diff options
Diffstat (limited to 'media/libyuv/source/scale.cc')
| -rw-r--r-- | media/libyuv/source/scale.cc | 1887 |
1 files changed, 1397 insertions, 490 deletions
diff --git a/media/libyuv/source/scale.cc b/media/libyuv/source/scale.cc index 36e3fe5281..830754ce64 100644 --- a/media/libyuv/source/scale.cc +++ b/media/libyuv/source/scale.cc @@ -17,6 +17,7 @@ #include "libyuv/planar_functions.h" // For CopyPlane #include "libyuv/row.h" #include "libyuv/scale_row.h" +#include "libyuv/scale_uv.h" // For UVScale #ifdef __cplusplus namespace libyuv { @@ -28,22 +29,31 @@ static __inline int Abs(int v) { } #define SUBSAMPLE(v, a, s) (v < 0) ? (-((-v + a) >> s)) : ((v + a) >> s) +#define CENTERSTART(dx, s) (dx < 0) ? -((-dx >> 1) + s) : ((dx >> 1) + s) // Scale plane, 1/2 // This is an optimized version for scaling down a plane to 1/2 of // its original size. -static void ScalePlaneDown2(int src_width, int src_height, - int dst_width, int dst_height, - int src_stride, int dst_stride, - const uint8* src_ptr, uint8* dst_ptr, +static void ScalePlaneDown2(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint8_t* src_ptr, + uint8_t* dst_ptr, enum FilterMode filtering) { int y; - void (*ScaleRowDown2)(const uint8* src_ptr, ptrdiff_t src_stride, - uint8* dst_ptr, int dst_width) = - filtering == kFilterNone ? ScaleRowDown2_C : - (filtering == kFilterLinear ? ScaleRowDown2Linear_C : ScaleRowDown2Box_C); - int row_stride = src_stride << 1; + void (*ScaleRowDown2)(const uint8_t* src_ptr, ptrdiff_t src_stride, + uint8_t* dst_ptr, int dst_width) = + filtering == kFilterNone + ? ScaleRowDown2_C + : (filtering == kFilterLinear ? ScaleRowDown2Linear_C + : ScaleRowDown2Box_C); + int row_stride = src_stride * 2; + (void)src_width; + (void)src_height; if (!filtering) { src_ptr += src_stride; // Point to odd rows. src_stride = 0; @@ -51,46 +61,78 @@ static void ScalePlaneDown2(int src_width, int src_height, #if defined(HAS_SCALEROWDOWN2_NEON) if (TestCpuFlag(kCpuHasNEON)) { - ScaleRowDown2 = filtering == kFilterNone ? ScaleRowDown2_Any_NEON : - (filtering == kFilterLinear ? ScaleRowDown2Linear_Any_NEON : - ScaleRowDown2Box_Any_NEON); + ScaleRowDown2 = + filtering == kFilterNone + ? ScaleRowDown2_Any_NEON + : (filtering == kFilterLinear ? ScaleRowDown2Linear_Any_NEON + : ScaleRowDown2Box_Any_NEON); if (IS_ALIGNED(dst_width, 16)) { - ScaleRowDown2 = filtering == kFilterNone ? ScaleRowDown2_NEON : - (filtering == kFilterLinear ? ScaleRowDown2Linear_NEON : - ScaleRowDown2Box_NEON); + ScaleRowDown2 = filtering == kFilterNone ? ScaleRowDown2_NEON + : (filtering == kFilterLinear + ? ScaleRowDown2Linear_NEON + : ScaleRowDown2Box_NEON); } } #endif #if defined(HAS_SCALEROWDOWN2_SSSE3) if (TestCpuFlag(kCpuHasSSSE3)) { - ScaleRowDown2 = filtering == kFilterNone ? ScaleRowDown2_Any_SSSE3 : - (filtering == kFilterLinear ? ScaleRowDown2Linear_Any_SSSE3 : - ScaleRowDown2Box_Any_SSSE3); + ScaleRowDown2 = + filtering == kFilterNone + ? ScaleRowDown2_Any_SSSE3 + : (filtering == kFilterLinear ? ScaleRowDown2Linear_Any_SSSE3 + : ScaleRowDown2Box_Any_SSSE3); if (IS_ALIGNED(dst_width, 16)) { - ScaleRowDown2 = filtering == kFilterNone ? ScaleRowDown2_SSSE3 : - (filtering == kFilterLinear ? ScaleRowDown2Linear_SSSE3 : - ScaleRowDown2Box_SSSE3); + ScaleRowDown2 = + filtering == kFilterNone + ? ScaleRowDown2_SSSE3 + : (filtering == kFilterLinear ? ScaleRowDown2Linear_SSSE3 + : ScaleRowDown2Box_SSSE3); } } #endif #if defined(HAS_SCALEROWDOWN2_AVX2) if (TestCpuFlag(kCpuHasAVX2)) { - ScaleRowDown2 = filtering == kFilterNone ? ScaleRowDown2_Any_AVX2 : - (filtering == kFilterLinear ? ScaleRowDown2Linear_Any_AVX2 : - ScaleRowDown2Box_Any_AVX2); + ScaleRowDown2 = + filtering == kFilterNone + ? ScaleRowDown2_Any_AVX2 + : (filtering == kFilterLinear ? ScaleRowDown2Linear_Any_AVX2 + : ScaleRowDown2Box_Any_AVX2); if (IS_ALIGNED(dst_width, 32)) { - ScaleRowDown2 = filtering == kFilterNone ? ScaleRowDown2_AVX2 : - (filtering == kFilterLinear ? ScaleRowDown2Linear_AVX2 : - ScaleRowDown2Box_AVX2); + ScaleRowDown2 = filtering == kFilterNone ? ScaleRowDown2_AVX2 + : (filtering == kFilterLinear + ? ScaleRowDown2Linear_AVX2 + : ScaleRowDown2Box_AVX2); } } #endif -#if defined(HAS_SCALEROWDOWN2_DSPR2) - if (TestCpuFlag(kCpuHasDSPR2) && IS_ALIGNED(src_ptr, 4) && - IS_ALIGNED(src_stride, 4) && IS_ALIGNED(row_stride, 4) && - IS_ALIGNED(dst_ptr, 4) && IS_ALIGNED(dst_stride, 4)) { - ScaleRowDown2 = filtering ? - ScaleRowDown2Box_DSPR2 : ScaleRowDown2_DSPR2; +#if defined(HAS_SCALEROWDOWN2_MSA) + if (TestCpuFlag(kCpuHasMSA)) { + ScaleRowDown2 = + filtering == kFilterNone + ? ScaleRowDown2_Any_MSA + : (filtering == kFilterLinear ? ScaleRowDown2Linear_Any_MSA + : ScaleRowDown2Box_Any_MSA); + if (IS_ALIGNED(dst_width, 32)) { + ScaleRowDown2 = filtering == kFilterNone ? ScaleRowDown2_MSA + : (filtering == kFilterLinear + ? ScaleRowDown2Linear_MSA + : ScaleRowDown2Box_MSA); + } + } +#endif +#if defined(HAS_SCALEROWDOWN2_LSX) + if (TestCpuFlag(kCpuHasLSX)) { + ScaleRowDown2 = + filtering == kFilterNone + ? ScaleRowDown2_Any_LSX + : (filtering == kFilterLinear ? ScaleRowDown2Linear_Any_LSX + : ScaleRowDown2Box_Any_LSX); + if (IS_ALIGNED(dst_width, 32)) { + ScaleRowDown2 = filtering == kFilterNone ? ScaleRowDown2_LSX + : (filtering == kFilterLinear + ? ScaleRowDown2Linear_LSX + : ScaleRowDown2Box_LSX); + } } #endif @@ -105,18 +147,25 @@ static void ScalePlaneDown2(int src_width, int src_height, } } -static void ScalePlaneDown2_16(int src_width, int src_height, - int dst_width, int dst_height, - int src_stride, int dst_stride, - const uint16* src_ptr, uint16* dst_ptr, +static void ScalePlaneDown2_16(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint16_t* src_ptr, + uint16_t* dst_ptr, enum FilterMode filtering) { int y; - void (*ScaleRowDown2)(const uint16* src_ptr, ptrdiff_t src_stride, - uint16* dst_ptr, int dst_width) = - filtering == kFilterNone ? ScaleRowDown2_16_C : - (filtering == kFilterLinear ? ScaleRowDown2Linear_16_C : - ScaleRowDown2Box_16_C); - int row_stride = src_stride << 1; + void (*ScaleRowDown2)(const uint16_t* src_ptr, ptrdiff_t src_stride, + uint16_t* dst_ptr, int dst_width) = + filtering == kFilterNone + ? ScaleRowDown2_16_C + : (filtering == kFilterLinear ? ScaleRowDown2Linear_16_C + : ScaleRowDown2Box_16_C); + int row_stride = src_stride * 2; + (void)src_width; + (void)src_height; if (!filtering) { src_ptr += src_stride; // Point to odd rows. src_stride = 0; @@ -124,23 +173,17 @@ static void ScalePlaneDown2_16(int src_width, int src_height, #if defined(HAS_SCALEROWDOWN2_16_NEON) if (TestCpuFlag(kCpuHasNEON) && IS_ALIGNED(dst_width, 16)) { - ScaleRowDown2 = filtering ? ScaleRowDown2Box_16_NEON : - ScaleRowDown2_16_NEON; + ScaleRowDown2 = + filtering ? ScaleRowDown2Box_16_NEON : ScaleRowDown2_16_NEON; } #endif #if defined(HAS_SCALEROWDOWN2_16_SSE2) if (TestCpuFlag(kCpuHasSSE2) && IS_ALIGNED(dst_width, 16)) { - ScaleRowDown2 = filtering == kFilterNone ? ScaleRowDown2_16_SSE2 : - (filtering == kFilterLinear ? ScaleRowDown2Linear_16_SSE2 : - ScaleRowDown2Box_16_SSE2); - } -#endif -#if defined(HAS_SCALEROWDOWN2_16_DSPR2) - if (TestCpuFlag(kCpuHasDSPR2) && IS_ALIGNED(src_ptr, 4) && - IS_ALIGNED(src_stride, 4) && IS_ALIGNED(row_stride, 4) && - IS_ALIGNED(dst_ptr, 4) && IS_ALIGNED(dst_stride, 4)) { - ScaleRowDown2 = filtering ? - ScaleRowDown2Box_16_DSPR2 : ScaleRowDown2_16_DSPR2; + ScaleRowDown2 = + filtering == kFilterNone + ? ScaleRowDown2_16_SSE2 + : (filtering == kFilterLinear ? ScaleRowDown2Linear_16_SSE2 + : ScaleRowDown2Box_16_SSE2); } #endif @@ -155,28 +198,79 @@ static void ScalePlaneDown2_16(int src_width, int src_height, } } +void ScalePlaneDown2_16To8(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint16_t* src_ptr, + uint8_t* dst_ptr, + int scale, + enum FilterMode filtering) { + int y; + void (*ScaleRowDown2)(const uint16_t* src_ptr, ptrdiff_t src_stride, + uint8_t* dst_ptr, int dst_width, int scale) = + (src_width & 1) + ? (filtering == kFilterNone + ? ScaleRowDown2_16To8_Odd_C + : (filtering == kFilterLinear ? ScaleRowDown2Linear_16To8_Odd_C + : ScaleRowDown2Box_16To8_Odd_C)) + : (filtering == kFilterNone + ? ScaleRowDown2_16To8_C + : (filtering == kFilterLinear ? ScaleRowDown2Linear_16To8_C + : ScaleRowDown2Box_16To8_C)); + int row_stride = src_stride * 2; + (void)dst_height; + if (!filtering) { + src_ptr += src_stride; // Point to odd rows. + src_stride = 0; + } + + if (filtering == kFilterLinear) { + src_stride = 0; + } + for (y = 0; y < src_height / 2; ++y) { + ScaleRowDown2(src_ptr, src_stride, dst_ptr, dst_width, scale); + src_ptr += row_stride; + dst_ptr += dst_stride; + } + if (src_height & 1) { + if (!filtering) { + src_ptr -= src_stride; // Point to last row. + } + ScaleRowDown2(src_ptr, 0, dst_ptr, dst_width, scale); + } +} + // Scale plane, 1/4 // This is an optimized version for scaling down a plane to 1/4 of // its original size. -static void ScalePlaneDown4(int src_width, int src_height, - int dst_width, int dst_height, - int src_stride, int dst_stride, - const uint8* src_ptr, uint8* dst_ptr, +static void ScalePlaneDown4(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint8_t* src_ptr, + uint8_t* dst_ptr, enum FilterMode filtering) { int y; - void (*ScaleRowDown4)(const uint8* src_ptr, ptrdiff_t src_stride, - uint8* dst_ptr, int dst_width) = + void (*ScaleRowDown4)(const uint8_t* src_ptr, ptrdiff_t src_stride, + uint8_t* dst_ptr, int dst_width) = filtering ? ScaleRowDown4Box_C : ScaleRowDown4_C; - int row_stride = src_stride << 2; + int row_stride = src_stride * 4; + (void)src_width; + (void)src_height; if (!filtering) { src_ptr += src_stride * 2; // Point to row 2. src_stride = 0; } #if defined(HAS_SCALEROWDOWN4_NEON) if (TestCpuFlag(kCpuHasNEON)) { - ScaleRowDown4 = filtering ? - ScaleRowDown4Box_Any_NEON : ScaleRowDown4_Any_NEON; + ScaleRowDown4 = + filtering ? ScaleRowDown4Box_Any_NEON : ScaleRowDown4_Any_NEON; if (IS_ALIGNED(dst_width, 8)) { ScaleRowDown4 = filtering ? ScaleRowDown4Box_NEON : ScaleRowDown4_NEON; } @@ -184,8 +278,8 @@ static void ScalePlaneDown4(int src_width, int src_height, #endif #if defined(HAS_SCALEROWDOWN4_SSSE3) if (TestCpuFlag(kCpuHasSSSE3)) { - ScaleRowDown4 = filtering ? - ScaleRowDown4Box_Any_SSSE3 : ScaleRowDown4_Any_SSSE3; + ScaleRowDown4 = + filtering ? ScaleRowDown4Box_Any_SSSE3 : ScaleRowDown4_Any_SSSE3; if (IS_ALIGNED(dst_width, 8)) { ScaleRowDown4 = filtering ? ScaleRowDown4Box_SSSE3 : ScaleRowDown4_SSSE3; } @@ -193,19 +287,29 @@ static void