8_t d0, d8;
uint16x8_t q12;
uint32x2_t d1;
uint64x1_t d3;
int i;
d0 = vld1_u8(src_ptr);
src_ptr += src_stride;
d8 = vld1_u8(ref_ptr);
ref_ptr += ref_stride;
q12 = vabdl_u8(d0, d8);
for (i = 0; i < 3; i++) {
d0 = vld1_u8(src_ptr);
src_ptr += src_stride;
d8 = vld1_u8(ref_ptr);
ref_ptr += ref_stride;
q12 = vabal_u8(q12, d0, d8);
}
d1 = vpaddl_u16(vget_low_u16(q12));
d3 = vpaddl_u32(d1);
return vget_lane_u32(vreinterpret_u32_u64(d3), 0);
}
unsigned int vp8_sad16x16_neon(
unsigned char *src_ptr,
int src_stride,
unsigned char *ref_ptr,
int ref_stride) {
uint8x16_t q0, q4;
uint16x8_t q12, q13;
uint32x4_t q1;
uint64x2_t q3;
uint32x2_t d5;
int i;
q0 = vld1q_u8(src_ptr);
src_ptr += src_stride;
q4 = vld1q_u8(ref_ptr);
ref_ptr += ref_stride;
q12 = vabdl_u8(vget_low_u8(q0), vget_low_u8(q4));
q13 = vabdl_u8(vget_high_u8(q0), vget_high_u8(q4));
for (i = 0; i < 15; i++) {
q0 = vld1q_u8(src_ptr);
src_ptr += src_stride;
q4 = vld1q_u8(ref_ptr);
ref_ptr += ref_stride;
q12 = vabal_u8(q12, vget_low_u8(q0), vget_low_u8(q4));
q13 = vabal_u8(q13, vget_high_u8(q0), vget_high_u8(q4));
}
q12 = vaddq_u16(q12, q13);
q1 = vpaddlq_u16(q12);
q3 = vpaddlq_u32(q1);
d5 = vadd_u32(vreinterpret_u32_u64(vget_low_u64(q3)),
vreinterpret_u32_u64(vget_high_u64(q3)));
return vget_lane_u32(d5, 0);
}
unsigned int vp8_sad16x8_neon(
unsigned char *src_ptr,
int src_stride,
unsigned char *ref_ptr,
int ref_stride) {
uint8x16_t q0, q4;
uint16x8_t q12, q13;
uint32x4_t q1;
uint64x2_t q3;
uint32x2_t d5;
int i;
q0 = vld1q_u8(src_ptr);
src_ptr += src_stride;
q4 = vld1q_u8(ref_ptr);
ref_ptr += ref_stride;
q12 = vabdl_u8(vget_low_u8(q0), vget_low_u8(q4));
q13 = vabdl_u8(vget_high_u8(q0), vget_high_u8(q4));
for (i = 0; i < 7; i++) {
q0 = vld1q_u8(src_ptr);
src_ptr += src_stride;
q4 = vld1q_u8(ref_ptr);
ref_ptr += ref_stride;
q12 = vabal_u8(q12, vget_low_u8(q0), vget_low_u8(q4));
q13 = vabal_u8(q13, vget_high_u8(q0), vget_high_u8(q4));
}
q12 = vaddq_u16(q12, q13);
q1 = vpaddlq_u16(q12);
q3 = vpaddlq_u32(q1);
d5 = vadd_u32(vreinterpret_u32_u64(vget_low_u64(q3)),
vreinterpret_u32_u64(vget_high_u64(q3)));
return vget_lane_u32(d5, 0);
}
/*
* Copyright (c) 2010 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "vpx_config.h"
#include "vpx/vpx_integer.h"
/* Copy 2 macroblocks to a buffer */
void vp8_copy32xn_c(unsigned char *src_ptr, int src_stride,
unsigned char *dst_ptr, int dst_stride,
int height)
{
int r;
for (r = 0; r < height; r++)
{
#if !(CONFIG_FAST_UNALIGNED)
dst_ptr[0] = src_ptr[0];
dst_ptr[1] = src_ptr[1];
dst_ptr[2] = src_ptr[2];
dst_ptr[3] = src_ptr[3];
dst_ptr[4] = src_ptr[4];
dst_ptr[5] = src_ptr[5];
dst_ptr[6] = src_ptr[6];
