Merge remote-tracking branch 'upstream/master'

Makefile

@@ -29,7 +29,7 @@ LIBDIR_NAME=lib
 SHAREDLIB_DIR=$(PREFIX)/lib
 PROJECT_NAME=openh264
 MODULE_NAME=gmpopenh264
-GMP_API_BRANCH=Firefox114_2
+GMP_API_BRANCH=Firefox135
 GTEST_VER=release-1.8.1
 CCASFLAGS=$(CFLAGS)
 STATIC_LDFLAGS=-lstdc++
@@ -38,8 +38,8 @@ USE_STACK_PROTECTOR = Yes
 USE_LOW_VERSION_NDK=No
 USE_ANT=No
 
-SHAREDLIB_MAJORVERSION=7
-FULL_VERSION := 2.5.0
+SHAREDLIB_MAJORVERSION=8
+FULL_VERSION := 2.6.0
 
 ifeq (,$(wildcard $(SRC_PATH)gmp-api))
 HAVE_GMP_API=No

RELEASES

@@ -1,6 +1,33 @@
 
 Releases
 -----------
+v2.6.0
+------
+- Initialize PSNR to 0 instead of NAN (Commit: cf3b514b, PR: #3836)
+- Enable PIC in x86 assembly code (Commit: c42fa040, PR: #3831)
+- Add QEMU test for LoongArch architecture (Commit: e49c0953, PR: #3833)
+- Add support for calculating PSNR for Y/U/V components (Commit: 33f7f486, PR: #3824)
+- Add support for PowerPC architecture in meson.build (Commit: 423eb2c3, PR: #3806)
+- Fix potential bug in the codebase (Commit: 63db555e, PR: #3818)
+- Pass the actual temporal ID to GMPVideoEncodedFrame (Commit: 1c47e719, PR: #3828)
+- Use HW_NCPUONLINE on OpenBSD / NetBSD for CPU detection (Commit: fc1605c1, PR: #3813)
+- Add Windows VSBuildTools VC17 compile support (Commit: 1ebb3f1a, PR: #3823)
+- Expose temporal layer info via GMP API at encoder initialization and on a per-frame basis (Commit: ad3f6904, PR: #3821)
+- Remove the use of -Werror to prevent warnings from being treated as errors (Commit: edbef4df, PR: #3809)
+- Use void casts to silence warnings about memcpy to a class (Commit: 6746bc48, PR: #3800)
+
+v2.5.0
+------
+- Fix some multi-thread decoding issues
+- Remove unused pSpatialLayer variable in DetermineTemporalSettings
+- Change picOrderCnt to 2
+- Fix cross compile from darwin arm64 to x86_64
+- Fix deadlock at the end of decoding
+- Fix frame-decode errors
+- Fix build with meson on riscv64
+- Enable pac bti support
+- Fix a bug in decoding H.264 streams encoded by Apple HWA encoder
+
 v2.4.1
 ------
 - Fix off by one regression in decoder

build/AutoBuildForWindows.bat

@@ -34,6 +34,7 @@ rem         --arm64 folder bin\ARM64
 rem
 rem      [winsdk_version] : full Windows 10 SDK number (e.g. 10.0.17763.0)
 rem      [vc_version] : Specify a VC++ version
+rem                     VC17 for VC++ 2022
 rem                     VC16 for VC++ 2019
 rem                     VC15 for VC++ 2017
 rem
@@ -97,6 +98,15 @@ goto :EOF
   set MinGWPath=C:\MinGW\bin
   set MsysPath=C:\MinGW\msys\1.0\bin
 
