Skip to content

L
libyuv
Commit 54f0b67c authored by fbarchard@google.com's avatar fbarchard@google.com
Browse files
Options

Scale up use pextrw to speed up a little 

BUG=208
TEST=out\release\libyuv_unittest --gtest_filter=*Scale*640*
Review URL: https://webrtc-codereview.appspot.com/1283008

git-svn-id: http://libyuv.googlecode.com/svn/trunk@649 16f28f9a-4ce2-e073-06de-1de4eb20be90
parent 98a1fbf5
Hide whitespace changes
Inline Side-by-side
Showing with 261 additions and 65 deletions
README.chromium
View file @ 54f0b67c
Name: libyuv
URL: http://code.google.com/p/libyuv/
Version: 648
Version: 649
License: BSD
License File: LICENSE
......
include/libyuv/version.h
View file @ 54f0b67c
......@@ -11,6 +11,6 @@
#ifndef INCLUDE_LIBYUV_VERSION_H_ // NOLINT
#define INCLUDE_LIBYUV_VERSION_H_
#define LIBYUV_VERSION 648
#define LIBYUV_VERSION 649
#endif // INCLUDE_LIBYUV_VERSION_H_ NOLINT
source/scale_argb.cc
View file @ 54f0b67c
......@@ -421,10 +421,63 @@ void ScaleARGBFilterRows_SSSE3(uint8* dst_argb, const uint8* src_argb,
}
}
// Column scaling unfiltered. SSSE3 version.
// TODO(fbarchard): Port to Neon
#define HAS_SCALEARGBCOLS_SSE2
__declspec(naked) __declspec(align(16))
static void ScaleARGBCols_SSE2(uint8* dst_argb, const uint8* src_argb,
int dst_width, int x, int dx) {
__asm {
push esi
push edi
mov edi, [esp + 8 + 4] // dst_argb
mov esi, [esp + 8 + 8] // src_argb
mov ecx, [esp + 8 + 12] // dst_width
movd xmm2, [esp + 8 + 16] // x
movd xmm3, [esp + 8 + 20] // dx
pextrw eax, xmm2, 1 // get x0 integer. preroll
sub ecx, 2
jl xloop29
movdqa xmm0, xmm2 // x1 = x0 + dx
paddd xmm0, xmm3
punpckldq xmm2, xmm0 // x0 x1
punpckldq xmm3, xmm3 // dx dx
paddd xmm3, xmm3 // dx * 2, dx * 2
pextrw edx, xmm2, 3 // get x1 integer. preroll
// 2 Pixel loop.
align 16
xloop2:
paddd xmm2, xmm3 // x += dx
movd xmm0, qword ptr [esi + eax * 4] // 1 source x0 pixels
movd xmm1, qword ptr [esi + edx * 4] // 1 source x1 pixels
punpckldq xmm0, xmm1 // x0 x1
pextrw eax, xmm2, 1 // get x0 integer. next iteration.
pextrw edx, xmm2, 3 // get x1 integer. next iteration.
movq qword ptr [edi], xmm0
lea edi, [edi + 8]
sub ecx, 2 // 2 pixels
jge xloop2
xloop29:
add ecx, 2 - 1
jl xloop99
// 1 pixel remainder
movd xmm0, qword ptr [esi + eax * 4] // 1 source x0 pixels
movd [edi], xmm0
xloop99:
pop edi
pop esi
ret
}
}
// Bilinear row filtering combines 2x1 -> 1x1. SSSE3 version.
// TODO(fbarchard): Port to Neon
// TODO(fbarchard): Port to Posix
// TODO(fbarchard): Consider lea to get 2nd pixel without incrementing.