dst_ptr[7] = src_ptr[7];
dst_ptr[8] = src_ptr[8];
dst_ptr[9] = src_ptr[9];
dst_ptr[10] = src_ptr[10];
dst_ptr[11] = src_ptr[11];
dst_ptr[12] = src_ptr[12];
dst_ptr[13] = src_ptr[13];
dst_ptr[14] = src_ptr[14];
dst_ptr[15] = src_ptr[15];
dst_ptr[16] = src_ptr[16];
dst_ptr[17] = src_ptr[17];
dst_ptr[18] = src_ptr[18];
dst_ptr[19] = src_ptr[19];
dst_ptr[20] = src_ptr[20];
dst_ptr[21] = src_ptr[21];
dst_ptr[22] = src_ptr[22];
dst_ptr[23] = src_ptr[23];
dst_ptr[24] = src_ptr[24];
dst_ptr[25] = src_ptr[25];
dst_ptr[26] = src_ptr[26];
dst_ptr[27] = src_ptr[27];
dst_ptr[28] = src_ptr[28];
dst_ptr[29] = src_ptr[29];
dst_ptr[30] = src_ptr[30];
dst_ptr[31] = src_ptr[31];
#else
((uint32_t *)dst_ptr)[0] = ((uint32_t *)src_ptr)[0] ;
((uint32_t *)dst_ptr)[1] = ((uint32_t *)src_ptr)[1] ;
((uint32_t *)dst_ptr)[2] = ((uint32_t *)src_ptr)[2] ;
((uint32_t *)dst_ptr)[3] = ((uint32_t *)src_ptr)[3] ;
((uint32_t *)dst_ptr)[4] = ((uint32_t *)src_ptr)[4] ;
((uint32_t *)dst_ptr)[5] = ((uint32_t *)src_ptr)[5] ;
((uint32_t *)dst_ptr)[6] = ((uint32_t *)src_ptr)[6] ;
((uint32_t *)dst_ptr)[7] = ((uint32_t *)src_ptr)[7] ;
#endif
src_ptr += src_stride;
dst_ptr += dst_stride;
}
}
......@@ -17,10 +17,11 @@
* higher quality.
*/
#include "postproc.h"
#include "variance.h"
#include "./vp8_rtcd.h"
#include "./vpx_dsp_rtcd.h"
#include "vp8/common/postproc.h"
#include "vp8/common/variance.h"
#include "vpx_mem/vpx_mem.h"
#include "vp8_rtcd.h"
#include "vpx_scale/yv12config.h"
#include <limits.h>
......@@ -160,9 +161,9 @@ static void multiframe_quality_enhance_block
vp8_variance8x8(v, uv_stride, vd, uvd_stride, &sse);
vsad = (sse + 32)>>6;
#else
sad = (vp8_sad16x16(y, y_stride, yd, yd_stride, UINT_MAX) + 128) >> 8;
usad = (vp8_sad8x8(u, uv_stride, ud, uvd_stride, UINT_MAX) + 32) >> 6;
vsad = (vp8_sad8x8(v, uv_stride, vd, uvd_stride, UINT_MAX)+ 32) >> 6;
sad = (vpx_sad16x16(y, y_stride, yd, yd_stride) + 128) >> 8;
usad = (vpx_sad8x8(u, uv_stride, ud, uvd_stride) + 32) >> 6;
vsad = (vpx_sad8x8(v, uv_stride, vd, uvd_stride)+ 32) >> 6;
#endif
}
else /* if (blksize == 8) */
......@@ -177,9 +178,9 @@ static void multiframe_quality_enhance_block
vp8_variance4x4(v, uv_stride, vd, uvd_stride, &sse);
vsad = (sse + 8)>>4;
#else
sad = (vp8_sad8x8(y, y_stride, yd, yd_stride, UINT_MAX) + 32) >> 6;
usad = (vp8_sad4x4(u, uv_stride, ud, uvd_stride, UINT_MAX) + 8) >> 4;
vsad = (vp8_sad4x4(v, uv_stride, vd, uvd_stride, UINT_MAX) + 8) >> 4;
sad = (vpx_sad8x8(y, y_stride, yd, yd_stride) + 32) >> 6;
usad = (vpx_sad4x4(u, uv_stride, ud, uvd_stride) + 8) >> 4;
vsad = (vpx_sad4x4(v, uv_stride, vd, uvd_stride) + 8) >> 4;
#endif
}
......
