Modify code style, remove trailing space.
This commit is contained in:
Licai Guo
parent
b7a25df13f
commit
7768cd0a98
@ -71,7 +71,7 @@
|
||||
4CE4474718BC61650017DF25 /* typedefs.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = typedefs.h; sourceTree = "<group>"; };
|
||||
4CE4474918BC61650017DF25 /* WelsThreadLib.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = WelsThreadLib.cpp; sourceTree = "<group>"; };
|
||||
4CE4474A18BC61650017DF25 /* WelsThreadLib.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = WelsThreadLib.h; sourceTree = "<group>"; };
|
||||
4CE447BC18C085320017DF25 /* deblocking_neon.S */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.asm; path = deblocking_neon.S; sourceTree = "<group>"; };
|
||||
4CE447BC18C085320017DF25 /* deblocking_neon.S */ = {isa = PBXFileReference; fileEncoding = 4; indentWidth = 4; lastKnownFileType = sourcecode.asm; path = deblocking_neon.S; sourceTree = "<group>"; tabWidth = 4; usesTabs = 0; };
|
||||
4CE447BE18C085900017DF25 /* arm_arch_common_macro.S */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.asm; path = arm_arch_common_macro.S; sourceTree = "<group>"; };
|
||||
/* End PBXFileReference section */
|
||||
|
||||
|
||||
@ -84,9 +84,9 @@
|
||||
4CE4464E18BC5EAA0017DF25 /* decoder_context.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = decoder_context.h; sourceTree = "<group>"; };
|
||||
4CE4464F18BC5EAA0017DF25 /* decoder_core.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = decoder_core.h; sourceTree = "<group>"; };
|
||||
4CE4465018BC5EAA0017DF25 /* error_code.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = error_code.h; sourceTree = "<group>"; };
|
||||
4CE4465118BC5EAA0017DF25 /* expand_pic.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = expand_pic.h; sourceTree = "<group>"; };
|
||||
4CE4465218BC5EAA0017DF25 /* fmo.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = fmo.h; sourceTree = "<group>"; };
|
||||
4CE4465318BC5EAA0017DF25 /* get_intra_predictor.h */ = {isa = PBXFileReference; fileEncoding = 4; indentWidth = 2; lastKnownFileType = sourcecode.c.h; path = get_intra_predictor.h; sourceTree = "<group>"; tabWidth = 2; usesTabs = 1; };
|
||||
4CE4465118BC5EAA0017DF25 /* expand_pic.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = expand_pic.h; sourceTree = "<group>"; usesTabs = 1; };
|
||||
4CE4465218BC5EAA0017DF25 /* fmo.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = fmo.h; sourceTree = "<group>"; usesTabs = 1; };
|
||||
4CE4465318BC5EAA0017DF25 /* get_intra_predictor.h */ = {isa = PBXFileReference; fileEncoding = 4; indentWidth = 4; lastKnownFileType = sourcecode.c.h; path = get_intra_predictor.h; sourceTree = "<group>"; tabWidth = 4; usesTabs = 0; };
|
||||
4CE4465418BC5EAA0017DF25 /* manage_dec_ref.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = manage_dec_ref.h; sourceTree = "<group>"; };
|
||||
4CE4465518BC5EAA0017DF25 /* mb_cache.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = mb_cache.h; sourceTree = "<group>"; };
|
||||
4CE4465618BC5EAA0017DF25 /* mc.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = mc.h; sourceTree = "<group>"; };
|
||||
@ -105,19 +105,19 @@
|
||||
4CE4466318BC5EAA0017DF25 /* vlc_decoder.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = vlc_decoder.h; sourceTree = "<group>"; };
|
||||
4CE4466418BC5EAA0017DF25 /* wels_common_basis.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = wels_common_basis.h; sourceTree = "<group>"; };
|
||||
4CE4466518BC5EAA0017DF25 /* wels_const.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = wels_const.h; sourceTree = "<group>"; };
|
||||
4CE4466718BC5EAA0017DF25 /* au_parser.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = au_parser.cpp; sourceTree = "<group>"; usesTabs = 1; };
|
||||
4CE4466818BC5EAA0017DF25 /* bit_stream.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = bit_stream.cpp; sourceTree = "<group>"; usesTabs = 1; };
|
||||
4CE4466918BC5EAA0017DF25 /* deblocking.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = deblocking.cpp; sourceTree = "<group>"; };
|
||||
4CE4466718BC5EAA0017DF25 /* au_parser.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = au_parser.cpp; sourceTree = "<group>"; usesTabs = 0; };
|
||||
4CE4466818BC5EAA0017DF25 /* bit_stream.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = bit_stream.cpp; sourceTree = "<group>"; usesTabs = 0; };
|
||||
4CE4466918BC5EAA0017DF25 /* deblocking.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = deblocking.cpp; sourceTree = "<group>"; tabWidth = 2; };
|
||||
4CE4466A18BC5EAA0017DF25 /* decode_mb_aux.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = decode_mb_aux.cpp; sourceTree = "<group>"; };
|
||||
4CE4466B18BC5EAA0017DF25 /* decode_slice.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = decode_slice.cpp; sourceTree = "<group>"; usesTabs = 1; };
|
||||
4CE4466C18BC5EAA0017DF25 /* decoder.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = decoder.cpp; sourceTree = "<group>"; tabWidth = 2; usesTabs = 1; };
|
||||
4CE4466B18BC5EAA0017DF25 /* decode_slice.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = decode_slice.cpp; sourceTree = "<group>"; usesTabs = 0; };
|
||||
4CE4466C18BC5EAA0017DF25 /* decoder.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = decoder.cpp; sourceTree = "<group>"; tabWidth = 4; usesTabs = 0; };
|
||||
4CE4466D18BC5EAA0017DF25 /* decoder_core.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = decoder_core.cpp; sourceTree = "<group>"; };
|
||||
4CE4466E18BC5EAA0017DF25 /* decoder_data_tables.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = decoder_data_tables.cpp; sourceTree = "<group>"; };
|
||||
4CE4466F18BC5EAA0017DF25 /* expand_pic.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = expand_pic.cpp; sourceTree = "<group>"; };
|
||||
4CE4467018BC5EAA0017DF25 /* fmo.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = fmo.cpp; sourceTree = "<group>"; };
|
||||
4CE4467118BC5EAA0017DF25 /* get_intra_predictor.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = get_intra_predictor.cpp; sourceTree = "<group>"; };
|
||||
4CE4467218BC5EAA0017DF25 /* manage_dec_ref.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = manage_dec_ref.cpp; sourceTree = "<group>"; };
|
||||
4CE4467318BC5EAA0017DF25 /* mc.cpp */ = {isa = PBXFileReference; fileEncoding = 4; indentWidth = 4; lastKnownFileType = sourcecode.cpp.cpp; path = mc.cpp; sourceTree = "<group>"; tabWidth = 1; usesTabs = 1; wrapsLines = 1; };
|
||||
4CE4467318BC5EAA0017DF25 /* mc.cpp */ = {isa = PBXFileReference; fileEncoding = 4; indentWidth = 4; lastKnownFileType = sourcecode.cpp.cpp; path = mc.cpp; sourceTree = "<group>"; tabWidth = 1; usesTabs = 0; wrapsLines = 1; };
|
||||
4CE4467418BC5EAA0017DF25 /* mem_align.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = mem_align.cpp; sourceTree = "<group>"; };
|
||||
4CE4467518BC5EAA0017DF25 /* memmgr_nal_unit.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = memmgr_nal_unit.cpp; sourceTree = "<group>"; };
|
||||
4CE4467618BC5EAA0017DF25 /* mv_pred.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = mv_pred.cpp; sourceTree = "<group>"; };
|
||||
|
||||
@ -218,25 +218,23 @@ void WelsXmmRegEmptyOp(void * pSrc) {
|
||||
#if defined(ANDROID_NDK)
|
||||
uint32_t WelsCPUFeatureDetectAndroid()
|
||||
{
|
||||
uint32_t uiCPU = 0;
|
||||
AndroidCpuFamily cpuFamily = ANDROID_CPU_FAMILY_UNKNOWN;
|
||||
uint64_t uiFeatures = 0;
|
||||
|
||||
cpuFamily = android_getCpuFamily();
|
||||
if (cpuFamily == ANDROID_CPU_FAMILY_ARM)
|
||||
{
|
||||
uiFeatures = android_getCpuFeatures();
|
||||
if (uiFeatures & ANDROID_CPU_ARM_FEATURE_ARMv7){
|
||||
uiCPU |= WELS_CPU_ARMv7;
|
||||
}
|
||||
if (uiFeatures & ANDROID_CPU_ARM_FEATURE_VFPv3){
|
||||
uiCPU |= WELS_CPU_VFPv3;
|
||||
}
|