ScalePlaneDown4(int src_width, int src_height, #endif #if defined(HAS_SCALEROWDOWN4_AVX2) if (TestCpuFlag(kCpuHasAVX2)) { - ScaleRowDown4 = filtering ? - ScaleRowDown4Box_Any_AVX2 : ScaleRowDown4_Any_AVX2; + ScaleRowDown4 = + filtering ? ScaleRowDown4Box_Any_AVX2 : ScaleRowDown4_Any_AVX2; if (IS_ALIGNED(dst_width, 16)) { ScaleRowDown4 = filtering ? ScaleRowDown4Box_AVX2 : ScaleRowDown4_AVX2; } } #endif -#if defined(HAS_SCALEROWDOWN4_DSPR2) - if (TestCpuFlag(kCpuHasDSPR2) && IS_ALIGNED(row_stride, 4) && - IS_ALIGNED(src_ptr, 4) && IS_ALIGNED(src_stride, 4) && - IS_ALIGNED(dst_ptr, 4) && IS_ALIGNED(dst_stride, 4)) { - ScaleRowDown4 = filtering ? - ScaleRowDown4Box_DSPR2 : ScaleRowDown4_DSPR2; +#if defined(HAS_SCALEROWDOWN4_MSA) + if (TestCpuFlag(kCpuHasMSA)) { + ScaleRowDown4 = + filtering ? ScaleRowDown4Box_Any_MSA : ScaleRowDown4_Any_MSA; + if (IS_ALIGNED(dst_width, 16)) { + ScaleRowDown4 = filtering ? ScaleRowDown4Box_MSA : ScaleRowDown4_MSA; + } + } +#endif +#if defined(HAS_SCALEROWDOWN4_LSX) + if (TestCpuFlag(kCpuHasLSX)) { + ScaleRowDown4 = + filtering ? ScaleRowDown4Box_Any_LSX : ScaleRowDown4_Any_LSX; + if (IS_ALIGNED(dst_width, 16)) { + ScaleRowDown4 = filtering ? ScaleRowDown4Box_LSX : ScaleRowDown4_LSX; + } } #endif @@ -219,38 +323,36 @@ static void ScalePlaneDown4(int src_width, int src_height, } } -static void ScalePlaneDown4_16(int src_width, int src_height, - int dst_width, int dst_height, - int src_stride, int dst_stride, - const uint16* src_ptr, uint16* dst_ptr, +static void ScalePlaneDown4_16(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint16_t* src_ptr, + uint16_t* dst_ptr, enum FilterMode filtering) { int y; - void (*ScaleRowDown4)(const uint16* src_ptr, ptrdiff_t src_stride, - uint16* dst_ptr, int dst_width) = + void (*ScaleRowDown4)(const uint16_t* src_ptr, ptrdiff_t src_stride, + uint16_t* dst_ptr, int dst_width) = filtering ? ScaleRowDown4Box_16_C : ScaleRowDown4_16_C; - int row_stride = src_stride << 2; + int row_stride = src_stride * 4; + (void)src_width; + (void)src_height; if (!filtering) { src_ptr += src_stride * 2; // Point to row 2. src_stride = 0; } #if defined(HAS_SCALEROWDOWN4_16_NEON) if (TestCpuFlag(kCpuHasNEON) && IS_ALIGNED(dst_width, 8)) { - ScaleRowDown4 = filtering ? ScaleRowDown4Box_16_NEON : - ScaleRowDown4_16_NEON; + ScaleRowDown4 = + filtering ? ScaleRowDown4Box_16_NEON : ScaleRowDown4_16_NEON; } #endif #if defined(HAS_SCALEROWDOWN4_16_SSE2) if (TestCpuFlag(kCpuHasSSE2) && IS_ALIGNED(dst_width, 8)) { - ScaleRowDown4 = filtering ? ScaleRowDown4Box_16_SSE2 : - ScaleRowDown4_16_SSE2; - } -#endif -#if defined(HAS_SCALEROWDOWN4_16_DSPR2) - if (TestCpuFlag(kCpuHasDSPR2) && IS_ALIGNED(row_stride, 4) && - IS_ALIGNED(src_ptr, 4) && IS_ALIGNED(src_stride, 4) && - IS_ALIGNED(dst_ptr, 4) && IS_ALIGNED(dst_stride, 4)) { - ScaleRowDown4 = filtering ? - ScaleRowDown4Box_16_DSPR2 : ScaleRowDown4_16_DSPR2; + ScaleRowDown4 = + filtering ? ScaleRowDown4Box_16_SSE2 : ScaleRowDown4_16_SSE2; } #endif @@ -265,18 +367,23 @@ static void ScalePlaneDown4_16(int src_width, int src_height, } // Scale plane down, 3/4 - -static void ScalePlaneDown34(int src_width, int src_height, - int dst_width, int dst_height, - int src_stride, int dst_stride, - const uint8* src_ptr, uint8* dst_ptr, +static void ScalePlaneDown34(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint8_t* src_ptr, + uint8_t* dst_ptr, enum FilterMode filtering) { int y; - void (*ScaleRowDown34_0)(const uint8* src_ptr, ptrdiff_t src_stride, - uint8* dst_ptr, int dst_width); - void (*ScaleRowDown34_1)(const uint8* src_ptr, ptrdiff_t src_stride, - uint8* dst_ptr, int dst_width); + void (*ScaleRowDown34_0)(const uint8_t* src_ptr, ptrdiff_t src_stride, + uint8_t* dst_ptr, int dst_width); + void (*ScaleRowDown34_1)(const uint8_t* src_ptr, ptrdiff_t src_stride, + uint8_t* dst_ptr, int dst_width); const int filter_stride = (filtering == kFilterLinear) ? 0 : src_stride; + (void)src_width; + (void)src_height; assert(dst_width % 3 == 0); if (!filtering) { ScaleRowDown34_0 = ScaleRowDown34_C; @@ -305,6 +412,46 @@ static void ScalePlaneDown34(int src_width, int src_height, } } #endif +#if defined(HAS_SCALEROWDOWN34_MSA) + if (TestCpuFlag(kCpuHasMSA)) { + if (!filtering) { + ScaleRowDown34_0 = ScaleRowDown34_Any_MSA; + ScaleRowDown34_1 = ScaleRowDown34_Any_MSA; + } else { + ScaleRowDown34_0 = ScaleRowDown34_0_Box_Any_MSA; + ScaleRowDown34_1 = ScaleRowDown34_1_Box_Any_MSA; + } + if (dst_width % 48 == 0) { + if (!filtering) { + ScaleRowDown34_0 = ScaleRowDown34_MSA; + ScaleRowDown34_1 = ScaleRowDown34_MSA; + } else { + ScaleRowDown34_0 = ScaleRowDown34_0_Box_MSA; + ScaleRowDown34_1 = ScaleRowDown34_1_Box_MSA; + } + } + } +#endif +#if defined(HAS_SCALEROWDOWN34_LSX) + if (TestCpuFlag(kCpuHasLSX)) { + if (!filtering) { + ScaleRowDown34_0 = ScaleRowDown34_Any_LSX; + ScaleRowDown34_1 = ScaleRowDown34_Any_LSX; + } else { + ScaleRowDown34_0 = ScaleRowDown34_0_Box_Any_LSX; + ScaleRowDown34_1 = ScaleRowDown34_1_Box_Any_LSX; + } + if (dst_width % 48 == 0) { + if (!filtering) { + ScaleRowDown34_0 = ScaleRowDown34_LSX; + ScaleRowDown34_1 = ScaleRowDown34_LSX; + } else { + ScaleRowDown34_0 = ScaleRowDown34_0_Box_LSX; + ScaleRowDown34_1 = ScaleRowDown34_1_Box_LSX; + } + } + } +#endif #if defined(HAS_SCALEROWDOWN34_SSSE3) if (TestCpuFlag(kCpuHasSSSE3)) { if (!filtering) { @@ -325,19 +472,6 @@ static void ScalePlaneDown34(int src_width, int src_height, } } #endif -#if defined(HAS_SCALEROWDOWN34_DSPR2) - if (TestCpuFlag(kCpuHasDSPR2) && (dst_width % 24 == 0) && - IS_ALIGNED(src_ptr, 4) && IS_ALIGNED(src_stride, 4) && - IS_ALIGNED(dst_ptr, 4) && IS_ALIGNED(dst_stride, 4)) { - if (!filtering) { - ScaleRowDown34_0 = ScaleRowDown34_DSPR2; - ScaleRowDown34_1 = ScaleRowDown34_DSPR2; - } else { - ScaleRowDown34_0 = ScaleRowDown34_0_Box_DSPR2; - ScaleRowDown34_1 = ScaleRowDown34_1_Box_DSPR2; - } - } -#endif for (y = 0; y < dst_height - 2; y += 3) { ScaleRowDown34_0(src_ptr, filter_stride, dst_ptr, dst_width); @@ -346,8 +480,7 @@ static void ScalePlaneDown34(int src_width, int src_height, ScaleRowDown34_1(src_ptr, filter_stride, dst_ptr, dst_width); src_ptr += src_stride; dst_ptr += dst_stride; - ScaleRowDown34_0(src_ptr + src_stride, -filter_stride, - dst_ptr, dst_width); + ScaleRowDown34_0(src_ptr + src_stride, -filter_stride, dst_ptr, dst_width); src_ptr += src_stride * 2; dst_ptr += dst_stride; } @@ -363,17 +496,23 @@ static void ScalePlaneDown34(int src_width, int src_height, } } -static void ScalePlaneDown34_16(int src_width, int src_height, - int dst_width, int dst_height, - int src_stride, int dst_stride, - const uint16* src_ptr, uint16* dst_ptr, +static void ScalePlaneDown34_16(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint16_t* src_ptr, + uint16_t* dst_ptr, enum FilterMode filtering) { int y; - void (*ScaleRowDown34_0)(const uint16* src_ptr, ptrdiff_t src_stride, - uint16* dst_ptr, int dst_width); - void (*ScaleRowDown34_1)(const uint16* src_ptr, ptrdiff_t src_stride, - uint16* dst_ptr, int dst_width); + void (*ScaleRowDown34_0)(const uint16_t* src_ptr, ptrdiff_t src_stride, + uint16_t* dst_ptr, int dst_width); + void (*ScaleRowDown34_1)(const uint16_t* src_ptr, ptrdiff_t src_stride, + uint16_t* dst_ptr, int dst_width); const int filter_stride = (filtering == kFilterLinear) ? 0 : src_stride; + (void)src_width; + (void)src_height; assert(dst_width % 3 == 0); if (!filtering) { ScaleRowDown34_0 = ScaleRowDown34_16_C; @@ -404,19 +543,6 @@ static void ScalePlaneDown34_16(int src_width, int src_height, } } #endif -#if defined(HAS_SCALEROWDOWN34_16_DSPR2) - if (TestCpuFlag(kCpuHasDSPR2) && (dst_width % 24 == 0) && - IS_ALIGNED(src_ptr, 4) && IS_ALIGNED(src_stride, 4) && - IS_ALIGNED(dst_ptr, 4) && IS_ALIGNED(dst_stride, 4)) { - if (!filtering) { - ScaleRowDown34_0 = ScaleRowDown34_16_DSPR2; - ScaleRowDown34_1 = ScaleRowDown34_16_DSPR2; - } else { - ScaleRowDown34_0 = ScaleRowDown34_0_Box_16_DSPR2; - ScaleRowDown34_1 = ScaleRowDown34_1_Box_16_DSPR2; - } - } -#endif for (y = 0; y < dst_height - 2; y += 3) { ScaleRowDown34_0(src_ptr, filter_stride, dst_ptr, dst_width); @@ -425,8 +551,7 @@ static void ScalePlaneDown34_16(int src_width, int src_height, ScaleRowDown34_1(src_ptr, filter_stride, dst_ptr, dst_width); src_ptr += src_stride; dst_ptr += dst_stride; - ScaleRowDown34_0(src_ptr + src_stride, -filter_stride, - dst_ptr, dst_width); + ScaleRowDown34_0(src_ptr + src_stride, -filter_stride, dst_ptr, dst_width); src_ptr += src_stride * 2; dst_ptr += dst_stride; } @@ -442,7 +567,6 @@ static void ScalePlaneDown34_16(int src_width, int src_height, } } - // Scale plane, 3/8 // This is an optimized version for scaling down a plane to 3/8 // of its original size. @@ -458,18 +582,24 @@ static void ScalePlaneDown34_16(int src_width, int src_height, // ggghhhii // Boxes are 3x3, 2x3, 3x2 and 2x2 -static void ScalePlaneDown38(int src_width, int src_height, - int dst_width, int dst_height, - int src_stride, int dst_stride, - const uint8* src_ptr, uint8* dst_ptr, +static void ScalePlaneDown38(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint8_t* src_ptr, + uint8_t* dst_ptr, enum FilterMode filtering) { int y; - void (*ScaleRowDown38_3)(const uint8* src_ptr, ptrdiff_t src_stride, - uint8* dst_ptr, int dst_width); - void (*ScaleRowDown38_2)(const uint8* src_ptr, ptrdiff_t src_stride, - uint8* dst_ptr, int dst_width); + void (*ScaleRowDown38_3)(const uint8_t* src_ptr, ptrdiff_t src_stride, + uint8_t* dst_ptr, int dst_width); + void (*ScaleRowDown38_2)(const uint8_t* src_ptr, ptrdiff_t src_stride, + uint8_t* dst_ptr, int dst_width); const int filter_stride = (filtering == kFilterLinear) ? 