......@@ -303,88 +303,6 @@ specialize qw/vp8_variance_halfpixvar16x16_hv mmx sse2 media neon/;
$vp8_variance_halfpixvar16x16_hv_sse2=vp8_variance_halfpixvar16x16_hv_wmt;
$vp8_variance_halfpixvar16x16_hv_media=vp8_variance_halfpixvar16x16_hv_armv6;
#
# Single block SAD
#
add_proto qw/unsigned int vp8_sad4x4/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp8_sad4x4 mmx sse2 neon/;
$vp8_sad4x4_sse2=vp8_sad4x4_wmt;
add_proto qw/unsigned int vp8_sad8x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp8_sad8x8 mmx sse2 neon/;
$vp8_sad8x8_sse2=vp8_sad8x8_wmt;
add_proto qw/unsigned int vp8_sad8x16/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp8_sad8x16 mmx sse2 neon/;
$vp8_sad8x16_sse2=vp8_sad8x16_wmt;
add_proto qw/unsigned int vp8_sad16x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp8_sad16x8 mmx sse2 neon/;
$vp8_sad16x8_sse2=vp8_sad16x8_wmt;
add_proto qw/unsigned int vp8_sad16x16/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad";
specialize qw/vp8_sad16x16 mmx sse2 sse3 media neon/;
$vp8_sad16x16_sse2=vp8_sad16x16_wmt;
$vp8_sad16x16_media=vp8_sad16x16_armv6;
#
# Multi-block SAD, comparing a reference to N blocks 1 pixel apart horizontally
#
add_proto qw/void vp8_sad4x4x3/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sad_array";
specialize qw/vp8_sad4x4x3 sse3/;
add_proto qw/void vp8_sad8x8x3/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sad_array";
specialize qw/vp8_sad8x8x3 sse3/;
add_proto qw/void vp8_sad8x16x3/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sad_array";
specialize qw/vp8_sad8x16x3 sse3/;
add_proto qw/void vp8_sad16x8x3/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sad_array";
specialize qw/vp8_sad16x8x3 sse3 ssse3/;
add_proto qw/void vp8_sad16x16x3/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int *sad_array";
specialize qw/vp8_sad16x16x3 sse3 ssse3/;
# Note the only difference in the following prototypes is that they return into
# an array of short
add_proto qw/void vp8_sad4x4x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned short *sad_array";
specialize qw/vp8_sad4x4x8 sse4_1/;
$vp8_sad4x4x8_sse4_1=vp8_sad4x4x8_sse4;
add_proto qw/void vp8_sad8x8x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned short *sad_array";
specialize qw/vp8_sad8x8x8 sse4_1/;
$vp8_sad8x8x8_sse4_1=vp8_sad8x8x8_sse4;
add_proto qw/void vp8_sad8x16x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned short *sad_array";
specialize qw/vp8_sad8x16x8 sse4_1/;
$vp8_sad8x16x8_sse4_1=vp8_sad8x16x8_sse4;
add_proto qw/void vp8_sad16x8x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned short *sad_array";
specialize qw/vp8_sad16x8x8 sse4_1/;
$vp8_sad16x8x8_sse4_1=vp8_sad16x8x8_sse4;
add_proto qw/void vp8_sad16x16x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned short *sad_array";
specialize qw/vp8_sad16x16x8 sse4_1/;
$vp8_sad16x16x8_sse4_1=vp8_sad16x16x8_sse4;
#
# Multi-block SAD, comparing a reference to N independent blocks
#
add_proto qw/void vp8_sad4x4x4d/, "const unsigned char *src_ptr, int src_stride, const unsigned char * const ref_ptr[], int ref_stride, unsigned int *sad_array";
specialize qw/vp8_sad4x4x4d sse3/;
add_proto qw/void vp8_sad8x8x4d/, "const unsigned char *src_ptr, int src_stride, const unsigned char * const ref_ptr[], int ref_stride, unsigned int *sad_array";
specialize qw/vp8_sad8x8x4d sse3/;
add_proto qw/void vp8_sad8x16x4d/, "const unsigned char *src_ptr, int src_stride, const unsigned char * const ref_ptr[], int ref_stride, unsigned int *sad_array";
specialize qw/vp8_sad8x16x4d sse3/;
add_proto qw/void vp8_sad16x8x4d/, "const unsigned char *src_ptr, int src_stride, const unsigned char * const ref_ptr[], int ref_stride, unsigned int *sad_array";
specialize qw/vp8_sad16x8x4d sse3/;
add_proto qw/void vp8_sad16x16x4d/, "const unsigned char *src_ptr, int src_stride, const unsigned char * const ref_ptr[], int ref_stride, unsigned int *sad_array";
specialize qw/vp8_sad16x16x4d sse3/;
#
# Encoder functions below this point.
#
......
This diff is collapsed.
