Add arm64 code for downsample.

zhiliang wang committed 12 years ago

Commit 5cb66bd1 parent f6385466 files · +292 −2Browse files

Showing the changes to codec/processing/src/downsample (3 of 6 files). Show all

Unified · Split

3 files

@@ -85,6 +85,17 @@ void CDownsampling::InitDownsampleFuncs (SDownsampleFuncs& sDownsampleFunc, int · 84 unmodified lines
85	85	sDownsampleFunc.pfGeneralRatioLuma = GeneralBilinearAccurateDownsamplerWrap_neon;
86	86	}
87	87	#endif
	88	+
	89	+#if defined(HAVE_NEON_AARCH64)
	90	+ if (iCpuFlag & WELS_CPU_NEON) {
	91	+ sDownsampleFunc.pfHalfAverage[0] = DyadicBilinearDownsamplerWidthx32_AArch64_neon;
	92	+ sDownsampleFunc.pfHalfAverage[1] = DyadicBilinearDownsampler_AArch64_neon;
	93	+ sDownsampleFunc.pfHalfAverage[2] = DyadicBilinearDownsampler_AArch64_neon;
	94	+ sDownsampleFunc.pfHalfAverage[3] = DyadicBilinearDownsampler_AArch64_neon;
	95	+ sDownsampleFunc.pfGeneralRatioChroma = GeneralBilinearAccurateDownsamplerWrap_AArch64_neon;
	96	+ sDownsampleFunc.pfGeneralRatioLuma = GeneralBilinearAccurateDownsamplerWrap_AArch64_neon;
	97	+ }
	98	+#endif
88	99	}
89	100
90	101	EResult CDownsampling::Process (int32_t iType, SPixMap* pSrcPixMap, SPixMap* pDstPixMap) {

@@ -120,6 +120,21 @@ void GeneralBilinearAccurateDownsampler_neon (uint8_t* pDst, const int32_t kiDst · 119 unmodified lines
120	120	WELSVP_EXTERN_C_END
121	121	#endif
122	122
	123	+#ifdef HAVE_NEON_AARCH64
	124	+WELSVP_EXTERN_C_BEGIN
	125	+// iSrcWidth no limitation
	126	+HalveDownsampleFunc DyadicBilinearDownsampler_AArch64_neon;
	127	+// iSrcWidth = x32 pixels
	128	+HalveDownsampleFunc DyadicBilinearDownsamplerWidthx32_AArch64_neon;
	129	+
	130	+GeneralDownsampleFunc GeneralBilinearAccurateDownsamplerWrap_AArch64_neon;
	131	+
	132	+void GeneralBilinearAccurateDownsampler_AArch64_neon (uint8_t* pDst, const int32_t kiDstStride, const int32_t kiDstWidth, const int32_t kiDstHeight,
	133	+ uint8_t* pSrc, const int32_t kiSrcStride, const uint32_t kuiScaleX, const uint32_t kuiScaleY);
	134	+
	135	+WELSVP_EXTERN_C_END
	136	+#endif
	137	+
123	138
124	139	class CDownsampling : public IStrategy {
125	140	public:

@@ -241,4 +241,15 @@ void GeneralBilinearAccurateDownsamplerWrap_neon (uint8_t* pDst, const int32_t k · 240 unmodified lines
241	241	uiScaley);
242	242	}
243	243	#endif
	244	+
	245	+#ifdef HAVE_NEON_AARCH64
	246	+void GeneralBilinearAccurateDownsamplerWrap_AArch64_neon (uint8_t* pDst, const int32_t kiDstStride, const int32_t kiDstWidth, const int32_t kiDstHeight,
	247	+ uint8_t* pSrc, const int32_t kiSrcStride, const int32_t kiSrcWidth, const int32_t kiSrcHeight) {
	248	+ const int32_t kiScaleBit = 15;
	249	+ const uint32_t kuiScale = (1 << kiScaleBit);
	250	+ uint32_t uiScalex = WELS_ROUND ((float)kiSrcWidth / (float)kiDstWidth * kuiScale);
	251	+ uint32_t uiScaley = WELS_ROUND ((float)kiSrcHeight / (float)kiDstHeight * kuiScale);
	252	+ GeneralBilinearAccurateDownsampler_AArch64_neon (pDst, kiDstStride, kiDstWidth, kiDstHeight, pSrc, kiSrcStride, uiScalex, uiScaley);
	253	+}
	254	+#endif
244	255	WELSVP_NAMESPACE_END

Keyboard shortcuts