Skip to content

L
libyuv
Commit 3e8a577b authored by fbarchard@google.com's avatar fbarchard@google.com
Browse files
Options

add ARGBRect, fix for 444 support in ConvertToI420 

BUG=none
TEST=none
Review URL: http://webrtc-codereview.appspot.com/289003

git-svn-id: http://libyuv.googlecode.com/svn/trunk@93 16f28f9a-4ce2-e073-06de-1de4eb20be90
parent 10f5556a
Hide whitespace changes
Inline Side-by-side
Showing with 182 additions and 149 deletions
README.chromium
View file @ 3e8a577b
Name: libyuv
URL: http://code.google.com/p/libyuv/
Version: 90
Version: 93
License: BSD
License File: LICENSE
......
include/libyuv/basic_types.h
View file @ 3e8a577b
......@@ -66,7 +66,6 @@ typedef char int8;
#define CPU_X86 1
#endif
#define IS_ALIGNED(p, a) (0==(reinterpret_cast<uintptr_t>(p) & ((a)-1)))
#define ALIGNP(p, t) \
(reinterpret_cast<uint8*>(((reinterpret_cast<uintptr_t>(p) + \
((t)-1)) & ~((t)-1))))
......
include/libyuv/planar_functions.h
View file @ 3e8a577b
......@@ -33,15 +33,7 @@ int I420Mirror(const uint8* src_y, int src_stride_y,
uint8* dst_v, int dst_stride_v,
int width, int height);
// Draw a rectangle into I420
int I420Rect(uint8* dst_y, int dst_stride_y,
uint8* dst_u, int dst_stride_u,
uint8* dst_v, int dst_stride_v,
int x, int y,
int width, int height,
int value_y, int value_u, int value_v);
// Convert I422 to I420. Used by MJPG.
// Convert I422 to I420.
int I422ToI420(const uint8* src_y, int src_stride_y,
const uint8* src_u, int src_stride_u,
const uint8* src_v, int src_stride_v,
......@@ -183,6 +175,20 @@ int ARGBToI400(const uint8* src_argb, int src_stride_argb,
uint8* dst_y, int dst_stride_y,
int width, int height);
// Draw a rectangle into I420
int I420Rect(uint8* dst_y, int dst_stride_y,
uint8* dst_u, int dst_stride_u,
uint8* dst_v, int dst_stride_v,
int x, int y,
int width, int height,
int value_y, int value_u, int value_v);
// Draw a rectangle into ARGB
int ARGBRect(uint8* dst_argb, int dst_stride_argb,
int x, int y,
int width, int height,
uint32 value);
} // namespace libyuv
#endif // INCLUDE_LIBYUV_PLANAR_FUNCTIONS_H_
source/convert.cc
View file @ 3e8a577b
......@@ -1171,8 +1171,8 @@ int ConvertToI420(const uint8* sample, size_t sample_size,
src_u = sample + src_width * (abs_src_height * 2 + crop_y) + crop_x;
}
I444ToI420(src_y, src_width,
src_u, halfwidth,
src_v, halfwidth,
src_u, src_width,
src_v, src_width,
y, y_stride,
u, u_stride,
v, v_stride,
......