||||
if (uiFeatures & ANDROID_CPU_ARM_FEATURE_NEON){
|
||||
uiCPU |= WELS_CPU_NEON;
|
||||
}
|
||||
}
|
||||
return uiCPU;
|
||||
uint32_t uiCPU = 0;
|
||||
AndroidCpuFamily cpuFamily = ANDROID_CPU_FAMILY_UNKNOWN;
|
||||
uint64_t uiFeatures = 0;
|
||||
cpuFamily = android_getCpuFamily();
|
||||
if (cpuFamily == ANDROID_CPU_FAMILY_ARM) {
|
||||
uiFeatures = android_getCpuFeatures();
|
||||
if (uiFeatures & ANDROID_CPU_ARM_FEATURE_ARMv7){
|
||||
uiCPU |= WELS_CPU_ARMv7;
|
||||
}
|
||||
if (uiFeatures & ANDROID_CPU_ARM_FEATURE_VFPv3){
|
||||
uiCPU |= WELS_CPU_VFPv3;
|
||||
}
|
||||
if (uiFeatures & ANDROID_CPU_ARM_FEATURE_NEON){
|
||||
uiCPU |= WELS_CPU_NEON;
|
||||
}
|
||||
}
|
||||
return uiCPU;
|
||||
}
|
||||
|
||||
#endif
|
||||
@ -246,9 +244,8 @@ uint32_t WelsCPUFeatureDetectIOS() //Need to be updated for the new device of AP
|
||||
{
|
||||
uint32_t uiCPU = 0;
|
||||
struct utsname sSystemInfo;
|
||||
|
||||
uname (&sSystemInfo);
|
||||
|
||||
|
||||
if ((0 != strcmp(sSystemInfo.machine, "iPhone1,1")) && //iPhone 2G
|
||||
(0 != strcmp(sSystemInfo.machine, "iPhone1,2")) && //iPhone 3G
|
||||
(0 != strcmp(sSystemInfo.machine, "iPod1,1")) && //iPod 1G
|
||||
|
||||
@ -82,12 +82,12 @@ void WelsXmmRegEmptyOp(void * pSrc);
|
||||
#if defined(ANDROID_NDK)
|
||||
uint32_t WelsCPUFeatureDetectAndroid();
|
||||
#endif
|
||||
|
||||
|
||||
#if defined(APPLE_IOS)
|
||||
uint32_t WelsCPUFeatureDetectIOS();
|
||||
#endif
|
||||
#endif
|
||||
|
||||
|
||||
#if defined(__cplusplus)
|
||||
}
|
||||
#endif//__cplusplus
|
||||
|
||||
@ -37,13 +37,13 @@ void DeblockChromaLt4H_ssse3 (uint8_t* pPixCb, uint8_t* pPixCr, int32_t iStride,
|
||||
#if defined(HAVE_NEON)
|
||||
void DeblockLumaLt4V_neon(uint8_t* pPixY, int32_t iStride, int32_t iAlpha, int32_t iBeta, int8_t* pTc);
|
||||
void DeblockLumaEq4V_neon(uint8_t* pPixY, int32_t iStride, int32_t iAlpha, int32_t iBeta);
|
||||
|
||||
|
||||
void DeblockLumaLt4H_neon(uint8_t* pPixY, int32_t iStride, int32_t iAlpha, int32_t iBeta, int8_t* pTc);
|
||||
void DeblockLumaEq4H_neon(uint8_t* pPixY, int32_t iStride, int32_t iAlpha, int32_t iBeta);
|
||||
|
||||
|
||||
void DeblockChromaLt4V_neon(uint8_t* pPixCb, uint8_t* pPixCr, int32_t iStride, int32_t iAlpha, int32_t iBeta, int8_t* pTC);
|
||||
void DeblockChromaEq4V_neon(uint8_t* pPixCb, uint8_t* pPixCr, int32_t iStride, int32_t iAlpha, int32_t iBeta);
|
||||
|
||||
|
||||
void DeblockChromaLt4H_neon(uint8_t* pPixCb, uint8_t* pPixCr, int32_t iStride, int32_t iAlpha, int32_t iBeta, int8_t* pTC);
|
||||
void DeblockChromaEq4H_neon(uint8_t* pPixCb, uint8_t* pPixCr, int32_t iStride, int32_t iAlpha, int32_t iBeta);
|
||||
#endif
|
||||
|
||||
File diff suppressed because it is too large
Load Diff
@ -40,72 +40,44 @@ extern "C" {
|
||||
#endif//__cplusplus
|
||||
|
||||
#if defined(HAVE_NEON)
|
||||
/*
|
||||
void McCopy_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride,
|
||||
int32_t iWidth, int32_t iHeight);
|
||||
void McHorVer20_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride,
|
||||
int32_t iWidth, int32_t iHeight);
|
||||
void McHorVer02_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride,
|
||||
int32_t iWidth, int32_t iHeight);
|
||||
void McHorVer22_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride,
|
||||
int32_t iWidth, int32_t iHeight);
|
||||
|
||||
void McHorVer01_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iWidth, int32_t iHeight);
|
||||
void McHorVer03_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iWidth, int32_t iHeight);
|
||||
void McHorVer10_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iWidth, int32_t iHeight);
|
||||
void McHorVer11_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iWidth, int32_t iHeight);
|
||||
void McHorVer12_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iWidth, int32_t iHeight);
|
||||
void McHorVer13_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iWidth, int32_t iHeight);
|
||||
void McHorVer21_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iWidth, int32_t iHeight);
|
||||
void McHorVer23_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iWidth, int32_t iHeight);
|
||||
void McHorVer30_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iWidth, int32_t iHeight);
|
||||
void McHorVer31_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iWidth, int32_t iHeight);
|
||||
void McHorVer32_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iWidth, int32_t iHeight);
|
||||
void McHorVer33_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iWidth, int32_t iHeight);
|
||||
|
||||
void McLuma_neon(uint8_t* pSrc, int32_t src_stride, uint8_t* dst, int32_t dst_stride,
|
||||
int16_t iMvX, int16_t iMvY, int32_t width, int32_t height);
|
||||
void McChroma_neon(uint8_t* pSrc, int32_t src_stride, uint8_t* dst, int32_t dst_stride,
|
||||
int16_t iMvX, int16_t iMvY, int32_t width, int32_t height);
|
||||
*/
|
||||
void McCopyWidthEq4_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
|
||||
|
||||
void McCopyWidthEq8_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
|
||||
|
||||
void McCopyWidthEq16_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
|
||||
|
||||
void McChromaWidthEq8_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t* pWeights, int32_t iHeight);
|
||||
|
||||
|
||||
void McChromaWidthEq4_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t* pWeights, int32_t iHeight);
|
||||
|
||||
|
||||
void PixelAvgWidthEq16_neon(uint8_t* pDst, int32_t iDstStride, uint8_t* pSrcA, uint8_t* pSrcB, int32_t iHeight);
|
||||
void PixelAvgWidthEq8_neon(uint8_t* pDst, int32_t iDstStride, uint8_t* pSrcA, uint8_t* pSrcB, int32_t iHeight);
|
||||
void PixelAvgWidthEq4_neon(uint8_t* pDst, int32_t iDstStride, uint8_t* pSrcA, uint8_t* pSrcB, int32_t iHeight);
|
||||
|
||||
|
||||
void McHorVer01WidthEq16_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
void McHorVer01WidthEq8_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
void McHorVer01WidthEq4_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
void McHorVer03WidthEq16_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
void McHorVer03WidthEq8_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
void McHorVer03WidthEq4_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
|
||||
|
||||
void McHorVer10WidthEq16_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
void McHorVer10WidthEq8_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
void McHorVer10WidthEq4_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
void McHorVer30WidthEq16_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
void McHorVer30WidthEq8_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
void McHorVer30WidthEq4_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
|
||||
|
||||
//horizontal filter to gain half sample, that is (2, 0) location in quarter sample
|
||||
void McHorVer20WidthEq16_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
void McHorVer20WidthEq8_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
void McHorVer20WidthEq4_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
|
||||
|
||||
//vertical filter to gain half sample, that is (0, 2) location in quarter sample
|
||||
void McHorVer02WidthEq16_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
void McHorVer02WidthEq8_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
void McHorVer02WidthEq4_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
|
||||
|
||||
//horizontal and vertical filter to gain half sample, that is (2, 2) location in quarter sample
|
||||
void McHorVer22WidthEq16_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
void McHorVer22WidthEq8_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride, int32_t iHeight);
|
||||
|
||||
@ -29,7 +29,7 @@
|
||||
* POSSIBILITY OF SUCH DAMAGE.