// Shuffle table for arranging 2 pixels into pairs for pmaddubsw
static const uvec8 kShuffleColARGB = {
......@@ -434,7 +487,7 @@ static const uvec8 kShuffleColARGB = {
// Shuffle table for duplicating 2 fractions into 8 bytes each
static const uvec8 kShuffleFractions = {
0u, 0u, 0u, 0u, 0u, 0u, 0u, 0u, 2u, 2u, 2u, 2u, 2u, 2u, 2u, 2u,
0u, 0u, 0u, 0u, 0u, 0u, 0u, 0u, 4u, 4u, 4u, 4u, 4u, 4u, 4u, 4u,
};
#define HAS_SCALEARGBFILTERCOLS_SSSE3
......@@ -442,44 +495,47 @@ __declspec(naked) __declspec(align(16))
static void ScaleARGBFilterCols_SSSE3(uint8* dst_argb, const uint8* src_argb,
int dst_width, int x, int dx) {
__asm {
push ebx
push ebp
push esi
push edi
mov edi, [esp + 16 + 4] // dst_argb
mov esi, [esp + 16 + 8] // src_argb
mov ecx, [esp + 16 + 12] // dst_width
mov edx, [esp + 16 + 16] // x
mov ebx, [esp + 16 + 20] // dx
movdqa xmm3, kShuffleFractions
mov edi, [esp + 8 + 4] // dst_argb
mov esi, [esp + 8 + 8] // src_argb
mov ecx, [esp + 8 + 12] // dst_width
movd xmm2, [esp + 8 + 16] // x
movd xmm3, [esp + 8 + 20] // dx
movdqa xmm4, kShuffleColARGB
pcmpeqb xmm5, xmm5 // generate 0x007f for inverting fraction.
psrlw xmm5, 9
movdqa xmm5, kShuffleFractions
pcmpeqb xmm6, xmm6 // generate 0x007f for inverting fraction.
psrlw xmm6, 9
pextrw eax, xmm2, 1 // get x0 integer. preroll
sub ecx, 2
jl xloop29
movdqa xmm0, xmm2 // x1 = x0 + dx
paddd xmm0, xmm3
punpckldq xmm2, xmm0 // x0 x1
punpckldq xmm3, xmm3 // dx dx
paddd xmm3, xmm3 // dx * 2, dx * 2
pextrw edx, xmm2, 3 // get x1 integer. preroll
// 2 Pixel loop.
align 16
xloop2:
mov eax, edx // get x0 integer
movd xmm1, edx // get x0 fraction
lea ebp, [edx + ebx] // get x1 integer (x + dx)
movd xmm2, ebp // get x1 fraction
shr eax, 16 // x0
punpcklwd xmm1, xmm2 // x0x1 fractions
lea edx, [edx + ebx * 2] // x += dx * 2
shr ebp, 16 // x1
movdqa xmm1, xmm2 // x0, x1 fractions.
paddd xmm2, xmm3 // x += dx
movq xmm0, qword ptr [esi + eax * 4] // 2 source x0 pixels
movhps xmm0, qword ptr [esi + ebp * 4] // 2 source x1 pixels
psrlw xmm1, 9 // 7 bit fractions.
pshufb xmm1, xmm3 // 0000000011111111
sub ecx, 2
movhps xmm0, qword ptr [esi + edx * 4] // 2 source x1 pixels
pshufb xmm1, xmm5 // 0000000011111111
pshufb xmm0, xmm4 // arrange pixels into pairs
pxor xmm1, xmm5 // 0..7f and 7f..0
pxor xmm1, xmm6 // 0..7f and 7f..0
pmaddubsw xmm0, xmm1 // argb_argb 16 bit, 2 pixels.
psrlw xmm0, 7
psrlw xmm0, 7 // argb 8.7 fixed point to low 8 bits.
pextrw eax, xmm2, 1 // get x0 integer. next iteration.
pextrw edx, xmm2, 3 // get x1 integer. next iteration.
packuswb xmm0, xmm0 // argb_argb 8 bits, 2 pixels.
movq qword ptr [edi], xmm0
lea edi, [edi + 8]
sub ecx, 2 // 2 pixels
jge xloop2
xloop29:
......@@ -487,15 +543,12 @@ static void ScaleARGBFilterCols_SSSE3(uint8* dst_argb, const uint8* src_argb,
jl xloop99
// 1 pixel remainder
mov eax, edx // get x0 integer
movd xmm1, edx // get x0 fraction
shr eax, 16 // x0
psrlw xmm2, 9 // 7 bit fractions.