+  set VC17BuildToolsPath=C:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC
+  set VC17CommunityPath=C:\Program Files\Microsoft Visual Studio\2022\Community\VC
+  set VC17ProfessionalPath=C:\Program Files\Microsoft Visual Studio\2022\Professional\VC
+  set VC17EnterprisePath=C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC
+  if exist "%VC17BuildToolsPath%" set VC17PATH=%VC17BuildToolsPath%
+  if exist "%VC17CommunityPath%" set VC17PATH=%VC17CommunityPath%
+  if exist "%VC17ProfessionalPath%" set VC17PATH=%VC17ProfessionalPath%
+  if exist "%VC17EnterprisePath%" set VC17PATH=%VC17EnterprisePath%
+
   set VC16BuildToolsPath=C:\Program Files (x86)\Microsoft Visual Studio\2019\BuildTools\VC
   set VC16CommunityPath=C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC
   set VC16ProfessionalPath=C:\Program Files (x86)\Microsoft Visual Studio\2019\Professional\VC
@@ -130,14 +140,21 @@ goto :EOF
   set VCVARSPATH=%VCPATH%
   if exist "%VC15Path%"    set VCVARSPATH=%VC15Path%\Auxiliary\Build
   if exist "%VC16Path%"    set VCVARSPATH=%VC16Path%\Auxiliary\Build
+  if exist "%VC17Path%"    set VCVARSPATH=%VC17Path%\Auxiliary\Build
 
-  if /I "%OPENH264_VC_VERSION%" == "VC15" (
+  if /I "%OPENH264_VC_VERSION%" == "VC17" (
     set VCPATH=
-    set "VCVARSPATH=%VC15Path%\Auxiliary\Build"
-) else if /I "%OPENH264_VC_VERSION%" == "VC12" (
-    set "VCPATH=%VC12Path%"
-	set "VCVARSPATH=%VCPATH%"
-)
+    set "VCVARSPATH=%VC17Path%\Auxiliary\Build"
+  ) else if /I "%OPENH264_VC_VERSION%" == "VC16" (
+      set VCPATH=
+      set "VCVARSPATH=%VC16Path%\Auxiliary\Build"
+  ) else if /I "%OPENH264_VC_VERSION%" == "VC15" (
+      set VCPATH=
+      set "VCVARSPATH=%VC15Path%\Auxiliary\Build"
+  ) else if /I "%OPENH264_VC_VERSION%" == "VC12" (
+      set "VCPATH=%VC12Path%"
+      set "VCVARSPATH=%VCPATH%"
+  )
 
   set GasScriptPath=%VCPATH%\bin
 
@@ -392,6 +409,7 @@ rem ***********************************************
   echo           Win32-Release-ASM
   echo      [winsdk_version] : full Windows 10 SDK number (e.g. 10.0.17763.0)
   echo      [vc_version] : Specify a VC++ version
+  echo                     VC17 for VC++ 2022
   echo                     VC16 for VC++ 2019
   echo                     VC15 for VC++ 2017
 

build/platform-darwin.mk

@@ -3,8 +3,8 @@ SHAREDLIB_DIR = $(PREFIX)/lib
 SHAREDLIBSUFFIX = dylib
 SHAREDLIBSUFFIXFULLVER=$(FULL_VERSION).$(SHAREDLIBSUFFIX)
 SHAREDLIBSUFFIXMAJORVER=$(SHAREDLIB_MAJORVERSION).$(SHAREDLIBSUFFIX)
-CURRENT_VERSION := 2.5.0
-COMPATIBILITY_VERSION := 2.5.0
+CURRENT_VERSION := 2.6.0
+COMPATIBILITY_VERSION := 2.6.0
 SHLDFLAGS = -dynamiclib -twolevel_namespace -undefined dynamic_lookup \
 	-fno-common -headerpad_max_install_names -install_name \
 	$(SHAREDLIB_DIR)/$(LIBPREFIX)$(PROJECT_NAME).$(SHAREDLIBSUFFIXMAJORVER)

build/platform-gnu-chain.mk

@@ -31,12 +31,3 @@ endif
 ifneq ($(filter %clang++,$(CXX)),)
 CXXFLAGS += -Wc++11-compat-reserved-user-defined-literal
 endif
-
-ifneq ($(filter %g++,$(CXX)),)
-ifeq ($(filter %clang++,$(CXX)),)
-GCCVER_GTEQ8 = $(shell echo $$(($$($(CXX) -dumpversion | awk -F "." '{print $$1}') >= 8)))
-ifeq ($(GCCVER_GTEQ8), 1)
-CXXFLAGS += -Wno-class-memaccess
-endif
-endif
-endif

codec/api/wels/codec_app_def.h

@@ -592,6 +592,9 @@ typedef struct TagEncParamExt {
   bool    bIsLosslessLink;             ///< LTR advanced setting
   bool    bFixRCOverShoot;             ///< fix rate control overshooting
   int     iIdrBitrateRatio;            ///< the target bits of IDR is (idr_bitrate_ratio/100) * average target bit per frame.
+  bool    bPsnrY;                      ///< get Y PSNR stats for the whole video sequence
+  bool    bPsnrU;                      ///< get U PSNR stats for the whole video sequence
+  bool    bPsnrV;                      ///< get V PSNR stats for the whole video sequence
 } SEncParamExt;
 