|
||||
*
|
||||
*/
|
||||
|
||||
|
||||
#ifdef HAVE_NEON
|
||||
.text
|
||||
#include "arm_arch_common_macro.S"
|
||||
@ -37,32 +37,32 @@
|
||||
|
||||
.macro ROW_TRANSFORM_1_STEP
|
||||
// { // input: src_d[0]~[3], output: e_q[0]~[3]; working: $8 $9
|
||||
vaddl.s16 $4, $0, $2 //int32 e[i][0] = src[0] + src[2];
|
||||
vsubl.s16 $5, $0, $2 //int32 e[i][1] = src[0] - src[2];
|
||||
vshr.s16 $8, $1, #1
|
||||
vshr.s16 $9, $3, #1
|
||||
vsubl.s16 $6, $8, $3 //int32 e[i][2] = (src[1]>>1)-src[3];
|
||||
vaddl.s16 $7, $1, $9 //int32 e[i][3] = src[1] + (src[3]>>1);
|
||||
vaddl.s16 $4, $0, $2 //int32 e[i][0] = src[0] + src[2];
|
||||
vsubl.s16 $5, $0, $2 //int32 e[i][1] = src[0] - src[2];
|
||||
vshr.s16 $8, $1, #1
|
||||
vshr.s16 $9, $3, #1
|
||||
vsubl.s16 $6, $8, $3 //int32 e[i][2] = (src[1]>>1)-src[3];
|
||||
vaddl.s16 $7, $1, $9 //int32 e[i][3] = src[1] + (src[3]>>1);
|
||||
// }
|
||||
.endm
|
||||
|
||||
.macro TRANSFORM_4BYTES // both row & col transform used
|
||||
// { // output: f_q[0]~[3], input: e_q[0]~[3];
|
||||
vadd.s32 $0, $4, $7 //int16 f[i][0] = e[i][0] + e[i][3];
|
||||
vadd.s32 $1, $5, $6 //int16 f[i][1] = e[i][1] + e[i][2];
|
||||
vsub.s32 $2, $5, $6 //int16 f[i][2] = e[i][1] - e[i][2];
|
||||
vsub.s32 $3, $4, $7 //int16 f[i][3] = e[i][0] - e[i][3];
|
||||
vadd.s32 $0, $4, $7 //int16 f[i][0] = e[i][0] + e[i][3];
|
||||
vadd.s32 $1, $5, $6 //int16 f[i][1] = e[i][1] + e[i][2];
|
||||
vsub.s32 $2, $5, $6 //int16 f[i][2] = e[i][1] - e[i][2];
|
||||
vsub.s32 $3, $4, $7 //int16 f[i][3] = e[i][0] - e[i][3];
|
||||
// }
|
||||
.endm
|
||||
|
||||
.macro COL_TRANSFORM_1_STEP
|
||||
// { // input: src_q[0]~[3], output: e_q[0]~[3];
|
||||
vadd.s32 $4, $0, $2 //int32 e[0][j] = f[0][j] + f[2][j];
|
||||
vsub.s32 $5, $0, $2 //int32 e[1][j] = f[0][j] - f[2][j];
|
||||
vshr.s32 $6, $1, #1
|
||||
vshr.s32 $7, $3, #1
|
||||
vsub.s32 $6, $6, $3 //int32 e[2][j] = (f[1][j]>>1) - f[3][j];
|
||||
vadd.s32 $7, $1, $7 //int32 e[3][j] = f[1][j] + (f[3][j]>>1);
|
||||
vadd.s32 $4, $0, $2 //int32 e[0][j] = f[0][j] + f[2][j];
|
||||
vsub.s32 $5, $0, $2 //int32 e[1][j] = f[0][j] - f[2][j];
|
||||
vshr.s32 $6, $1, #1
|
||||
vshr.s32 $7, $3, #1
|
||||
vsub.s32 $6, $6, $3 //int32 e[2][j] = (f[1][j]>>1) - f[3][j];
|
||||
vadd.s32 $7, $1, $7 //int32 e[3][j] = f[1][j] + (f[3][j]>>1);
|
||||
// }
|
||||
.endm
|
||||
|
||||
@ -70,101 +70,101 @@
|
||||
|
||||
.macro ROW_TRANSFORM_1_STEP arg0, arg1, arg2, arg3, arg4, arg5, arg6, arg7, arg8, arg9
|
||||
// { // input: src_d[0]~[3], output: e_q[0]~[3]; working: $8 $9
|
||||
vaddl.s16 \arg4, \arg0, \arg2 //int32 e[i][0] = src[0] + src[2];
|
||||
vsubl.s16 \arg5, \arg0, \arg2 //int32 e[i][1] = src[0] - src[2];
|
||||
vshr.s16 \arg8, \arg1, #1
|
||||
vshr.s16 \arg9, \arg3, #1
|
||||
vsubl.s16 \arg6, \arg8, \arg3 //int32 e[i][2] = (src[1]>>1)-src[3];
|
||||
vaddl.s16 \arg7, \arg1, \arg9 //int32 e[i][3] = src[1] + (src[3]>>1);
|
||||
vaddl.s16 \arg4, \arg0, \arg2 //int32 e[i][0] = src[0] + src[2];
|
||||
vsubl.s16 \arg5, \arg0, \arg2 //int32 e[i][1] = src[0] - src[2];
|
||||
vshr.s16 \arg8, \arg1, #1
|
||||
vshr.s16 \arg9, \arg3, #1
|
||||
vsubl.s16 \arg6, \arg8, \arg3 //int32 e[i][2] = (src[1]>>1)-src[3];
|
||||
vaddl.s16 \arg7, \arg1, \arg9 //int32 e[i][3] = src[1] + (src[3]>>1);
|
||||
// }
|
||||
.endm
|
||||
|
||||
.macro TRANSFORM_4BYTES arg0, arg1, arg2, arg3, arg4, arg5, arg6, arg7 // both row & col transform used
|
||||
// { // output: f_q[0]~[3], input: e_q[0]~[3];
|
||||
vadd.s32 \arg0, \arg4, \arg7 //int16 f[i][0] = e[i][0] + e[i][3];
|
||||
vadd.s32 \arg1, \arg5, \arg6 //int16 f[i][1] = e[i][1] + e[i][2];
|
||||
vsub.s32 \arg2, \arg5, \arg6 //int16 f[i][2] = e[i][1] - e[i][2];
|
||||
vsub.s32 \arg3, \arg4, \arg7 //int16 f[i][3] = e[i][0] - e[i][3];
|
||||
vadd.s32 \arg0, \arg4, \arg7 //int16 f[i][0] = e[i][0] + e[i][3];
|
||||
vadd.s32 \arg1, \arg5, \arg6 //int16 f[i][1] = e[i][1] + e[i][2];
|
||||
vsub.s32 \arg2, \arg5, \arg6 //int16 f[i][2] = e[i][1] - e[i][2];
|
||||
vsub.s32 \arg3, \arg4, \arg7 //int16 f[i][3] = e[i][0] - e[i][3];
|
||||
// }
|
||||
.endm
|
||||
|
||||
.macro COL_TRANSFORM_1_STEP arg0, arg1, arg2, arg3, arg4, arg5, arg6, arg7
|
||||
// { // input: src_q[0]~[3], output: e_q[0]~[3];
|
||||
vadd.s32 \arg4, \arg0, \arg2 //int32 e[0][j] = f[0][j] + f[2][j];
|
||||
vsub.s32 \arg5, \arg0, \arg2 //int32 e[1][j] = f[0][j] - f[2][j];
|
||||
vshr.s32 \arg6, \arg1, #1
|
||||
vshr.s32 \arg7, \arg3, #1
|
||||
vsub.s32 \arg6, \arg6, \arg3 //int32 e[2][j] = (f[1][j]>>1) - f[3][j];
|
||||
vadd.s32 \arg7, \arg1, \arg7 //int32 e[3][j] = f[1][j] + (f[3][j]>>1);
|
||||
vadd.s32 \arg4, \arg0, \arg2 //int32 e[0][j] = f[0][j] + f[2][j];
|
||||
vsub.s32 \arg5, \arg0, \arg2 //int32 e[1][j] = f[0][j] - f[2][j];
|
||||
vshr.s32 \arg6, \arg1, #1
|
||||
vshr.s32 \arg7, \arg3, #1
|
||||
vsub.s32 \arg6, \arg6, \arg3 //int32 e[2][j] = (f[1][j]>>1) - f[3][j];
|
||||
vadd.s32 \arg7, \arg1, \arg7 //int32 e[3][j] = f[1][j] + (f[3][j]>>1);
|
||||
// }
|
||||
.endm
|
||||
#endif
|
||||
// r0 int16_t* block,
|
||||
// r1 int8_t* non_zero_count,
|
||||
WELS_ASM_FUNC_BEGIN SetNonZeroCount_neon
|
||||
|
||||
WELS_ASM_FUNC_BEGIN SetNonZeroCount_neon
|
||||
|
||||
vld1.64 {d0-d2}, [r1]
|
||||
|
||||
|
||||
vceq.s8 q0, q0, #0
|
||||
vceq.s8 d2, d2, #0
|
||||
vmvn q0, q0
|
||||
vmvn d2, d2
|
||||
vabs.s8 q0, q0
|
||||
vabs.s8 d2, d2
|
||||
|
||||
|
||||
vst1.64 {d0-d2}, [r1]
|
||||
WELS_ASM_FUNC_END
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
|
||||
// r0 int16_t * block,
|
||||
// r0 int16_t * block,
|
||||
// r1 int32_t stride
|
||||
WELS_ASM_FUNC_BEGIN WelsResBlockZero16x16_neon// can use for 256*sizeof(int16_t)
|
||||
WELS_ASM_FUNC_BEGIN WelsResBlockZero16x16_neon// can use for 256*sizeof(int16_t)
|
||||
push {r2}
|
||||
mov r2, #16
|
||||
// each row 16 elements, 16*sizeof(int16_t)
|
||||
// memset(ptr_dest, 0, 16*sizeof(int16_t));
|
||||
// ptr_dest += stride;
|
||||
// ptr_dest += stride;
|
||||
lsl r1, r1, #1 // r1 = 2*r1
|
||||
veor.i16 q0, q0, q0
|
||||
veor.i16 q1, q1, q1
|
||||
|
||||
block_zero_16x16_luma_loop:
|
||||
|
||||
block_zero_16x16_luma_loop:
|
||||
vst1.i16 {q0, q1}, [r0], r1
|
||||
subs r2, r2, #2
|
||||
vst1.i16 {q0, q1}, [r0], r1
|
||||
vst1.i16 {q0, q1}, [r0], r1
|
||||
bne block_zero_16x16_luma_loop
|
||||
|
||||
|
||||
pop {r2}
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsResBlockZero8x8_neon// can use for 64*sizeof(int16_t)
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsResBlockZero8x8_neon// can use for 64*sizeof(int16_t)
|
||||
push {r2}
|
||||
mov r2, #8
|
||||
// each row 8 elements, 8*sizeof(int16_t)
|
||||
// memset(ptr_dest, 0, 8*sizeof(int16_t));
|
||||
// ptr_dest += stride;
|
||||
// ptr_dest += stride;
|
||||
lsl r1, r1, #1
|
||||
veor.i16 q0, q0, q0
|
||||
|
||||
block_zero_8x8_chma_loop:
|
||||
|
||||
block_zero_8x8_chma_loop:
|
||||
vst1.i16 {q0}, [r0], r1
|
||||
subs r2, r2, #2
|
||||
vst1.i16 {q0}, [r0], r1
|
||||
vst1.i16 {q0}, [r0], r1
|
||||
bne block_zero_8x8_chma_loop
|
||||
|
||||
|
||||
pop {r2}
|
||||
WELS_ASM_FUNC_END
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
|
||||
// uint8_t *pred, const int32_t stride, int16_t *rs
|
||||
WELS_ASM_FUNC_BEGIN IdctResAddPred_neon
|
||||
WELS_ASM_FUNC_BEGIN IdctResAddPred_neon
|
||||
|
||||
vld4.s16 {d0, d1, d2, d3}, [r2] // cost 3 cycles!
|
||||
|
||||
vld4.s16 {d0, d1, d2, d3}, [r2] // cost 3 cycles!
