ref: 72af533f795098ea5e4952a1cc3947ddc8831c81
parent: 6b2792b0e0189fa3f7d75b2ac885417606225656
parent: 153eec46e04e4bd24c67ed3ca58374c856231cce
author: Yunqing Wang <yunqingwang@google.com>
date: Tue Dec 13 05:39:44 EST 2011
Merge "Align image buffer in multiple-resolution encoder"
--- a/third_party/libyuv/README.webm
+++ b/third_party/libyuv/README.webm
@@ -1,6 +1,6 @@
Name: libyuv
URL: http://code.google.com/p/libyuv/
-Version: 90
+Version: 102
License: BSD
License File: LICENSE
--- a/third_party/libyuv/include/libyuv/basic_types.h
+++ b/third_party/libyuv/include/libyuv/basic_types.h
@@ -13,7 +13,7 @@
#include <stddef.h> // for NULL, size_t
-#ifndef WIN32
+#if !(defined(_MSC_VER) && (_MSC_VER < 1600))
#include <stdint.h> // for uintptr_t
#endif
@@ -20,15 +20,6 @@
#ifndef INT_TYPES_DEFINED
#define INT_TYPES_DEFINED
#ifdef COMPILER_MSVC
-typedef __int64 int64;
-#else
-typedef long long int64;
-#endif /* COMPILER_MSVC */
-typedef int int32;
-typedef short int16;
-typedef char int8;
-
-#ifdef COMPILER_MSVC
typedef unsigned __int64 uint64;
typedef __int64 int64;
#ifndef INT64_C
@@ -38,9 +29,20 @@
#define UINT64_C(x) x ## UI64
#endif
#define INT64_F "I64"
-#else
+#else // COMPILER_MSVC
+#ifdef __LP64__
+typedef unsigned long uint64;
+typedef long int64;
+#ifndef INT64_C
+#define INT64_C(x) x ## L
+#endif
+#ifndef UINT64_C
+#define UINT64_C(x) x ## UL
+#endif
+#define INT64_F "l"
+#else // __LP64__
typedef unsigned long long uint64;
-//typedef long long int64;
+typedef long long int64;
#ifndef INT64_C
#define INT64_C(x) x ## LL
#endif
@@ -48,10 +50,14 @@
#define UINT64_C(x) x ## ULL
#endif
#define INT64_F "ll"
-#endif /* COMPILER_MSVC */
+#endif // __LP64__
+#endif // COMPILER_MSVC
typedef unsigned int uint32;
+typedef int int32;
typedef unsigned short uint16;
+typedef short int16;
typedef unsigned char uint8;
+typedef char int8;
#endif // INT_TYPES_DEFINED
// Detect compiler is for x86 or x64.
@@ -60,7 +66,6 @@
#define CPU_X86 1
#endif
-#define IS_ALIGNED(p, a) (0==((uintptr_t)(p) & ((a)-1)))
#define ALIGNP(p, t) \
((uint8*)((((uintptr_t)(p) + \
((t)-1)) & ~((t)-1))))
--- a/third_party/libyuv/include/libyuv/cpu_id.h
+++ b/third_party/libyuv/include/libyuv/cpu_id.h
@@ -11,21 +11,39 @@
#ifndef INCLUDE_LIBYUV_CPU_ID_H_
#define INCLUDE_LIBYUV_CPU_ID_H_
-//namespace libyuv {+#ifdef __cplusplus
+namespace libyuv {+extern "C" {+#endif
// These flags are only valid on x86 processors
static const int kCpuHasSSE2 = 1;
static const int kCpuHasSSSE3 = 2;
-// SIMD support on ARM processors
+// These flags are only valid on ARM processors
static const int kCpuHasNEON = 4;
+// Internal flag to indicate cpuid is initialized.
+static const int kCpuInitialized = 8;
+
// Detect CPU has SSE2 etc.
-int TestCpuFlag(int flag);
+// test_flag parameter should be one of kCpuHas constants above
+// returns non-zero if instruction set is detected
+static __inline int TestCpuFlag(int test_flag) {+ extern int cpu_info_;
+ extern int InitCpuFlags();
+ return (cpu_info_ ? cpu_info_ : InitCpuFlags()) & test_flag;
+}
// For testing, allow CPU flags to be disabled.
-void MaskCpuFlagsForTest(int enable_flags);
+// ie MaskCpuFlags(~kCpuHasSSSE3) to disable SSSE3.
+// -1 to enable all cpu specific optimizations.
+// 0 to disable all cpu specific optimizations.
+void MaskCpuFlags(int enable_flags);
-//} // namespace libyuv
+#ifdef __cplusplus
+} // extern "C"
+} // namespace libyuv
+#endif
#endif // INCLUDE_LIBYUV_CPU_ID_H_
--- a/third_party/libyuv/include/libyuv/scale.h
+++ b/third_party/libyuv/include/libyuv/scale.h
@@ -13,7 +13,10 @@
#include "third_party/libyuv/include/libyuv/basic_types.h"
-//namespace libyuv {+#ifdef __cplusplus
+namespace libyuv {+extern "C" {+#endif
// Supported filtering
typedef enum {@@ -42,16 +45,8 @@
int dst_width, int dst_height,
FilterMode filtering);
-// Legacy API
-// If dst_height_offset is non-zero, the image is offset by that many pixels
-// and stretched to (dst_height - dst_height_offset * 2) pixels high,
-// instead of dst_height.
-int Scale_1(const uint8* src, int src_width, int src_height,
- uint8* dst, int dst_width, int dst_height, int dst_height_offset,
- int interpolate);
-
-// Same, but specified src terms of each plane location and stride.