movq xmm0, qword ptr [esi + eax * 4] // 2 source x0 pixels
psrlw xmm1, 9 // 7 bit fractions.
pshufb xmm1, xmm3 // 00000000
pshufb xmm2, xmm5 // 00000000
pshufb xmm0, xmm4 // arrange pixels into pairs
pxor xmm1, xmm5 // 0..7f and 7f..0
pmaddubsw xmm0, xmm1 // argb 16 bit, 1 pixel.
pxor xmm2, xmm6 // 0..7f and 7f..0
pmaddubsw xmm0, xmm2 // argb 16 bit, 1 pixel.
psrlw xmm0, 7
packuswb xmm0, xmm0 // argb 8 bits, 1 pixel.
movd [edi], xmm0
......@@ -503,8 +556,6 @@ static void ScaleARGBFilterCols_SSSE3(uint8* dst_argb, const uint8* src_argb,
pop edi
pop esi
pop ebp
pop ebx
ret
}
}
......@@ -877,6 +928,149 @@ void ScaleARGBFilterRows_SSSE3(uint8* dst_argb, const uint8* src_argb,
#endif
);
}
#define HAS_SCALEARGBCOLS_SSE2
static void ScaleARGBCols_SSE2(uint8* dst_argb, const uint8* src_argb,
int dst_width, int x, int dx) {
intptr_t x0, x1;
asm volatile (
"movd %5,%%xmm2 \n"
"movd %6,%%xmm3 \n"
"pextrw $0x1,%%xmm2,%3 \n"
"sub $0x2,%2 \n"
"jl 29f \n"
"movdqa %%xmm2,%%xmm0 \n"
"paddd %%xmm3,%%xmm0 \n"
"punpckldq %%xmm0,%%xmm2 \n"
"punpckldq %%xmm3,%%xmm3 \n"
"paddd %%xmm3,%%xmm3 \n"
"pextrw $0x3,%%xmm2,%4 \n"
"lea 0x0(%2),%2 \n"
".p2align 4 \n"
"2: \n"
"paddd %%xmm3,%%xmm2 \n"
"movd (%1,%3,4),%%xmm0 \n"
"movd (%1,%4,4),%%xmm1 \n"
"punpckldq %%xmm1,%%xmm0 \n"
"pextrw $0x1,%%xmm2,%3 \n"
"pextrw $0x3,%%xmm2,%4 \n"
"movq %%xmm0,(%0) \n"
"lea 0x8(%0),%0 \n"
"sub $0x2,%2 \n"
"jge 2b \n"
"29: \n"
"add $0x1,%2 \n"
"jl 99f \n"
"movd (%1,%3,4),%%xmm0 \n"
"movd %%xmm0,(%0) \n"
"99: \n"
: "+r"(dst_argb), // %0
"+r"(src_argb), // %1
"+rm"(dst_width), // %2
"+r"(x0), // %3
"+r"(x1) // %4
: "rm"(x), // %5
"rm"(dx) // %6
: "memory", "cc"
#if defined(__SSE2__)
, "xmm0", "xmm1", "xmm2", "xmm3"
#endif
);
}
#ifdef __APPLE__
#define CONST
#else
#define CONST static const
#endif
// Shuffle table for arranging 2 pixels into pairs for pmaddubsw
CONST uvec8 kShuffleColARGB = {
0u, 4u, 1u, 5u, 2u, 6u, 3u, 7u, // bbggrraa 1st pixel
8u, 12u, 9u, 13u, 10u, 14u, 11u, 15u // bbggrraa 2nd pixel
};
// Shuffle table for duplicating 2 fractions into 8 bytes each
CONST uvec8 kShuffleFractions = {
0u, 0u, 0u, 0u, 0u, 0u, 0u, 0u, 4u, 4u, 4u, 4u, 4u, 4u, 4u, 4u,
};
// Bilinear row filtering combines 4x2 -> 4x1. SSSE3 version
#define HAS_SCALEARGBFILTERCOLS_SSSE3
static void ScaleARGBFilterCols_SSSE3(uint8* dst_argb, const uint8* src_argb,