|
||||
|
||||
ROW_TRANSFORM_1_STEP d0, d1, d2, d3, q4, q5, q6, q7, d4, d5
|
||||
|
||||
TRANSFORM_4BYTES q0, q1, q2, q3, q4, q5, q6, q7
|
||||
|
||||
|
||||
TRANSFORM_4BYTES q0, q1, q2, q3, q4, q5, q6, q7
|
||||
|
||||
// transform element 32bits
|
||||
vtrn.s32 q0, q1 //[0 1 2 3]+[4 5 6 7]-->[0 4 2 6]+[1 5 3 7]
|
||||
vtrn.s32 q2, q3 //[8 9 10 11]+[12 13 14 15]-->[8 12 10 14]+[9 13 11 15]
|
||||
@ -172,9 +172,9 @@ block_zero_8x8_chma_loop:
|
||||
vswp d3, d6 //[1 5 3 7]+[9 13 11 15]-->[1 5 9 13]+[3 7 11 15]
|
||||
|
||||
COL_TRANSFORM_1_STEP q0, q1, q2, q3, q4, q5, q6, q7
|
||||
|
||||
TRANSFORM_4BYTES q0, q1, q2, q3, q4, q5, q6, q7
|
||||
|
||||
|
||||
TRANSFORM_4BYTES q0, q1, q2, q3, q4, q5, q6, q7
|
||||
|
||||
//after clip_table[MAX_NEG_CROP] into [0, 255]
|
||||
mov r2, r0
|
||||
vld1.32 {d12[0]},[r0],r1
|
||||
@ -186,7 +186,7 @@ block_zero_8x8_chma_loop:
|
||||
vrshrn.s32 d9, q1, #6
|
||||
vrshrn.s32 d10, q2, #6
|
||||
vrshrn.s32 d11, q3, #6
|
||||
|
||||
|
||||
vmovl.u8 q0,d12
|
||||
vmovl.u8 q1,d14
|
||||
vadd.s16 q0,q4
|
||||
@ -199,5 +199,5 @@ block_zero_8x8_chma_loop:
|
||||
vst1.32 {d12[1]},[r2],r1
|
||||
vst1.32 {d14[0]},[r2],r1
|
||||
vst1.32 {d14[1]},[r2]
|
||||
WELS_ASM_FUNC_END
|
||||
WELS_ASM_FUNC_END
|
||||
#endif
|
||||
|
||||
@ -29,7 +29,7 @@
|
||||
* POSSIBILITY OF SUCH DAMAGE.
|
||||
*
|
||||
*/
|
||||
|
||||
|
||||
#ifdef HAVE_NEON
|
||||
//Global macro
|
||||
.text
|
||||
@ -61,79 +61,60 @@
|
||||
.endm
|
||||
#endif
|
||||
|
||||
/*
|
||||
* void get_i16x16_luma_pred_v(uint8_t *pred, const int32_t stride)
|
||||
* r0 --- pred
|
||||
* r1 --- stride
|
||||
* return --- void
|
||||
*/
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI16x16LumaPredV_neon
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI16x16LumaPredV_neon
|
||||
//Get the top line data to 'q0'
|
||||
sub r2, r0, r1
|
||||
vldm r2, {d0, d1}
|
||||
|
||||
|
||||
mov r2, r0
|
||||
mov r3, #4
|
||||
//Set the top line to the each line of MB(16*16)
|
||||
//Set the top line to the each line of MB(16*16)
|
||||
loop_0_get_i16x16_luma_pred_v:
|
||||
vst1.8 {d0,d1}, [r2], r1
|
||||
vst1.8 {d0,d1}, [r2], r1
|
||||
vst1.8 {d0,d1}, [r2], r1
|
||||
vst1.8 {d0,d1}, [r2], r1
|
||||
subs r3, #1
|
||||
bne loop_0_get_i16x16_luma_pred_v
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
bne loop_0_get_i16x16_luma_pred_v
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
|
||||
/*
|
||||
* void get_i16x16_luma_pred_h(uint8_t *pred, const int32_t stride)
|
||||
* r0 --- pred
|
||||
* r1 --- stride
|
||||
* return --- void
|
||||
*/
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI16x16LumaPredH_neon
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI16x16LumaPredH_neon
|
||||
sub r2, r0, #1
|
||||
mov r3, #4
|
||||
loop_0_get_i16x16_luma_pred_h:
|
||||
//Get one byte data from left side
|
||||
vld1.8 {d0[],d1[]}, [r2], r1
|
||||
vld1.8 {d2[],d3[]}, [r2], r1
|
||||
vld1.8 {d4[],d5[]}, [r2], r1
|
||||
vld1.8 {d2[],d3[]}, [r2], r1
|
||||
vld1.8 {d4[],d5[]}, [r2], r1
|
||||
vld1.8 {d6[],d7[]}, [r2], r1
|
||||
|
||||
|
||||
//Set the line of MB using the left side byte data
|
||||
vst1.8 {d0,d1}, [r0], r1
|
||||
vst1.8 {d2,d3}, [r0], r1
|
||||
vst1.8 {d4,d5}, [r0], r1
|
||||
vst1.8 {d6,d7}, [r0], r1
|
||||
|
||||
|
||||
subs r3, #1
|
||||
bne loop_0_get_i16x16_luma_pred_h
|
||||
bne loop_0_get_i16x16_luma_pred_h
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
|
||||
/*
|
||||
* void get_i16x16_luma_pred_dc_both(uint8_t *pred, const int32_t stride)
|
||||
* r0 --- pred
|
||||
* r1 --- stride
|
||||
* return --- void
|
||||
*/
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI16x16LumaPredDc_neon
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI16x16LumaPredDc_neon
|
||||
//stmdb sp!, { r2-r5, lr}
|
||||
//Get the left vertical line data
|
||||
sub r2, r0, #1
|
||||
GET_8BYTE_DATA d0, r2, r1
|
||||
GET_8BYTE_DATA d1, r2, r1
|
||||
|
||||
|
||||
//Get the top horizontal line data
|
||||
sub r2, r0, r1
|
||||
sub r2, r0, r1
|
||||
vldm r2, {d2, d3}
|
||||
|
||||
|
||||
//Calculate the sum of top horizontal line data and vertical line data
|
||||
vpaddl.u8 q0, q0
|
||||
vpaddl.u8 q1, q1
|
||||
@ -141,11 +122,11 @@ loop_0_get_i16x16_luma_pred_h:
|
||||
vadd.u16 d0, d0, d1
|
||||
vpaddl.u16 d0, d0
|
||||
vpaddl.u32 d0, d0
|
||||
|
||||
//Calculate the mean value
|
||||
|
||||
//Calculate the mean value
|
||||
vrshr.u16 d0, d0, #5
|
||||
vdup.8 q0, d0[0]
|
||||
|
||||
|
||||
//Set the mean value to the all of member of MB
|
||||
mov r2, #4
|
||||
loop_0_get_i16x16_luma_pred_dc_both:
|
||||
@ -154,28 +135,22 @@ loop_0_get_i16x16_luma_pred_dc_both:
|
||||
vst1.8 {d0,d1}, [r0], r1
|
||||
vst1.8 {d0,d1}, [r0], r1
|
||||
subs r2, #1
|
||||
bne loop_0_get_i16x16_luma_pred_dc_both
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
bne loop_0_get_i16x16_luma_pred_dc_both
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
|
||||
/*
|
||||
* void get_i16x16_luma_pred_plane(uint8_t *pred, const int32_t stride)
|
||||
* r0 --- pred
|
||||
* r1 --- stride
|
||||
* return --- void
|
||||
*/
|
||||
|
||||
//The table for SIMD instruction {(8,7,6,5,4,3,2,1) * 5}
|
||||
CONST0_GET_I16X16_LUMA_PRED_PLANE: .long 0x191e2328, 0x050a0f14
|
||||
|
||||
//The table for SIMD instruction {-7,-6,-5,-4,-3,-2,-1,0}
|
||||
//The table for SIMD instruction {-7,-6,-5,-4,-3,-2,-1,0}
|
||||
CONST1_GET_I16X16_LUMA_PRED_PLANE: .long 0xfcfbfaf9, 0x00fffefd
|
||||
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI16x16LumaPredPlane_neon
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI16x16LumaPredPlane_neon
|
||||
//stmdb sp!, { r2-r5, lr}
|
||||
|
||||
|
||||
//Load the table {(8,7,6,5,4,3,2,1) * 5}
|
||||
adr r2, CONST0_GET_I16X16_LUMA_PRED_PLANE
|
||||
vldr d0, [r2]
|
||||
@ -184,51 +159,51 @@ CONST1_GET_I16X16_LUMA_PRED_PLANE: .long 0xfcfbfaf9, 0x00fffefd
|
||||
sub r2, r0, r1
|
||||
sub r3, r2, #1
|
||||
vld1.8 d1, [r3]
|
||||
|
||||
|
||||
//Pack the top[8] ~ top[15] to d2
|
||||
add r3, #9
|
||||
vld1.8 d2, [r3]
|
||||
|
||||
|
||||
//Save the top[15] to d6 for next step
|
||||
vdup.u8 d6, d2[7]
|
||||
|
||||
|
||||
//Get and pack left[-1] ~ left[6] to d4
|
||||
sub r3, r2, #1
|
||||
GET_8BYTE_DATA d4, r3, r1
|
||||
|
||||
|
||||
//Get and pack left[8] ~ left[15] to d3
|
||||
add r3, r1
|
||||
GET_8BYTE_DATA d3, r3, r1
|
||||
|
||||
|
||||
//Save the left[15] to d7 for next step
|
||||
vdup.u8 d7, d3[7]
|
||||
|
||||
|
||||
//revert the sequence of d2,d3
|
||||
vrev64.8 q1, q1
|
||||
|
||||
vsubl.u8 q2, d3, d4 //q2={left[8]-left[6],left[9]-left[5],left[10]-left[4], ...}
|
||||
vsubl.u8 q1, d2, d1 //q1={top[8]-top[6],top[9]-top[5],top[10]-top[4], ...}
|
||||
|
||||
|
||||
|
||||
vmovl.u8 q0, d0
|
||||
vmul.s16 q1, q0, q1 //q1 = q1*{(8,7,6,5,4,3,2,1) * 5}
|
||||
vmul.s16 q2, q0, q2 //q2 = q2*{(8,7,6,5,4,3,2,1) * 5}
|
||||
|
||||
|
||||
//Calculate the sum of items of q1, q2
|
||||
vpadd.s16 d0, d2, d3
|
||||
vpadd.s16 d1, d4, d5
|
||||
vpaddl.s16 q0, q0
|
||||
vpaddl.s32 q0, q0
|
||||
|
||||
|
||||
//Get the value of 'b', 'c' and extend to q1, q2.