 /**
@@ -635,6 +638,7 @@ typedef struct {
   int   iNalCount;              ///< count number of NAL coded already
   int*  pNalLengthInByte;       ///< length of NAL size in byte from 0 to iNalCount-1
   unsigned char*  pBsBuf;       ///< buffer of bitstream contained
+  float rPsnr[3];               ///< PSNR values for Y/U/V
 } SLayerBSInfo, *PLayerBSInfo;
 
 /**
@@ -659,7 +663,11 @@ typedef struct Source_Picture_s {
   int       iPicWidth;             ///< luma picture width in x coordinate
   int       iPicHeight;            ///< luma picture height in y coordinate
   long long uiTimeStamp;           ///< timestamp of the source picture, unit: millisecond
+  bool      bPsnrY;                ///< get Y PSNR for this frame
+  bool      bPsnrU;                ///< get U PSNR for this frame
+  bool      bPsnrV;                ///< get V PSNR for this frame
 } SSourcePicture;
+
 /**
 * @brief Structure for bit rate info
 */

codec/api/wels/codec_ver.h

@@ -4,12 +4,12 @@
 
 #include "codec_app_def.h"
 
-static const OpenH264Version g_stCodecVersion  = {2, 5, 0, 2411};
-static const char* const g_strCodecVer  = "OpenH264 version:2.5.0.2411";
+static const OpenH264Version g_stCodecVersion  = {2, 6, 0, 2502};
+static const char* const g_strCodecVer  = "OpenH264 version:2.6.0.2502";
 
 #define OPENH264_MAJOR (2)
-#define OPENH264_MINOR (5)
+#define OPENH264_MINOR (6)
 #define OPENH264_REVISION (0)
-#define OPENH264_RESERVED (2411)
+#define OPENH264_RESERVED (2502)
 
 #endif  // CODEC_VER_H

codec/common/loongarch/mc_horver_lsx.c

@@ -483,7 +483,7 @@ void McHorVer22WidthEq4_lsx(const uint8_t *pSrc, int32_t iSrcStride, uint8_t *pD
   uint8_t *psrc = (uint8_t*)pSrc - 2;
   __m128i src0, src1, src2, src3, src4, src5;
   __m128i tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, out0_l, out0_h;
-  __m128i dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7;
+  __m128i dst0, dst1, dst2, dst3, dst4={0}, dst5={0}, dst6={0}, dst7={0};
   v8i16 mask1 = {3, 4, 5, 6, 7, 8, 9, 10};
   for (int i = 0; i < iHeight; i++) {
     DUP4_ARG2(__lsx_vldx,
@@ -565,7 +565,7 @@ void McHorVer22WidthEq5_lsx(const uint8_t *pSrc, int32_t iSrcStride, uint8_t *pD
   uint8_t *psrc = (uint8_t*)pSrc - 2;
   __m128i src0, src1, src2, src3, src4, src5;
   __m128i tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, out0_l, out0_h;
-  __m128i dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7;
+  __m128i dst0, dst1, dst2, dst3, dst4={0}, dst5={0}, dst6={0}, dst7={0};
   v8i16 mask1 = {3, 4, 5, 6, 7, 8, 9, 10};
   for (int i = 0; i < iHeight; i++) {
     DUP4_ARG2(__lsx_vldx,

codec/common/meson.build

@@ -71,6 +71,10 @@ elif cpu_family in ['loongarch32', 'loongarch64']
   cpp_sources += asm_sources
 elif cpu_family == 'riscv64'
   asm_sources = []
+elif cpu_family == 'ppc'
+  asm_sources = []
+elif cpu_family == 'ppc64'
+  asm_sources = []
 else
   error('Unsupported cpu_family @0@'.format(cpu_family))
 endif

codec/common/src/WelsThreadLib.cpp

@@ -55,6 +55,8 @@
 #endif
 #ifdef __APPLE__
 #define HW_NCPU_NAME "hw.logicalcpu"
+#elif defined(HW_NCPUONLINE)
+#define HW_NCPU_NAME "hw.ncpuonline"
 #else
 #define HW_NCPU_NAME "hw.ncpu"
 #endif
@@ -522,7 +524,7 @@ WELS_THREAD_ERROR_CODE    WelsQueryLogicalProcessInfo (WelsLogicalProcessInfo* p
   size_t len = sizeof (pInfo->ProcessorCount);
 