-int Scale_2(const uint8* src_y, const uint8* src_u, const uint8* src_v,
+// Legacy API. Deprecated
+int Scale(const uint8* src_y, const uint8* src_u, const uint8* src_v,
int src_stride_y, int src_stride_u, int src_stride_v,
int src_width, int src_height,
uint8* dst_y, uint8* dst_u, uint8* dst_v,
@@ -59,9 +54,17 @@
int dst_width, int dst_height,
int interpolate);
+// Legacy API. Deprecated
+int ScaleOffset(const uint8* src, int src_width, int src_height,
+ uint8* dst, int dst_width, int dst_height, int dst_yoffset,
+ int interpolate);
+
// For testing, allow disabling of optimizations.
void SetUseReferenceImpl(int use);
-//} // namespace libyuv
+#ifdef __cplusplus
+} // extern "C"
+} // namespace libyuv
+#endif
#endif // INCLUDE_LIBYUV_SCALE_H_
--- a/third_party/libyuv/source/cpu_id.c
+++ b/third_party/libyuv/source/cpu_id.c
@@ -9,19 +9,23 @@
*/
#include "third_party/libyuv/include/libyuv/cpu_id.h"
-#include "third_party/libyuv/include/libyuv/basic_types.h" // for CPU_X86
#ifdef _MSC_VER
#include <intrin.h>
#endif
+#ifdef __ANDROID__
+#include <cpu-features.h>
+#endif
+#include "third_party/libyuv/include/libyuv/basic_types.h" // for CPU_X86
+
// TODO(fbarchard): Use cpuid.h when gcc 4.4 is used on OSX and Linux.
#if (defined(__pic__) || defined(__APPLE__)) && defined(__i386__)
static inline void __cpuid(int cpu_info[4], int info_type) {- __asm__ volatile (
- "mov %%ebx, %%edi\n"
- "cpuid\n"
- "xchg %%edi, %%ebx\n"
+ asm volatile (
+ "mov %%ebx, %%edi \n"
+ "cpuid \n"
+ "xchg %%edi, %%ebx \n"
: "=a"(cpu_info[0]), "=D"(cpu_info[1]), "=c"(cpu_info[2]), "=d"(cpu_info[3])
: "a"(info_type)
);
@@ -28,8 +32,8 @@
}
#elif defined(__i386__) || defined(__x86_64__)
static inline void __cpuid(int cpu_info[4], int info_type) {- __asm__ volatile (
- "cpuid\n"
+ asm volatile (
+ "cpuid \n"
: "=a"(cpu_info[0]), "=b"(cpu_info[1]), "=c"(cpu_info[2]), "=d"(cpu_info[3])
: "a"(info_type)
);
@@ -36,39 +40,42 @@
}
#endif
-//namespace libyuv {+#ifdef __cplusplus
+namespace libyuv {+extern "C" {+#endif
// CPU detect function for SIMD instruction sets.
-static int cpu_info_initialized_ = 0;
-static int cpu_info_ = 0;
+int cpu_info_ = 0;
-// Global lock for cpu initialization.
-static void InitCpuFlags() {+int InitCpuFlags() {#ifdef CPU_X86
int cpu_info[4];
__cpuid(cpu_info, 1);
- cpu_info_ = (cpu_info[2] & 0x00000200 ? kCpuHasSSSE3 : 0) |
- (cpu_info[3] & 0x04000000 ? kCpuHasSSE2 : 0);
+ cpu_info_ = (cpu_info[3] & 0x04000000 ? kCpuHasSSE2 : 0) |
+ (cpu_info[2] & 0x00000200 ? kCpuHasSSSE3 : 0) |
+ kCpuInitialized;
+#elif defined(__ANDROID__) && defined(__ARM_NEON__)
+ uint64_t features = android_getCpuFeatures();
+ cpu_info_ = ((features & ANDROID_CPU_ARM_FEATURE_NEON) ? kCpuHasNEON : 0) |
+ kCpuInitialized;
#elif defined(__ARM_NEON__)
// gcc -mfpu=neon defines __ARM_NEON__
- // if code is specifically built for Neon-only, enable the flag.
- cpu_info_ |= kCpuHasNEON;
+ // Enable Neon if you want support for Neon and Arm, and use MaskCpuFlags
+ // to disable Neon on devices that do not have it.
+ cpu_info_ = kCpuHasNEON | kCpuInitialized;
#else
- cpu_info_ = 0;
+ cpu_info_ = kCpuInitialized;
#endif
- cpu_info_initialized_ = 1;
+ return cpu_info_;
}
-void MaskCpuFlagsForTest(int enable_flags) {+void MaskCpuFlags(int enable_flags) {InitCpuFlags();
- cpu_info_ &= enable_flags;
+ cpu_info_ = (cpu_info_ & enable_flags) | kCpuInitialized;
}
-int TestCpuFlag(int flag) {- if (!cpu_info_initialized_) {- InitCpuFlags();
- }
- return cpu_info_ & flag ? 1 : 0;
-}
-
-//} // namespace libyuv
+#ifdef __cplusplus
+} // extern "C"
+} // namespace libyuv
+#endif
--- a/third_party/libyuv/source/row.h
+++ b/third_party/libyuv/source/row.h
@@ -14,7 +14,7 @@
#include "third_party/libyuv/include/libyuv/basic_types.h"
#define kMaxStride (2048 * 4)
-//#define IS_ALIGNED(p, a) (!((uintptr_t)(p) & ((a) - 1)))
+#define IS_ALIGNED(p, a) (!((uintptr_t)(p) & ((a) - 1)))
#if defined(COVERAGE_ENABLED) || defined(TARGET_IPHONE_SIMULATOR)
#define YUV_DISABLE_ASM
@@ -72,7 +72,10 @@
#define HAS_REVERSE_ROW_NEON
#endif
-//extern "C" {+#ifdef __cplusplus
+namespace libyuv {+extern "C" {+#endif
#ifdef HAS_ARGBTOYROW_SSSE3
void ARGBToYRow_SSSE3(const uint8* src_argb, uint8* dst_y, int pix);
@@ -253,6 +256,9 @@
#endif
-//} // extern "C"
+#ifdef __cplusplus
+} // extern "C"
+} // namespace libyuv
+#endif
#endif // LIBYUV_SOURCE_ROW_H_
--- a/third_party/libyuv/source/scale.c
+++ b/third_party/libyuv/source/scale.c
@@ -15,6 +15,17 @@
#include "third_party/libyuv/include/libyuv/cpu_id.h"
#include "third_party/libyuv/source/row.h"
+
+#ifdef __cplusplus
+namespace libyuv {+extern "C" {+#endif
+
+/*
+ * Note: Defining YUV_DISABLE_ASM allows to use c version.
