add option for enable/disable AVX2

Guangwei Wang committed 10 years ago

Commit 7d00e8b1 parent 898073114 files · +56 −1Browse files

Showing the changes to codec/processing/src (5 of 14 files). Show all

Unified · Split

5 files

@@ -107,10 +107,12 @@ void CDownsampling::InitDownsampleFuncs (SDownsampleFuncs& sDownsampleFunc, int · 106 unmodified lines
107	107	sDownsampleFunc.pfQuarterDownsampler = DyadicBilinearQuarterDownsampler_sse4;
108	108	sDownsampleFunc.pfGeneralRatioChroma = GeneralBilinearAccurateDownsamplerWrap_sse41;
109	109	}
	110	+#ifdef HAVE_AVX2
110	111	if (iCpuFlag & WELS_CPU_AVX2) {
111	112	sDownsampleFunc.pfGeneralRatioChroma = GeneralBilinearAccurateDownsamplerWrap_avx2;
112	113	sDownsampleFunc.pfGeneralRatioLuma = GeneralBilinearFastDownsamplerWrap_avx2;
113	114	}
	115	+#endif
114	116	#endif//X86_ASM
115	117
116	118	#if defined(HAVE_NEON)

@@ -99,8 +99,10 @@ GeneralDownsampleFunc GeneralBilinearFastDownsamplerWrap_sse2; · 98 unmodified lines
99	99	GeneralDownsampleFunc GeneralBilinearAccurateDownsamplerWrap_sse2;
100	100	GeneralDownsampleFunc GeneralBilinearFastDownsamplerWrap_ssse3;
101	101	GeneralDownsampleFunc GeneralBilinearAccurateDownsamplerWrap_sse41;
	102	+#ifdef HAVE_AVX2
102	103	GeneralDownsampleFunc GeneralBilinearFastDownsamplerWrap_avx2;
103	104	GeneralDownsampleFunc GeneralBilinearAccurateDownsamplerWrap_avx2;
	105	+#endif
104	106
105	107	SpecificDownsampleFunc DyadicBilinearOneThirdDownsampler_ssse3;
106	108	SpecificDownsampleFunc DyadicBilinearOneThirdDownsampler_sse4;
@@ -120,12 +122,14 @@ void GeneralBilinearFastDownsampler_ssse3 (uint8_t* pDst, int32_t iDstStride, in · 13 unmodified lines
120	122	void GeneralBilinearAccurateDownsampler_sse41 (uint8_t* pDst, int32_t iDstStride, int32_t iDstWidth,
121	123	int32_t iDstHeight, uint8_t* pSrc, int32_t iSrcStride, uint32_t uiScaleX,
122	124	uint32_t uiScaleY);
	125	+#ifdef HAVE_AVX2
123	126	void GeneralBilinearFastDownsampler_avx2 (uint8_t* pDst, int32_t iDstStride, int32_t iDstWidth,
124	127	int32_t iDstHeight, uint8_t* pSrc, int32_t iSrcStride, uint32_t uiScaleX,
125	128	uint32_t uiScaleY);
126	129	void GeneralBilinearAccurateDownsampler_avx2 (uint8_t* pDst, int32_t iDstStride, int32_t iDstWidth,
127	130	int32_t iDstHeight, uint8_t* pSrc, int32_t iSrcStride, uint32_t uiScaleX,
128	131	uint32_t uiScaleY);
	132	+#endif
129	133
130	134	WELSVP_EXTERN_C_END
131	135	#endif

@@ -284,8 +284,10 @@ DEFINE_GENERAL_BILINEAR_FAST_DOWNSAMPLER_WRAP (sse2) · 283 unmodified lines
284	284	DEFINE_GENERAL_BILINEAR_ACCURATE_DOWNSAMPLER_WRAP (sse2)
285	285	DEFINE_GENERAL_BILINEAR_FAST_DOWNSAMPLER_WRAP (ssse3)
286	286	DEFINE_GENERAL_BILINEAR_ACCURATE_DOWNSAMPLER_WRAP (sse41)
	287	+#ifdef HAVE_AVX2
287	288	DEFINE_GENERAL_BILINEAR_FAST_DOWNSAMPLER_WRAP (avx2)
288	289	DEFINE_GENERAL_BILINEAR_ACCURATE_DOWNSAMPLER_WRAP (avx2)
	290	+#endif
289	291	#endif //X86_ASM
290	292
291	293	#ifdef HAVE_NEON

@@ -3254,6 +3254,7 @@ WELS_EXTERN GeneralBilinearAccurateDownsampler_sse41 · 3253 unmodified lines
3254	3254	%undef xmm_xfrac1_begin
3255	3255	%undef xmm_xfrac_inc
3256	3256
	3257	+%ifdef HAVE_AVX2
3257	3258	; xpos_int=%1 xpos_frac=%2 inc_int+1=%3 inc_frac=%4 tmp=%5
3258	3259	%macro AVX2_BilinearIncXposuw 5
3259	3260	vpaddusw %5, %2, %4
@@ -4552,3 +4553,5 @@ WELS_EXTERN GeneralBilinearAccurateDownsampler_avx2 · 1292 unmodified lines
4552	4553	%undef ymm_xfrac0_begin
4553	4554	%undef ymm_xfrac1_begin
4554	4555	%undef ymm_xfrac_inc
	4556	+%endif
	4557	+

@@ -2088,6 +2088,7 @@ sqdiff_bgd_width_loop: · 2087 unmodified lines
2088	2088	%assign push_num push_num - stack_alloc_num
2089	2089	%endmacro
2090	2090
	2091	+%ifdef HAVE_AVX2
2091	2092	; Max unsigned byte per quadword
2092	2093	; out=%1 in=%2 tmp=%3
2093	2094	%macro AVX2_Maxubq 3
@@ -3557,3 +3558,6 @@ WELS_EXTERN VAACalcSadSsdBgd_avx2 · 1463 unmodified lines
3557	3558	%undef p_sd8x8
3558	3559	%undef p_mad8x8
3559	3560	ret
	3561	+
	3562	+%endif
	3563	+

Keyboard shortcuts