int dst_width, int x, int dx) {
intptr_t x0, x1;
asm volatile (
"movdqa %0,%%xmm4 \n"
"movdqa %1,%%xmm5 \n"
:
: "m"(kShuffleColARGB), // %0
"m"(kShuffleFractions) // %1
);
asm volatile (
"movd %5,%%xmm2 \n"
"movd %6,%%xmm3 \n"
"pcmpeqb %%xmm6,%%xmm6 \n"
"psrlw $0x9,%%xmm6 \n"
"pextrw $0x1,%%xmm2,%3 \n"
"sub $0x2,%2 \n"
"jl 29f \n"
"movdqa %%xmm2,%%xmm0 \n"
"paddd %%xmm3,%%xmm0 \n"
"punpckldq %%xmm0,%%xmm2 \n"
"punpckldq %%xmm3,%%xmm3 \n"
"paddd %%xmm3,%%xmm3 \n"
"pextrw $0x3,%%xmm2,%4 \n"
".p2align 4 \n"
"2: \n"
"movdqa %%xmm2,%%xmm1 \n"
"paddd %%xmm3,%%xmm2 \n"
"movq (%1,%3,4),%%xmm0 \n"
"psrlw $0x9,%%xmm1 \n"
"movhps (%1,%4,4),%%xmm0 \n"
"pshufb %%xmm5,%%xmm1 \n"
"pshufb %%xmm4,%%xmm0 \n"
"pxor %%xmm6,%%xmm1 \n"
"pmaddubsw %%xmm1,%%xmm0 \n"
"psrlw $0x7,%%xmm0 \n"
"pextrw $0x1,%%xmm2,%3 \n"
"pextrw $0x3,%%xmm2,%4 \n"
"packuswb %%xmm0,%%xmm0 \n"
"movq %%xmm0,(%0) \n"
"lea 0x8(%0),%0 \n"
"sub $0x2,%2 \n"
"jge 2b \n"
"29: \n"
"add $0x1,%2 \n"
"jl 99f \n"
"psrlw $0x9,%%xmm2 \n"
"movq (%1,%3,4),%%xmm0 \n"
"pshufb %%xmm5,%%xmm2 \n"
"pshufb %%xmm4,%%xmm0 \n"
"pxor %%xmm6,%%xmm2 \n"
"pmaddubsw %%xmm2,%%xmm0 \n"
"psrlw $0x7,%%xmm0 \n"
"packuswb %%xmm0,%%xmm0 \n"
"movd %%xmm0,(%0) \n"
"99: \n"
: "+r"(dst_argb), // %0
"+r"(src_argb), // %1
"+rm"(dst_width), // %2
"+r"(x0), // %3
"+r"(x1) // %4
: "rm"(x), // %5
"rm"(dx) // %6
: "memory", "cc"
#if defined(__SSE2__)
, "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6"
#endif
);
}
#endif // defined(__x86_64__) || defined(__i386__)
static void ScaleARGBRowDown2_C(const uint8* src_argb,
......@@ -947,14 +1141,10 @@ static void ScaleARGBRowDownEvenInt_C(const uint8* src_argb,
}
}
// (1-f)a + fb can be replaced with a + f(b-a)
#define BLENDER1(a, b, f) (static_cast<int>(a) + \
((f) * (static_cast<int>(b) - static_cast<int>(a)) >> 16))
// Mimics SSSE3 blender
#define BLENDER1(a, b, f) ((a) * (0x7f ^ f) + (b) * f) >> 7
#define BLENDERC(a, b, f, s) static_cast<uint32>( \
BLENDER1(((a) >> s) & 255, ((b) >> s) & 255, f) << s)
#define BLENDER(a, b, f) \
BLENDERC(a, b, f, 24) | BLENDERC(a, b, f, 16) | \
BLENDERC(a, b, f, 8) | BLENDERC(a, b, f, 0)
......@@ -965,22 +1155,25 @@ static void ScaleARGBFilterCols_C(uint8* dst_argb, const uint8* src_argb,
uint32* dst = reinterpret_cast<uint32*>(dst_argb);
for (int j = 0; j < dst_width - 1; j += 2) {
int xi = x >> 16;