+ */
+//#define YUV_DISABLE_ASM
+
#if defined(_MSC_VER)
#define ALIGN16(var) __declspec(align(16)) var
#else
@@ -26,8 +37,6 @@
// Note: Some SSE2 reference manuals
// cpuvol1.pdf agner_instruction_tables.pdf 253666.pdf 253667.pdf
-//namespace libyuv {-
// Set the following flag to true to revert to only
// using the reference implementation ScalePlaneBox(), and
// NOT the optimized versions. Useful for debugging and
@@ -40,9 +49,7 @@
use_reference_impl_ = use;
}
-// TODO: The preprocessor definitions for Win64 are not right in build system.
-// Disable optimized code for now.
-#define YUV_DISABLE_ASM
+// ScaleRowDown2Int also used by planar functions
/**
* NEON downscalers with interpolation.
@@ -511,83 +518,116 @@
!defined(YUV_DISABLE_ASM)
#if defined(_MSC_VER)
#define TALIGN16(t, var) __declspec(align(16)) t _ ## var
-#elif defined(OSX) && defined(__i386__)
+#elif (defined(__APPLE__) || defined(__MINGW32__) || defined(__CYGWIN__)) && defined(__i386__)
#define TALIGN16(t, var) t var __attribute__((aligned(16)))
#else
#define TALIGN16(t, var) t _ ## var __attribute__((aligned(16)))
#endif
+#if (defined(__APPLE__) || defined(__MINGW32__) || defined(__CYGWIN__)) && \
+ defined(__i386__)
+#define DECLARE_FUNCTION(name) \
+ ".text \n" \
+ ".globl _" #name " \n" \
+"_" #name ": \n"
+#else
+#define DECLARE_FUNCTION(name) \
+ ".text \n" \
+ ".global " #name " \n" \
+#name ": \n"
+#endif
+
+
// Offsets for source bytes 0 to 9
+//extern "C"
TALIGN16(const uint8, shuf0[16]) =
{ 0, 1, 3, 4, 5, 7, 8, 9, 128, 128, 128, 128, 128, 128, 128, 128 };// Offsets for source bytes 11 to 20 with 8 subtracted = 3 to 12.
+//extern "C"
TALIGN16(const uint8, shuf1[16]) =
{ 3, 4, 5, 7, 8, 9, 11, 12, 128, 128, 128, 128, 128, 128, 128, 128 };// Offsets for source bytes 21 to 31 with 16 subtracted = 5 to 31.
+//extern "C"
TALIGN16(const uint8, shuf2[16]) =
{ 5, 7, 8, 9, 11, 12, 13, 15, 128, 128, 128, 128, 128, 128, 128, 128 };// Offsets for source bytes 0 to 10
+//extern "C"
TALIGN16(const uint8, shuf01[16]) =
{ 0, 1, 1, 2, 2, 3, 4, 5, 5, 6, 6, 7, 8, 9, 9, 10 };// Offsets for source bytes 10 to 21 with 8 subtracted = 3 to 13.
+//extern "C"
TALIGN16(const uint8, shuf11[16]) =
{ 2, 3, 4, 5, 5, 6, 6, 7, 8, 9, 9, 10, 10, 11, 12, 13 };// Offsets for source bytes 21 to 31 with 16 subtracted = 5 to 31.
+//extern "C"
TALIGN16(const uint8, shuf21[16]) =
{ 5, 6, 6, 7, 8, 9, 9, 10, 10, 11, 12, 13, 13, 14, 14, 15 };// Coefficients for source bytes 0 to 10
+//extern "C"
TALIGN16(const uint8, madd01[16]) =
{ 3, 1, 2, 2, 1, 3, 3, 1, 2, 2, 1, 3, 3, 1, 2, 2 };// Coefficients for source bytes 10 to 21
+//extern "C"
TALIGN16(const uint8, madd11[16]) =
{ 1, 3, 3, 1, 2, 2, 1, 3, 3, 1, 2, 2, 1, 3, 3, 1 };// Coefficients for source bytes 21 to 31
+//extern "C"
TALIGN16(const uint8, madd21[16]) =
{ 2, 2, 1, 3, 3, 1, 2, 2, 1, 3, 3, 1, 2, 2, 1, 3 };// Coefficients for source bytes 21 to 31
+//extern "C"
TALIGN16(const int16, round34[8]) =
{ 2, 2, 2, 2, 2, 2, 2, 2 };+//extern "C"
TALIGN16(const uint8, shuf38a[16]) =
{ 0, 3, 6, 8, 11, 14, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128 };+//extern "C"
TALIGN16(const uint8, shuf38b[16]) =
{ 128, 128, 128, 128, 128, 128, 0, 3, 6, 8, 11, 14, 128, 128, 128, 128 };// Arrange words 0,3,6 into 0,1,2
+//extern "C"
TALIGN16(const uint8, shufac0[16]) =
{ 0, 1, 6, 7, 12, 13, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128 };// Arrange words 0,3,6 into 3,4,5
+//extern "C"
TALIGN16(const uint8, shufac3[16]) =