|
||||
vrshr.s64 q0, #6
|
||||
vdup.s16 q1, d0[0]
|
||||
vdup.s16 q2, d1[0]
|
||||
|
||||
|
||||
//Load the table {-7,-6,-5,-4,-3,-2,-1,0} to d0
|
||||
adr r2, CONST1_GET_I16X16_LUMA_PRED_PLANE
|
||||
vld1.32 {d0}, [r2]
|
||||
|
||||
|
||||
//Get the value of 'a' and save to q3
|
||||
vaddl.u8 q3, d6, d7
|
||||
vshl.u16 q3, #4
|
||||
@ -237,156 +212,132 @@ CONST1_GET_I16X16_LUMA_PRED_PLANE: .long 0xfcfbfaf9, 0x00fffefd
|
||||
vmovl.s8 q0, d0
|
||||
vmla.s16 q3, q0, q1
|
||||
vmla.s16 q3, q2, d0[0]
|
||||
|
||||
|
||||
//Calculate a+'b'*{1,2,3,4,5,6,7,8} + c*{-7}
|
||||
vshl.s16 q5, q1, #3
|
||||
vadd.s16 q5, q3
|
||||
|
||||
|
||||
//right shift 5 bits and rounding
|
||||
vqrshrun.s16 d0, q3, #5
|
||||
vqrshrun.s16 d1, q5, #5
|
||||
|
||||
|
||||
//Set the line of MB
|
||||
vst1.u32 {d0,d1}, [r0], r1
|
||||
|
||||
|
||||
|
||||
|
||||
//Do the same processing for setting other lines
|
||||
mov r2, #15
|
||||
loop_0_get_i16x16_luma_pred_plane:
|
||||
loop_0_get_i16x16_luma_pred_plane:
|
||||
vadd.s16 q3, q2
|
||||
vadd.s16 q5, q2
|
||||
vqrshrun.s16 d0, q3, #5
|
||||
vqrshrun.s16 d1, q5, #5
|
||||
vst1.u32 {d0,d1}, [r0], r1
|
||||
subs r2, #1
|
||||
bne loop_0_get_i16x16_luma_pred_plane
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
bne loop_0_get_i16x16_luma_pred_plane
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
/*
|
||||
* void get_i4x4_luma_pred_v(uint8_t *pred, const int32_t stride);
|
||||
* r0 --- pred
|
||||
* r1 --- stride
|
||||
* return --- void
|
||||
*/
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI4x4LumaPredV_neon
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI4x4LumaPredV_neon
|
||||
//stmdb sp!, { r2-r5, lr}
|
||||
//Load the top row (4 bytes)
|
||||
sub r2, r0, r1
|
||||
ldr r2, [r2]
|
||||
|
||||
|
||||
//Set the luma MB using top line
|
||||
str r2, [r0], r1
|
||||
str r2, [r0], r1
|
||||
str r2, [r0], r1
|
||||
str r2, [r0]
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
/*
|
||||
* void get_i4x4_luma_pred_h(uint8_t *pred, const int32_t stride);
|
||||
* r0 --- pred
|
||||
* r1 --- stride
|
||||
* return --- void
|
||||
*/
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI4x4LumaPredH_neon
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI4x4LumaPredH_neon
|
||||
//stmdb sp!, { r2-r5, lr}
|
||||
//Load the left column (4 bytes)
|
||||
sub r2, r0, #1
|
||||
vld1.8 {d0[]}, [r2], r1
|
||||
vld1.8 {d1[]}, [r2], r1
|
||||
vld1.8 {d2[]}, [r2], r1
|
||||
vld1.8 {d1[]}, [r2], r1
|
||||
vld1.8 {d2[]}, [r2], r1
|
||||
vld1.8 {d3[]}, [r2]
|
||||
|
||||
|
||||
//Set the luma MB using the left side byte
|
||||
vst1.32 {d0[0]}, [r0], r1
|
||||
vst1.32 {d1[0]}, [r0], r1
|
||||
vst1.32 {d2[0]}, [r0], r1
|
||||
vst1.32 {d3[0]}, [r0]
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
/*
|
||||
* void get_i4x4_luma_pred_d_l(uint8_t *pred, const int32_t stride);
|
||||
* r0 --- pred
|
||||
* r1 --- stride
|
||||
* return --- void
|
||||
*/
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI4x4LumaPredDDL_neon
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI4x4LumaPredDDL_neon
|
||||
//stmdb sp!, { r2-r5, lr}
|
||||
//Load the top row data(8 bytes)
|
||||
sub r2, r0, r1
|
||||
vld1.32 {d0}, [r2]
|
||||
|
||||
|
||||
//For "t7 + (t7<<1)"
|
||||
vdup.8 d1, d0[7]
|
||||
|
||||
|
||||
//calculate "t0+t1,t1+t2,t2+t3...t6+t7,t7+t7"
|
||||
vext.8 d1, d0, d1, #1
|
||||
vaddl.u8 q1, d1, d0
|
||||
|
||||
|
||||
//calculate "x,t0+t1+t1+t2,t1+t2+t2+t3,...t5+t6+t6+t7,t6+t7+t7+t7"
|
||||
vext.8 q2, q1, q1, #14
|
||||
vadd.u16 q0, q1, q2
|
||||
|
||||
|
||||
//right shift 2 bits and rounding
|
||||
vqrshrn.u16 d0, q0, #2
|
||||
|
||||
|
||||
//Save "ddl0, ddl1, ddl2, ddl3"
|
||||
vext.8 d1, d0, d0, #1
|
||||
vst1.32 d1[0], [r0], r1
|
||||
|
||||
|
||||
//Save "ddl1, ddl2, ddl3, ddl4"
|
||||
vext.8 d1, d0, d0, #2
|
||||
vst1.32 d1[0], [r0], r1
|
||||
|
||||
|
||||
//Save "ddl2, ddl3, ddl4, ddl5"
|
||||
vext.8 d1, d0, d0, #3
|
||||
vst1.32 d1[0], [r0], r1
|
||||
|
||||
vst1.32 d1[0], [r0], r1
|
||||
|
||||
//Save "ddl3, ddl4, ddl5, ddl6"
|
||||
vst1.32 d0[1], [r0]
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
vst1.32 d0[1], [r0]
|
||||
|
||||
/*
|
||||
* void get_i4x4_luma_pred_d_r(uint8_t *pred, const int32_t stride);
|
||||
* r0 --- pred
|
||||
* r1 --- stride
|
||||
* return --- void
|
||||
*/
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI4x4LumaPredDDR_neon
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI4x4LumaPredDDR_neon
|
||||
//stmdb sp!, { r2-r5, lr}
|
||||
//Load the top row (4 bytes)
|
||||
sub r2, r0, r1
|
||||
vld1.32 {d0[1]}, [r2]
|
||||
|
||||
|
||||
//Load the left column (5 bytes)
|
||||
sub r2, #1
|
||||
vld1.8 {d0[3]}, [r2], r1
|
||||
vld1.8 {d0[2]}, [r2], r1
|
||||
vld1.8 {d0[2]}, [r2], r1
|
||||
vld1.8 {d0[1]}, [r2], r1
|
||||
vld1.8 {d0[0]}, [r2], r1
|
||||
vld1.8 {d0[0]}, [r2], r1
|
||||
vld1.8 {d1[7]}, [r2] //For packing the right sequence to do SIMD processing
|
||||
|
||||
|
||||
|
||||
|
||||
vext.8 d2, d1, d0, #7 //d0:{L2,L1,L0,LT,T0,T1,T2,T3}
|
||||
//d2:{L3,L2,L1,L0,LT,T0,T1,T2}
|
||||
|
||||
|
||||
//q2:{L2+L3,L1+L2,L0+L1...T1+T2,T2+T3}
|
||||
vaddl.u8 q2, d2, d0
|
||||
|
||||
|
||||
//q1:{TL0+LT0,LT0+T01,...L12+L23}
|
||||
vext.8 q3, q3, q2, #14
|
||||
vadd.u16 q1, q2, q3
|
||||
|
||||
|
||||
//right shift 2 bits and rounding
|
||||
vqrshrn.u16 d0, q1, #2
|
||||
|
||||
|
||||
//Adjust the data sequence for setting luma MB of 'pred'
|
||||
vst1.32 d0[1], [r0], r1
|
||||
vext.8 d0, d0, d0, #7
|
||||
@ -396,34 +347,29 @@ loop_0_get_i16x16_luma_pred_plane:
|
||||
vext.8 d0, d0, d0, #7
|
||||
vst1.32 d0[1], [r0]
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
|
||||
/*
|
||||
* void get_i4x4_luma_pred_v_l(uint8_t *pred, const int32_t stride);
|
||||
* r0 --- pred
|
||||
* r1 --- stride
|
||||
* return --- void
|
||||
*/
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI4x4LumaPredVL_neon
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI4x4LumaPredVL_neon
|
||||
//stmdb sp!, { r2-r5, lr}
|
||||
//Load the top row (8 bytes)
|
||||
sub r2, r0, r1
|
||||
vld1.32 {d0}, [r2]
|
||||
|
||||
|
||||
|
||||
vext.8 d1, d0, d0, #1
|
||||
vaddl.u8 q1, d1, d0 //q1:{t0+t1,t1+t2,t2+t3...t5+t6,x,x}
|
||||
|
||||
|
||||
vext.8 q2, q1, q1, #2
|
||||
vadd.u16 q2, q1, q2 //q2:{t0+t1+t1+t2,t1+t2+t2+t3,...t4+t5+t5+t6,x,x}
|
||||
|
||||
|
||||
//calculate the "vl0,vl1,vl2,vl3,vl4"
|
||||
vqrshrn.u16 d0, q1, #1
|
||||
|
||||
|
||||
//calculate the "vl5,vl6,vl7,vl8,vl9"
|
||||
vqrshrn.u16 d1, q2, #2
|
||||
|
||||
|
||||
//Adjust the data sequence for setting the luma MB
|
||||
vst1.32 d0[0], [r0], r1
|
||||
vst1.32 d1[0], [r0], r1
|
||||
@ -431,121 +377,104 @@ loop_0_get_i16x16_luma_pred_plane:
|
||||
vext.8 d1, d1, d1, #1
|
||||
vst1.32 d0[0], [r0], r1
|
||||
vst1.32 d1[0], [r0]
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
|
||||
/*
|
||||
* void get_i4x4_luma_pred_v_r(uint8_t *pred, const int32_t stride);
|
||||
* r0 --- pred
|
||||
* r1 --- stride
|
||||
* return --- void
|
||||
*/
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI4x4LumaPredVR_neon
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI4x4LumaPredVR_neon
|
||||
//stmdb sp!, { r2-r5, lr}
|
||||
//Load the top row (4 bytes)
|
||||
sub r2, r0, r1
|
||||
vld1.32 {d0[1]}, [r2]
|
||||
|
||||
|
||||
//Load the left column (4 bytes)
|
||||
sub r2, #1
|
||||
vld1.8 {d0[3]}, [r2], r1
|
||||
vld1.8 {d0[3]}, [r2], r1
|
||||
vld1.8 {d0[2]}, [r2], r1
|
||||
vld1.8 {d0[1]}, [r2], r1
|
||||
vld1.8 {d0[0]}, [r2]
|
||||
vld1.8 {d0[1]}, [r2], r1
|
||||
vld1.8 {d0[0]}, [r2]
|
||||
|
||||
|
||||
|
||||
vext.8 d1, d0, d0, #7
|
||||
vaddl.u8 q1, d0, d1 //q1:{X,L2+L1,L1+L0,L0+LT,LT+T0,T0+T1,T1+T2,T2+T3}
|
||||
|
||||
|
||||
vext.u8 q2, q1, q1, #14
|
||||
vadd.u16 q2, q2, q1 //q2:{X,L2+L1+L1+L0,L1+L0+L0+LT,...T1+T2+T2+T3}
|
||||
|
||||
|
||||
//Calculate the vr0 ~ vr9
|
||||
vqrshrn.u16 d1, q2, #2
|
||||
vqrshrn.u16 d0, q1, #1
|
||||
|
||||
|
||||
//Adjust the data sequence for setting the luma MB
|
||||
vst1.32 d0[1], [r0], r1
|
||||
vst1.32 d1[1], [r0], r1
|
||||
add r2, r0, r1
|
||||
vst1.8 d1[3], [r0]!