int xf = (x >> 9) & 0x7f;
uint32 a = src[xi];
uint32 b = src[xi + 1];
dst[0] = BLENDER(a, b, x & 0xffff);
dst[0] = BLENDER(a, b, xf);
x += dx;
xi = x >> 16;
xf = (x >> 9) & 0x7f;
a = src[xi];
b = src[xi + 1];
dst[1] = BLENDER(a, b, x & 0xffff);
dst[1] = BLENDER(a, b, xf);
x += dx;
dst += 2;
}
if (dst_width & 1) {
int xi = x >> 16;
int xf = (x >> 9) & 0x7f;
uint32 a = src[xi];
uint32 b = src[xi + 1];
dst[0] = BLENDER(a, b, x & 0xffff);
dst[0] = BLENDER(a, b, xf);
}
}
......@@ -1115,13 +1308,8 @@ static void ScaleARGBBilinear(int src_width, int src_height,
assert(src_width * 4 <= kMaxStride);
SIMD_ALIGNED(uint8 row[kMaxStride + 16]);
void (*ScaleARGBFilterRows)(uint8* dst_argb, const uint8* src_argb,
ptrdiff_t src_stride,
int dst_width, int source_y_fraction) =
ptrdiff_t src_stride, int dst_width, int source_y_fraction) =
ScaleARGBFilterRows_C;
void (*ScaleARGBFilterCols)(uint8* dst_argb, const uint8* src_argb,
int dst_width, int x, int dx) =
ScaleARGBFilterCols_C;
#if defined(HAS_SCALEARGBFILTERROWS_SSE2)
if (TestCpuFlag(kCpuHasSSE2) && IS_ALIGNED(src_width, 4) &&
IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride, 16)) {
......@@ -1134,6 +1322,8 @@ static void ScaleARGBBilinear(int src_width, int src_height,
ScaleARGBFilterRows = ScaleARGBFilterRows_SSSE3;
}
#endif
void (*ScaleARGBFilterCols)(uint8* dst_argb, const uint8* src_argb,
int dst_width, int x, int dx) = ScaleARGBFilterCols_C;
#if defined(HAS_SCALEARGBFILTERCOLS_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3)) {
ScaleARGBFilterCols = ScaleARGBFilterCols_SSSE3;
......@@ -1166,8 +1356,8 @@ static void ScaleARGBBilinear(int src_width, int src_height,
// Scales a single row of pixels using point sampling.
// Code is adapted from libyuv bilinear yuv scaling, but with bilinear
// interpolation off, and argb pixels instead of yuv.
static void ScaleARGBCols(uint8* dst_argb, const uint8* src_argb,
int dst_width, int x, int dx) {
static void ScaleARGBCols_C(uint8* dst_argb, const uint8* src_argb,
int dst_width, int x, int dx) {
const uint32* src = reinterpret_cast<const uint32*>(src_argb);
uint32* dst = reinterpret_cast<uint32*>(dst_argb);
for (int j = 0; j < dst_width - 1; j += 2) {
......@@ -1182,7 +1372,6 @@ static void ScaleARGBCols(uint8* dst_argb, const uint8* src_argb,
}
}
// ScaleARGB ARGB to/from any dimensions, without interpolation.