{ 128, 128, 128, 128, 128, 128, 0, 1, 6, 7, 12, 13, 128, 128, 128, 128 };// Scaling values for boxes of 3x3 and 2x3
+//extern "C"
TALIGN16(const uint16, scaleac3[8]) =
{ 65536 / 9, 65536 / 9, 65536 / 6, 65536 / 9, 65536 / 9, 65536 / 6, 0, 0 };// Arrange first value for pixels 0,1,2,3,4,5
+//extern "C"
TALIGN16(const uint8, shufab0[16]) =
{ 0, 128, 3, 128, 6, 128, 8, 128, 11, 128, 14, 128, 128, 128, 128, 128 };// Arrange second value for pixels 0,1,2,3,4,5
+//extern "C"
TALIGN16(const uint8, shufab1[16]) =
{ 1, 128, 4, 128, 7, 128, 9, 128, 12, 128, 15, 128, 128, 128, 128, 128 };// Arrange third value for pixels 0,1,2,3,4,5
+//extern "C"
TALIGN16(const uint8, shufab2[16]) =
{ 2, 128, 5, 128, 128, 128, 10, 128, 13, 128, 128, 128, 128, 128, 128, 128 };// Scaling values for boxes of 3x2 and 2x2
+//extern "C"
TALIGN16(const uint16, scaleab2[8]) =
{ 65536 / 3, 65536 / 3, 65536 / 2, 65536 / 3, 65536 / 3, 65536 / 2, 0, 0 };#endif
@@ -1620,14 +1660,7 @@
void ScaleRowDown8Int_SSE2(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width);
asm(
- ".text \n"
-#if defined(OSX)
- ".globl _ScaleRowDown8Int_SSE2 \n"
-"_ScaleRowDown8Int_SSE2: \n"
-#else
- ".global ScaleRowDown8Int_SSE2 \n"
-"ScaleRowDown8Int_SSE2: \n"
-#endif
+ DECLARE_FUNCTION(ScaleRowDown8Int_SSE2)
"pusha \n"
"mov 0x24(%esp),%esi \n"
"mov 0x28(%esp),%ebx \n"
@@ -1691,14 +1724,7 @@
void ScaleRowDown34_SSSE3(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width);
asm(
- ".text \n"
-#if defined(OSX)
- ".globl _ScaleRowDown34_SSSE3 \n"
-"_ScaleRowDown34_SSSE3: \n"
-#else
- ".global ScaleRowDown34_SSSE3 \n"
-"ScaleRowDown34_SSSE3: \n"
-#endif
+ DECLARE_FUNCTION(ScaleRowDown34_SSSE3)
"pusha \n"
"mov 0x24(%esp),%esi \n"
"mov 0x2c(%esp),%edi \n"
@@ -1729,14 +1755,7 @@
void ScaleRowDown34_1_Int_SSSE3(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width);
asm(
- ".text \n"
-#if defined(OSX)
- ".globl _ScaleRowDown34_1_Int_SSSE3 \n"
-"_ScaleRowDown34_1_Int_SSSE3: \n"
-#else
- ".global ScaleRowDown34_1_Int_SSSE3 \n"
-"ScaleRowDown34_1_Int_SSSE3: \n"
-#endif
+ DECLARE_FUNCTION(ScaleRowDown34_1_Int_SSSE3)
"pusha \n"
"mov 0x24(%esp),%esi \n"
"mov 0x28(%esp),%ebp \n"
@@ -1790,14 +1809,7 @@
void ScaleRowDown34_0_Int_SSSE3(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width);
asm(
- ".text \n"
-#if defined(OSX)
- ".globl _ScaleRowDown34_0_Int_SSSE3 \n"
-"_ScaleRowDown34_0_Int_SSSE3: \n"
-#else
- ".global ScaleRowDown34_0_Int_SSSE3 \n"
-"ScaleRowDown34_0_Int_SSSE3: \n"
-#endif
+ DECLARE_FUNCTION(ScaleRowDown34_0_Int_SSSE3)
"pusha \n"
"mov 0x24(%esp),%esi \n"
"mov 0x28(%esp),%ebp \n"
@@ -1854,14 +1866,7 @@
void ScaleRowDown38_SSSE3(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width);
asm(
- ".text \n"
-#if defined(OSX)
- ".globl _ScaleRowDown38_SSSE3 \n"
-"_ScaleRowDown38_SSSE3: \n"
-#else
- ".global ScaleRowDown38_SSSE3 \n"
-"ScaleRowDown38_SSSE3: \n"
-#endif
+ DECLARE_FUNCTION(ScaleRowDown38_SSSE3)
"pusha \n"
"mov 0x24(%esp),%esi \n"
"mov 0x28(%esp),%edx \n"
@@ -1890,14 +1895,7 @@
void ScaleRowDown38_3_Int_SSSE3(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width);
asm(
- ".text \n"
-#if defined(OSX)
- ".globl _ScaleRowDown38_3_Int_SSSE3 \n"
-"_ScaleRowDown38_3_Int_SSSE3: \n"
-#else
- ".global ScaleRowDown38_3_Int_SSSE3 \n"
-"ScaleRowDown38_3_Int_SSSE3: \n"
-#endif
+ DECLARE_FUNCTION(ScaleRowDown38_3_Int_SSSE3)
"pusha \n"
"mov 0x24(%esp),%esi \n"
"mov 0x28(%esp),%edx \n"
@@ -1954,14 +1952,7 @@