|
||||
vst1.16 d0[2], [r0]!
|
||||
vst1.16 d0[2], [r0]!
|
||||
vst1.8 d0[6], [r0]!
|
||||
vst1.8 d1[2], [r2]!
|
||||
vst1.16 d1[2], [r2]!
|
||||
vst1.16 d1[2], [r2]!
|
||||
vst1.8 d1[6], [r2]
|
||||
WELS_ASM_FUNC_END
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
|
||||
/*
|
||||
* get_i4x4_luma_pred_h_u(uint8_t *pred, const int32_t stride);
|
||||
* r0 --- pred
|
||||
* r1 --- stride
|
||||
* return --- void
|
||||
*/
|
||||
//NO TEST
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI4x4LumaPredHU_neon
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI4x4LumaPredHU_neon
|
||||
//stmdb sp!, { r2-r5, lr}
|
||||
//Load the left column data
|
||||
sub r2, r0, #1
|
||||
mov r3, #3
|
||||
mul r3, r1
|
||||
add r3, r2
|
||||
vld1.8 {d0[]}, [r3]
|
||||
vld1.8 {d0[4]}, [r2], r1
|
||||
vld1.8 {d0[]}, [r3]
|
||||
vld1.8 {d0[4]}, [r2], r1
|
||||
vld1.8 {d0[5]}, [r2], r1
|
||||
vld1.8 {d0[6]}, [r2], r1 //d0:{L3,L3,L3,L3,L0,L1,L2,L3}
|
||||
vld1.8 {d0[6]}, [r2], r1 //d0:{L3,L3,L3,L3,L0,L1,L2,L3}
|
||||
|
||||
vext.8 d1, d0, d0, #1
|
||||
vaddl.u8 q2, d0, d1 //q2:{L3+L3,L3+L3,L3+L3,L3+L0,L0+L1,L1+L2,L2+L3,L3+L3}
|
||||
|
||||
vaddl.u8 q2, d0, d1 //q2:{L3+L3,L3+L3,L3+L3,L3+L0,L0+L1,L1+L2,L2+L3,L3+L3}
|
||||
|
||||
vext.u8 d2, d5, d4, #2
|
||||
vadd.u16 d3, d2, d5 //d3:{L0+L1+L1+L2,L1+L2+L2+L3,L2+L3+L3+L3,L3+L3+L3+L3}
|
||||
|
||||
vadd.u16 d3, d2, d5 //d3:{L0+L1+L1+L2,L1+L2+L2+L3,L2+L3+L3+L3,L3+L3+L3+L3}
|
||||
|
||||
//Calculate the hu0 ~ hu5
|
||||
vqrshrn.u16 d2, q2, #1
|
||||
vqrshrn.u16 d1, q1, #2
|
||||
|
||||
|
||||
//Adjust the data sequence for setting the luma MB
|
||||
vzip.8 d2, d1
|
||||
vst1.32 d1[0], [r0], r1
|
||||
vext.8 d2, d1, d1, #2
|
||||
vext.8 d2, d1, d1, #2
|
||||
vst1.32 d2[0], [r0], r1
|
||||
vst1.32 d1[1], [r0], r1
|
||||
vst1.32 d0[0], [r0]
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
/*
|
||||
* void get_i4x4_luma_pred_h_d(uint8_t *pred, const int32_t stride);
|
||||
* r0 --- pred
|
||||
* r1 --- stride
|
||||
* return --- void
|
||||
*/
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI4x4LumaPredHD_neon
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderI4x4LumaPredHD_neon
|
||||
//stmdb sp!, { r2-r5, lr}
|
||||
//Load the data
|
||||
sub r2, r0, r1
|
||||
sub r2, #1
|
||||
vld1.32 {d0[1]}, [r2], r1
|
||||
vld1.8 {d0[3]}, [r2], r1
|
||||
vld1.8 {d0[3]}, [r2], r1
|
||||
vld1.8 {d0[2]}, [r2], r1
|
||||
vld1.8 {d0[1]}, [r2], r1
|
||||
vld1.8 {d0[1]}, [r2], r1
|
||||
vld1.8 {d0[0]}, [r2] //d0:{L3,L2,L1,L0,LT,T0,T1,T2}
|
||||
|
||||
|
||||
vext.8 d1, d0, d0, #7
|
||||
vaddl.u8 q1, d0, d1 //q1:{x,L3+L2,L2+L1,L1+L0,L0+LT,LT+T0,T0+T1,T1+T2}
|
||||
|
||||
|
||||
vext.u8 q2, q1, q1, #14 //q2:{x,x, L3+L2,L2+L1,L1+L0,L0+LT,LT+T0,T0+T1}
|
||||
vadd.u16 q3, q2, q1 //q3:{x,x,L3+L2+L2+L1,L2+L1+L1+L0,L1+L0+L0+LT,L0+LT+LT+T0,LT+T0+T0+T1,T0+T1+T1+T2}
|
||||
|
||||
|
||||
//Calculate the hd0~hd9
|
||||
vqrshrn.u16 d1, q3, #2
|
||||
vqrshrn.u16 d0, q2, #1
|
||||
|
||||
|
||||
//Adjust the data sequence for setting the luma MB
|
||||
vmov d3, d1
|
||||
vtrn.8 d0, d1
|
||||
@ -556,17 +485,10 @@ loop_0_get_i16x16_luma_pred_plane:
|
||||
vst2.16 {d2[2], d3[2]}, [r0], r1
|
||||
vst2.16 {d0[1], d1[1]}, [r0]
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
|
||||
/*
|
||||
* void get_i_chroma_pred_v(uint8_t *pred, const int32_t stride);
|
||||
* r0 --- pred
|
||||
* r1 --- stride
|
||||
* return --- void
|
||||
*/
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderIChromaPredV_neon
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderIChromaPredV_neon
|
||||
//stmdb sp!, { r2-r5, lr}
|
||||
//Get the top row (8 byte)
|
||||
sub r2, r0, r1
|
||||
@ -580,32 +502,25 @@ loop_0_get_i16x16_luma_pred_plane:
|
||||
vst1.8 {d0}, [r0], r1
|
||||
vst1.8 {d0}, [r0], r1
|
||||
vst1.8 {d0}, [r0], r1
|
||||
vst1.8 {d0}, [r0]
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
vst1.8 {d0}, [r0]
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
|
||||
/*
|
||||
* void get_i_chroma_pred_h(uint8_t *pred, const int32_t stride);
|
||||
* r0 --- pred
|
||||
* r1 --- stride
|
||||
* return --- void
|
||||
*/
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderIChromaPredH_neon
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderIChromaPredH_neon
|
||||
//stmdb sp!, { r2-r5, lr}
|
||||
////Get the left column (8 byte)
|
||||
sub r2, r0, #1
|
||||
vld1.8 {d0[]}, [r2], r1
|
||||
vld1.8 {d1[]}, [r2], r1
|
||||
vld1.8 {d2[]}, [r2], r1
|
||||
vld1.8 {d1[]}, [r2], r1
|
||||
vld1.8 {d2[]}, [r2], r1
|
||||
vld1.8 {d3[]}, [r2], r1
|
||||
vld1.8 {d4[]}, [r2], r1
|
||||
vld1.8 {d5[]}, [r2], r1
|
||||
vld1.8 {d6[]}, [r2], r1
|
||||
vld1.8 {d5[]}, [r2], r1
|
||||
vld1.8 {d6[]}, [r2], r1
|
||||
vld1.8 {d7[]}, [r2]
|
||||
|
||||
//Set the chroma MB using left column data
|
||||
|
||||
//Set the chroma MB using left column data
|