// Fixed point math is used for performance: The upper 16 bits
// of x and dx is the integer part of the source position and
......@@ -1192,13 +1381,20 @@ static void ScaleARGBSimple(int src_width, int src_height,
int dst_width, int dst_height,
int src_stride, int dst_stride,
const uint8* src_argb, uint8* dst_argb) {
void (*ScaleARGBCols)(uint8* dst_argb, const uint8* src_argb,
int dst_width, int x, int dx) = ScaleARGBCols_C;
#if defined(HAS_SCALEARGBCOLS_SSE2)
if (TestCpuFlag(kCpuHasSSE2)) {
ScaleARGBCols = ScaleARGBCols_SSE2;
}
#endif
int dx = (src_width << 16) / dst_width;
int dy = (src_height << 16) / dst_height;
int x = (dx >= 65536) ? ((dx >> 1) - 32768) : (dx >> 1);
int y = (dy >= 65536) ? ((dy >> 1) - 32768) : (dy >> 1);
for (int i = 0; i < dst_height; ++i) {
ScaleARGBCols(dst_argb, src_argb + (y >> 16) * src_stride, dst_width, x,
dx);
ScaleARGBCols(dst_argb, src_argb + (y >> 16) * src_stride,
dst_width, x, dx);
dst_argb += dst_stride;
y += dy;
}
......
unit_test/scale_argb_test.cc
View file @ 54f0b67c
......@@ -149,7 +149,7 @@ TEST_F(libyuvTest, ARGBScaleDownBy4_None) {
dst_width, dst_height,
kFilterNone,
benchmark_iterations_);
EXPECT_LE(max_diff, 2); // This is the only scale factor with error of 2.
EXPECT_LE(max_diff, 1);
}
TEST_F(libyuvTest, ARGBScaleDownBy4_Bilinear) {
......@@ -162,7 +162,7 @@ TEST_F(libyuvTest, ARGBScaleDownBy4_Bilinear) {
dst_width, dst_height,
kFilterBilinear,
benchmark_iterations_);
EXPECT_LE(max_diff, 2); // This is the only scale factor with error of 2.
EXPECT_LE(max_diff, 1);
}
TEST_F(libyuvTest, ARGBScaleDownBy4_Box) {
......@@ -175,7 +175,7 @@ TEST_F(libyuvTest, ARGBScaleDownBy4_Box) {
dst_width, dst_height,
kFilterBox,
benchmark_iterations_);
EXPECT_LE(max_diff, 2); // This is the only scale factor with error of 2.
EXPECT_LE(max_diff, 1);
}
TEST_F(libyuvTest, ARGBScaleDownBy5_None) {
......@@ -357,7 +357,7 @@ TEST_F(libyuvTest, ARGBScaleTo1366x768_Bilinear) {
dst_width, dst_height,
kFilterBilinear,
benchmark_iterations_);
EXPECT_LE(max_diff, 2);
EXPECT_LE(max_diff, 1);
}
......@@ -384,7 +384,7 @@ TEST_F(libyuvTest, ARGBScaleTo1280x720_Bilinear) {
dst_width, dst_height,
kFilterBilinear,
benchmark_iterations_);
EXPECT_LE(max_diff, 2);
EXPECT_LE(max_diff, 1);
}
TEST_F(libyuvTest, ARGBScaleTo853x480_None) {
......@@ -410,7 +410,7 @@ TEST_F(libyuvTest, ARGBScaleTo853x480_Bilinear) {
dst_width, dst_height,
kFilterBilinear,
benchmark_iterations_);
EXPECT_LE(max_diff, 3);
EXPECT_LE(max_diff, 1);
}
TEST_F(libyuvTest, ARGBScaleFrom640x360_None) {
......@@ -423,7 +423,7 @@ TEST_F(libyuvTest, ARGBScaleFrom640x360_None) {
dst_width, dst_height,
kFilterNone,
benchmark_iterations_);
EXPECT_LE(max_diff, 2);
EXPECT_LE(max_diff, 1);
}
TEST_F(libyuvTest, ARGBScaleFrom640x360_Bilinear) {
......@@ -436,7 +436,7 @@ TEST_F(libyuvTest, ARGBScaleFrom640x360_Bilinear) {
dst_width, dst_height,
kFilterBilinear,
benchmark_iterations_);
EXPECT_LE(max_diff, 3);
EXPECT_LE(max_diff, 1);
}
} // namespace libyuv
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