void ScaleRowDown38_2_Int_SSSE3(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width);
asm(
- ".text \n"
-#if defined(OSX)
- ".globl _ScaleRowDown38_2_Int_SSSE3 \n"
-"_ScaleRowDown38_2_Int_SSSE3: \n"
-#else
- ".global ScaleRowDown38_2_Int_SSSE3 \n"
-"ScaleRowDown38_2_Int_SSSE3: \n"
-#endif
+ DECLARE_FUNCTION(ScaleRowDown38_2_Int_SSSE3)
"pusha \n"
"mov 0x24(%esp),%esi \n"
"mov 0x28(%esp),%edx \n"
@@ -2001,14 +1992,7 @@
uint16* dst_ptr, int src_width,
int src_height);
asm(
- ".text \n"
-#if defined(OSX)
- ".globl _ScaleAddRows_SSE2 \n"
-"_ScaleAddRows_SSE2: \n"
-#else
- ".global ScaleAddRows_SSE2 \n"
-"ScaleAddRows_SSE2: \n"
-#endif
+ DECLARE_FUNCTION(ScaleAddRows_SSE2)
"pusha \n"
"mov 0x24(%esp),%esi \n"
"mov 0x28(%esp),%edx \n"
@@ -2052,14 +2036,7 @@
const uint8* src_ptr, int src_stride,
int dst_width, int source_y_fraction);
asm(
- ".text \n"
-#if defined(OSX)
- ".globl _ScaleFilterRows_SSE2 \n"
-"_ScaleFilterRows_SSE2: \n"
-#else
- ".global ScaleFilterRows_SSE2 \n"
-"ScaleFilterRows_SSE2: \n"
-#endif
+ DECLARE_FUNCTION(ScaleFilterRows_SSE2)
"push %esi \n"
"push %edi \n"
"mov 0xc(%esp),%edi \n"
@@ -2147,14 +2124,7 @@
const uint8* src_ptr, int src_stride,
int dst_width, int source_y_fraction);
asm(
- ".text \n"
-#if defined(OSX)
- ".globl _ScaleFilterRows_SSSE3 \n"
-"_ScaleFilterRows_SSSE3: \n"
-#else
- ".global ScaleFilterRows_SSSE3 \n"
-"ScaleFilterRows_SSSE3: \n"
-#endif
+ DECLARE_FUNCTION(ScaleFilterRows_SSSE3)
"push %esi \n"
"push %edi \n"
"mov 0xc(%esp),%edi \n"
@@ -2318,7 +2288,7 @@
static void ScaleRowDown34_1_Int_SSSE3(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) {- asm volatile(
+ asm volatile (
"movdqa (%4),%%xmm2 \n" // _shuf01
"movdqa (%5),%%xmm3 \n" // _shuf11
"movdqa (%6),%%xmm4 \n" // _shuf21
@@ -2436,7 +2406,7 @@
#define HAS_SCALEROWDOWN38_SSSE3
static void ScaleRowDown38_SSSE3(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) {- asm volatile(
+ asm volatile (
"movdqa (%3),%%xmm4 \n"
"movdqa (%4),%%xmm5 \n"
"1:"
@@ -2560,7 +2530,7 @@
static void ScaleAddRows_SSE2(const uint8* src_ptr, int src_stride,
uint16* dst_ptr, int src_width,
int src_height) {- asm volatile(
+ asm volatile (
"pxor %%xmm5,%%xmm5 \n"
"1:"
"movdqa (%0),%%xmm2 \n"
@@ -2602,7 +2572,7 @@
const uint8* src_ptr, int src_stride,
int dst_width, int source_y_fraction) { if (source_y_fraction == 0) {- asm volatile(
+ asm volatile (
"1:"
"movdqa (%1),%%xmm0 \n"
"lea 0x10(%1),%1 \n"
@@ -2620,7 +2590,7 @@
);
return;
} else if (source_y_fraction == 128) {- asm volatile(
+ asm volatile (
"1:"
"movdqa (%1),%%xmm0 \n"
"movdqa (%1,%3,1),%%xmm2 \n"
@@ -2640,7 +2610,7 @@
);
return;
} else {- asm volatile(
+ asm volatile (
"mov %3,%%eax \n"
"movd %%eax,%%xmm6 \n"
"punpcklwd %%xmm6,%%xmm6 \n"
@@ -2693,7 +2663,7 @@
const uint8* src_ptr, int src_stride,
int dst_width, int source_y_fraction) { if (source_y_fraction == 0) {- asm volatile(
+ asm volatile (
"1:"
"movdqa (%1),%%xmm0 \n"
"lea 0x10(%1),%1 \n"
@@ -2711,7 +2681,7 @@
);
return;
} else if (source_y_fraction == 128) {- asm volatile(
+ asm volatile (
"1:"
"movdqa (%1),%%xmm0 \n"
"movdqa (%1,%3,1),%%xmm2 \n"
@@ -2731,7 +2701,7 @@
);
return;
} else {- asm volatile(
+ asm volatile (
"mov %3,%%eax \n"
"shr %%eax \n"
"mov %%al,%%ah \n"
@@ -3095,10 +3065,7 @@
ScaleRowDown2 = filtering ? ScaleRowDown2Int_NEON : ScaleRowDown2_NEON;
} else
#endif
-/* TODO: Force to call C version all the time in ordert to get matching results
- * in multi-resolution encoder example.