source/planar_functions.cc
View file @ 3e8a577b
......@@ -304,138 +304,6 @@ int I420Mirror(const uint8* src_y, int src_stride_y,
return 0;
}
// SetRows32 writes 'count' bytes using a 32 bit value repeated
#if defined(__ARM_NEON__) && !defined(YUV_DISABLE_ASM)
#define HAS_SETROW_NEON
static void SetRow32_NEON(uint8* dst, uint32 v32, int count) {
asm volatile (
"vdup.u32 q0, %2 \n" // duplicate 4 ints
"1: \n"
"vst1.u32 {q0}, [%0]! \n" // store
"subs %1, %1, #16 \n" // 16 processed per loop
"bhi 1b \n"
: "+r"(dst), // %0
"+r"(count) // %1
: "r"(v32) // %2
: "q0", "memory", "cc"
);
}
#elif defined(_M_IX86) && !defined(YUV_DISABLE_ASM)
#define HAS_SETROW_SSE2
__declspec(naked)
static void SetRow32_SSE2(uint8* dst, uint32 v32, int count) {
__asm {
mov eax, [esp + 4] // dst
movd xmm5, [esp + 8] // v32
mov ecx, [esp + 12] // count
pshufd xmm5, xmm5, 0
convertloop:
movdqa [eax], xmm5
lea eax, [eax + 16]
sub ecx, 16
ja convertloop
ret
}
}
#elif (defined(__x86_64__) || defined(__i386__)) && !defined(YUV_DISABLE_ASM)
#define HAS_SETROW_SSE2
static void SetRow32_SSE2(uint8* dst, uint32 v32, int count) {
asm volatile (
"movd %2, %%xmm5 \n"
"pshufd $0x0,%%xmm5,%%xmm5 \n"
"1: \n"
"movdqa %%xmm5,(%0) \n"
"lea 0x10(%0),%0 \n"
"sub $0x10,%1 \n"
"ja 1b \n"
: "+r"(dst), // %0
"+r"(count) // %1
: "r"(v32) // %2
: "memory", "cc"
#if defined(__SSE2__)
, "xmm5"
#endif
);
}
#endif
static void SetRow8_C(uint8* dst, uint32 v8, int count) {
memset(dst, v8, count);
}
static void I420SetPlane(uint8* dst_y, int dst_stride_y,
int width, int height,
int value) {
void (*SetRow)(uint8* dst, uint32 value, int pix);
#if defined(HAS_SETROW_NEON)
if (TestCpuFlag(kCpuHasNEON) &&
IS_ALIGNED(width, 16) &&
IS_ALIGNED(dst_y, 16) && IS_ALIGNED(dst_stride_y, 16)) {
SetRow = SetRow32_NEON;
} else
#elif defined(HAS_SETROW_SSE2)
if (TestCpuFlag(kCpuHasSSE2) &&
IS_ALIGNED(width, 16) &&
IS_ALIGNED(dst_y, 16) && IS_ALIGNED(dst_stride_y, 16)) {
SetRow = SetRow32_SSE2;
} else
#endif
{
SetRow = SetRow8_C;
}
uint32 v32 = value | (value << 8) | (value << 16) | (value << 24);
// Set plane
for (int y = 0; y < height; ++y) {
SetRow(dst_y, v32, width);
dst_y += dst_stride_y;
}
}
// Draw a rectangle into I420
int I420Rect(uint8* dst_y, int dst_stride_y,
uint8* dst_u, int dst_stride_u,
uint8* dst_v, int dst_stride_v,
int x, int y,
int width, int height,
int value_y, int value_u, int value_v) {
if (!dst_y || !dst_u || !dst_v ||
width <= 0 || height == 0 ||
x < 0 || y < 0 ||
value_y < 0 || value_y > 255 ||
value_u < 0 || value_u > 255 ||
value_v < 0 || value_v > 255) {
return -1;
}
// Negative height means invert the image.