0 : src_stride; assert(dst_width % 3 == 0); + (void)src_width; + (void)src_height; if (!filtering) { ScaleRowDown38_3 = ScaleRowDown38_C; ScaleRowDown38_2 = ScaleRowDown38_C; @@ -517,16 +647,43 @@ static void ScalePlaneDown38(int src_width, int src_height, } } #endif -#if defined(HAS_SCALEROWDOWN38_DSPR2) - if (TestCpuFlag(kCpuHasDSPR2) && (dst_width % 12 == 0) && - IS_ALIGNED(src_ptr, 4) && IS_ALIGNED(src_stride, 4) && - IS_ALIGNED(dst_ptr, 4) && IS_ALIGNED(dst_stride, 4)) { +#if defined(HAS_SCALEROWDOWN38_MSA) + if (TestCpuFlag(kCpuHasMSA)) { + if (!filtering) { + ScaleRowDown38_3 = ScaleRowDown38_Any_MSA; + ScaleRowDown38_2 = ScaleRowDown38_Any_MSA; + } else { + ScaleRowDown38_3 = ScaleRowDown38_3_Box_Any_MSA; + ScaleRowDown38_2 = ScaleRowDown38_2_Box_Any_MSA; + } + if (dst_width % 12 == 0) { + if (!filtering) { + ScaleRowDown38_3 = ScaleRowDown38_MSA; + ScaleRowDown38_2 = ScaleRowDown38_MSA; + } else { + ScaleRowDown38_3 = ScaleRowDown38_3_Box_MSA; + ScaleRowDown38_2 = ScaleRowDown38_2_Box_MSA; + } + } + } +#endif +#if defined(HAS_SCALEROWDOWN38_LSX) + if (TestCpuFlag(kCpuHasLSX)) { if (!filtering) { - ScaleRowDown38_3 = ScaleRowDown38_DSPR2; - ScaleRowDown38_2 = ScaleRowDown38_DSPR2; + ScaleRowDown38_3 = ScaleRowDown38_Any_LSX; + ScaleRowDown38_2 = ScaleRowDown38_Any_LSX; } else { - ScaleRowDown38_3 = ScaleRowDown38_3_Box_DSPR2; - ScaleRowDown38_2 = ScaleRowDown38_2_Box_DSPR2; + ScaleRowDown38_3 = ScaleRowDown38_3_Box_Any_LSX; + ScaleRowDown38_2 = ScaleRowDown38_2_Box_Any_LSX; + } + if (dst_width % 12 == 0) { + if (!filtering) { + ScaleRowDown38_3 = ScaleRowDown38_LSX; + ScaleRowDown38_2 = ScaleRowDown38_LSX; + } else { + ScaleRowDown38_3 = ScaleRowDown38_3_Box_LSX; + ScaleRowDown38_2 = ScaleRowDown38_2_Box_LSX; + } } } #endif @@ -554,17 +711,23 @@ static void ScalePlaneDown38(int src_width, int src_height, } } -static void ScalePlaneDown38_16(int src_width, int src_height, - int dst_width, int dst_height, - int src_stride, int dst_stride, - const uint16* src_ptr, uint16* dst_ptr, +static void ScalePlaneDown38_16(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint16_t* src_ptr, + uint16_t* dst_ptr, enum FilterMode filtering) { int y; - void (*ScaleRowDown38_3)(const uint16* src_ptr, ptrdiff_t src_stride, - uint16* dst_ptr, int dst_width); - void (*ScaleRowDown38_2)(const uint16* src_ptr, ptrdiff_t src_stride, - uint16* dst_ptr, int dst_width); + void (*ScaleRowDown38_3)(const uint16_t* src_ptr, ptrdiff_t src_stride, + uint16_t* dst_ptr, int dst_width); + void (*ScaleRowDown38_2)(const uint16_t* src_ptr, ptrdiff_t src_stride, + uint16_t* dst_ptr, int dst_width); const int filter_stride = (filtering == kFilterLinear) ? 0 : src_stride; + (void)src_width; + (void)src_height; assert(dst_width % 3 == 0); if (!filtering) { ScaleRowDown38_3 = ScaleRowDown38_16_C; @@ -595,19 +758,6 @@ static void ScalePlaneDown38_16(int src_width, int src_height, } } #endif -#if defined(HAS_SCALEROWDOWN38_16_DSPR2) - if (TestCpuFlag(kCpuHasDSPR2) && (dst_width % 12 == 0) && - IS_ALIGNED(src_ptr, 4) && IS_ALIGNED(src_stride, 4) && - IS_ALIGNED(dst_ptr, 4) && IS_ALIGNED(dst_stride, 4)) { - if (!filtering) { - ScaleRowDown38_3 = ScaleRowDown38_16_DSPR2; - ScaleRowDown38_2 = ScaleRowDown38_16_DSPR2; - } else { - ScaleRowDown38_3 = ScaleRowDown38_3_Box_16_DSPR2; - ScaleRowDown38_2 = ScaleRowDown38_2_Box_16_DSPR2; - } - } -#endif for (y = 0; y < dst_height - 2; y += 3) { ScaleRowDown38_3(src_ptr, filter_stride, dst_ptr, dst_width); @@ -634,8 +784,8 @@ static void ScalePlaneDown38_16(int src_width, int src_height, #define MIN1(x) ((x) < 1 ? 1 : (x)) -static __inline uint32 SumPixels(int iboxwidth, const uint16* src_ptr) { - uint32 sum = 0u; +static __inline uint32_t SumPixels(int iboxwidth, const uint16_t* src_ptr) { + uint32_t sum = 0u; int x; assert(iboxwidth > 0); for (x = 0; x < iboxwidth; ++x) { @@ -644,8 +794,8 @@ static __inline uint32 SumPixels(int iboxwidth, const uint16* src_ptr) { return sum; } -static __inline uint32 SumPixels_16(int iboxwidth, const uint32* src_ptr) { - uint32 sum = 0u; +static __inline uint32_t SumPixels_16(int iboxwidth, const uint32_t* src_ptr) { + uint32_t sum = 0u; int x; assert(iboxwidth > 0); for (x = 0; x < iboxwidth; ++x) { @@ -654,8 +804,12 @@ static __inline uint32 SumPixels_16(int iboxwidth, const uint32* src_ptr) { return sum; } -static void ScaleAddCols2_C(int dst_width, int boxheight, int x, int dx, - const uint16* src_ptr, uint8* dst_ptr) { +static void ScaleAddCols2_C(int dst_width, + int boxheight, + int x, + int dx, + const uint16_t* src_ptr, + uint8_t* dst_ptr) { int i; int scaletbl[2]; int minboxwidth = dx >> 16; @@ -666,13 +820,18 @@ static void ScaleAddCols2_C(int dst_width, int boxheight, int x, int dx, int ix = x >> 16; x += dx; boxwidth = MIN1((x >> 16) - ix); - *dst_ptr++ = SumPixels(boxwidth, src_ptr + ix) * - scaletbl[boxwidth - minboxwidth] >> 16; + *dst_ptr++ = (uint8_t)(SumPixels(boxwidth, src_ptr + ix) * + scaletbl[boxwidth - minboxwidth] >> + 16); } } -static void ScaleAddCols2_16_C(int dst_width, int boxheight, int x, int dx, - const uint32* src_ptr, uint16* dst_ptr) { +static void ScaleAddCols2_16_C(int dst_width, + int boxheight, + int x, + int dx, + const uint32_t* src_ptr, + uint16_t* dst_ptr) { int i; int scaletbl[2]; int minboxwidth = dx >> 16; @@ -684,34 +843,48 @@ static void ScaleAddCols2_16_C(int dst_width, int boxheight, int x, int dx, x += dx; boxwidth = MIN1((x >> 16) - ix); *dst_ptr++ = SumPixels_16(boxwidth, src_ptr + ix) * - scaletbl[boxwidth - minboxwidth] >> 16; + scaletbl[boxwidth - minboxwidth] >> + 16; } } -static void ScaleAddCols0_C(int dst_width, int boxheight, int x, int, - const uint16* src_ptr, uint8* dst_ptr) { +static void ScaleAddCols0_C(int dst_width, + int boxheight, + int x, + int dx, + const uint16_t* src_ptr, + uint8_t* dst_ptr) { int scaleval = 65536 / boxheight; int i; + (void)dx; src_ptr += (x >> 16); for (i = 0; i < dst_width; ++i) { - *dst_ptr++ = src_ptr[i] * scaleval >> 16; + *dst_ptr++ = (uint8_t)(src_ptr[i] * scaleval >> 16); } } -static void ScaleAddCols1_C(int dst_width, int boxheight, int x, int dx, - const uint16* src_ptr, uint8* dst_ptr) { +static void ScaleAddCols1_C(int dst_width, + int boxheight, + int x, + int dx, + const uint16_t* src_ptr, + uint8_t* dst_ptr) { int boxwidth = MIN1(dx >> 16); int scaleval = 65536 / (boxwidth * boxheight); int i; x >>= 16; for (i = 0; i < dst_width; ++i) { - *dst_ptr++ = SumPixels(boxwidth, src_ptr + x) * scaleval >> 16; + *dst_ptr++ = (uint8_t)(SumPixels(boxwidth, src_ptr + x) * scaleval >> 16); x += boxwidth; } } -static void ScaleAddCols1_16_C(int dst_width, int boxheight, int x, int dx, - const uint32* src_ptr, uint16* dst_ptr) { +static void ScaleAddCols1_16_C(int dst_width, + int boxheight, + int x, + int dx, + const uint32_t* src_ptr, + uint16_t* dst_ptr) { int boxwidth = MIN1(dx >> 16); int scaleval = 65536 / (boxwidth * boxheight); int i; @@ -728,10 +901,14 @@ static void ScaleAddCols1_16_C(int dst_width, int boxheight, int x, int dx, // one pixel of destination using fixed point (16.16) to step // through source, sampling a box of pixel with simple // averaging. -static void ScalePlaneBox(int src_width, int src_height, - int dst_width, int dst_height, - int src_stride, int dst_stride, - const uint8* src_ptr, uint8* dst_ptr) { +static void ScalePlaneBox(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint8_t* src_ptr, + uint8_t* dst_ptr) { int j, k; // Initial source x/y coordinate and step values as 16.16 fixed point. int x = 0; @@ -739,18 +916,18 @@ static void ScalePlaneBox(int src_width, int src_height, int dx = 0; int dy = 0; const int max_y = (src_height << 16); - ScaleSlope(src_width, src_height, dst_width, dst_height, kFilterBox, - &x, &y, &dx, &dy); + ScaleSlope(src_width, src_height, dst_width, dst_height, kFilterBox, &x, &y, + &dx, &dy); src_width = Abs(src_width); { - // Allocate a row buffer of uint16. + // Allocate a row buffer of uint16_t. align_buffer_64(row16, src_width * 2); void (*ScaleAddCols)(int dst_width, int boxheight, int x, int dx, - const uint16* src_ptr, uint8* dst_ptr) = - (dx & 0xffff) ? ScaleAddCols2_C: - ((dx != 0x10000) ? ScaleAddCols1_C : ScaleAddCols0_C); - void (*ScaleAddRow)(const uint8* src_ptr, uint16* dst_ptr, int src_width) = - ScaleAddRow_C; + const uint16_t* src_ptr, uint8_t* dst_ptr) = + (dx & 0xffff) ? ScaleAddCols2_C + : ((dx != 0x10000) ? ScaleAddCols1_C : ScaleAddCols0_C); + void (*ScaleAddRow)(const uint8_t* src_ptr, uint16_t* dst_ptr, + int src_width) = ScaleAddRow_C; #if defined(HAS_SCALEADDROW_SSE2) if (TestCpuFlag(kCpuHasSSE2)) { ScaleAddRow = ScaleAddRow_Any_SSE2; @@ -775,11 +952,27 @@ static void ScalePlaneBox(int src_width, int src_height, } } #endif +#if defined(HAS_SCALEADDROW_MSA) + if (TestCpuFlag(kCpuHasMSA)) { + ScaleAddRow = ScaleAddRow_Any_MSA; + if (IS_ALIGNED(src_width, 16)) { + ScaleAddRow = ScaleAddRow_MSA; + } + } +#endif +#if defined(HAS_SCALEADDROW_LSX) + if (TestCpuFlag(kCpuHasLSX)) { + ScaleAddRow = ScaleAddRow_Any_LSX; + if (IS_ALIGNED(src_width, 16)) { + ScaleAddRow = ScaleAddRow_LSX; + } + } +#endif for (j = 0; j < dst_height; ++j) { int boxheight; int iy = y >> 16; - const uint8* src = src_ptr + iy * src_stride; + const uint8_t* src = src_ptr + iy * (int64_t)src_stride; y += dy; if (y > max_y) { y = max_y; @@ -787,20 +980,24 @@ static void ScalePlaneBox(int src_width, int src_height, boxheight = MIN1((y >> 16) - iy); memset(row16, 0, src_width * 2); for (k = 0; k < boxheight; ++k) { - ScaleAddRow(src, (uint16 *)(row16), src_width); + ScaleAddRow(src, (uint16_t*)(row16), src_width); src += src_stride; } - ScaleAddCols(dst_width, boxheight, x, dx, (uint16*)(row16), dst_ptr); + ScaleAddCols(dst_width, boxheight, x, dx, (uint16_t*)(row16), dst_ptr); dst_ptr += dst_stride; } free_aligned_buffer_64(row16); } } -static void ScalePlaneBox_16(int src_width, int src_height, - int dst_width, int dst_height, - int src_stride, int dst_stride, - const uint16* src_ptr, uint16* dst_ptr) { +static void ScalePlaneBox_16(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint16_t* src_ptr, + uint16_t* dst_ptr) { int j, k; // Initial source x/y coordinate and step values as 16.16 fixed point. int x = 0; @@ -808,17 +1005,17 @@ static void ScalePlaneBox_16(int src_width, int src_height, int dx = 0; int dy = 0; const int max_y = (src_height << 16); - ScaleSlope(src_width, src_height, dst_width, dst_height, kFilterBox, - &x, &y, &dx, &dy); + ScaleSlope(src_width, src_height, dst_width, dst_height, kFilterBox, &x, &y, + &dx, &dy); src_width = Abs(src_width); { - // Allocate a row buffer of uint32. + // Allocate a row buffer of uint32_t. align_buffer_64(row32, src_width * 4); void (*ScaleAddCols)(int dst_width, int boxheight, int x, int dx, - const uint32* src_ptr, uint16* dst_ptr) = - (dx & 0xffff) ? ScaleAddCols2_16_C: ScaleAddCols1_16_C; - void (*ScaleAddRow)(const uint16* src_ptr, uint32* dst_ptr, int src_width) = - ScaleAddRow_16_C; + const uint32_t* src_ptr, uint16_t* dst_ptr) = + (dx & 0xffff) ? ScaleAddCols2_16_C : ScaleAddCols1_16_C; + void (*ScaleAddRow)(const uint16_t* src_ptr, uint32_t* dst_ptr, + int src_width) = ScaleAddRow_16_C; #if defined(HAS_SCALEADDROW_16_SSE2) if (TestCpuFlag(kCpuHasSSE2) && IS_ALIGNED(src_width, 16)) { @@ -829,7 +1026,7 @@ static void ScalePlaneBox_16(int