- */
-#if 0 //defined(HAS_SCALEROWDOWN2_SSE2)
+#if defined(HAS_SCALEROWDOWN2_SSE2)
if (TestCpuFlag(kCpuHasSSE2) &&
IS_ALIGNED(dst_width, 16) &&
IS_ALIGNED(src_ptr, 16) && IS_ALIGNED(src_stride, 16) &&
@@ -3267,33 +3234,33 @@
}
}
{- int src_row = 0;
+ int src_row = 0;
int y;
for (y = 0; y < dst_height; ++y) {- switch (src_row) {- case 0:
- ScaleRowDown34_0(src_ptr, src_stride, dst_ptr, dst_width);
- break;
+ switch (src_row) {+ case 0:
+ ScaleRowDown34_0(src_ptr, src_stride, dst_ptr, dst_width);
+ break;
- case 1:
- ScaleRowDown34_1(src_ptr, src_stride, dst_ptr, dst_width);
- break;
+ case 1:
+ ScaleRowDown34_1(src_ptr, src_stride, dst_ptr, dst_width);
+ break;
- case 2:
- ScaleRowDown34_0(src_ptr + src_stride, -src_stride,
- dst_ptr, dst_width);
- break;
- }
- ++src_row;
+ case 2:
+ ScaleRowDown34_0(src_ptr + src_stride, -src_stride,
+ dst_ptr, dst_width);
+ break;
+ }
+ ++src_row;
+ src_ptr += src_stride;
+ dst_ptr += dst_stride;
+ if (src_row >= 3) {src_ptr += src_stride;
- dst_ptr += dst_stride;
- if (src_row >= 3) {- src_ptr += src_stride;
- src_row = 0;
- }
+ src_row = 0;
}
}
}
+}
/**
* Scale plane, 3/8
@@ -3350,27 +3317,27 @@
}
}
{- int src_row = 0;
+ int src_row = 0;
int y;
for (y = 0; y < dst_height; ++y) {- switch (src_row) {- case 0:
- case 1:
- ScaleRowDown38_3(src_ptr, src_stride, dst_ptr, dst_width);
- src_ptr += src_stride * 3;
- ++src_row;
- break;
+ switch (src_row) {+ case 0:
+ case 1:
+ ScaleRowDown38_3(src_ptr, src_stride, dst_ptr, dst_width);
+ src_ptr += src_stride * 3;
+ ++src_row;
+ break;
- case 2:
- ScaleRowDown38_2(src_ptr, src_stride, dst_ptr, dst_width);
- src_ptr += src_stride * 2;
- src_row = 0;
- break;
- }
- dst_ptr += dst_stride;
+ case 2:
+ ScaleRowDown38_2(src_ptr, src_stride, dst_ptr, dst_width);
+ src_ptr += src_stride * 2;
+ src_row = 0;
+ break;
}
+ dst_ptr += dst_stride;
}
}
+}
__inline static uint32 SumBox(int iboxwidth, int iboxheight,
int src_stride, const uint8* src_ptr) {@@ -3421,15 +3388,15 @@
scaletbl[0] = 65536 / (minboxwidth * boxheight);
scaletbl[1] = 65536 / ((minboxwidth + 1) * boxheight);
{- int *scaleptr = scaletbl - minboxwidth;
- int x = 0;
+ int *scaleptr = scaletbl - minboxwidth;
+ int x = 0;
int i;
for (i = 0; i < dst_width; ++i) {- int ix = x >> 16;
+ int ix = x >> 16;
int boxwidth;
- x += dx;
+ x += dx;
boxwidth = (x >> 16) - ix;
- *dst_ptr++ = SumPixels(boxwidth, src_ptr + ix) * scaleptr[boxwidth] >> 16;
+ *dst_ptr++ = SumPixels(boxwidth, src_ptr + ix) * scaleptr[boxwidth] >> 16;
}
}
}
@@ -3509,20 +3476,20 @@
}
{- int y = 0;
+ int y = 0;
int j;
for (j = 0; j < dst_height; ++j) {- int iy = y >> 16;
- const uint8* const src = src_ptr + iy * src_stride;
+ int iy = y >> 16;
+ const uint8* const src = src_ptr + iy * src_stride;
int boxheight;
- y += dy;
- if (y > (src_height << 16)) {- y = (src_height << 16);
- }
+ y += dy;
+ if (y > (src_height << 16)) {+ y = (src_height << 16);
+ }
boxheight = (y >> 16) - iy;
- ScaleAddRows(src, src_stride, row, src_width, boxheight);
- ScaleAddCols(dst_width, boxheight, dx, row, dst_ptr);
- dst_ptr += dst_stride;
+ ScaleAddRows(src, src_stride, row, src_width, boxheight);
+ ScaleAddCols(dst_width, boxheight, dx, row, dst_ptr);
+ dst_ptr += dst_stride;
}
}
}
@@ -3614,24 +3581,24 @@
ScaleFilterCols = ScaleFilterCols_C;
{- int y = 0;
- int maxy = ((src_height - 1) << 16) - 1; // max is filter of last 2 rows.
+ int y = 0;
+ int maxy = ((src_height - 1) << 16) - 1; // max is filter of last 2 rows.
int j;
for (j = 0; j < dst_height; ++j) {- int iy = y >> 16;
- int fy = (y >> 8) & 255;
- const uint8* const src = src_ptr + iy * src_stride;
- ScaleFilterRows(row, src, src_stride, src_width, fy);
- ScaleFilterCols(dst_ptr, row, dst_width, dx);
- dst_ptr += dst_stride;
- y += dy;
- if (y > maxy) {- y = maxy;
- }
+ int iy = y >> 16;
+ int fy = (y >> 8) & 255;
+ const uint8* const src = src_ptr + iy * src_stride;
+ ScaleFilterRows(row, src, src_stride, src_width, fy);
+ ScaleFilterCols(dst_ptr, row, dst_width, dx);
+ dst_ptr += dst_stride;
+ y += dy;
+ if (y > maxy) {+ y = maxy;
}
}
}
}
+}
/**
* Scale plane to/from any dimensions, without interpolation.