 #if defined(__OpenBSD__)
-  int scname[] = { CTL_HW, HW_NCPU };
+  int scname[] = { CTL_HW, HW_NCPUONLINE };
   if (sysctl (scname, 2, &pInfo->ProcessorCount, &len, NULL, 0) == -1)
 #else
   if (sysctlbyname (HW_NCPU_NAME, &pInfo->ProcessorCount, &len, NULL, 0) == -1)

codec/decoder/core/src/decoder.cpp

@@ -846,6 +846,10 @@ int32_t WelsDecodeBs (PWelsDecoderContext pCtx, const uint8_t* kpBsBuf, const in
             return pCtx->iErrorCode;
           }
 
+          if (pCtx->iErrorCode != ERR_NONE && !(pCtx->iErrorCode & dsDataErrorConcealed)) {
+              return pCtx->iErrorCode;
+          }
+
           pDstNal += (iDstIdx + 4); //init, increase 4 reserved zero bytes, used to store the next NAL
           if ((iSrcLength - iSrcConsumed + 4) > (pRawData->pEnd - pDstNal)) {
             pDstNal = pRawData->pCurPos = pRawData->pHead;

codec/decoder/meson.build

@@ -56,6 +56,10 @@ elif cpu_family in ['loongarch32', 'loongarch64']
   cpp_sources += asm_sources
 elif cpu_family == 'riscv64'
   asm_sources = []
+elif cpu_family == 'ppc'
+  asm_sources = []
+elif cpu_family == 'ppc64'
+  asm_sources = []
 else
   error('Unsupported cpu family @0@'.format(cpu_family))
 endif

codec/encoder/core/inc/as264_common.h

@@ -74,7 +74,6 @@
 #endif//ENABLE_FRAME_DUMP
 #endif//__UNITTEST__
 
-//#define ENABLE_PSNR_CALC
 //#define STAT_OUTPUT
 //#define MB_TYPES_CHECK
 //
@@ -88,10 +87,6 @@
 //@if !FRAME_INFO_OUTPUT
 #if !defined(FRAME_INFO_OUTPUT)
 
-#if defined(ENABLE_PSNR_CALC)
-#undef ENABLE_PSNR_CALC
-#endif//ENABLE_PSNR_CALC
-
 //#if defined(STAT_OUTPUT)
 //#undef STAT_OUTPUT
 //#endif//STAT_OUTPUT

codec/encoder/core/inc/param_svc.h

@@ -180,6 +180,9 @@ typedef struct TagWelsSvcCodingParam: SEncParamExt {
     param.bIsLosslessLink = false;
     param.bFixRCOverShoot = true;
     param.iIdrBitrateRatio = IDR_BITRATE_RATIO * 100;
+    param.bPsnrY = false;
+    param.bPsnrU = false;
+    param.bPsnrV = false;
     for (int32_t iLayer = 0; iLayer < MAX_SPATIAL_LAYER_NUM; iLayer++) {
       param.sSpatialLayers[iLayer].uiProfileIdc = PRO_UNKNOWN;
       param.sSpatialLayers[iLayer].uiLevelIdc = LEVEL_UNKNOWN;
@@ -345,6 +348,9 @@ typedef struct TagWelsSvcCodingParam: SEncParamExt {
     bIsLosslessLink = pCodingParam.bIsLosslessLink;
     bFixRCOverShoot = pCodingParam.bFixRCOverShoot;
     iIdrBitrateRatio = pCodingParam.iIdrBitrateRatio;
+    bPsnrY = pCodingParam.bPsnrY;
+    bPsnrU = pCodingParam.bPsnrU;
+    bPsnrV = pCodingParam.bPsnrV;
     if (iUsageType == SCREEN_CONTENT_REAL_TIME && !bIsLosslessLink && bEnableLongTermReference) {
       bEnableLongTermReference = false;
     }