src_width, int src_height, for (j = 0; j < dst_height; ++j) { int boxheight; int iy = y >> 16; - const uint16* src = src_ptr + iy * src_stride; + const uint16_t* src = src_ptr + iy * (int64_t)src_stride; y += dy; if (y > max_y) { y = max_y; @@ -837,10 +1034,10 @@ static void ScalePlaneBox_16(int src_width, int src_height, boxheight = MIN1((y >> 16) - iy); memset(row32, 0, src_width * 4); for (k = 0; k < boxheight; ++k) { - ScaleAddRow(src, (uint32 *)(row32), src_width); + ScaleAddRow(src, (uint32_t*)(row32), src_width); src += src_stride; } - ScaleAddCols(dst_width, boxheight, x, dx, (uint32*)(row32), dst_ptr); + ScaleAddCols(dst_width, boxheight, x, dx, (uint32_t*)(row32), dst_ptr); dst_ptr += dst_stride; } free_aligned_buffer_64(row32); @@ -848,10 +1045,14 @@ static void ScalePlaneBox_16(int src_width, int src_height, } // Scale plane down with bilinear interpolation. -void ScalePlaneBilinearDown(int src_width, int src_height, - int dst_width, int dst_height, - int src_stride, int dst_stride, - const uint8* src_ptr, uint8* dst_ptr, +void ScalePlaneBilinearDown(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint8_t* src_ptr, + uint8_t* dst_ptr, enum FilterMode filtering) { // Initial source x/y coordinate and step values as 16.16 fixed point. int x = 0; @@ -864,14 +1065,14 @@ void ScalePlaneBilinearDown(int src_width, int src_height, const int max_y = (src_height - 1) << 16; int j; - void (*ScaleFilterCols)(uint8* dst_ptr, const uint8* src_ptr, - int dst_width, int x, int dx) = + void (*ScaleFilterCols)(uint8_t* dst_ptr, const uint8_t* src_ptr, + int dst_width, int x, int dx) = (src_width >= 32768) ? ScaleFilterCols64_C : ScaleFilterCols_C; - void (*InterpolateRow)(uint8* dst_ptr, const uint8* src_ptr, - ptrdiff_t src_stride, int dst_width, int source_y_fraction) = - InterpolateRow_C; - ScaleSlope(src_width, src_height, dst_width, dst_height, filtering, - &x, &y, &dx, &dy); + void (*InterpolateRow)(uint8_t* dst_ptr, const uint8_t* src_ptr, + ptrdiff_t src_stride, int dst_width, + int source_y_fraction) = InterpolateRow_C; + ScaleSlope(src_width, src_height, dst_width, dst_height, filtering, &x, &y, + &dx, &dy); src_width = Abs(src_width); #if defined(HAS_INTERPOLATEROW_SSSE3) @@ -898,15 +1099,22 @@ void ScalePlaneBilinearDown(int src_width, int src_height, } } #endif -#if defined(HAS_INTERPOLATEROW_DSPR2) - if (TestCpuFlag(kCpuHasDSPR2)) { - InterpolateRow = InterpolateRow_Any_DSPR2; - if (IS_ALIGNED(src_width, 4)) { - InterpolateRow = InterpolateRow_DSPR2; +#if defined(HAS_INTERPOLATEROW_MSA) + if (TestCpuFlag(kCpuHasMSA)) { + InterpolateRow = InterpolateRow_Any_MSA; + if (IS_ALIGNED(src_width, 32)) { + InterpolateRow = InterpolateRow_MSA; + } + } +#endif +#if defined(HAS_INTERPOLATEROW_LSX) + if (TestCpuFlag(kCpuHasLSX)) { + InterpolateRow = InterpolateRow_Any_LSX; + if (IS_ALIGNED(src_width, 32)) { + InterpolateRow = InterpolateRow_LSX; } } #endif - #if defined(HAS_SCALEFILTERCOLS_SSSE3) if (TestCpuFlag(kCpuHasSSSE3) && src_width < 32768) { @@ -921,13 +1129,29 @@ void ScalePlaneBilinearDown(int src_width, int src_height, } } #endif +#if defined(HAS_SCALEFILTERCOLS_MSA) + if (TestCpuFlag(kCpuHasMSA) && src_width < 32768) { + ScaleFilterCols = ScaleFilterCols_Any_MSA; + if (IS_ALIGNED(dst_width, 16)) { + ScaleFilterCols = ScaleFilterCols_MSA; + } + } +#endif +#if defined(HAS_SCALEFILTERCOLS_LSX) + if (TestCpuFlag(kCpuHasLSX) && src_width < 32768) { + ScaleFilterCols = ScaleFilterCols_Any_LSX; + if (IS_ALIGNED(dst_width, 16)) { + ScaleFilterCols = ScaleFilterCols_LSX; + } + } +#endif if (y > max_y) { y = max_y; } for (j = 0; j < dst_height; ++j) { int yi = y >> 16; - const uint8* src = src_ptr + yi * src_stride; + const uint8_t* src = src_ptr + yi * (int64_t)src_stride; if (filtering == kFilterLinear) { ScaleFilterCols(dst_ptr, src, dst_width, x, dx); } else { @@ -944,10 +1168,14 @@ void ScalePlaneBilinearDown(int src_width, int src_height, free_aligned_buffer_64(row); } -void ScalePlaneBilinearDown_16(int src_width, int src_height, - int dst_width, int dst_height, - int src_stride, int dst_stride, - const uint16* src_ptr, uint16* dst_ptr, +void ScalePlaneBilinearDown_16(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint16_t* src_ptr, + uint16_t* dst_ptr, enum FilterMode filtering) { // Initial source x/y coordinate and step values as 16.16 fixed point. int x = 0; @@ -960,19 +1188,19 @@ void ScalePlaneBilinearDown_16(int src_width, int src_height, const int max_y = (src_height - 1) << 16; int j; - void (*ScaleFilterCols)(uint16* dst_ptr, const uint16* src_ptr, - int dst_width, int x, int dx) = + void (*ScaleFilterCols)(uint16_t* dst_ptr, const uint16_t* src_ptr, + int dst_width, int x, int dx) = (src_width >= 32768) ? ScaleFilterCols64_16_C : ScaleFilterCols_16_C; - void (*InterpolateRow)(uint16* dst_ptr, const uint16* src_ptr, - ptrdiff_t src_stride, int dst_width, int source_y_fraction) = - InterpolateRow_16_C; - ScaleSlope(src_width, src_height, dst_width, dst_height, filtering, - &x, &y, &dx, &dy); + void (*InterpolateRow)(uint16_t* dst_ptr, const uint16_t* src_ptr, + ptrdiff_t src_stride, int dst_width, + int source_y_fraction) = InterpolateRow_16_C; + ScaleSlope(src_width, src_height, dst_width, dst_height, filtering, &x, &y, + &dx, &dy); src_width = Abs(src_width); #if defined(HAS_INTERPOLATEROW_16_SSE2) if (TestCpuFlag(kCpuHasSSE2)) { - InterpolateRow = InterpolateRow_Any_16_SSE2; + InterpolateRow = InterpolateRow_16_Any_SSE2; if (IS_ALIGNED(src_width, 16)) { InterpolateRow = InterpolateRow_16_SSE2; } @@ -980,7 +1208,7 @@ void ScalePlaneBilinearDown_16(int src_width, int src_height, #endif #if defined(HAS_INTERPOLATEROW_16_SSSE3) if (TestCpuFlag(kCpuHasSSSE3)) { - InterpolateRow = InterpolateRow_Any_16_SSSE3; + InterpolateRow = InterpolateRow_16_Any_SSSE3; if (IS_ALIGNED(src_width, 16)) { InterpolateRow = InterpolateRow_16_SSSE3; } @@ -988,7 +1216,7 @@ void ScalePlaneBilinearDown_16(int src_width, int src_height, #endif #if defined(HAS_INTERPOLATEROW_16_AVX2) if (TestCpuFlag(kCpuHasAVX2)) { - InterpolateRow = InterpolateRow_Any_16_AVX2; + InterpolateRow = InterpolateRow_16_Any_AVX2; if (IS_ALIGNED(src_width, 32)) { InterpolateRow = InterpolateRow_16_AVX2; } @@ -996,21 +1224,12 @@ void ScalePlaneBilinearDown_16(int src_width, int src_height, #endif #if defined(HAS_INTERPOLATEROW_16_NEON) if (TestCpuFlag(kCpuHasNEON)) { - InterpolateRow = InterpolateRow_Any_16_NEON; + InterpolateRow = InterpolateRow_16_Any_NEON; if (IS_ALIGNED(src_width, 16)) { InterpolateRow = InterpolateRow_16_NEON; } } #endif -#if defined(HAS_INTERPOLATEROW_16_DSPR2) - if (TestCpuFlag(kCpuHasDSPR2)) { - InterpolateRow = InterpolateRow_Any_16_DSPR2; - if (IS_ALIGNED(src_width, 4)) { - InterpolateRow = InterpolateRow_16_DSPR2; - } - } -#endif - #if defined(HAS_SCALEFILTERCOLS_16_SSSE3) if (TestCpuFlag(kCpuHasSSSE3) && src_width < 32768) { @@ -1023,13 +1242,13 @@ void ScalePlaneBilinearDown_16(int src_width, int src_height, for (j = 0; j < dst_height; ++j) { int yi = y >> 16; - const uint16* src = src_ptr + yi * src_stride; + const uint16_t* src = src_ptr + yi * (int64_t)src_stride; if (filtering == kFilterLinear) { ScaleFilterCols(dst_ptr, src, dst_width, x, dx); } else { int yf = (y >> 8) & 255; - InterpolateRow((uint16*)row, src, src_stride, src_width, yf); - ScaleFilterCols(dst_ptr, (uint16*)row, dst_width, x, dx); + InterpolateRow((uint16_t*)row, src, src_stride, src_width, yf); + ScaleFilterCols(dst_ptr, (uint16_t*)row, dst_width, x, dx); } dst_ptr += dst_stride; y += dy; @@ -1041,10 +1260,14 @@ void ScalePlaneBilinearDown_16(int src_width, int src_height, } // Scale up down with bilinear interpolation. -void ScalePlaneBilinearUp(int src_width, int src_height, - int dst_width, int dst_height, - int src_stride, int dst_stride, - const uint8* src_ptr, uint8* dst_ptr, +void ScalePlaneBilinearUp(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint8_t* src_ptr, + uint8_t* dst_ptr, enum FilterMode filtering) { int j; // Initial source x/y coordinate and step values as 16.16 fixed point. @@ -1053,14 +1276,14 @@ void ScalePlaneBilinearUp(int src_width, int src_height, int dx = 0; int dy = 0; const int max_y = (src_height - 1) << 16; - void (*InterpolateRow)(uint8* dst_ptr, const uint8* src_ptr, - ptrdiff_t src_stride, int dst_width, int source_y_fraction) = - InterpolateRow_C; - void (*ScaleFilterCols)(uint8* dst_ptr, const uint8* src_ptr, - int dst_width, int x, int dx) = + void (*InterpolateRow)(uint8_t* dst_ptr, const uint8_t* src_ptr, + ptrdiff_t src_stride, int dst_width, + int source_y_fraction) = InterpolateRow_C; + void (*ScaleFilterCols)(uint8_t* dst_ptr, const uint8_t* src_ptr, + int dst_width, int x, int dx) = filtering ? ScaleFilterCols_C : ScaleCols_C; - ScaleSlope(src_width, src_height, dst_width, dst_height, filtering, - &x, &y, &dx, &dy); + ScaleSlope(src_width, src_height, dst_width, dst_height, filtering, &x, &y, + &dx, &dy); src_width = Abs(src_width); #if defined(HAS_INTERPOLATEROW_SSSE3) @@ -1087,14 +1310,6 @@ void ScalePlaneBilinearUp(int src_width, int src_height, } } #endif -#if defined(HAS_INTERPOLATEROW_DSPR2) - if (TestCpuFlag(kCpuHasDSPR2)) { - InterpolateRow = InterpolateRow_Any_DSPR2; - if (IS_ALIGNED(dst_width, 4)) { - InterpolateRow = InterpolateRow_DSPR2; - } - } -#endif if (filtering && src_width >= 32768) { ScaleFilterCols = ScaleFilterCols64_C; @@ -1112,6 +1327,22 @@ void ScalePlaneBilinearUp(int src_width, int src_height, } } #endif +#if defined(HAS_SCALEFILTERCOLS_MSA) + if (filtering && TestCpuFlag(kCpuHasMSA) && src_width < 32768) { + ScaleFilterCols = ScaleFilterCols_Any_MSA; + if (IS_ALIGNED(dst_width, 16)) { + ScaleFilterCols = ScaleFilterCols_MSA; + } + } +#endif +#if defined(HAS_SCALEFILTERCOLS_LSX) + if (filtering && TestCpuFlag(kCpuHasLSX) && src_width < 32768) { + ScaleFilterCols = ScaleFilterCols_Any_LSX; + if (IS_ALIGNED(dst_width, 16)) { + ScaleFilterCols = ScaleFilterCols_LSX; + } + } +#endif if (!filtering && src_width * 2 == dst_width && x < 0x8000) { ScaleFilterCols = ScaleColsUp2_C; #if defined(HAS_SCALECOLS_SSE2) @@ -1126,14 +1357,14 @@ void ScalePlaneBilinearUp(int src_width, int src_height, } { int yi = y >> 16; - const uint8* src = src_ptr + yi * src_stride; + const uint8_t* src = src_ptr + yi * (int64_t)src_stride; // Allocate 2 row buffers. - const int kRowSize = (dst_width + 31) & ~31; - align_buffer_64(row, kRowSize * 2); + const int row_size = (dst_width + 31) & ~31; + align_buffer_64(row, row_size * 2); - uint8* rowptr = row; - int rowstride = kRowSize; + uint8_t* rowptr = row; + int rowstride = row_size; int lasty = yi; ScaleFilterCols(rowptr, src, dst_width, x, dx); @@ -1141,7 +1372,9 @@ void ScalePlaneBilinearUp(int src_width, int src_height, src += src_stride; } ScaleFilterCols(rowptr + rowstride, src, dst_width, x, dx); - src += src_stride; + if (src_height > 2) { + src += src_stride; + } for (j = 0; j < dst_height; ++j) { yi = y >> 16; @@ -1149,14 +1382,16 @@ void ScalePlaneBilinearUp(int src_width, int src_height, if (y > max_y) { y = max_y; yi = y >> 16; - src = src_ptr + yi * src_stride; + src = src_ptr + yi * (int64_t)src_stride; } if (yi != lasty) { ScaleFilterCols(rowptr, src, dst_width, x, dx); rowptr += rowstride; rowstride = -rowstride; lasty = yi; - src += src_stride; + if ((y + 65536) < max_y) { + src += src_stride; + } } } if (filtering == kFilterLinear) { @@ -1172,10 +1407,335 @@ void ScalePlaneBilinearUp(int src_width, int src_height, } } -void ScalePlaneBilinearUp_16(int src_width, int src_height, - int dst_width, int dst_height, - int src_stride, int dst_stride, - const uint16* src_ptr, uint16* dst_ptr, +// Scale plane, horizontally up by 2 times. +// Uses linear filter horizontally, nearest vertically. +// This is an optimized version for scaling up a plane to 2 times of +// its original width, using linear interpolation. +// This is used to scale U and V planes of I422 to I444. +void ScalePlaneUp2_Linear(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint8_t* src_ptr, + uint8_t* dst_ptr) { + void (*ScaleRowUp)(const uint8_t* src_ptr, uint8_t* dst_ptr, int dst_width) = + ScaleRowUp2_Linear_Any_C; + int i; + int y; + int dy; + + // This function can only scale up by 2 times horizontally. + assert(src_width == ((dst_width + 1) / 2)); + +#ifdef HAS_SCALEROWUP2_LINEAR_SSE2 + if (TestCpuFlag(kCpuHasSSE2)) { + ScaleRowUp = ScaleRowUp2_Linear_Any_SSE2; + } +#endif + +#ifdef HAS_SCALEROWUP2_LINEAR_SSSE3 + if (TestCpuFlag(kCpuHasSSSE3)) { + ScaleRowUp = ScaleRowUp2_Linear_Any_SSSE3; + } +#endif + +#ifdef HAS_SCALEROWUP2_LINEAR_AVX2 + if (TestCpuFlag(kCpuHasAVX2)) { + ScaleRowUp = ScaleRowUp2_Linear_Any_AVX2; + } +#endif + +#ifdef HAS_SCALEROWUP2_LINEAR_NEON + if (TestCpuFlag(kCpuHasNEON)) { + ScaleRowUp = ScaleRowUp2_Linear_Any_NEON; + } +#endif + + if (dst_height == 1) { + ScaleRowUp(src_ptr + ((src_height - 1) / 2) * (int64_t)src_stride, dst_ptr, + dst_width); + } else { + dy = FixedDiv(src_height - 1, dst_height - 1); + y = (1 << 15) - 1; + for (i = 0; i < dst_height; ++i) { + ScaleRowUp(src_ptr + (y >> 16) * (int64_t)src_stride, dst_ptr, dst_width); + dst_ptr += dst_stride; + y += dy; + } + } +} + +// Scale plane, up by 2 times. +// This is an optimized version for scaling up a plane to 2 times of +// its original size, using bilinear interpolation. +// This is used to scale U and V planes of I420 to I444. +void ScalePlaneUp2_Bilinear(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint8_t* src_ptr, + uint8_t* dst_ptr) { + void (*Scale2RowUp)(const uint8_t* src_ptr, ptrdiff_t src_stride, + uint8_t* dst_ptr, ptrdiff_t dst_stride, int dst_width) = + ScaleRowUp2_Bilinear_Any_C; + int x; + + // This function can only scale up by 2 times. + assert(src_width == ((dst_width + 1) / 2)); + assert(src_height == ((dst_height + 1) / 2)); + +#ifdef HAS_SCALEROWUP2_BILINEAR_SSE2 + if (TestCpuFlag(kCpuHasSSE2)) { + Scale2RowUp = ScaleRowUp2_Bilinear_Any_SSE2; + } +#endif + +#ifdef HAS_SCALEROWUP2_BILINEAR_SSSE3 + if (TestCpuFlag(kCpuHasSSSE3)) { + Scale2RowUp = ScaleRowUp2_Bilinear_Any_SSSE3; + } +#endif + +#ifdef HAS_SCALEROWUP2_BILINEAR_AVX2 + if (TestCpuFlag(kCpuHasAVX2)) { + Scale2RowUp = ScaleRowUp2_Bilinear_Any_AVX2; + } +#endif + +#ifdef HAS_SCALEROWUP2_BILINEAR_NEON + if (TestCpuFlag(kCpuHasNEON)) { + Scale2RowUp = ScaleRowUp2_Bilinear_Any_NEON; + } +#endif + + Scale2RowUp(src_ptr, 0, dst_ptr, 0, dst_width); + dst_ptr += dst_stride; + for (x = 0; x < src_height - 1; ++x) { + Scale2RowUp(src_ptr, src_stride, dst_ptr, dst_stride, dst_width); + src_ptr += src_stride; + // TODO(fbarchard): Test performance of writing one row of destination at a + // time. + dst_ptr += 2 * dst_stride; + } + if (!(dst_height & 1)) { + Scale2RowUp(src_ptr, 0, dst_ptr, 0, dst_width); + } +} + +// Scale at most 14 bit plane, horizontally up by 2 times. +// This is an optimized version for scaling up a plane to 2 times of +// its original width, using linear interpolation. +// stride is in count of uint16_t. +// This is used to scale U and V planes of I210 to I410 and I212 to I412. +void ScalePlaneUp2_12_Linear(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint16_t* src_ptr, + uint16_t* dst_ptr) { + void (*ScaleRowUp)(const uint16_t* src_ptr, uint16_t* dst_ptr, + int dst_width) = ScaleRowUp2_Linear_16_Any_C; + int i; + int y; + int dy; + + // This function can only scale up by 2 times horizontally. + assert(src_width == ((dst_width + 1) / 2)); + +#ifdef HAS_SCALEROWUP2_LINEAR_12_SSSE3 + if (TestCpuFlag(kCpuHasSSSE3)) { + ScaleRowUp = ScaleRowUp2_Linear_12_Any_SSSE3; + } +#endif + +#ifdef HAS_SCALEROWUP2_LINEAR_12_AVX2 + if (TestCpuFlag(kCpuHasAVX2)) { + ScaleRowUp = ScaleRowUp2_Linear_12_Any_AVX2; + } +#endif + +#ifdef HAS_SCALEROWUP2_LINEAR_12_NEON + if (TestCpuFlag(kCpuHasNEON)) { + ScaleRowUp = ScaleRowUp2_Linear_12_Any_NEON; + } +#endif + + if (dst_height == 1) { + ScaleRowUp(src_ptr + ((src_height - 1) / 2) * (int64_t)src_stride, dst_ptr, + dst_width); + } else { + dy = FixedDiv(src_height - 1, dst_height - 1); + y = (1 << 15) - 1; + for (i = 0; i < dst_height; ++i) { + ScaleRowUp(src_ptr + (y >> 16) * (int64_t)src_stride, dst_ptr, dst_width); + dst_ptr += dst_stride; + y += dy; + } + } +} + +// Scale at most 12 bit plane, up by 2 times. +// This is an optimized version for scaling up a plane to 2 times of +// its original size, using bilinear interpolation. +// stride is in count of uint16_t. +// This is used to scale U and V planes of I010 to I410 and I012 to I412. +void ScalePlaneUp2_12_Bilinear(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint16_t* src_ptr, + uint16_t* dst_ptr) { + void (*Scale2RowUp)(const uint16_t* src_ptr, ptrdiff_t src_stride, + uint16_t* dst_ptr, ptrdiff_t dst_stride, int dst_width) = + ScaleRowUp2_Bilinear_16_Any_C; + int x; + + // This function can only scale up by 2 times. + assert(src_width == ((dst_width + 1) / 2)); + assert(src_height == ((dst_height + 1) / 2)); + +#ifdef HAS_SCALEROWUP2_BILINEAR_12_SSSE3 + if (TestCpuFlag(kCpuHasSSSE3)) { + Scale2RowUp = ScaleRowUp2_Bilinear_12_Any_SSSE3; + } +#endif + +#ifdef HAS_SCALEROWUP2_BILINEAR_12_AVX2 + if (TestCpuFlag(kCpuHasAVX2)) { + Scale2RowUp = ScaleRowUp2_Bilinear_12_Any_AVX2; + } +#endif + +#ifdef HAS_SCALEROWUP2_BILINEAR_12_NEON + if (TestCpuFlag(kCpuHasNEON)) { + Scale2RowUp = ScaleRowUp2_Bilinear_12_Any_NEON; + } +#endif + + Scale2RowUp(src_ptr, 0, dst_ptr, 0, dst_width); + dst_ptr += dst_stride; + for (x = 0; x < src_height - 1; ++x) { + Scale2RowUp(src_ptr, src_stride, dst_ptr, dst_stride, dst_width); + src_ptr += src_stride; + dst_ptr += 2 * dst_stride; + } + if (!(dst_height & 1)) { + Scale2RowUp(src_ptr, 0, dst_ptr, 0, dst_width); + } +} + +void ScalePlaneUp2_16_Linear(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint16_t* src_ptr, + uint16_t* dst_ptr) { + void (*ScaleRowUp)(const uint16_t* src_ptr, uint16_t* dst_ptr, + int dst_width) = ScaleRowUp2_Linear_16_Any_C; + int i; + int y; + int dy; + + // This function can only scale up by 2 times horizontally. + assert(src_width == ((dst_width + 1) / 2)); + +#ifdef HAS_SCALEROWUP2_LINEAR_16_SSE2 + if (TestCpuFlag(kCpuHasSSE2)) { + ScaleRowUp = ScaleRowUp2_Linear_16_Any_SSE2; + } +#endif + +#ifdef HAS_SCALEROWUP2_LINEAR_16_AVX2 + if (TestCpuFlag(kCpuHasAVX2)) { + ScaleRowUp = ScaleRowUp2_Linear_16_Any_AVX2; + } +#endif + +#ifdef HAS_SCALEROWUP2_LINEAR_16_NEON + if (TestCpuFlag(kCpuHasNEON)) { + ScaleRowUp = ScaleRowUp2_Linear_16_Any_NEON; + } +#endif + + if (dst_height == 1) { + ScaleRowUp(src_ptr + ((src_height - 1) / 2) * (int64_t)src_stride, dst_ptr, + dst_width); + } else { + dy = FixedDiv(src_height - 1, dst_height - 1); + y = (1 << 15) - 1; + for (i = 0; i < dst_height; ++i) { + ScaleRowUp(src_ptr + (y >> 16) * (int64_t)src_stride, dst_ptr, dst_width); + dst_ptr += dst_stride; + y += dy; + } + } +} + +void ScalePlaneUp2_16_Bilinear(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint16_t* src_ptr, + uint16_t* dst_ptr) { + void (*Scale2RowUp)(const uint16_t* src_ptr, ptrdiff_t src_stride, + uint16_t* dst_ptr, ptrdiff_t dst_stride, int dst_width) = + ScaleRowUp2_Bilinear_16_Any_C; + int x; + + // This function can only scale up by 2 times. + assert(src_width == ((dst_width + 1) / 2)); + assert(src_height == ((dst_height + 1) / 2)); + +#ifdef HAS_SCALEROWUP2_BILINEAR_16_SSE2 + if (TestCpuFlag(kCpuHasSSE2)) { + Scale2RowUp = ScaleRowUp2_Bilinear_16_Any_SSE2; + } +#endif + +#ifdef HAS_SCALEROWUP2_BILINEAR_16_AVX2 + if (TestCpuFlag(kCpuHasAVX2)) { + Scale2RowUp = ScaleRowUp2_Bilinear_16_Any_AVX2; + } +#endif + +#ifdef HAS_SCALEROWUP2_BILINEAR_16_NEON + if (TestCpuFlag(kCpuHasNEON)) { + Scale2RowUp = ScaleRowUp2_Bilinear_16_Any_NEON; + } +#endif + + Scale2RowUp(src_ptr, 0, dst_ptr, 0, dst_width); + dst_ptr += dst_stride; + for (x = 0; x < src_height - 1; ++x) { + Scale2RowUp(src_ptr, src_stride, dst_ptr, dst_stride, dst_width); + src_ptr += src_stride; + dst_ptr += 2 * dst_stride; + } + if (!