@@ -3818,25 +3785,26 @@
src_stride_v = -src_stride_v;
}
{- int halfsrc_width = (src_width + 1) >> 1;
- int halfsrc_height = (src_height + 1) >> 1;
- int halfdst_width = (dst_width + 1) >> 1;
- int halfoheight = (dst_height + 1) >> 1;
+ int src_halfwidth = (src_width + 1) >> 1;
+ int src_halfheight = (src_height + 1) >> 1;
+ int dst_halfwidth = (dst_width + 1) >> 1;
+ int dst_halfheight = (dst_height + 1) >> 1;
ScalePlane(src_y, src_stride_y, src_width, src_height,
dst_y, dst_stride_y, dst_width, dst_height,
filtering, use_reference_impl_);
- ScalePlane(src_u, src_stride_u, halfsrc_width, halfsrc_height,
- dst_u, dst_stride_u, halfdst_width, halfoheight,
- filtering, use_reference_impl_);
- ScalePlane(src_v, src_stride_v, halfsrc_width, halfsrc_height,
- dst_v, dst_stride_v, halfdst_width, halfoheight,
- filtering, use_reference_impl_);
+ ScalePlane(src_u, src_stride_u, src_halfwidth, src_halfheight,
+ dst_u, dst_stride_u, dst_halfwidth, dst_halfheight,
+ filtering, use_reference_impl_);
+ ScalePlane(src_v, src_stride_v, src_halfwidth, src_halfheight,
+ dst_v, dst_stride_v, dst_halfwidth, dst_halfheight,
+ filtering, use_reference_impl_);
}
return 0;
}
-int Scale_2(const uint8* src_y, const uint8* src_u, const uint8* src_v,
+// Deprecated api
+int Scale(const uint8* src_y, const uint8* src_u, const uint8* src_v,
int src_stride_y, int src_stride_u, int src_stride_v,
int src_width, int src_height,
uint8* dst_y, uint8* dst_u, uint8* dst_v,
@@ -3843,11 +3811,6 @@
int dst_stride_y, int dst_stride_u, int dst_stride_v,
int dst_width, int dst_height,
int interpolate) {- int halfsrc_width;
- int halfsrc_height;
- int halfdst_width;
- int halfoheight;
- FilterMode filtering;
if (!src_y || !src_u || !src_v || src_width <= 0 || src_height == 0 ||
!dst_y || !dst_u || !dst_v || dst_width <= 0 || dst_height <= 0) {return -1;
@@ -3864,51 +3827,58 @@
src_stride_u = -src_stride_u;
src_stride_v = -src_stride_v;
}
- halfsrc_width = (src_width + 1) >> 1;
- halfsrc_height = (src_height + 1) >> 1;
- halfdst_width = (dst_width + 1) >> 1;
- halfoheight = (dst_height + 1) >> 1;
- filtering = interpolate ? kFilterBox : kFilterNone;
+ {+ int src_halfwidth = (src_width + 1) >> 1;
+ int src_halfheight = (src_height + 1) >> 1;
+ int dst_halfwidth = (dst_width + 1) >> 1;
+ int dst_halfheight = (dst_height + 1) >> 1;
+ FilterMode filtering = interpolate ? kFilterBox : kFilterNone;
ScalePlane(src_y, src_stride_y, src_width, src_height,
dst_y, dst_stride_y, dst_width, dst_height,
filtering, use_reference_impl_);
- ScalePlane(src_u, src_stride_u, halfsrc_width, halfsrc_height,
- dst_u, dst_stride_u, halfdst_width, halfoheight,
+ ScalePlane(src_u, src_stride_u, src_halfwidth, src_halfheight,
+ dst_u, dst_stride_u, dst_halfwidth, dst_halfheight,
filtering, use_reference_impl_);
- ScalePlane(src_v, src_stride_v, halfsrc_width, halfsrc_height,
- dst_v, dst_stride_v, halfdst_width, halfoheight,
+ ScalePlane(src_v, src_stride_v, src_halfwidth, src_halfheight,
+ dst_v, dst_stride_v, dst_halfwidth, dst_halfheight,
filtering, use_reference_impl_);
+ }
return 0;
}
-int Scale_1(const uint8* src, int src_width, int src_height,
- uint8* dst, int dst_width, int dst_height, int ooffset,
+// Deprecated api
+int ScaleOffset(const uint8* src, int src_width, int src_height,
+ uint8* dst, int dst_width, int dst_height, int dst_yoffset,
int interpolate) {if (!src || src_width <= 0 || src_height <= 0 ||
- !dst || dst_width <= 0 || dst_height <= 0 || ooffset < 0 ||
- ooffset >= dst_height) {+ !dst || dst_width <= 0 || dst_height <= 0 || dst_yoffset < 0 ||
+ dst_yoffset >= dst_height) {return -1;
}
- ooffset = ooffset & ~1; // chroma requires offset to multiple of 2.
+ dst_yoffset = dst_yoffset & ~1; // chroma requires offset to multiple of 2.
{- int halfsrc_width = (src_width + 1) >> 1;
- int halfsrc_height = (src_height + 1) >> 1;
- int halfdst_width = (dst_width + 1) >> 1;
- int halfoheight = (dst_height + 1) >> 1;
- int aheight = dst_height - ooffset * 2; // actual output height
- const uint8* const iyptr = src;
- uint8* oyptr = dst + ooffset * dst_width;
- const uint8* const iuptr = src + src_width * src_height;
- uint8* ouptr = dst + dst_width * dst_height + (ooffset >> 1) * halfdst_width;
- const uint8* const ivptr = src + src_width * src_height +
- halfsrc_width * halfsrc_height;
- uint8* ovptr = dst + dst_width * dst_height + halfdst_width * halfoheight +
- (ooffset >> 1) * halfdst_width;
- return Scale_2(iyptr, iuptr, ivptr, src_width, halfsrc_width, halfsrc_width,
- src_width, src_height, oyptr, ouptr, ovptr, dst_width,
- halfdst_width, halfdst_width, dst_width, aheight, interpolate);
+ int src_halfwidth = (src_width + 1) >> 1;
+ int src_halfheight = (src_height + 1) >> 1;
+ int dst_halfwidth = (dst_width + 1) >> 1;
+ int dst_halfheight = (dst_height + 1) >> 1;
+ int aheight = dst_height - dst_yoffset * 2; // actual output height
+ const uint8* const src_y = src;
+ const uint8* const src_u = src + src_width * src_height;
+ const uint8* const src_v = src + src_width * src_height +
+ src_halfwidth * src_halfheight;
+ uint8* dst_y = dst + dst_yoffset * dst_width;
+ uint8* dst_u = dst + dst_width * dst_height +
+ (dst_yoffset >> 1) * dst_halfwidth;
+ uint8* dst_v = dst + dst_width * dst_height + dst_halfwidth * dst_halfheight +
+ (dst_yoffset >> 1) * dst_halfwidth;
+ return Scale(src_y, src_u, src_v, src_width, src_halfwidth, src_halfwidth,
+ src_width, src_height, dst_y, dst_u, dst_v, dst_width,
+ dst_halfwidth, dst_halfwidth, dst_width, aheight, interpolate);
}
}
-//} // namespace libyuv
+#ifdef __cplusplus
+} // extern "C"
+} // namespace libyuv
+#endif
--- a/vp8_multi_resolution_encoder.c
+++ b/vp8_multi_resolution_encoder.c
@@ -78,6 +78,8 @@
exit(EXIT_FAILURE);
}
+int (*read_frame_p)(FILE *f, vpx_image_t *img);
+
static int read_frame(FILE *f, vpx_image_t *img) {size_t nbytes, to_read;
int res = 1;
@@ -92,6 +94,55 @@
return res;
}
+static int read_frame_by_row(FILE *f, vpx_image_t *img) {+ size_t nbytes, to_read;
+ int res = 1;
+ int plane;
+
+ for (plane = 0; plane < 3; plane++)
+ {+ unsigned char *ptr;
+ int w = (plane ? (1 + img->d_w) / 2 : img->d_w);
+ int h = (plane ? (1 + img->d_h) / 2 : img->d_h);
+ int r;
+
+ /* Determine the correct plane based on the image format. The for-loop
+ * always counts in Y,U,V order, but this may not match the order of
+ * the data on disk.
