diff options
Diffstat (limited to 'include/libyuv/row.h')
| -rw-r--r-- | include/libyuv/row.h | 99 |
1 files changed, 78 insertions, 21 deletions
diff --git a/include/libyuv/row.h b/include/libyuv/row.h index 5b244d77..0455b4cc 100644 --- a/include/libyuv/row.h +++ b/include/libyuv/row.h @@ -31,7 +31,10 @@ extern "C" { #endif // MemorySanitizer does not support assembly code yet. http://crbug.com/344505 #if defined(__has_feature) -#if __has_feature(memory_sanitizer) +#if __has_feature(memory_sanitizer) && !defined(LIBYUV_DISABLE_NEON) +#define LIBYUV_DISABLE_NEON +#endif +#if __has_feature(memory_sanitizer) && !defined(LIBYUV_DISABLE_X86) #define LIBYUV_DISABLE_X86 #endif #endif @@ -161,7 +164,6 @@ extern "C" { #define HAS_ARGBSEPIAROW_SSSE3 #define HAS_ARGBSHADEROW_SSE2 #define HAS_ARGBSUBTRACTROW_SSE2 -#define HAS_ARGBUNATTENUATEROW_SSE2 #define HAS_BLENDPLANEROW_SSSE3 #define HAS_COMPUTECUMULATIVESUMROW_SSE2 #define HAS_CUMULATIVESUMTOAVERAGEROW_SSE2 @@ -171,9 +173,6 @@ extern "C" { #define HAS_SOBELXROW_SSE2 #define HAS_SOBELXYROW_SSE2 #define HAS_SOBELYROW_SSE2 -#if !defined(LIBYUV_BIT_EXACT) -#define HAS_ARGBATTENUATEROW_SSSE3 -#endif // The following functions fail on gcc/clang 32 bit with fpic and framepointer. // caveat: clangcl uses row_win.cc which works. @@ -241,11 +240,7 @@ extern "C" { #define HAS_ARGBADDROW_AVX2 #define HAS_ARGBMULTIPLYROW_AVX2 #define HAS_ARGBSUBTRACTROW_AVX2 -#define HAS_ARGBUNATTENUATEROW_AVX2 #define HAS_BLENDPLANEROW_AVX2 -#if !defined(LIBYUV_BIT_EXACT) -#define HAS_ARGBATTENUATEROW_AVX2 -#endif #if defined(__x86_64__) || !defined(__pic__) || defined(__clang__) || \ defined(_MSC_VER) @@ -285,14 +280,15 @@ extern "C" { #define HAS_ABGRTOAR30ROW_SSSE3 #define HAS_ABGRTOYJROW_SSSE3 #define HAS_AR64TOARGBROW_SSSE3 +#define HAS_ARGBATTENUATEROW_SSSE3 #define HAS_ARGBTOAB64ROW_SSSE3 #define HAS_ARGBTOAR30ROW_SSSE3 #define HAS_ARGBTOAR64ROW_SSSE3 +#define HAS_ARGBUNATTENUATEROW_SSE2 #define HAS_CONVERT16TO8ROW_SSSE3 #define HAS_CONVERT8TO16ROW_SSE2 -#define HAS_DETILEROW_SSE2 #define HAS_DETILEROW_16_SSE2 -#define HAS_DETILEROW_16_AVX +#define HAS_DETILEROW_SSE2 #define HAS_DETILESPLITUVROW_SSSE3 #define HAS_DETILETOYUY2_SSE2 #define HAS_HALFMERGEUVROW_SSSE3 @@ -345,13 +341,16 @@ extern "C" { #define HAS_ABGRTOYJROW_AVX2 #define HAS_ABGRTOYROW_AVX2 #define HAS_AR64TOARGBROW_AVX2 +#define HAS_ARGBATTENUATEROW_AVX2 #define HAS_ARGBTOAB64ROW_AVX2 #define HAS_ARGBTOAR30ROW_AVX2 #define HAS_ARGBTOAR64ROW_AVX2 #define HAS_ARGBTORAWROW_AVX2 #define HAS_ARGBTORGB24ROW_AVX2 +#define HAS_ARGBUNATTENUATEROW_AVX2 #define HAS_CONVERT16TO8ROW_AVX2 #define HAS_CONVERT8TO16ROW_AVX2 +#define HAS_DETILEROW_16_AVX #define HAS_DIVIDEROW_16_AVX2 #define HAS_HALFMERGEUVROW_AVX2 #define HAS_I210TOAR30ROW_AVX2 @@ -795,19 +794,25 @@ extern "C" { #endif #if !defined(LIBYUV_DISABLE_RVV) && defined(__riscv_vector) +#define HAS_COPYROW_RVV +#if __riscv_v_intrinsic == 11000 #define HAS_AB64TOARGBROW_RVV +#define HAS_ABGRTOYJROW_RVV +#define HAS_ABGRTOYROW_RVV #define HAS_AR64TOARGBROW_RVV #define HAS_ARGBATTENUATEROW_RVV +#define HAS_ARGBBLENDROW_RVV +#define HAS_ARGBCOPYYTOALPHAROW_RVV +#define HAS_ARGBEXTRACTALPHAROW_RVV #define HAS_ARGBTOAB64ROW_RVV #define HAS_ARGBTOAR64ROW_RVV #define HAS_ARGBTORAWROW_RVV #define HAS_ARGBTORGB24ROW_RVV -#define HAS_ARGBTOYROW_RVV #define HAS_ARGBTOYJROW_RVV -#define HAS_ABGRTOYROW_RVV -#define HAS_ABGRTOYJROW_RVV +#define HAS_ARGBTOYMATRIXROW_RVV +#define HAS_ARGBTOYROW_RVV #define HAS_BGRATOYROW_RVV -#define HAS_COPYROW_RVV +#define HAS_BLENDPLANEROW_RVV #define HAS_I400TOARGBROW_RVV #define HAS_I422ALPHATOARGBROW_RVV #define HAS_I422TOARGBROW_RVV @@ -822,10 +827,10 @@ extern "C" { #define HAS_MERGERGBROW_RVV #define HAS_MERGEUVROW_RVV #define HAS_MERGEXRGBROW_RVV -#define HAS_SPLITARGBROW_RVV -#define HAS_SPLITRGBROW_RVV -#define HAS_SPLITUVROW_RVV -#define HAS_SPLITXRGBROW_RVV +#define HAS_NV12TOARGBROW_RVV +#define HAS_NV12TORGB24ROW_RVV +#define HAS_NV21TOARGBROW_RVV +#define HAS_NV21TORGB24ROW_RVV #define HAS_RAWTOARGBROW_RVV #define HAS_RAWTORGB24ROW_RVV #define HAS_RAWTORGBAROW_RVV @@ -834,8 +839,15 @@ extern "C" { #define HAS_RGB24TOARGBROW_RVV #define HAS_RGB24TOYJROW_RVV #define HAS_RGB24TOYROW_RVV -#define HAS_RGBATOYROW_RVV #define HAS_RGBATOYJROW_RVV +#define HAS_RGBATOYMATRIXROW_RVV +#define HAS_RGBATOYROW_RVV +#define HAS_RGBTOYMATRIXROW_RVV +#define HAS_SPLITARGBROW_RVV +#define HAS_SPLITRGBROW_RVV +#define HAS_SPLITUVROW_RVV +#define HAS_SPLITXRGBROW_RVV +#endif #endif #if defined(_MSC_VER) && !defined(__CLR_VER) && !defined(__clang__) @@ -1351,6 +1363,26 @@ void UYVYToARGBRow_LSX(const uint8_t* src_uyvy, uint8_t* dst_argb, const struct YuvConstants* yuvconstants, int width); +void NV12ToARGBRow_RVV(const uint8_t* src_y, + const uint8_t* src_uv, + uint8_t* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV21ToARGBRow_RVV(const uint8_t* src_y, + const uint8_t* src_vu, + uint8_t* dst_argb, + const struct YuvConstants* yuvconstants, + int width); +void NV12ToRGB24Row_RVV(const uint8_t* src_y, + const uint8_t* src_uv, + uint8_t* dst_rgb24, + const struct YuvConstants* yuvconstants, + int width); +void NV21ToRGB24Row_RVV(const uint8_t* src_y, + const uint8_t* src_vu, + uint8_t* dst_rgb24, + const struct YuvConstants* yuvconstants, + int width); void ARGBToYRow_AVX2(const uint8_t* src_argb, uint8_t* dst_y, int width); void ARGBToYRow_Any_AVX2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width); @@ -3081,6 +3113,9 @@ void ARGBExtractAlphaRow_MSA(const uint8_t* src_argb, void ARGBExtractAlphaRow_LSX(const uint8_t* src_argb, uint8_t* dst_a, int width); +void ARGBExtractAlphaRow_RVV(const uint8_t* src_argb, + uint8_t* dst_a, + int width); void ARGBExtractAlphaRow_Any_SSE2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width); @@ -3100,6 +3135,7 @@ void ARGBExtractAlphaRow_Any_LSX(const uint8_t* src_ptr, void ARGBCopyYToAlphaRow_C(const uint8_t* src, uint8_t* dst, int width); void ARGBCopyYToAlphaRow_SSE2(const uint8_t* src, uint8_t* dst, int width); void ARGBCopyYToAlphaRow_AVX2(const uint8_t* src, uint8_t* dst, int width); +void ARGBCopyYToAlphaRow_RVV(const uint8_t* src, uint8_t* dst, int width); void ARGBCopyYToAlphaRow_Any_SSE2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width); @@ -4515,6 +4551,10 @@ void ARGBBlendRow_LSX(const uint8_t* src_argb0, const uint8_t* src_argb1, uint8_t* dst_argb, int width); +void ARGBBlendRow_RVV(const uint8_t* src_argb0, + const uint8_t* src_argb1, + uint8_t* dst_argb, + int width); void ARGBBlendRow_C(const uint8_t* src_argb, const uint8_t* src_argb1, uint8_t* dst_argb, @@ -4541,6 +4581,11 @@ void BlendPlaneRow_Any_AVX2(const uint8_t* y_buf, const uint8_t* v_buf, uint8_t* dst_ptr, int width); +void BlendPlaneRow_RVV(const uint8_t* src0, + const uint8_t* src1, + const uint8_t* alpha, + uint8_t* dst, + int width); void BlendPlaneRow_C(const uint8_t* src0, const uint8_t* src1, const uint8_t* alpha, @@ -6180,7 +6225,19 @@ void ByteToFloatRow_Any_NEON(const uint8_t* src_ptr, float* dst_ptr, float param, int width); - +// Convert FP16 Half Floats to FP32 Floats +void ConvertFP16ToFP32Row_NEON(const uint16_t* src, // fp16 + float* dst, + int width); +// Convert a column of FP16 Half Floats to a row of FP32 Floats +void ConvertFP16ToFP32Column_NEON(const uint16_t* src, // fp16 + int src_stride, // stride in elements + float* dst, + int width); +// Convert FP32 Floats to FP16 Half Floats +void ConvertFP32ToFP16Row_NEON(const float* src, + uint16_t* dst, // fp16 + int width); void ARGBLumaColorTableRow_C(const uint8_t* src_argb, uint8_t* dst_argb, int width, |