(dst_height & 1)) { + Scale2RowUp(src_ptr, 0, dst_ptr, 0, dst_width); + } +} + +void ScalePlaneBilinearUp_16(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint16_t* src_ptr, + uint16_t* dst_ptr, enum FilterMode filtering) { int j; // Initial source x/y coordinate and step values as 16.16 fixed point. @@ -1184,19 +1744,19 @@ void ScalePlaneBilinearUp_16(int src_width, int src_height, int dx = 0; int dy = 0; const int max_y = (src_height - 1) << 16; - void (*InterpolateRow)(uint16* dst_ptr, const uint16* src_ptr, - ptrdiff_t src_stride, int dst_width, int source_y_fraction) = - InterpolateRow_16_C; - void (*ScaleFilterCols)(uint16* dst_ptr, const uint16* src_ptr, - int dst_width, int x, int dx) = + void (*InterpolateRow)(uint16_t* dst_ptr, const uint16_t* src_ptr, + ptrdiff_t src_stride, int dst_width, + int source_y_fraction) = InterpolateRow_16_C; + void (*ScaleFilterCols)(uint16_t* dst_ptr, const uint16_t* src_ptr, + int dst_width, int x, int dx) = filtering ? ScaleFilterCols_16_C : ScaleCols_16_C; - ScaleSlope(src_width, src_height, dst_width, dst_height, filtering, - &x, &y, &dx, &dy); + ScaleSlope(src_width, src_height, dst_width, dst_height, filtering, &x, &y, + &dx, &dy); src_width = Abs(src_width); #if defined(HAS_INTERPOLATEROW_16_SSE2) if (TestCpuFlag(kCpuHasSSE2)) { - InterpolateRow = InterpolateRow_Any_16_SSE2; + InterpolateRow = InterpolateRow_16_Any_SSE2; if (IS_ALIGNED(dst_width, 16)) { InterpolateRow = InterpolateRow_16_SSE2; } @@ -1204,7 +1764,7 @@ void ScalePlaneBilinearUp_16(int src_width, int src_height, #endif #if defined(HAS_INTERPOLATEROW_16_SSSE3) if (TestCpuFlag(kCpuHasSSSE3)) { - InterpolateRow = InterpolateRow_Any_16_SSSE3; + InterpolateRow = InterpolateRow_16_Any_SSSE3; if (IS_ALIGNED(dst_width, 16)) { InterpolateRow = InterpolateRow_16_SSSE3; } @@ -1212,7 +1772,7 @@ void ScalePlaneBilinearUp_16(int src_width, int src_height, #endif #if defined(HAS_INTERPOLATEROW_16_AVX2) if (TestCpuFlag(kCpuHasAVX2)) { - InterpolateRow = InterpolateRow_Any_16_AVX2; + InterpolateRow = InterpolateRow_16_Any_AVX2; if (IS_ALIGNED(dst_width, 32)) { InterpolateRow = InterpolateRow_16_AVX2; } @@ -1220,20 +1780,12 @@ void ScalePlaneBilinearUp_16(int src_width, int src_height, #endif #if defined(HAS_INTERPOLATEROW_16_NEON) if (TestCpuFlag(kCpuHasNEON)) { - InterpolateRow = InterpolateRow_Any_16_NEON; + InterpolateRow = InterpolateRow_16_Any_NEON; if (IS_ALIGNED(dst_width, 16)) { InterpolateRow = InterpolateRow_16_NEON; } } #endif -#if defined(HAS_INTERPOLATEROW_16_DSPR2) - if (TestCpuFlag(kCpuHasDSPR2)) { - InterpolateRow = InterpolateRow_Any_16_DSPR2; - if (IS_ALIGNED(dst_width, 4)) { - InterpolateRow = InterpolateRow_16_DSPR2; - } - } -#endif if (filtering && src_width >= 32768) { ScaleFilterCols = ScaleFilterCols64_16_C; @@ -1251,20 +1803,19 @@ void ScalePlaneBilinearUp_16(int src_width, int src_height, } #endif } - if (y > max_y) { y = max_y; } { int yi = y >> 16; - const uint16* src = src_ptr + yi * src_stride; + const uint16_t* src = src_ptr + yi * (int64_t)src_stride; // Allocate 2 row buffers. - const int kRowSize = (dst_width + 31) & ~31; - align_buffer_64(row, kRowSize * 4); + const int row_size = (dst_width + 31) & ~31; + align_buffer_64(row, row_size * 4); - uint16* rowptr = (uint16*)row; - int rowstride = kRowSize; + uint16_t* rowptr = (uint16_t*)row; + int rowstride = row_size; int lasty = yi; ScaleFilterCols(rowptr, src, dst_width, x, dx); @@ -1272,7 +1823,9 @@ void ScalePlaneBilinearUp_16(int src_width, int src_height, src += src_stride; } ScaleFilterCols(rowptr + rowstride, src, dst_width, x, dx); - src += src_stride; + if (src_height > 2) { + src += src_stride; + } for (j = 0; j < dst_height; ++j) { yi = y >> 16; @@ -1280,14 +1833,16 @@ void ScalePlaneBilinearUp_16(int src_width, int src_height, if (y > max_y) { y = max_y; yi = y >> 16; - src = src_ptr + yi * src_stride; + src = src_ptr + yi * (int64_t)src_stride; } if (yi != lasty) { ScaleFilterCols(rowptr, src, dst_width, x, dx); rowptr += rowstride; rowstride = -rowstride; lasty = yi; - src += src_stride; + if ((y + 65536) < max_y) { + src += src_stride; + } } } if (filtering == kFilterLinear) { @@ -1308,20 +1863,24 @@ void ScalePlaneBilinearUp_16(int src_width, int src_height, // of x and dx is the integer part of the source position and // the lower 16 bits are the fixed decimal part. -static void ScalePlaneSimple(int src_width, int src_height, - int dst_width, int dst_height, - int src_stride, int dst_stride, - const uint8* src_ptr, uint8* dst_ptr) { +static void ScalePlaneSimple(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint8_t* src_ptr, + uint8_t* dst_ptr) { int i; - void (*ScaleCols)(uint8* dst_ptr, const uint8* src_ptr, - int dst_width, int x, int dx) = ScaleCols_C; + void (*ScaleCols)(uint8_t* dst_ptr, const uint8_t* src_ptr, int dst_width, + int x, int dx) = ScaleCols_C; // Initial source x/y coordinate and step values as 16.16 fixed point. int x = 0; int y = 0; int dx = 0; int dy = 0; - ScaleSlope(src_width, src_height, dst_width, dst_height, kFilterNone, - &x, &y, &dx, &dy); + ScaleSlope(src_width, src_height, dst_width, dst_height, kFilterNone, &x, &y, + &dx, &dy); src_width = Abs(src_width); if (src_width * 2 == dst_width && x < 0x8000) { @@ -1334,26 +1893,31 @@ static void ScalePlaneSimple(int src_width, int src_height, } for (i = 0; i < dst_height; ++i) { - ScaleCols(dst_ptr, src_ptr + (y >> 16) * src_stride, dst_width, x, dx); + ScaleCols(dst_ptr, src_ptr + (y >> 16) * (int64_t)src_stride, dst_width, x, + dx); dst_ptr += dst_stride; y += dy; } } -static void ScalePlaneSimple_16(int src_width, int src_height, - int dst_width, int dst_height, - int src_stride, int dst_stride, - const uint16* src_ptr, uint16* dst_ptr) { +static void ScalePlaneSimple_16(int src_width, + int src_height, + int dst_width, + int dst_height, + int src_stride, + int dst_stride, + const uint16_t* src_ptr, + uint16_t* dst_ptr) { int i; - void (*ScaleCols)(uint16* dst_ptr, const uint16* src_ptr, - int dst_width, int x, int dx) = ScaleCols_16_C; + void (*ScaleCols)(uint16_t* dst_ptr, const uint16_t* src_ptr, int dst_width, + int x, int dx) = ScaleCols_16_C; // Initial source x/y coordinate and step values as 16.16 fixed point. int x = 0; int y = 0; int dx = 0; int dy = 0; - ScaleSlope(src_width, src_height, dst_width, dst_height, kFilterNone, - &x, &y, &dx, &dy); + ScaleSlope(src_width, src_height, dst_width, dst_height, kFilterNone, &x, &y, + &dx, &dy); src_width = Abs(src_width); if (src_width * 2 == dst_width && x < 0x8000) { @@ -1366,8 +1930,8 @@ static void ScalePlaneSimple_16(int src_width, int src_height, } for (i = 0; i < dst_height; ++i) { - ScaleCols(dst_ptr, src_ptr + (y >> 16) * src_stride, - dst_width, x, dx); + ScaleCols(dst_ptr, src_ptr + (y >> 16) * (int64_t)src_stride, dst_width, x, + dx); dst_ptr += dst_stride; y += dy; } @@ -1375,24 +1939,26 @@ static void ScalePlaneSimple_16(int src_width, int src_height, // Scale a plane. // This function dispatches to a specialized scaler based on scale factor. - LIBYUV_API -void ScalePlane(const uint8* src, int src_stride, - int src_width, int src_height, - uint8* dst, int dst_stride, - int dst_width, int dst_height, +void ScalePlane(const uint8_t* src, + int src_stride, + int src_width, + int src_height, + uint8_t* dst, + int dst_stride, + int dst_width, + int dst_height, enum FilterMode filtering) { // Simplify filtering when possible. - filtering = ScaleFilterReduce(src_width, src_height, - dst_width, dst_height, filtering); + filtering = ScaleFilterReduce(src_width, src_height, dst_width, dst_height, + filtering); // Negative height means invert the image. if (src_height < 0) { src_height = -src_height; - src = src + (src_height - 1) * src_stride; + src = src + (src_height - 1) * (int64_t)src_stride; src_stride = -src_stride; } - // Use specialized scales to improve performance for common resolutions. // For example, all the 1/2 scalings will use ScalePlaneDown2() if (dst_width == src_width && dst_height == src_height) { @@ -1401,48 +1967,64 @@ void ScalePlane(const uint8* src, int src_stride, return; } if (dst_width == src_width && filtering != kFilterBox) { - int dy = FixedDiv(src_height, dst_height); + int dy = 0; + int y = 0; + // When scaling down, use the center 2 rows to filter. + // When scaling up, last row of destination uses the last 2 source rows. + if (dst_height <= src_height) { + dy = FixedDiv(src_height, dst_height); + y = CENTERSTART(dy, -32768); // Subtract 0.5 (32768) to center filter. + } else if (src_height > 1 && dst_height > 1) { + dy = FixedDiv1(src_height, dst_height); + } // Arbitrary scale vertically, but unscaled horizontally. - ScalePlaneVertical(src_height, - dst_width, dst_height, - src_stride, dst_stride, src, dst, - 0, 0, dy, 1, filtering); + ScalePlaneVertical(src_height, dst_width, dst_height, src_stride, + dst_stride, src, dst, 0, y, dy, /*bpp=*/1, filtering); return; } if (dst_width <= Abs(src_width) && dst_height <= src_height) { // Scale down. - if (4 * dst_width == 3 * src_width && - 4 * dst_height == 3 * src_height) { + if (4 * dst_width == 3 * src_width && 4 * dst_height == 3 * src_height) { // optimized, 3/4 - ScalePlaneDown34(src_width, src_height, dst_width, dst_height, - src_stride, dst_stride, src, dst, filtering); + ScalePlaneDown34(src_width, src_height, dst_width, dst_height, src_stride, + dst_stride, src, dst, filtering); return; } if (2 * dst_width == src_width && 2 * dst_height == src_height) { // optimized, 1/2 - ScalePlaneDown2(src_width, src_height, dst_width, dst_height, - src_stride, dst_stride, src, dst, filtering); + ScalePlaneDown2(src_width, src_height, dst_width, dst_height, src_stride, + dst_stride, src, dst, filtering); return; } // 3/8 rounded up for odd sized chroma height. - if (8 * dst_width == 3 * src_width && - dst_height == ((src_height * 3 + 7) / 8)) { + if (8 * dst_width == 3 * src_width && 8 * dst_height == 3 * src_height) { // optimized, 3/8 - ScalePlaneDown38(src_width, src_height, dst_width, dst_height, - src_stride, dst_stride, src, dst, filtering); + ScalePlaneDown38(src_width, src_height, dst_width, dst_height, src_stride, + dst_stride, src, dst, filtering); return; } if (4 * dst_width == src_width && 4 * dst_height == src_height && (filtering == kFilterBox || filtering == kFilterNone)) { // optimized, 1/4 - ScalePlaneDown4(src_width, src_height, dst_width, dst_height, - src_stride, dst_stride, src, dst, filtering); + ScalePlaneDown4(src_width, src_height, dst_width, dst_height, src_stride, + dst_stride, src, dst, filtering); return; } } if (filtering == kFilterBox && dst_height * 2 < src_height) { - ScalePlaneBox(src_width, src_height, dst_width, dst_height, - src_stride, dst_stride, src, dst); + ScalePlaneBox(src_width, src_height, dst_width, dst_height, src_stride, + dst_stride, src, dst); + return; + } + if ((dst_width + 1) / 2 == src_width && filtering == kFilterLinear) { + ScalePlaneUp2_Linear(src_width, src_height, dst_width, dst_height, + src_stride, dst_stride, src, dst); + return; + } + if ((dst_height + 1) / 2 == src_height && (dst_width + 1) / 2 == src_width && + (filtering == kFilterBilinear || filtering == kFilterBox)) { + ScalePlaneUp2_Bilinear(src_width, src_height, dst_width, dst_height, + src_stride, dst_stride, src, dst); return; } if (filtering && dst_height > src_height) { @@ -1455,27 +2037,30 @@ void ScalePlane(const uint8* src, int