+ */
+ switch (plane)
+ {+ case 1:
+ ptr = img->planes[img->fmt==VPX_IMG_FMT_YV12? VPX_PLANE_V : VPX_PLANE_U];
+ break;
+ case 2:
+ ptr = img->planes[img->fmt==VPX_IMG_FMT_YV12?VPX_PLANE_U : VPX_PLANE_V];
+ break;
+ default:
+ ptr = img->planes[plane];
+ }
+
+ for (r = 0; r < h; r++)
+ {+ to_read = w;
+
+ nbytes = fread(ptr, 1, to_read, f);
+ if(nbytes != to_read) {+ res = 0;
+ if(nbytes > 0)
+ printf("Warning: Read partial frame. Check your width & height!\n");+ break;
+ }
+
+ ptr += img->stride[plane];
+ }
+ if (!res)
+ break;
+ }
+
+ return res;
+}
+
static void write_ivf_file_header(FILE *outfile,
const vpx_codec_enc_cfg_t *cfg,
int frame_cnt) {@@ -262,9 +313,14 @@
/* Allocate image for each encoder */
for (i=0; i< NUM_ENCODERS; i++)
- if(!vpx_img_alloc(&raw[i], VPX_IMG_FMT_I420, cfg[i].g_w, cfg[i].g_h, 1))
+ if(!vpx_img_alloc(&raw[i], VPX_IMG_FMT_I420, cfg[i].g_w, cfg[i].g_h, 32))
die("Failed to allocate image", cfg[i].g_w, cfg[i].g_h);+ if (raw[0].stride[VPX_PLANE_Y] == raw[0].d_w)
+ read_frame_p = read_frame;
+ else
+ read_frame_p = read_frame_by_row;
+
for (i=0; i< NUM_ENCODERS; i++)
write_ivf_file_header(outfile[i], &cfg[i], 0);
@@ -305,35 +361,22 @@
const vpx_codec_cx_pkt_t *pkt[NUM_ENCODERS];
flags = 0;
- frame_avail = read_frame(infile, &raw[0]);
+ frame_avail = read_frame_p(infile, &raw[0]);
- for ( i=1; i<NUM_ENCODERS; i++)
+ if(frame_avail)
{- if(frame_avail)
+ for ( i=1; i<NUM_ENCODERS; i++)
{/*Scale the image down a number of times by downsampling factor*/
- int src_uvwidth = (raw[i-1].d_w + 1) >> 1;
- int src_uvheight = (raw[i-1].d_h + 1) >> 1;
- const unsigned char* src_y = raw[i-1].planes[VPX_PLANE_Y];
- const unsigned char* src_u = raw[i-1].planes[VPX_PLANE_Y]
- + raw[i-1].d_w*raw[i-1].d_h;
- const unsigned char* src_v = raw[i-1].planes[VPX_PLANE_Y]
- + raw[i-1].d_w*raw[i-1].d_h
- + src_uvwidth*src_uvheight;
- int dst_uvwidth = (raw[i].d_w + 1) >> 1;
- int dst_uvheight = (raw[i].d_h + 1) >> 1;
- unsigned char* dst_y = raw[i].planes[VPX_PLANE_Y];
- unsigned char* dst_u = raw[i].planes[VPX_PLANE_Y]
- + raw[i].d_w*raw[i].d_h;
- unsigned char* dst_v = raw[i].planes[VPX_PLANE_Y]
- + raw[i].d_w*raw[i].d_h
- + dst_uvwidth*dst_uvheight;
-
/* FilterMode 1 or 2 give better psnr than FilterMode 0. */
- I420Scale(src_y, raw[i-1].d_w, src_u, src_uvwidth, src_v,
- src_uvwidth, raw[i-1].d_w, raw[i-1].d_h,
- dst_y, raw[i].d_w, dst_u, dst_uvwidth,
- dst_v, dst_uvwidth, raw[i].d_w, raw[i].d_h, 1);
+ I420Scale(raw[i-1].planes[VPX_PLANE_Y], raw[i-1].stride[VPX_PLANE_Y],
+ raw[i-1].planes[VPX_PLANE_U], raw[i-1].stride[VPX_PLANE_U],
+ raw[i-1].planes[VPX_PLANE_V], raw[i-1].stride[VPX_PLANE_V],
+ raw[i-1].d_w, raw[i-1].d_h,
+ raw[i].planes[VPX_PLANE_Y], raw[i].stride[VPX_PLANE_Y],
+ raw[i].planes[VPX_PLANE_U], raw[i].stride[VPX_PLANE_U],
+ raw[i].planes[VPX_PLANE_V], raw[i].stride[VPX_PLANE_V],
+ raw[i].d_w, raw[i].d_h, 1);
}
}
--- a/vpx/src/vpx_encoder.c
+++ b/vpx/src/vpx_encoder.c
@@ -243,6 +243,7 @@
ctx--;
if (img) img--;
}
+ ctx++;
}
FLOATING_POINT_RESTORE();