Skip to content

L
libyuv
Commit af137b61 authored by fbarchard@google.com's avatar fbarchard@google.com
Browse files
Options

Interpolate for SSE2 with specialization for 1/4, 1/2, 3/4 and 1 

BUG=177
TESTED=try bots
Review URL: https://webrtc-codereview.appspot.com/1102004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@559 16f28f9a-4ce2-e073-06de-1de4eb20be90
parent 50d7bf4a
Show whitespace changes
Inline Side-by-side
Showing with 115 additions and 4 deletions
README.chromium
View file @ af137b61
Name: libyuv
URL: http://code.google.com/p/libyuv/
Version: 558
Version: 559
License: BSD
License File: LICENSE
......
include/libyuv/row.h
View file @ af137b61
......@@ -106,6 +106,7 @@ extern "C" {
#define HAS_ARGBBLENDROW_SSSE3
#define HAS_ARGBCOLORMATRIXROW_SSSE3
#define HAS_ARGBGRAYROW_SSSE3
#define HAS_ARGBINTERPOLATEROW_SSE2
#define HAS_ARGBINTERPOLATEROW_SSSE3
#define HAS_ARGBMIRRORROW_SSSE3
#define HAS_ARGBMULTIPLYROW_SSE2
......@@ -122,7 +123,6 @@ extern "C" {
// TODO(fbarchard): Port to gcc.
#if !defined(YUV_DISABLE_ASM) && defined(_M_IX86)
#define HAS_ARGBCOLORTABLEROW_X86
#define HAS_ARGBINTERPOLATEROW_SSE2
#endif
// The following are Yasm x86 only.
......
include/libyuv/version.h
View file @ af137b61
......@@ -11,6 +11,6 @@
#ifndef INCLUDE_LIBYUV_VERSION_H_ // NOLINT
#define INCLUDE_LIBYUV_VERSION_H_
#define LIBYUV_VERSION 558
#define LIBYUV_VERSION 559
#endif // INCLUDE_LIBYUV_VERSION_H_ NOLINT
source/row_posix.cc
View file @ af137b61
......@@ -4559,7 +4559,6 @@ void ARGBInterpolateRow_SSSE3(uint8* dst_argb, const uint8* src_argb,
"lea 0x10(%1),%1 \n"
"jg 100b \n"
// Extrude last pixel.
"99: \n"
: "+r"(dst_argb), // %0
"+r"(src_argb), // %1
......@@ -4573,6 +4572,118 @@ void ARGBInterpolateRow_SSSE3(uint8* dst_argb, const uint8* src_argb,
);
}
// Bilinear image filtering.
// Same as ScaleARGBFilterRows_SSSE3 but without last pixel duplicated.
void ARGBInterpolateRow_SSE2(uint8* dst_argb, const uint8* src_argb,
ptrdiff_t src_stride, int dst_width,
int source_y_fraction) {
asm volatile (
"sub %1,%0 \n"
"shr %3 \n"
"cmp $0x0,%3 \n"
"je 100f \n"
"cmp $0x20,%3 \n"
"je 75f \n"
"cmp $0x40,%3 \n"
"je 50f \n"
"cmp $0x60,%3 \n"
"je 25f \n"
"movd %3,%%xmm0 \n"
"neg %3 \n"
"add $0x80,%3 \n"
"movd %3,%%xmm5 \n"
"punpcklbw %%xmm0,%%xmm5 \n"
"punpcklwd %%xmm5,%%xmm5 \n"
"pshufd $0x0,%%xmm5,%%xmm5 \n"
"pxor %%xmm4,%%xmm4 \n"
// General purpose row blend.
".p2align 4 \n"
"1: \n"
"movdqa (%1),%%xmm0 \n"
"movdqa (%1,%4,1),%%xmm2 \n"
"movdqa %%xmm0,%%xmm1 \n"
"movdqa %%xmm2,%%xmm3 \n"
"punpcklbw %%xmm4,%%xmm2 \n"
"punpckhbw %%xmm4,%%xmm3 \n"
"punpcklbw %%xmm4,%%xmm0 \n"
"punpckhbw %%xmm4,%%xmm1 \n"
"psubw %%xmm0,%%xmm2 \n"
"psubw %%xmm1,%%xmm3 \n"
"paddw %%xmm2,%%xmm2 \n"
"paddw %%xmm3,%%xmm3 \n"
"pmulhw %%xmm5,%%xmm2 \n"
"pmulhw %%xmm5,%%xmm3 \n"
"paddw %%xmm2,%%xmm0 \n"
"paddw %%xmm3,%%xmm1 \n"
"packuswb %%xmm1,%%xmm0 \n"
"sub $0x4,%2 \n"
"movdqa %%xmm0,(%1,%0,1) \n"
"lea 0x10(%1),%1 \n"
"jg 1b \n"
"jmp 99f \n"
// Blend 25 / 75.
".p2align 4 \n"
"25: \n"
"movdqa (%1),%%xmm0 \n"
"movdqa (%1,%4,1),%%xmm1 \n"
"pavgb %%xmm1,%%xmm0 \n"
"pavgb %%xmm1,%%xmm0 \n"
"sub $0x4,%2 \n"
"movdqa %%xmm0,(%1,%0,1) \n"
"lea 0x10(%1),%1 \n"
"jg 25b \n"
"jmp 99f \n"
// Blend 50 / 50.
".p2align 4 \n"
"50: \n"
"movdqa (%1),%%xmm0 \n"
"movdqa (%1,%4,1),%%xmm1 \n"
"pavgb %%xmm1,%%xmm0 \n"
"sub $0x4,%2 \n"
"movdqa %%xmm0,(%1,%0,1) \n"
"lea 0x10(%1),%1 \n"
"jg 50b \n"
"jmp 99f \n"
// Blend 75 / 25.
".p2align 4 \n"
"75: \n"
"movdqa (%1),%%xmm1 \n"
"movdqa (%1,%4,1),%%xmm0 \n"
"pavgb %%xmm1,%%xmm0 \n"
"pavgb %%xmm1,%%xmm0 \n"
"sub $0x4,%2 \n"
"movdqa %%xmm0,(%1,%0,1) \n"
"lea 0x10(%1),%1 \n"
"jg 75b \n"
"jmp 99f \n"
// Blend 100 / 0 - Copy row unchanged.
".p2align 4 \n"
"100: \n"
"movdqa (%1),%%xmm0 \n"
"sub $0x4,%2 \n"
"movdqa %%xmm0,(%1,%0,1) \n"
"lea 0x10(%1),%1 \n"
"jg 100b \n"
"99: \n"
: "+r"(dst_argb), // %0
"+r"(src_argb), // %1
"+r"(dst_width), // %2
"+r"(source_y_fraction) // %3
: "r"(static_cast<intptr_t>(src_stride)) // %4
: "memory", "cc"
#if defined(__SSE2__)
, "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
#endif
);
}
void HalfRow_SSE2(const uint8* src_uv, int src_uv_stride,
uint8* dst_uv, int pix) {
asm volatile (
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment