generic-library/openh264
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Move the qpel function table into the luma function itself

Browse Source 
This unifies the luma MC interface to match the decoder side.
This commit is contained in:
Martin Storsjö 2015-01-27 11:47:48 +02:00
parent a8262dd9e6
commit 1cf68cad13
5 changed files with 51 additions and 47 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
codec/encoder/core/inc/wels_func_ptr_def.h
View File

@ -89,7 +89,7 @@ typedef struct TagMcFunc {
PWelsLumaHalfpelMcFunc pfLumaHalfpelCen;
PWelsMcFunc pfChromaMc;
PWelsLumaQuarpelMcFunc pfLumaQuarpelMc[16];
PWelsMcFunc pfLumaMc;
PWelsSampleAveragingFunc pfSampleAveraging[2];
} SMcFunc;

79
codec/encoder/core/src/mc.cpp
View File

@ -395,6 +395,17 @@ void McChroma_c (const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t
}
}
}
void McLuma_c (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) {
static const PWelsLumaQuarpelMcFunc pWelsMcFuncWidthEq16[16] = { //[y*4+x]
McCopyWidthEq16_c, McHorVer10WidthEq16_c, McHorVer20WidthEq16_c, McHorVer30WidthEq16_c,
McHorVer01WidthEq16_c, McHorVer11WidthEq16_c, McHorVer21WidthEq16_c, McHorVer31WidthEq16_c,
McHorVer02WidthEq16_c, McHorVer12WidthEq16_c, McHorVer22WidthEq16_c, McHorVer32WidthEq16_c,
McHorVer03WidthEq16_c, McHorVer13WidthEq16_c, McHorVer23WidthEq16_c, McHorVer33WidthEq16_c
};
uint8_t uiMvpIdx = ((iMvY & 0x03) << 2) + (iMvX & 0x03);
pWelsMcFuncWidthEq16[uiMvpIdx] (pSrc, iSrcStride, pDst, iDstStride, iHeight);
}
//***************************************************************************//
// MMXEXT and SSE2 implementation //
//***************************************************************************//
@ -575,6 +586,17 @@ void McChroma_ssse3 (const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int
}
void McLuma_sse2 (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) {
static const PWelsLumaQuarpelMcFunc pWelsMcFuncWidthEq16_sse2[16] = {
McCopyWidthEq16_sse2, McHorVer10WidthEq16_sse2, McHorVer20WidthEq16_sse2, McHorVer30WidthEq16_sse2,
McHorVer01WidthEq16_sse2, McHorVer11WidthEq16_sse2, McHorVer21WidthEq16_sse2, McHorVer31WidthEq16_sse2,
McHorVer02WidthEq16_sse2, McHorVer12WidthEq16_sse2, McHorVer22WidthEq16_sse2, McHorVer32WidthEq16_sse2,
McHorVer03WidthEq16_sse2, McHorVer13WidthEq16_sse2, McHorVer23WidthEq16_sse2, McHorVer33WidthEq16_sse2
};
uint8_t uiMvpIdx = ((iMvY & 0x03) << 2) + (iMvX & 0x03);
pWelsMcFuncWidthEq16_sse2[uiMvpIdx] (pSrc, iSrcStride, pDst, iDstStride, iHeight);
}
#endif //X86_ASM
//***************************************************************************//
@ -666,6 +688,17 @@ void EncMcChroma_neon (const uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, i
McChromaWidthEq4_neon (pSrc, iSrcStride, pDst, iDstStride, (int32_t*) (g_kuiABCD[kiD8y][kiD8x]), iHeight);
}
}
void EncMcLuma_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) {
static const PWelsLumaQuarpelMcFunc pWelsMcFuncWidthEq16_neon[16] = { //[x][y]
McCopyWidthEq16_neon, McHorVer10WidthEq16_neon, McHorVer20WidthEq16_neon, McHorVer30WidthEq16_neon,
McHorVer01WidthEq16_neon, EncMcHorVer11_neon, EncMcHorVer21_neon, EncMcHorVer31_neon,
McHorVer02WidthEq16_neon, EncMcHorVer12_neon, McHorVer22WidthEq16_neon, EncMcHorVer32_neon,
McHorVer03WidthEq16_neon, EncMcHorVer13_neon, EncMcHorVer23_neon, EncMcHorVer33_neon
};
uint8_t uiMvpIdx = ((iMvY & 0x03) << 2) + (iMvX & 0x03);
pWelsMcFuncWidthEq16_neon[uiMvpIdx] (pSrc, iSrcStride, pDst, iDstStride, iHeight);
}
#endif
#if defined(HAVE_NEON_AARCH64)
@ -763,47 +796,28 @@ void EncMcChroma_AArch64_neon (const uint8_t* pSrc, int32_t iSrcStride, uint8_t*
McChromaWidthEq4_AArch64_neon (pSrc, iSrcStride, pDst, iDstStride, (int32_t*) (g_kuiABCD[kiD8y][kiD8x]), iHeight);
}
}
#endif
void WelsInitMcFuncs (SMcFunc* pMcFuncs, uint32_t uiCpuFlag) {
static const PWelsSampleAveragingFunc pfPixAvgFunc[2] = {PixelAvgWidthEq8_c, PixelAvgWidthEq16_c};
static const PWelsLumaQuarpelMcFunc pWelsMcFuncWidthEq16[16] = { //[y*4+x]
McCopyWidthEq16_c, McHorVer10WidthEq16_c, McHorVer20WidthEq16_c, McHorVer30WidthEq16_c,
McHorVer01WidthEq16_c, McHorVer11WidthEq16_c, McHorVer21WidthEq16_c, McHorVer31WidthEq16_c,
McHorVer02WidthEq16_c, McHorVer12WidthEq16_c, McHorVer22WidthEq16_c, McHorVer32WidthEq16_c,
McHorVer03WidthEq16_c, McHorVer13WidthEq16_c, McHorVer23WidthEq16_c, McHorVer33WidthEq16_c
};
#if defined (X86_ASM)
static const PWelsLumaQuarpelMcFunc pWelsMcFuncWidthEq16_sse2[16] = {
McCopyWidthEq16_sse2, McHorVer10WidthEq16_sse2, McHorVer20WidthEq16_sse2, McHorVer30WidthEq16_sse2,
McHorVer01WidthEq16_sse2, McHorVer11WidthEq16_sse2, McHorVer21WidthEq16_sse2, McHorVer31WidthEq16_sse2,
McHorVer02WidthEq16_sse2, McHorVer12WidthEq16_sse2, McHorVer22WidthEq16_sse2, McHorVer32WidthEq16_sse2,
McHorVer03WidthEq16_sse2, McHorVer13WidthEq16_sse2, McHorVer23WidthEq16_sse2, McHorVer33WidthEq16_sse2
};
#endif
#if defined(HAVE_NEON)
static const PWelsLumaQuarpelMcFunc pWelsMcFuncWidthEq16_neon[16] = { //[x][y]
McCopyWidthEq16_neon, McHorVer10WidthEq16_neon, McHorVer20WidthEq16_neon, McHorVer30WidthEq16_neon,
McHorVer01WidthEq16_neon, EncMcHorVer11_neon, EncMcHorVer21_neon, EncMcHorVer31_neon,
McHorVer02WidthEq16_neon, EncMcHorVer12_neon, McHorVer22WidthEq16_neon, EncMcHorVer32_neon,
McHorVer03WidthEq16_neon, EncMcHorVer13_neon, EncMcHorVer23_neon, EncMcHorVer33_neon
};
#endif
#if defined(HAVE_NEON_AARCH64)
void EncMcLuma_AArch64_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) {