src_stride, src_stride, dst_stride, src, dst, filtering); return; } - ScalePlaneSimple(src_width, src_height, dst_width, dst_height, - src_stride, dst_stride, src, dst); + ScalePlaneSimple(src_width, src_height, dst_width, dst_height, src_stride, + dst_stride, src, dst); } LIBYUV_API -void ScalePlane_16(const uint16* src, int src_stride, - int src_width, int src_height, - uint16* dst, int dst_stride, - int dst_width, int dst_height, - enum FilterMode filtering) { +void ScalePlane_16(const uint16_t* src, + int src_stride, + int src_width, + int src_height, + uint16_t* dst, + int dst_stride, + int dst_width, + int dst_height, + enum FilterMode filtering) { // Simplify filtering when possible. - filtering = ScaleFilterReduce(src_width, src_height, - dst_width, dst_height, filtering); + filtering = ScaleFilterReduce(src_width, src_height, dst_width, dst_height, + filtering); // Negative height means invert the image. if (src_height < 0) { src_height = -src_height; - src = src + (src_height - 1) * src_stride; + src = src + (src_height - 1) * (int64_t)src_stride; src_stride = -src_stride; } - // Use specialized scales to improve performance for common resolutions. // For example, all the 1/2 scalings will use ScalePlaneDown2() if (dst_width == src_width && dst_height == src_height) { @@ -1483,19 +2068,28 @@ void ScalePlane_16(const uint16* src, int src_stride, CopyPlane_16(src, src_stride, dst, dst_stride, dst_width, dst_height); return; } - if (dst_width == src_width) { - int dy = FixedDiv(src_height, dst_height); - // Arbitrary scale vertically, but unscaled vertically. - ScalePlaneVertical_16(src_height, - dst_width, dst_height, - src_stride, dst_stride, src, dst, - 0, 0, dy, 1, filtering); + if (dst_width == src_width && filtering != kFilterBox) { + int dy = 0; + int y = 0; + // When scaling down, use the center 2 rows to filter. + // When scaling up, last row of destination uses the last 2 source rows. + if (dst_height <= src_height) { + dy = FixedDiv(src_height, dst_height); + y = CENTERSTART(dy, -32768); // Subtract 0.5 (32768) to center filter. + // When scaling up, ensure the last row of destination uses the last + // source. Avoid divide by zero for dst_height but will do no scaling + // later. + } else if (src_height > 1 && dst_height > 1) { + dy = FixedDiv1(src_height, dst_height); + } + // Arbitrary scale vertically, but unscaled horizontally. + ScalePlaneVertical_16(src_height, dst_width, dst_height, src_stride, + dst_stride, src, dst, 0, y, dy, /*bpp=*/1, filtering); return; } if (dst_width <= Abs(src_width) && dst_height <= src_height) { // Scale down. - if (4 * dst_width == 3 * src_width && - 4 * dst_height == 3 * src_height) { + if (4 * dst_width == 3 * src_width && 4 * dst_height == 3 * src_height) { // optimized, 3/4 ScalePlaneDown34_16(src_width, src_height, dst_width, dst_height, src_stride, dst_stride, src, dst, filtering); @@ -1508,15 +2102,14 @@ void ScalePlane_16(const uint16* src, int src_stride, return; } // 3/8 rounded up for odd sized chroma height. - if (8 * dst_width == 3 * src_width && - dst_height == ((src_height * 3 + 7) / 8)) { + if (8 * dst_width == 3 * src_width && 8 * dst_height == 3 * src_height) { // optimized, 3/8 ScalePlaneDown38_16(src_width, src_height, dst_width, dst_height, src_stride, dst_stride, src, dst, filtering); return; } if (4 * dst_width == src_width && 4 * dst_height == src_height && - filtering != kFilterBilinear) { + (filtering == kFilterBox || filtering == kFilterNone)) { // optimized, 1/4 ScalePlaneDown4_16(src_width, src_height, dst_width, dst_height, src_stride, dst_stride, src, dst, filtering); @@ -1524,8 +2117,19 @@ void ScalePlane_16(const uint16* src, int src_stride, } } if (filtering == kFilterBox && dst_height * 2 < src_height) { - ScalePlaneBox_16(src_width, src_height, dst_width, dst_height, - src_stride, dst_stride, src, dst); + ScalePlaneBox_16(src_width, src_height, dst_width, dst_height, src_stride, + dst_stride, src, dst); + return; + } + if ((dst_width + 1) / 2 == src_width && filtering == kFilterLinear) { + ScalePlaneUp2_16_Linear(src_width, src_height, dst_width, dst_height, + src_stride, dst_stride, src, dst); + return; + } + if ((dst_height + 1) / 2 == src_height && (dst_width + 1) / 2 == src_width && + (filtering == kFilterBilinear || filtering == kFilterBox)) { + ScalePlaneUp2_16_Bilinear(src_width, src_height, dst_width, dst_height, + src_stride, dst_stride, src, dst); return; } if (filtering && dst_height > src_height) { @@ -1538,132 +2142,435 @@ void ScalePlane_16(const uint16* src, int src_stride, src_stride, dst_stride, src, dst, filtering); return; } - ScalePlaneSimple_16(src_width, src_height, dst_width, dst_height, - src_stride, dst_stride, src, dst); + ScalePlaneSimple_16(src_width, src_height, dst_width, dst_height, src_stride, + dst_stride, src, dst); +} + +LIBYUV_API +void ScalePlane_12(const uint16_t* src, + int src_stride, + int src_width, + int src_height, + uint16_t* dst, + int dst_stride, + int dst_width, + int dst_height, + enum FilterMode filtering) { + // Simplify filtering when possible. + filtering = ScaleFilterReduce(src_width, src_height, dst_width, dst_height, + filtering); + + // Negative height means invert the image. + if (src_height < 0) { + src_height = -src_height; + src = src + (src_height - 1) * (int64_t)src_stride; + src_stride = -src_stride; + } + + if ((dst_width + 1) / 2 == src_width && filtering == kFilterLinear) { + ScalePlaneUp2_12_Linear(src_width, src_height, dst_width, dst_height, + src_stride, dst_stride, src, dst); + return; + } + if ((dst_height + 1) / 2 == src_height && (dst_width + 1) / 2 == src_width && + (filtering == kFilterBilinear || filtering == kFilterBox)) { + ScalePlaneUp2_12_Bilinear(src_width, src_height, dst_width, dst_height, + src_stride, dst_stride, src, dst); + return; + } + + ScalePlane_16(src, src_stride, src_width, src_height, dst, dst_stride, + dst_width, dst_height, filtering); } // Scale an I420 image. // This function in turn calls a scaling function for each plane. LIBYUV_API -int I420Scale(const uint8* src_y, int src_stride_y, - const uint8* src_u, int src_stride_u, - const uint8* src_v, int src_stride_v, - int src_width, int src_height, - uint8* dst_y, int dst_stride_y, - uint8* dst_u, int dst_stride_u, - uint8* dst_v, int dst_stride_v, - int dst_width, int dst_height, +int I420Scale(const uint8_t* src_y, + int src_stride_y, + const uint8_t* src_u, + int src_stride_u, + const uint8_t* src_v, + int src_stride_v, + int src_width, + int src_height, + uint8_t* dst_y, + int dst_stride_y, + uint8_t* dst_u, + int dst_stride_u, + uint8_t* dst_v, + int dst_stride_v, + int dst_width, + int dst_height, enum FilterMode filtering) { int src_halfwidth = SUBSAMPLE(src_width, 1, 1); int src_halfheight = SUBSAMPLE(src_height, 1, 1); int dst_halfwidth = SUBSAMPLE(dst_width, 1, 1); int dst_halfheight = SUBSAMPLE(dst_height, 1, 1); - if (!src_y || !src_u || !src_v || src_width == 0 || src_height == 0 || - src_width > 32768 || src_height > 32768 || - !dst_y || !dst_u || !dst_v || dst_width <= 0 || dst_height <= 0) { + + if (!src_y || !src_u || !src_v || src_width <= 0 || src_height == 0 || + src_width > 32768 || src_height > 32768 || !dst_y || !dst_u || !dst_v || + dst_width <= 0 || dst_height <= 0) { return -1; } - ScalePlane(src_y, src_stride_y, src_width, src_height, - dst_y, dst_stride_y, dst_width, dst_height, - filtering); - ScalePlane(src_u, src_stride_u, src_halfwidth, src_halfheight, - dst_u, dst_stride_u, dst_halfwidth, dst_halfheight, - filtering); - ScalePlane(src_v, src_stride_v, src_halfwidth, src_halfheight, - dst_v, dst_stride_v, dst_halfwidth, dst_halfheight, - filtering); + ScalePlane(src_y, src_stride_y, src_width, src_height, dst_y, dst_stride_y, + dst_width, dst_height, filtering); + ScalePlane(src_u, src_stride_u, src_halfwidth, src_halfheight, dst_u, + dst_stride_u, dst_halfwidth, dst_halfheight, filtering); + ScalePlane(src_v, src_stride_v, src_halfwidth, src_halfheight, dst_v, + dst_stride_v, dst_halfwidth, dst_halfheight, filtering); return 0; } LIBYUV_API -int I420Scale_16(const uint16* src_y, int src_stride_y, - const uint16* src_u, int src_stride_u, - const uint16* src_v, int src_stride_v, - int src_width, int src_height, - uint16* dst_y, int dst_stride_y, - uint16* dst_u, int dst_stride_u, - uint16* dst_v, int dst_stride_v, - int dst_width, int dst_height, +int I420Scale_16(const uint16_t* src_y, + int src_stride_y, + const uint16_t* src_u, + int src_stride_u, + const uint16_t* src_v, + int src_stride_v, + int src_width, + int src_height, + uint16_t* dst_y, + int dst_stride_y, + uint16_t* dst_u, + int dst_stride_u, + uint16_t* dst_v, + int dst_stride_v, + int dst_width, + int dst_height, enum FilterMode filtering) { int src_halfwidth = SUBSAMPLE(src_width, 1, 1); int src_halfheight = SUBSAMPLE(src_height, 1, 1); int dst_halfwidth = SUBSAMPLE(dst_width, 1, 1); int dst_halfheight = SUBSAMPLE(dst_height, 1, 1); - if (!src_y || !src_u || !src_v || src_width == 0 || src_height == 0 || - src_width > 32768 || src_height > 32768 || - !dst_y || !dst_u || !dst_v || dst_width <= 0 || dst_height <= 0) { + + if (!src_y || !src_u || !src_v || src_width <= 0 || src_height == 0 || + src_width > 32768 || src_height > 32768 || !dst_y || !dst_u || !dst_v || + dst_width <= 0 || dst_height <= 0) { return -1; } - ScalePlane_16(src_y, src_stride_y, src_width, src_height, - dst_y, dst_stride_y, dst_width, dst_height, - filtering); - ScalePlane_16(src_u, src_stride_u, src_halfwidth, src_halfheight, - dst_u, dst_stride_u, dst_halfwidth, dst_halfheight, - filtering); - ScalePlane_16(src_v, src_stride_v, src_halfwidth, src_halfheight, - dst_v, dst_stride_v, dst_halfwidth, dst_halfheight, - filtering); + ScalePlane_16(src_y, src_stride_y, src_width, src_height, dst_y, dst_stride_y, + dst_width, dst_height, filtering); + ScalePlane_16(src_u, src_stride_u, src_halfwidth, src_halfheight, dst_u, + dst_stride_u, dst_halfwidth, dst_halfheight, filtering); + ScalePlane_16(src_v, src_stride_v, src_halfwidth, src_halfheight, dst_v, + dst_stride_v, dst_halfwidth, dst_halfheight, filtering); return 0; } -// Deprecated api LIBYUV_API -int Scale(const uint8* src_y, const uint8* src_u, const uint8* src_v, - int src_stride_y, int src_stride_u, int src_stride_v, - int src_width, int src_height, - uint8* dst_y, uint8* dst_u, uint8* dst_v, - int dst_stride_y, int dst_stride_u, int dst_stride_v, - int dst_width, int dst_height, - LIBYUV_BOOL interpolate) { - return