......@@ -14,16 +14,17 @@
#include "vpx_config.h"
#include "vpx/vpx_integer.h"
#ifdef __cplusplus
extern "C" {
#endif
typedef unsigned int(*vp8_sad_fn_t)(
const unsigned char *src_ptr,
typedef unsigned int(*vpx_sad_fn_t)(
const uint8_t *src_ptr,
int source_stride,
const unsigned char *ref_ptr,
int ref_stride,
unsigned int max_sad);
const uint8_t *ref_ptr,
int ref_stride);
typedef void (*vp8_copy32xn_fn_t)(
const unsigned char *src_ptr,
......@@ -32,27 +33,17 @@ typedef void (*vp8_copy32xn_fn_t)(
int ref_stride,
int n);
typedef void (*vp8_sad_multi_fn_t)(
typedef void (*vpx_sad_multi_fn_t)(
const unsigned char *src_ptr,
int source_stride,
const unsigned char *ref_ptr,
const unsigned char *ref_array,
int ref_stride,
unsigned int *sad_array);
typedef void (*vp8_sad_multi1_fn_t)
(
const unsigned char *src_ptr,
int source_stride,
const unsigned char *ref_ptr,
int ref_stride,
unsigned short *sad_array
);
typedef void (*vp8_sad_multi_d_fn_t)
typedef void (*vpx_sad_multi_d_fn_t)
(
const unsigned char *src_ptr,
int source_stride,
const unsigned char * const ref_ptr[],
const unsigned char * const ref_array[],
int ref_stride,
unsigned int *sad_array
);
......@@ -102,15 +93,15 @@ typedef unsigned int (*vp8_get16x16prederror_fn_t)
typedef struct variance_vtable
{
vp8_sad_fn_t sdf;
vpx_sad_fn_t sdf;
vp8_variance_fn_t vf;
vp8_subpixvariance_fn_t svf;
vp8_variance_fn_t svf_halfpix_h;
vp8_variance_fn_t svf_halfpix_v;
vp8_variance_fn_t svf_halfpix_hv;
vp8_sad_multi_fn_t sdx3f;
vp8_sad_multi1_fn_t sdx8f;
vp8_sad_multi_d_fn_t sdx4df;
vpx_sad_multi_fn_t sdx3f;
vpx_sad_multi_fn_t sdx8f;
vpx_sad_multi_d_fn_t sdx4df;
#if ARCH_X86 || ARCH_X86_64
vp8_copy32xn_fn_t copymem;
#endif
......
;
; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
;
; Use of this source code is governed by a BSD-style license
; that can be found in the LICENSE file in the root of the source
; tree. An additional intellectual property rights grant can be found
; in the file PATENTS. All contributing project authors may
; be found in the AUTHORS file in the root of the source tree.
;
%include "vpx_ports/x86_abi_support.asm"
;void vp8_copy32xn_sse2(
; unsigned char *src_ptr,
; int src_stride,
; unsigned char *dst_ptr,
; int dst_stride,
; int height);
global sym(vp8_copy32xn_sse2) PRIVATE
sym(vp8_copy32xn_sse2):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 5
SAVE_XMM 7
push rsi
push rdi
; end prolog
mov rsi, arg(0) ;src_ptr
mov rdi, arg(2) ;dst_ptr
movsxd rax, dword ptr arg(1) ;src_stride
movsxd rdx, dword ptr arg(3) ;dst_stride
movsxd rcx, dword ptr arg(4) ;height
.block_copy_sse2_loopx4:
movdqu xmm0, XMMWORD PTR [rsi]
movdqu xmm1, XMMWORD PTR [rsi + 16]
movdqu xmm2, XMMWORD PTR [rsi + rax]
movdqu xmm3, XMMWORD PTR [rsi + rax + 16]
lea rsi, [rsi+rax*2]
movdqu xmm4, XMMWORD PTR [rsi]
movdqu xmm5, XMMWORD PTR [rsi + 16]
movdqu xmm6, XMMWORD PTR [rsi + rax]
movdqu xmm7, XMMWORD PTR [rsi + rax + 16]
lea rsi, [rsi+rax*2]
movdqa XMMWORD PTR [rdi], xmm0
movdqa XMMWORD PTR [rdi + 16], xmm1
movdqa XMMWORD PTR [rdi + rdx], xmm2
movdqa XMMWORD PTR [rdi + rdx + 16], xmm3
lea rdi, [rdi+rdx*2]
movdqa XMMWORD PTR [rdi], xmm4
movdqa XMMWORD PTR [rdi + 16], xmm5
movdqa XMMWORD PTR [rdi + rdx], xmm6
movdqa XMMWORD PTR [rdi + rdx + 16], xmm7
lea rdi, [rdi+rdx*2]
sub rcx, 4
cmp rcx, 4
jge .block_copy_sse2_loopx4
cmp rcx, 0
je .copy_is_done
.block_copy_sse2_loop:
movdqu xmm0, XMMWORD PTR [rsi]
movdqu xmm1, XMMWORD PTR [rsi + 16]
lea rsi, [rsi+rax]
movdqa XMMWORD PTR [rdi], xmm0
movdqa XMMWORD PTR [rdi + 16], xmm1
lea rdi, [rdi+rdx]
sub rcx, 1
jne .block_copy_sse2_loop
.copy_is_done:
; begin epilog
pop rdi
pop rsi
RESTORE_XMM
UNSHADOW_ARGS
pop rbp
ret
;
; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
;
; Use of this source code is governed by a BSD-style license
; that can be found in the LICENSE file in the root of the source
; tree. An additional intellectual property rights grant can be found
; in the file PATENTS. All contributing project authors may
; be found in the AUTHORS file in the root of the source tree.