static const PWelsLumaQuarpelMcFunc pWelsMcFuncWidthEq16_AArch64_neon[16] = { //[x][y]
McCopyWidthEq16_AArch64_neon, McHorVer10WidthEq16_AArch64_neon, McHorVer20WidthEq16_AArch64_neon, McHorVer30WidthEq16_AArch64_neon,
McHorVer01WidthEq16_AArch64_neon, EncMcHorVer11_AArch64_neon, EncMcHorVer21_AArch64_neon, EncMcHorVer31_AArch64_neon,
McHorVer02WidthEq16_AArch64_neon, EncMcHorVer12_AArch64_neon, McHorVer22WidthEq16_AArch64_neon, EncMcHorVer32_AArch64_neon,
McHorVer03WidthEq16_AArch64_neon, EncMcHorVer13_AArch64_neon, EncMcHorVer23_AArch64_neon, EncMcHorVer33_AArch64_neon
};
uint8_t uiMvpIdx = ((iMvY & 0x03) << 2) + (iMvX & 0x03);
pWelsMcFuncWidthEq16_AArch64_neon[uiMvpIdx] (pSrc, iSrcStride, pDst, iDstStride, iHeight);
}
#endif
void WelsInitMcFuncs (SMcFunc* pMcFuncs, uint32_t uiCpuFlag) {
static const PWelsSampleAveragingFunc pfPixAvgFunc[2] = {PixelAvgWidthEq8_c, PixelAvgWidthEq16_c};
pMcFuncs->pfLumaHalfpelHor = McHorVer20_c;
pMcFuncs->pfLumaHalfpelVer = McHorVer02_c;
pMcFuncs->pfLumaHalfpelCen = McHorVer22_c;
memcpy (pMcFuncs->pfSampleAveraging, pfPixAvgFunc, sizeof (pfPixAvgFunc));
pMcFuncs->pfChromaMc = McChroma_c;
memcpy (pMcFuncs->pfLumaQuarpelMc, pWelsMcFuncWidthEq16, sizeof (pWelsMcFuncWidthEq16));
pMcFuncs->pfLumaMc = McLuma_c;
#if defined (X86_ASM)
if (uiCpuFlag & WELS_CPU_SSE2) {
pMcFuncs->pfLumaHalfpelHor = McHorVer20Width9Or17_sse2;
@ -812,7 +826,7 @@ void WelsInitMcFuncs (SMcFunc* pMcFuncs, uint32_t uiCpuFlag) {
pMcFuncs->pfSampleAveraging[0] = PixelAvgWidthEq8_mmx;
pMcFuncs->pfSampleAveraging[1] = PixelAvgWidthEq16_sse2;
pMcFuncs->pfChromaMc = McChroma_sse2;
memcpy (pMcFuncs->pfLumaQuarpelMc, pWelsMcFuncWidthEq16_sse2, sizeof (pWelsMcFuncWidthEq16_sse2));
pMcFuncs->pfLumaMc = McLuma_sse2;
}
if (uiCpuFlag & WELS_CPU_SSSE3) {
@ -823,7 +837,7 @@ void WelsInitMcFuncs (SMcFunc* pMcFuncs, uint32_t uiCpuFlag) {
#if defined(HAVE_NEON)
if (uiCpuFlag & WELS_CPU_NEON) {
memcpy (pMcFuncs->pfLumaQuarpelMc, pWelsMcFuncWidthEq16_neon, sizeof (pWelsMcFuncWidthEq16_neon));
pMcFuncs->pfLumaMc = EncMcLuma_neon;
pMcFuncs->pfChromaMc = EncMcChroma_neon;
pMcFuncs->pfSampleAveraging[0] = PixStrideAvgWidthEq8_neon;
pMcFuncs->pfSampleAveraging[1] = PixStrideAvgWidthEq16_neon;
@ -834,8 +848,7 @@ void WelsInitMcFuncs (SMcFunc* pMcFuncs, uint32_t uiCpuFlag) {
#endif
#if defined(HAVE_NEON_AARCH64)
if (uiCpuFlag & WELS_CPU_NEON) {
memcpy (pMcFuncs->pfLumaQuarpelMc, pWelsMcFuncWidthEq16_AArch64_neon,
sizeof (pWelsMcFuncWidthEq16_AArch64_neon));
pMcFuncs->pfLumaMc = EncMcLuma_AArch64_neon;
pMcFuncs->pfChromaMc = EncMcChroma_AArch64_neon;
pMcFuncs->pfSampleAveraging[0] = PixStrideAvgWidthEq8_AArch64_neon;
pMcFuncs->pfSampleAveraging[1] = PixStrideAvgWidthEq16_AArch64_neon;