I420Scale(src_y, src_stride_y, - src_u, src_stride_u, - src_v, src_stride_v, - src_width, src_height, - dst_y, dst_stride_y, - dst_u, dst_stride_u, - dst_v, dst_stride_v, - dst_width, dst_height, - interpolate ? kFilterBox : kFilterNone); +int I420Scale_12(const uint16_t* src_y, + int src_stride_y, + const uint16_t* src_u, + int src_stride_u, + const uint16_t* src_v, + int src_stride_v, + int src_width, + int src_height, + uint16_t* dst_y, + int dst_stride_y, + uint16_t* dst_u, + int dst_stride_u, + uint16_t* dst_v, + int dst_stride_v, + int dst_width, + int dst_height, + enum FilterMode filtering) { + int src_halfwidth = SUBSAMPLE(src_width, 1, 1); + int src_halfheight = SUBSAMPLE(src_height, 1, 1); + int dst_halfwidth = SUBSAMPLE(dst_width, 1, 1); + int dst_halfheight = SUBSAMPLE(dst_height, 1, 1); + + if (!src_y || !src_u || !src_v || src_width <= 0 || src_height == 0 || + src_width > 32768 || src_height > 32768 || !dst_y || !dst_u || !dst_v || + dst_width <= 0 || dst_height <= 0) { + return -1; + } + + ScalePlane_12(src_y, src_stride_y, src_width, src_height, dst_y, dst_stride_y, + dst_width, dst_height, filtering); + ScalePlane_12(src_u, src_stride_u, src_halfwidth, src_halfheight, dst_u, + dst_stride_u, dst_halfwidth, dst_halfheight, filtering); + ScalePlane_12(src_v, src_stride_v, src_halfwidth, src_halfheight, dst_v, + dst_stride_v, dst_halfwidth, dst_halfheight, filtering); + return 0; } -// Deprecated api +// Scale an I444 image. +// This function in turn calls a scaling function for each plane. + +LIBYUV_API +int I444Scale(const uint8_t* src_y, + int src_stride_y, + const uint8_t* src_u, + int src_stride_u, + const uint8_t* src_v, + int src_stride_v, + int src_width, + int src_height, + uint8_t* dst_y, + int dst_stride_y, + uint8_t* dst_u, + int dst_stride_u, + uint8_t* dst_v, + int dst_stride_v, + int dst_width, + int dst_height, + enum FilterMode filtering) { + if (!src_y || !src_u || !src_v || src_width <= 0 || src_height == 0 || + src_width > 32768 || src_height > 32768 || !dst_y || !dst_u || !dst_v || + dst_width <= 0 || dst_height <= 0) { + return -1; + } + + ScalePlane(src_y, src_stride_y, src_width, src_height, dst_y, dst_stride_y, + dst_width, dst_height, filtering); + ScalePlane(src_u, src_stride_u, src_width, src_height, dst_u, dst_stride_u, + dst_width, dst_height, filtering); + ScalePlane(src_v, src_stride_v, src_width, src_height, dst_v, dst_stride_v, + dst_width, dst_height, filtering); + return 0; +} + +LIBYUV_API +int I444Scale_16(const uint16_t* src_y, + int src_stride_y, + const uint16_t* src_u, + int src_stride_u, + const uint16_t* src_v, + int src_stride_v, + int src_width, + int src_height, + uint16_t* dst_y, + int dst_stride_y, + uint16_t* dst_u, + int dst_stride_u, + uint16_t* dst_v, + int dst_stride_v, + int dst_width, + int dst_height, + enum FilterMode filtering) { + if (!src_y || !src_u || !src_v || src_width <= 0 || src_height == 0 || + src_width > 32768 || src_height > 32768 || !dst_y || !dst_u || !dst_v || + dst_width <= 0 || dst_height <= 0) { + return -1; + } + + ScalePlane_16(src_y, src_stride_y, src_width, src_height, dst_y, dst_stride_y, + dst_width, dst_height, filtering); + ScalePlane_16(src_u, src_stride_u, src_width, src_height, dst_u, dst_stride_u, + dst_width, dst_height, filtering); + ScalePlane_16(src_v, src_stride_v, src_width, src_height, dst_v, dst_stride_v, + dst_width, dst_height, filtering); + return 0; +} + +LIBYUV_API +int I444Scale_12(const uint16_t* src_y, + int src_stride_y, + const uint16_t* src_u, + int src_stride_u, + const uint16_t* src_v, + int src_stride_v, + int src_width, + int src_height, + uint16_t* dst_y, + int dst_stride_y, + uint16_t* dst_u, + int dst_stride_u, + uint16_t* dst_v, + int dst_stride_v, + int dst_width, + int dst_height, + enum FilterMode filtering) { + if (!src_y || !src_u || !src_v || src_width <= 0 || src_height == 0 || + src_width > 32768 || src_height > 32768 || !dst_y || !dst_u || !dst_v || + dst_width <= 0 || dst_height <= 0) { + return -1; + } + + ScalePlane_12(src_y, src_stride_y, src_width, src_height, dst_y, dst_stride_y, + dst_width, dst_height, filtering); + ScalePlane_12(src_u, src_stride_u, src_width, src_height, dst_u, dst_stride_u, + dst_width, dst_height, filtering); + ScalePlane_12(src_v, src_stride_v, src_width, src_height, dst_v, dst_stride_v, + dst_width, dst_height, filtering); + return 0; +} + +// Scale an I422 image. +// This function in turn calls a scaling function for each plane. + +LIBYUV_API +int I422Scale(const uint8_t* src_y, + int src_stride_y, + const uint8_t* src_u, + int src_stride_u, + const uint8_t* src_v, + int src_stride_v, + int src_width, + int src_height, + uint8_t* dst_y, + int dst_stride_y, + uint8_t* dst_u, + int dst_stride_u, + uint8_t* dst_v, + int dst_stride_v, + int dst_width, + int dst_height, + enum FilterMode filtering) { + int src_halfwidth = SUBSAMPLE(src_width, 1, 1); + int dst_halfwidth = SUBSAMPLE(dst_width, 1, 1); + + if (!src_y || !src_u || !src_v || src_width <= 0 || src_height == 0 || + src_width > 32768 || src_height > 32768 || !dst_y || !dst_u || !dst_v || + dst_width <= 0 || dst_height <= 0) { + return -1; + } + + ScalePlane(src_y, src_stride_y, src_width, src_height, dst_y, dst_stride_y, + dst_width, dst_height, filtering); + ScalePlane(src_u, src_stride_u, src_halfwidth, src_height, dst_u, + dst_stride_u, dst_halfwidth, dst_height, filtering); + ScalePlane(src_v, src_stride_v, src_halfwidth, src_height, dst_v, + dst_stride_v, dst_halfwidth, dst_height, filtering); + return 0; +} + +LIBYUV_API +int I422Scale_16(const uint16_t* src_y, + int src_stride_y, + const uint16_t* src_u, + int src_stride_u, + const uint16_t* src_v, + int src_stride_v, + int src_width, + int src_height, + uint16_t* dst_y, + int dst_stride_y, + uint16_t* dst_u, + int dst_stride_u, + uint16_t* dst_v, + int dst_stride_v, + int dst_width, + int dst_height, + enum FilterMode filtering) { + int src_halfwidth = SUBSAMPLE(src_width, 1, 1); + int dst_halfwidth = SUBSAMPLE(dst_width, 1, 1); + + if (!src_y || !src_u || !src_v || src_width <= 0 || src_height == 0 || + src_width > 32768 || src_height > 32768 || !dst_y || !dst_u || !dst_v || + dst_width <= 0 || dst_height <= 0) { + return -1; + } + + ScalePlane_16(src_y, src_stride_y, src_width, src_height, dst_y, dst_stride_y, + dst_width, dst_height, filtering); + ScalePlane_16(src_u, src_stride_u, src_halfwidth, src_height, dst_u, + dst_stride_u, dst_halfwidth, dst_height, filtering); + ScalePlane_16(src_v, src_stride_v, src_halfwidth, src_height, dst_v, + dst_stride_v, dst_halfwidth, dst_height, filtering); + return 0; +} + +LIBYUV_API +int I422Scale_12(const uint16_t* src_y, + int src_stride_y, + const uint16_t* src_u, + int src_stride_u, + const uint16_t* src_v, + int src_stride_v, + int src_width, + int src_height, + uint16_t* dst_y, + int dst_stride_y, + uint16_t* dst_u, + int dst_stride_u, + uint16_t* dst_v, + int dst_stride_v, + int dst_width, + int dst_height, + enum FilterMode filtering) { + int src_halfwidth = SUBSAMPLE(src_width, 1, 1); + int dst_halfwidth = SUBSAMPLE(dst_width, 1, 1); + + if (!src_y || !src_u || !src_v || src_width <= 0 || src_height == 0 || + src_width > 32768 || src_height > 32768 || !dst_y || !dst_u || !dst_v || + dst_width <= 0 || dst_height <= 0) { + return -1; + } + + ScalePlane_12(src_y, src_stride_y, src_width, src_height, dst_y, dst_stride_y, + dst_width, dst_height, filtering); + ScalePlane_12(src_u, src_stride_u, src_halfwidth, src_height, dst_u, + dst_stride_u, dst_halfwidth, dst_height, filtering); + ScalePlane_12(src_v, src_stride_v, src_halfwidth, src_height, dst_v, + dst_stride_v, dst_halfwidth, dst_height, filtering); + return 0; +} + +// Scale an NV12 image. +// This function in turn calls a scaling function for each plane. + LIBYUV_API -int ScaleOffset(const uint8* src, int src_width, int src_height, - uint8* dst, int dst_width, int dst_height, int dst_yoffset, - LIBYUV_BOOL interpolate) { - // Chroma requires offset to multiple of 2. - int dst_yoffset_even = dst_yoffset & ~1; +int NV12Scale(const uint8_t* src_y, + int src_stride_y, + const uint8_t* src_uv, + int src_stride_uv, + int src_width, + int src_height, + uint8_t* dst_y, + int dst_stride_y, + uint8_t* dst_uv, + int dst_stride_uv, + int dst_width, + int dst_height, + enum FilterMode filtering) { int src_halfwidth = SUBSAMPLE(src_width, 1, 1); int src_halfheight = SUBSAMPLE(src_height, 1, 1); int dst_halfwidth = SUBSAMPLE(dst_width, 1, 1); int dst_halfheight = SUBSAMPLE(dst_height, 1, 1); - int aheight = dst_height - dst_yoffset_even * 2; // actual output height - const uint8* src_y = src; - const uint8* src_u = src + src_width * src_height; - const uint8* src_v = src + src_width * src_height + - src_halfwidth * src_halfheight; - uint8* dst_y = dst + dst_yoffset_even * dst_width; - uint8* dst_u = dst + dst_width * dst_height + - (dst_yoffset_even >> 1) * dst_halfwidth; - uint8* dst_v = dst + dst_width * dst_height + dst_halfwidth * dst_halfheight + - (dst_yoffset_even >> 1) * dst_halfwidth; - if (!src || src_width <= 0 || src_height <= 0 || - !dst || dst_width <= 0 || dst_height <= 0 || dst_yoffset_even < 0 || - dst_yoffset_even >= dst_height) { + + if (!src_y || !src_uv || src_width <= 0 || src_height == 0 || + src_width > 32768 || src_height > 32768 || !dst_y || !dst_uv || + dst_width <= 0 || dst_height <= 0) { return -1; } - return I420Scale(src_y, src_width, - src_u, src_halfwidth, - src_v, src_halfwidth, - src_width, src_height, - dst_y, dst_width, - dst_u, dst_halfwidth, - dst_v, dst_halfwidth, - dst_width, aheight, - interpolate ? kFilterBox : kFilterNone); + + ScalePlane(src_y, src_stride_y, src_width, src_height, dst_y, dst_stride_y, + dst_width, dst_height, filtering); + UVScale(src_uv, src_stride_uv, src_halfwidth, src_halfheight, dst_uv, + dst_stride_uv, dst_halfwidth, dst_halfheight, filtering); + return 0; +} + +// Deprecated api +LIBYUV_API +int Scale(const uint8_t* src_y, + const uint8_t* src_u, + const uint8_t* src_v, + int src_stride_y, + int src_stride_u, + int src_stride_v, + int src_width, + int src_height, + uint8_t* dst_y, + uint8_t* dst_u, + uint8_t* dst_v, + int dst_stride_y, + int dst_stride_u, + int dst_stride_v, + int dst_width, + int dst_height, + LIBYUV_BOOL interpolate) { + return I420Scale(src_y, src_stride_y, src_u, src_stride_u, src_v, + src_stride_v, src_width, src_height, dst_y, dst_stride_y, + dst_u, dst_stride_u, dst_v, dst_stride_v, dst_width, + dst_height, interpolate ? kFilterBox : kFilterNone); } #ifdef __cplusplus |