if (height < 0) {
height = -height;
int halfheight = (height + 1) >> 1;
dst_y = dst_y + (height - 1) * dst_stride_y;
dst_u = dst_u + (halfheight - 1) * dst_stride_u;
dst_v = dst_v + (halfheight - 1) * dst_stride_v;
dst_stride_y = -dst_stride_y;
dst_stride_u = -dst_stride_u;
dst_stride_v = -dst_stride_v;
}
int halfwidth = (width + 1) >> 1;
int halfheight = (height + 1) >> 1;
uint8* start_y = dst_y + y * dst_stride_y + x;
uint8* start_u = dst_u + (y / 2) * dst_stride_u + (x / 2);
uint8* start_v = dst_v + (y / 2) * dst_stride_v + (x / 2);
I420SetPlane(start_y, dst_stride_y, width, height, value_y);
I420SetPlane(start_u, dst_stride_u, halfwidth, halfheight, value_u);
I420SetPlane(start_v, dst_stride_v, halfwidth, halfheight, value_v);
return 0;
}
#if defined(_M_IX86) && !defined(YUV_DISABLE_ASM)
#define HAS_HALFROW_SSE2
__declspec(naked)
......@@ -1785,5 +1653,169 @@ int BG24ToARGB(const uint8* src_bg24, int src_stride_bg24,
return 0;
}
// SetRows32 writes 'count' bytes using a 32 bit value repeated
#if defined(__ARM_NEON__) && !defined(YUV_DISABLE_ASM)
#define HAS_SETROW_NEON
static void SetRow32_NEON(uint8* dst, uint32 v32, int count) {
asm volatile (
"vdup.u32 q0, %2 \n" // duplicate 4 ints
"1: \n"
"vst1.u32 {q0}, [%0]! \n" // store
"subs %1, %1, #16 \n" // 16 processed per loop
"bhi 1b \n"
: "+r"(dst), // %0
"+r"(count) // %1
: "r"(v32) // %2
: "q0", "memory", "cc"
);
}
#elif defined(_M_IX86) && !defined(YUV_DISABLE_ASM)
#define HAS_SETROW_SSE2
__declspec(naked)
static void SetRow32_SSE2(uint8* dst, uint32 v32, int count) {
__asm {
mov eax, [esp + 4] // dst
movd xmm5, [esp + 8] // v32
mov ecx, [esp + 12] // count
pshufd xmm5, xmm5, 0
convertloop:
movdqa [eax], xmm5
lea eax, [eax + 16]
sub ecx, 16
ja convertloop
ret
}
}
#elif (defined(__x86_64__) || defined(__i386__)) && !defined(YUV_DISABLE_ASM)
#define HAS_SETROW_SSE2
static void SetRow32_SSE2(uint8* dst, uint32 v32, int count) {
asm volatile (
"movd %2, %%xmm5 \n"
"pshufd $0x0,%%xmm5,%%xmm5 \n"
"1: \n"
"movdqa %%xmm5,(%0) \n"
"lea 0x10(%0),%0 \n"
"sub $0x10,%1 \n"
"ja 1b \n"
: "+r"(dst), // %0
"+r"(count) // %1
: "r"(v32) // %2
: "memory", "cc"
#if defined(__SSE2__)
, "xmm5"
#endif
);
}
#endif
static void SetRow8_C(uint8* dst, uint32 v8, int count) {
memset(dst, v8, count);
}
static void SetPlane(uint8* dst_y, int dst_stride_y,
int width, int height,
uint32 value) {
void (*SetRow)(uint8* dst, uint32 value, int pix);
#if defined(HAS_SETROW_NEON)
if (TestCpuFlag(kCpuHasNEON) &&
IS_ALIGNED(width, 16) &&
IS_ALIGNED(dst_y, 16) && IS_ALIGNED(dst_stride_y, 16)) {
SetRow = SetRow32_NEON;
} else
#elif defined(HAS_SETROW_SSE2)
if (TestCpuFlag(kCpuHasSSE2) &&
IS_ALIGNED(width, 16) &&
IS_ALIGNED(dst_y, 16) && IS_ALIGNED(dst_stride_y, 16)) {
SetRow = SetRow32_SSE2;
} else
#endif
{
SetRow = SetRow8_C;
}
uint32 v32 = value | (value << 8) | (value << 16) | (value << 24);
// Set plane
for (int y = 0; y < height; ++y) {
SetRow(dst_y, v32, width);
dst_y += dst_stride_y;
}
}
// Draw a rectangle into I420
int I420Rect(uint8* dst_y, int dst_stride_y,