6
codec/encoder/core/src/svc_base_layer_md.cpp
View File

@ -1247,7 +1247,7 @@ void WelsMdBackgroundMbEnc (sWelsEncCtx* pEncCtx, SWelsMD* pWelsMd, SMB* pCurMb,
pDstCr = pMbCache->pMemPredChroma + 64;
}
//MC
pFunc->sMcFuncs.pfLumaQuarpelMc[0] (pRefLuma, iLineSizeY, pDstLuma, 16, 16);
pFunc->sMcFuncs.pfLumaMc (pRefLuma, iLineSizeY, pDstLuma, 16, 0, 0, 16, 16);
pFunc->sMcFuncs.pfChromaMc (pRefCb, iLineSizeUV, pDstCb, 8, sMvp.iMvX, sMvp.iMvY, 8, 8); //Cb
pFunc->sMcFuncs.pfChromaMc (pRefCr, iLineSizeUV, pDstCr, 8, sMvp.iMvX, sMvp.iMvY, 8, 8); //Cr
@ -1313,7 +1313,6 @@ bool WelsMdPSkipEnc (sWelsEncCtx* pEncCtx, SWelsMD* pWelsMd, SMB* pCurMb, SMbCac
uint8_t* pDstCr = pMbCache->pSkipMb + 256 + 64;
SMVUnitXY sMvp = { 0 };
uint8_t uiMvpIdx;
int32_t n;
int32_t iEncStride = pCurLayer->iEncStride[0];
@ -1343,8 +1342,7 @@ bool WelsMdPSkipEnc (sWelsEncCtx* pEncCtx, SWelsMD* pWelsMd, SMB* pCurMb, SMbCac
//luma
pRefLuma += sQpelMvp.iMvY * iLineSizeY + sQpelMvp.iMvX;
uiMvpIdx = ((sMvp.iMvY & 0x03) << 2) + (sMvp.iMvX & 0x03);
pFunc->sMcFuncs.pfLumaQuarpelMc[uiMvpIdx] (pRefLuma, iLineSizeY, pDstLuma, 16, 16);
pFunc->sMcFuncs.pfLumaMc (pRefLuma, iLineSizeY, pDstLuma, 16, sMvp.iMvX, sMvp.iMvY, 16, 16);
iSadCostLuma = pFunc->sSampleDealingFuncs.pfSampleSad[BLOCK_16x16] (pMbCache->SPicData.pEncMb[0],
pCurLayer->iEncStride[0], pDstLuma, 16);

2
codec/encoder/core/src/svc_mode_decision.cpp
View File

@ -414,7 +414,7 @@ void SvcMdSCDMbEnc (sWelsEncCtx* pEncCtx, SWelsMD* pWelsMd, SMB* pCurMb, SMbCach
pDstCr = pMbCache->pMemPredChroma + 64;
}
//MC
pFunc->sMcFuncs.pfLumaQuarpelMc[0] (pRefLuma + iOffsetY, iLineSizeY, pDstLuma, 16, 16);
pFunc->sMcFuncs.pfLumaMc (pRefLuma + iOffsetY, iLineSizeY, pDstLuma, 16, 0, 0, 16, 16);
pFunc->sMcFuncs.pfChromaMc (pRefCb + iOffsetUV, iLineSizeUV, pDstCb, 8, sMvp.iMvX, sMvp.iMvY, 8, 8);
pFunc->sMcFuncs.pfChromaMc (pRefCr + iOffsetUV, iLineSizeUV, pDstCr, 8, sMvp.iMvX, sMvp.iMvY, 8, 8);

9
test/encoder/EncUT_MotionCompensation.cpp
View File

@ -4,17 +4,10 @@
#include "cpu.h"
using namespace WelsEnc;
static void McLumaFunc (SMcFunc* pFuncs, 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) {
uint8_t uiMvpIdx = ((iMvY & 0x03) << 2) + (iMvX & 0x03);
ASSERT_EQ (iWidth, 16);
pFuncs->pfLumaQuarpelMc[uiMvpIdx] (pSrc, iSrcStride, pDst, iDstStride, iHeight);
}
#define InitMcFunc WelsInitMcFuncs
#define LUMA_FUNC(funcs, src, srcstride, dst, dststride, mvx, mvy, width, height) \
McLumaFunc (funcs, src, srcstride, dst, dststride, mvx, mvy, width, height)
sMcFunc.pfLumaMc (src, srcstride, dst, dststride, mvx, mvy, width, height)
#define CHROMA_FUNC sMcFunc.pfChromaMc