;
%include "vpx_ports/x86_abi_support.asm"
%macro STACK_FRAME_CREATE_X3 0
%if ABI_IS_32BIT
%define src_ptr rsi
%define src_stride rax
%define ref_ptr rdi
%define ref_stride rdx
%define end_ptr rcx
%define ret_var rbx
%define result_ptr arg(4)
%define max_sad arg(4)
%define height dword ptr arg(4)
push rbp
mov rbp, rsp
push rsi
push rdi
push rbx
mov rsi, arg(0) ; src_ptr
mov rdi, arg(2) ; ref_ptr
movsxd rax, dword ptr arg(1) ; src_stride
movsxd rdx, dword ptr arg(3) ; ref_stride
%else
%if LIBVPX_YASM_WIN64
SAVE_XMM 7, u
%define src_ptr rcx
%define src_stride rdx
%define ref_ptr r8
%define ref_stride r9
%define end_ptr r10
%define ret_var r11
%define result_ptr [rsp+xmm_stack_space+8+4*8]
%define max_sad [rsp+xmm_stack_space+8+4*8]
%define height dword ptr [rsp+xmm_stack_space+8+4*8]
%else
%define src_ptr rdi
%define src_stride rsi
%define ref_ptr rdx
%define ref_stride rcx
%define end_ptr r9
%define ret_var r10
%define result_ptr r8
%define max_sad r8
%define height r8
%endif
%endif
%endmacro
%macro STACK_FRAME_DESTROY_X3 0
%define src_ptr
%define src_stride
%define ref_ptr
%define ref_stride
%define end_ptr
%define ret_var
%define result_ptr
%define max_sad
%define height
%if ABI_IS_32BIT
pop rbx
pop rdi
pop rsi
pop rbp
%else
%if LIBVPX_YASM_WIN64
RESTORE_XMM
%endif
%endif
ret
%endmacro
;void vp8_copy32xn_sse3(
; unsigned char *src_ptr,
; int src_stride,
; unsigned char *dst_ptr,
; int dst_stride,
; int height);
global sym(vp8_copy32xn_sse3) PRIVATE
sym(vp8_copy32xn_sse3):
STACK_FRAME_CREATE_X3
.block_copy_sse3_loopx4:
lea end_ptr, [src_ptr+src_stride*2]
movdqu xmm0, XMMWORD PTR [src_ptr]
movdqu xmm1, XMMWORD PTR [src_ptr + 16]
movdqu xmm2, XMMWORD PTR [src_ptr + src_stride]
movdqu xmm3, XMMWORD PTR [src_ptr + src_stride + 16]
movdqu xmm4, XMMWORD PTR [end_ptr]
movdqu xmm5, XMMWORD PTR [end_ptr + 16]
movdqu xmm6, XMMWORD PTR [end_ptr + src_stride]
movdqu xmm7, XMMWORD PTR [end_ptr + src_stride + 16]
lea src_ptr, [src_ptr+src_stride*4]
lea end_ptr, [ref_ptr+ref_stride*2]
movdqa XMMWORD PTR [ref_ptr], xmm0
movdqa XMMWORD PTR [ref_ptr + 16], xmm1
movdqa XMMWORD PTR [ref_ptr + ref_stride], xmm2
movdqa XMMWORD PTR [ref_ptr + ref_stride + 16], xmm3
movdqa XMMWORD PTR [end_ptr], xmm4
movdqa XMMWORD PTR [end_ptr + 16], xmm5
movdqa XMMWORD PTR [end_ptr + ref_stride], xmm6
movdqa XMMWORD PTR [end_ptr + ref_stride + 16], xmm7
lea ref_ptr, [ref_ptr+ref_stride*4]
sub height, 4
cmp height,