uint8* dst_u, int dst_stride_u,
uint8* dst_v, int dst_stride_v,
int x, int y,
int width, int height,
int value_y, int value_u, int value_v) {
if (!dst_y || !dst_u || !dst_v ||
width <= 0 || height <= 0 ||
x < 0 || y < 0 ||
value_y < 0 || value_y > 255 ||
value_u < 0 || value_u > 255 ||
value_v < 0 || value_v > 255) {
return -1;
}
int halfwidth = (width + 1) >> 1;
int halfheight = (height + 1) >> 1;
uint8* start_y = dst_y + y * dst_stride_y + x;
uint8* start_u = dst_u + (y / 2) * dst_stride_u + (x / 2);
uint8* start_v = dst_v + (y / 2) * dst_stride_v + (x / 2);
SetPlane(start_y, dst_stride_y, width, height, value_y);
SetPlane(start_u, dst_stride_u, halfwidth, halfheight, value_u);
SetPlane(start_v, dst_stride_v, halfwidth, halfheight, value_v);
return 0;
}
// count measured in bytes
static void SetRow32_C(uint8* dst, uint32 v8, int count) {
uint32* d = reinterpret_cast<uint32*>(dst);
for (int x = 0; x < count; x += 4) {
*d++ = v8;
}
}
// Draw a rectangle into ARGB
int ARGBRect(uint8* dst_argb, int dst_stride_argb,
int x, int y,
int width, int height,
uint32 value) {
if (!dst_argb ||
width <= 0 || height <= 0 ||
x < 0 || y < 0) {
return -1;
}
void (*SetRow)(uint8* dst, uint32 value, int count);
#if defined(HAS_SETROW_NEON)
if (TestCpuFlag(kCpuHasNEON) &&
IS_ALIGNED(width, 16) &&
IS_ALIGNED(dst_argb, 16) && IS_ALIGNED(dst_stride_argb, 16)) {
SetRow = SetRow32_NEON;
} else
#elif defined(HAS_SETROW_SSE2)
if (TestCpuFlag(kCpuHasSSE2) &&
IS_ALIGNED(width, 16) &&
IS_ALIGNED(dst_argb, 16) && IS_ALIGNED(dst_stride_argb, 16)) {
SetRow = SetRow32_SSE2;
} else
#endif
{
SetRow = SetRow32_C;
}
int w32 = width << 2;
for (int y = 0; y < height; ++y) {
SetRow(dst_argb, value, w32);
dst_argb += dst_stride_argb;
}
return 0;
}
} // namespace libyuv
source/scale.cc
View file @ 3e8a577b
......@@ -3060,7 +3060,6 @@ static void ScalePlaneDown2(int src_width, int src_height,
assert(IS_ALIGNED(src_height, 2));
void (*ScaleRowDown2)(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width);
#if defined(HAS_SCALEROWDOWN2_NEON)
if (TestCpuFlag(kCpuHasNEON) &&
IS_ALIGNED(dst_width, 16)) {
......@@ -3078,7 +3077,6 @@ static void ScalePlaneDown2(int src_width, int src_height,
{
ScaleRowDown2 = filtering ? ScaleRowDown2Int_C : ScaleRowDown2_C;
}
for (int y = 0; y < dst_height; ++y) {
ScaleRowDown2(src_ptr, src_stride, dst_ptr, dst_width);
src_ptr += (src_stride << 1);
......@@ -3101,7 +3099,6 @@ static void ScalePlaneDown4(int src_width, int src_height,
assert(IS_ALIGNED(src_height, 4));
void (*ScaleRowDown4)(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width);
#if defined(HAS_SCALEROWDOWN4_NEON)
if (TestCpuFlag(kCpuHasNEON) &&
IS_ALIGNED(dst_width, 4)) {
......@@ -3119,7 +3116,6 @@ static void ScalePlaneDown4(int src_width, int src_height,
{
ScaleRowDown4 = filtering ? ScaleRowDown4Int_C : ScaleRowDown4_C;
}
for (int y = 0; y < dst_height; ++y) {
ScaleRowDown4(src_ptr, src_stride, dst_ptr, dst_width);
src_ptr += (src_stride << 2);
......
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