||||
vst1.8 {d0}, [r0], r1
|
||||
vst1.8 {d1}, [r0], r1
|
||||
vst1.8 {d2}, [r0], r1
|
||||
@ -613,100 +528,88 @@ loop_0_get_i16x16_luma_pred_plane:
|
||||
vst1.8 {d4}, [r0], r1
|
||||
vst1.8 {d5}, [r0], r1
|
||||
vst1.8 {d6}, [r0], r1
|
||||
vst1.8 {d7}, [r0]
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
vst1.8 {d7}, [r0]
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
|
||||
/*
|
||||
* void get_i_chroma_pred_dc_both(uint8_t *pred, const int32_t stride);
|
||||
* r0 --- pred
|
||||
* r1 --- stride
|
||||
* return --- void
|
||||
*/
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderIChromaPredDC_neon
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderIChromaPredDC_neon
|
||||
//stmdb sp!, { r2-r5, lr}
|
||||
//Load the left column data (8 bytes)
|
||||
sub r2, r0, #1
|
||||
GET_8BYTE_DATA d0, r2, r1
|
||||
|
||||
GET_8BYTE_DATA d0, r2, r1
|
||||
|
||||
//Load the top row data (8 bytes)
|
||||
sub r2, r0, r1
|
||||
sub r2, r0, r1
|
||||
vldr d1, [r2]
|
||||
|
||||
|
||||
//Calculate the sum of left column and top row
|
||||
vpaddl.u8 q0, q0
|
||||
vpaddl.u16 q0, q0
|
||||
vadd.u32 d2, d0, d1 //'m1' save to d2
|
||||
|
||||
vrshr.u32 q0, q0, #2 //calculate 'm2','m3'
|
||||
vrshr.u32 d2, d2, #3 //calculate 'm4'
|
||||
|
||||
|
||||
vrshr.u32 q0, q0, #2 //calculate 'm2','m3'
|
||||
vrshr.u32 d2, d2, #3 //calculate 'm4'
|
||||
|
||||
//duplicate the 'mx' to a vector line
|
||||
vdup.8 d4, d2[0]
|
||||
vdup.8 d5, d1[4]
|
||||
vdup.8 d6, d0[4]
|
||||
vdup.8 d7, d2[4]
|
||||
|
||||
//Set the chroma MB
|
||||
vst2.32 {d4[0],d5[0]}, [r0], r1
|
||||
vst2.32 {d4[0],d5[0]}, [r0], r1
|
||||
vst2.32 {d4[0],d5[0]}, [r0], r1
|
||||
vst2.32 {d4[0],d5[0]}, [r0], r1
|
||||
vst2.32 {d6[0],d7[0]}, [r0], r1
|
||||
vst2.32 {d6[0],d7[0]}, [r0], r1
|
||||
vst2.32 {d6[0],d7[0]}, [r0], r1
|
||||
vst2.32 {d6[0],d7[0]}, [r0]
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
/*
|
||||
* void get_i_chroma_pred_plane(uint8_t *pred, const int32_t stride);
|
||||
* r0 --- pred
|
||||
* r1 --- stride
|
||||
* return --- void
|
||||
*/
|
||||
//Set the chroma MB
|
||||
vst2.32 {d4[0],d5[0]}, [r0], r1
|
||||
vst2.32 {d4[0],d5[0]}, [r0], r1
|
||||
vst2.32 {d4[0],d5[0]}, [r0], r1
|
||||
vst2.32 {d4[0],d5[0]}, [r0], r1
|
||||
vst2.32 {d6[0],d7[0]}, [r0], r1
|
||||
vst2.32 {d6[0],d7[0]}, [r0], r1
|
||||
vst2.32 {d6[0],d7[0]}, [r0], r1
|
||||
vst2.32 {d6[0],d7[0]}, [r0]
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
|
||||
//Table {{1,2,3,4,1,2,3,4}*17}
|
||||
CONST0_GET_I_CHROMA_PRED_PLANE: .long 0x44332211, 0x44332211//0x140f0a05, 0x28231e19
|
||||
//Table {-3,-2,-1,0,1,2,3,4}
|
||||
CONST1_GET_I_CHROMA_PRED_PLANE: .long 0xfffefffd, 0x0000ffff,0x00020001,0x00040003
|
||||
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderIChromaPredPlane_neon
|
||||
WELS_ASM_FUNC_BEGIN WelsDecoderIChromaPredPlane_neon
|
||||
//stmdb sp!, { r2-r5, lr}
|
||||
//Load the top row data
|
||||
sub r2, r0, #1
|
||||
sub r2, r1
|
||||
vld1.32 {d1[0]}, [r2]
|
||||
vld1.32 {d1[0]}, [r2]
|
||||
add r2, #5
|
||||
vld1.32 {d0[0]}, [r2]
|
||||
|
||||
|
||||
//Load the left column data
|
||||
sub r2, #5
|
||||
vld1.8 {d1[4]}, [r2], r1
|
||||
vld1.8 {d1[5]}, [r2], r1
|
||||
vld1.8 {d1[5]}, [r2], r1
|
||||
vld1.8 {d1[6]}, [r2], r1
|
||||
vld1.8 {d1[7]}, [r2], r1 //d1:{LT,T0,T1,T2,LT,L0,L1,L2}
|
||||
vld1.8 {d1[7]}, [r2], r1 //d1:{LT,T0,T1,T2,LT,L0,L1,L2}
|
||||
add r2, r1
|
||||
vld1.8 {d0[4]}, [r2], r1
|
||||
vld1.8 {d0[5]}, [r2], r1
|
||||
vld1.8 {d0[6]}, [r2], r1
|
||||
vld1.8 {d0[7]}, [r2] //d0:{T4,T5,T6,T7,L4,L5,L6.L7}
|
||||
|
||||
|
||||
|
||||
|
||||
//Save T7 to d3 for next step
|
||||
vdup.u8 d3, d0[3]
|
||||
//Save L7 to d4 for next step
|
||||
vdup.u8 d4, d0[7]
|
||||
|
||||
|
||||
//Calculate the value of 'a' and save to q2
|
||||
vaddl.u8 q2, d3, d4
|
||||
vshl.u16 q2, #4
|
||||
|
||||
|
||||
//Load the table {{1,2,3,4,1,2,3,4}*17}
|
||||
adr r2, CONST0_GET_I_CHROMA_PRED_PLANE
|
||||
vld1.32 {d2}, [r2]
|
||||
|
||||
|
||||
//Calculate the 'b','c', and save to q0
|
||||
vrev32.8 d1, d1
|
||||
vsubl.u8 q0, d0, d1
|
||||
@ -715,32 +618,32 @@ CONST1_GET_I_CHROMA_PRED_PLANE: .long 0xfffefffd, 0x0000ffff,0x00020001,0x000400
|
||||
vpaddl.s16 q0, q0
|
||||
vpaddl.s32 q0, q0
|
||||
vrshr.s64 q0, #5
|
||||
|
||||
|
||||
//Load the table {-3,-2,-1,0,1,2,3,4} to q3
|
||||
adr r2, CONST1_GET_I_CHROMA_PRED_PLANE
|
||||
vld1.32 {d6, d7}, [r2]
|
||||
|
||||
|
||||
//Duplicate the 'b','c' to q0, q1 for SIMD instruction
|
||||
vdup.s16 q1, d1[0]
|
||||
vdup.s16 q0, d0[0]
|
||||
|
||||
|
||||
//Calculate the "(a + b * (j - 3) + c * (- 3) + 16) >> 5;"
|
||||
vmla.s16 q2, q0, q3
|
||||
vmla.s16 q2, q1, d6[0]
|
||||
vqrshrun.s16 d0, q2, #5
|
||||
|
||||
|
||||
//Set a line of chroma MB
|
||||
vst1.u32 {d0}, [r0], r1
|
||||
|
||||
|
||||
//Do the same processing for each line.
|
||||
mov r2, #7
|
||||
loop_0_get_i_chroma_pred_plane:
|
||||
loop_0_get_i_chroma_pred_plane:
|
||||
vadd.s16 q2, q1
|
||||
vqrshrun.s16 d0, q2, #5
|
||||
vst1.u32 {d0}, [r0], r1
|
||||
subs r2, #1
|
||||
bne loop_0_get_i_chroma_pred_plane
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
bne loop_0_get_i_chroma_pred_plane
|
||||
|
||||
WELS_ASM_FUNC_END
|
||||
|
||||
#endif
|
||||
|
||||
File diff suppressed because it is too large
Load Diff
@ -112,7 +112,7 @@ void WelsDecoderI16x16LumaPredV_neon(uint8_t *pPred, const int32_t kiStride);
|
||||
void WelsDecoderI16x16LumaPredH_neon(uint8_t *pPred, const int32_t kiStride);
|
||||
void WelsDecoderI16x16LumaPredDc_neon(uint8_t *pPred, const int32_t kiStride);
|
||||
void WelsDecoderI16x16LumaPredPlane_neon(uint8_t *pPred, const int32_t kiStride);
|
||||
|
||||
|
||||
void WelsDecoderI4x4LumaPredV_neon(uint8_t *pPred, const int32_t kiStride);
|
||||
void WelsDecoderI4x4LumaPredH_neon(uint8_t *pPred, const int32_t kiStride);
|
||||
void WelsDecoderI4x4LumaPredDDL_neon(uint8_t *pPred, const int32_t kiStride);
|
||||
@ -121,11 +121,11 @@ void WelsDecoderI4x4LumaPredVL_neon(uint8_t *pPred, const int32_t kiStride);
|
||||
void WelsDecoderI4x4LumaPredVR_neon(uint8_t *pPred, const int32_t kiStride);
|
||||
void WelsDecoderI4x4LumaPredHU_neon(uint8_t *pPred, const int32_t kiStride);
|
||||
void WelsDecoderI4x4LumaPredHD_neon(uint8_t *pPred, const int32_t kiStride);
|
||||
|
||||
|
||||
void WelsDecoderIChromaPredV_neon(uint8_t *pPred, const int32_t kiStride);
|
||||
void WelsDecoderIChromaPredH_neon(uint8_t *pPred, const int32_t kiStride);
|
||||
void WelsDecoderIChromaPredDC_neon(uint8_t *pPred, const int32_t kiStride);
|
||||
void WelsDecoderIChromaPredPlane_neon(uint8_t *pPred, const int32_t kiStride);
|
||||
void WelsDecoderIChromaPredPlane_neon(uint8_t *pPred, const int32_t kiStride);
|
||||
#endif//HAVE_NEON
|
||||
|
||||
#if defined(__cplusplus)
|
||||
|
||||
@ -720,13 +720,13 @@ void DeblockingInit (SDeblockingFunc* pFunc, int32_t iCpu) {
|
||||
#endif
|
||||
|
||||
#if defined(HAVE_NEON)
|
||||
if ( iCpu & WELS_CPU_NEON )
|
||||
if ( iCpu & WELS_CPU_NEON )
|
||||
{
|
||||
pFunc->pfLumaDeblockingLT4Ver = DeblockLumaLt4V_neon;
|
||||
pFunc->pfLumaDeblockingEQ4Ver = DeblockLumaEq4V_neon;
|
||||
pFunc->pfLumaDeblockingLT4Hor = DeblockLumaLt4H_neon;
|
||||
pFunc->pfLumaDeblockingEQ4Hor = DeblockLumaEq4H_neon;
|
||||
|
||||
|
||||
pFunc->pfChromaDeblockingLT4Ver = DeblockChromaLt4V_neon;
|
||||
pFunc->pfChromaDeblockingEQ4Ver = DeblockChromaEq4V_neon;
|
||||
pFunc->pfChromaDeblockingLT4Hor = DeblockChromaLt4H_neon;
|
||||
|
||||
@ -1152,8 +1152,8 @@ void WelsBlockFuncInit (SBlockFunc* pFunc, int32_t iCpu) {
|
||||
#ifdef HAVE_NEON
|
||||
if ( iCpu & WELS_CPU_NEON ) {
|
||||
pFunc->pWelsBlockZero16x16Func = WelsResBlockZero16x16_neon;
|
||||
pFunc->pWelsBlockZero8x8Func = WelsResBlockZero8x8_neon;
|
||||
pFunc->pWelsSetNonZeroCountFunc = SetNonZeroCount_neon;
|
||||
pFunc->pWelsBlockZero8x8Func = WelsResBlockZero8x8_neon;
|
||||
pFunc->pWelsSetNonZeroCountFunc = SetNonZeroCount_neon;
|
||||
}
|
||||
#endif
|
||||
}
|
||||
|
||||
@ -662,30 +662,30 @@ void AssignFuncPointerForRec (PWelsDecoderContext pCtx) {
|
||||
|
||||
InitDctClipTable();
|
||||
pCtx->pIdctResAddPredFunc = IdctResAddPred_c;
|
||||
|
||||
|
||||
#if defined(HAVE_NEON)
|
||||
if ( pCtx->uiCpuFlag & WELS_CPU_NEON ) {
|
||||
pCtx->pIdctResAddPredFunc = IdctResAddPred_neon;
|
||||
|
||||
pCtx->pGetI16x16LumaPredFunc[I16_PRED_DC] = WelsDecoderI16x16LumaPredDc_neon;
|
||||
pCtx->pGetI16x16LumaPredFunc[I16_PRED_P] = WelsDecoderI16x16LumaPredPlane_neon;
|
||||
pCtx->pGetI16x16LumaPredFunc[I16_PRED_H] = WelsDecoderI16x16LumaPredH_neon;
|
||||
pCtx->pGetI16x16LumaPredFunc[I16_PRED_V] = WelsDecoderI16x16LumaPredV_neon;
|
||||
|
||||
pCtx->pGetI4x4LumaPredFunc[I4_PRED_V ] = WelsDecoderI4x4LumaPredV_neon;
|
||||
pCtx->pGetI4x4LumaPredFunc[I4_PRED_H ] = WelsDecoderI4x4LumaPredH_neon;
|
||||
pCtx->pGetI4x4LumaPredFunc[I4_PRED_DDL ] = WelsDecoderI4x4LumaPredDDL_neon;
|
||||
pCtx->pGetI4x4LumaPredFunc[I4_PRED_DDR ] = WelsDecoderI4x4LumaPredDDR_neon;
|
||||
pCtx->pGetI4x4LumaPredFunc[I4_PRED_VL ] = WelsDecoderI4x4LumaPredVL_neon;
|
||||
pCtx->pGetI4x4LumaPredFunc[I4_PRED_VR ] = WelsDecoderI4x4LumaPredVR_neon;
|
||||
pCtx->pGetI4x4LumaPredFunc[I4_PRED_HU ] = WelsDecoderI4x4LumaPredHU_neon;
|
||||
pCtx->pGetI4x4LumaPredFunc[I4_PRED_HD ] = WelsDecoderI4x4LumaPredHD_neon;
|
||||
|
||||
pCtx->pGetIChromaPredFunc[C_PRED_H] = WelsDecoderIChromaPredH_neon;
|
||||
pCtx->pGetIChromaPredFunc[C_PRED_V] = WelsDecoderIChromaPredV_neon;
|
||||
pCtx->pGetIChromaPredFunc[C_PRED_P ] = WelsDecoderIChromaPredPlane_neon;
|
||||
pCtx->pGetIChromaPredFunc[C_PRED_DC] = WelsDecoderIChromaPredDC_neon;
|
||||
}
|
||||
|
||||
pCtx->pGetI16x16LumaPredFunc[I16_PRED_DC] = WelsDecoderI16x16LumaPredDc_neon;
|
||||
pCtx->pGetI16x16LumaPredFunc[I16_PRED_P] = WelsDecoderI16x16LumaPredPlane_neon;
|
||||
pCtx->pGetI16x16LumaPredFunc[I16_PRED_H] = WelsDecoderI16x16LumaPredH_neon;
|
||||
pCtx->pGetI16x16LumaPredFunc[I16_PRED_V] = WelsDecoderI16x16LumaPredV_neon;
|
||||
|
||||
pCtx->pGetI4x4LumaPredFunc[I4_PRED_V ] = WelsDecoderI4x4LumaPredV_neon;
|
||||
pCtx->pGetI4x4LumaPredFunc[I4_PRED_H ] = WelsDecoderI4x4LumaPredH_neon;
|
||||
pCtx->pGetI4x4LumaPredFunc[I4_PRED_DDL ] = WelsDecoderI4x4LumaPredDDL_neon;
|
||||
pCtx->pGetI4x4LumaPredFunc[I4_PRED_DDR ] = WelsDecoderI4x4LumaPredDDR_neon;
|
||||
pCtx->pGetI4x4LumaPredFunc[I4_PRED_VL ] = WelsDecoderI4x4LumaPredVL_neon;
|
||||
pCtx->pGetI4x4LumaPredFunc[I4_PRED_VR ] = WelsDecoderI4x4LumaPredVR_neon;
|
||||
pCtx->pGetI4x4LumaPredFunc[I4_PRED_HU ] = WelsDecoderI4x4LumaPredHU_neon;
|
||||
pCtx->pGetI4x4LumaPredFunc[I4_PRED_HD ] = WelsDecoderI4x4LumaPredHD_neon;
|
||||
|
||||
pCtx->pGetIChromaPredFunc[C_PRED_H] = WelsDecoderIChromaPredH_neon;
|
||||
pCtx->pGetIChromaPredFunc[C_PRED_V] = WelsDecoderIChromaPredV_neon;
|
||||
pCtx->pGetIChromaPredFunc[C_PRED_P ] = WelsDecoderIChromaPredPlane_neon;
|
||||
pCtx->pGetIChromaPredFunc[C_PRED_DC] = WelsDecoderIChromaPredDC_neon;
|
||||
}
|
||||
#endif//HAVE_NEON
|
||||
|
||||
|
||||
|
||||
@ -669,8 +669,8 @@ void McCopy_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t
|
||||
void McHorVer20_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride,
|
||||
int32_t iWidth, int32_t iHeight)
|
||||
{
|
||||
if (iWidth == 16)
|
||||
McHorVer20WidthEq16_neon(pSrc, iSrcStride, pDst, iDstStride, iHeight);
|
||||
if (iWidth == 16)
|
||||
McHorVer20WidthEq16_neon(pSrc, iSrcStride, pDst, iDstStride, iHeight);
|
||||
else if (iWidth == 8)
|
||||
McHorVer20WidthEq8_neon(pSrc, iSrcStride, pDst, iDstStride, iHeight);
|
||||
else if (iWidth == 4)
|
||||
@ -690,13 +690,13 @@ void McHorVer22_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int
|
||||
int32_t iWidth, int32_t iHeight)
|
||||
{
|
||||
if (iWidth == 16)
|
||||
McHorVer22WidthEq16_neon(pSrc, iSrcStride, pDst, iDstStride, iHeight);
|
||||
McHorVer22WidthEq16_neon(pSrc, iSrcStride, pDst, iDstStride, iHeight);
|
||||
else if (iWidth == 8)
|
||||
McHorVer22WidthEq8_neon(pSrc, iSrcStride, pDst, iDstStride, iHeight);
|
||||
else if (iWidth == 4)
|
||||
McHorVer22WidthEq4_neon(pSrc, iSrcStride, pDst, iDstStride, iHeight);
|
||||
}
|
||||
|
||||
|
||||
void McHorVer01_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride,
|
||||
int32_t iWidth, int32_t iHeight)
|
||||
{
|
||||
@ -927,7 +927,7 @@ void McHorVer33_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int
|
||||
PixelAvgWidthEq4_neon(pDst, iDstStride, pHorTmp, pVerTmp, iHeight);
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
void McLuma_neon(const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride,
|
||||
int16_t iMvX, int16_t iMvY, int32_t iWidth, int32_t iHeight)
|
||||
{
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user