michael@0: /*
michael@0:  *  Copyright 2012 The LibYuv Project Authors. All rights reserved.
michael@0:  *
michael@0:  *  Use of this source code is governed by a BSD-style license
michael@0:  *  that can be found in the LICENSE file in the root of the source
michael@0:  *  tree. An additional intellectual property rights grant can be found
michael@0:  *  in the file PATENTS. All contributing project authors may
michael@0:  *  be found in the AUTHORS file in the root of the source tree.
michael@0:  */
michael@0: 
michael@0: #include "libyuv/row.h"
michael@0: 
michael@0: #include "libyuv/basic_types.h"
michael@0: 
michael@0: #ifdef __cplusplus
michael@0: namespace libyuv {
michael@0: extern "C" {
michael@0: #endif
michael@0: 
michael@0: // TODO(fbarchard): Consider 'any' functions handling any quantity of pixels.
michael@0: // TODO(fbarchard): Consider 'any' functions handling odd alignment.
michael@0: // YUV to RGB does multiple of 8 with SIMD and remainder with C.
michael@0: #define YANY(NAMEANY, I420TORGB_SIMD, I420TORGB_C, UV_SHIFT, BPP, MASK)        \
michael@0:     void NAMEANY(const uint8* y_buf,                                           \
michael@0:                  const uint8* u_buf,                                           \
michael@0:                  const uint8* v_buf,                                           \
michael@0:                  uint8* rgb_buf,                                               \
michael@0:                  int width) {                                                  \
michael@0:       int n = width & ~MASK;                                                   \
michael@0:       I420TORGB_SIMD(y_buf, u_buf, v_buf, rgb_buf, n);                         \
michael@0:       I420TORGB_C(y_buf + n,                                                   \
michael@0:                   u_buf + (n >> UV_SHIFT),                                     \
michael@0:                   v_buf + (n >> UV_SHIFT),                                     \
michael@0:                   rgb_buf + n * BPP, width & MASK);                            \
michael@0:     }
michael@0: 
michael@0: #ifdef HAS_I422TOARGBROW_SSSE3
michael@0: YANY(I444ToARGBRow_Any_SSSE3, I444ToARGBRow_Unaligned_SSSE3, I444ToARGBRow_C,
michael@0:      0, 4, 7)
michael@0: YANY(I422ToARGBRow_Any_SSSE3, I422ToARGBRow_Unaligned_SSSE3, I422ToARGBRow_C,
michael@0:      1, 4, 7)
michael@0: YANY(I411ToARGBRow_Any_SSSE3, I411ToARGBRow_Unaligned_SSSE3, I411ToARGBRow_C,
michael@0:      2, 4, 7)
michael@0: YANY(I422ToBGRARow_Any_SSSE3, I422ToBGRARow_Unaligned_SSSE3, I422ToBGRARow_C,
michael@0:      1, 4, 7)
michael@0: YANY(I422ToABGRRow_Any_SSSE3, I422ToABGRRow_Unaligned_SSSE3, I422ToABGRRow_C,
michael@0:      1, 4, 7)
michael@0: YANY(I422ToRGBARow_Any_SSSE3, I422ToRGBARow_Unaligned_SSSE3, I422ToRGBARow_C,
michael@0:      1, 4, 7)
michael@0: // I422ToRGB565Row_SSSE3 is unaligned.
michael@0: YANY(I422ToARGB4444Row_Any_SSSE3, I422ToARGB4444Row_SSSE3, I422ToARGB4444Row_C,
michael@0:      1, 2, 7)
michael@0: YANY(I422ToARGB1555Row_Any_SSSE3, I422ToARGB1555Row_SSSE3, I422ToARGB1555Row_C,
michael@0:      1, 2, 7)
michael@0: YANY(I422ToRGB565Row_Any_SSSE3, I422ToRGB565Row_SSSE3, I422ToRGB565Row_C,
michael@0:      1, 2, 7)
michael@0: // I422ToRGB24Row_SSSE3 is unaligned.
michael@0: YANY(I422ToRGB24Row_Any_SSSE3, I422ToRGB24Row_SSSE3, I422ToRGB24Row_C, 1, 3, 7)
michael@0: YANY(I422ToRAWRow_Any_SSSE3, I422ToRAWRow_SSSE3, I422ToRAWRow_C, 1, 3, 7)
michael@0: YANY(I422ToYUY2Row_Any_SSE2, I422ToYUY2Row_SSE2, I422ToYUY2Row_C, 1, 2, 15)
michael@0: YANY(I422ToUYVYRow_Any_SSE2, I422ToUYVYRow_SSE2, I422ToUYVYRow_C, 1, 2, 15)
michael@0: #endif  // HAS_I422TOARGBROW_SSSE3
michael@0: #ifdef HAS_I422TOARGBROW_AVX2
michael@0: YANY(I422ToARGBRow_Any_AVX2, I422ToARGBRow_AVX2, I422ToARGBRow_C, 1, 4, 15)
michael@0: #endif  // HAS_I422TOARGBROW_AVX2
michael@0: #ifdef HAS_I422TOARGBROW_NEON
michael@0: YANY(I444ToARGBRow_Any_NEON, I444ToARGBRow_NEON, I444ToARGBRow_C, 0, 4, 7)
michael@0: YANY(I422ToARGBRow_Any_NEON, I422ToARGBRow_NEON, I422ToARGBRow_C, 1, 4, 7)
michael@0: YANY(I411ToARGBRow_Any_NEON, I411ToARGBRow_NEON, I411ToARGBRow_C, 2, 4, 7)
michael@0: YANY(I422ToBGRARow_Any_NEON, I422ToBGRARow_NEON, I422ToBGRARow_C, 1, 4, 7)
michael@0: YANY(I422ToABGRRow_Any_NEON, I422ToABGRRow_NEON, I422ToABGRRow_C, 1, 4, 7)
michael@0: YANY(I422ToRGBARow_Any_NEON, I422ToRGBARow_NEON, I422ToRGBARow_C, 1, 4, 7)
michael@0: YANY(I422ToRGB24Row_Any_NEON, I422ToRGB24Row_NEON, I422ToRGB24Row_C, 1, 3, 7)
michael@0: YANY(I422ToRAWRow_Any_NEON, I422ToRAWRow_NEON, I422ToRAWRow_C, 1, 3, 7)
michael@0: YANY(I422ToARGB4444Row_Any_NEON, I422ToARGB4444Row_NEON, I422ToARGB4444Row_C,
michael@0:      1, 2, 7)
michael@0: YANY(I422ToARGB1555Row_Any_NEON, I422ToARGB1555Row_NEON, I422ToARGB1555Row_C,
michael@0:      1, 2, 7)
michael@0: YANY(I422ToRGB565Row_Any_NEON, I422ToRGB565Row_NEON, I422ToRGB565Row_C, 1, 2, 7)
michael@0: YANY(I422ToYUY2Row_Any_NEON, I422ToYUY2Row_NEON, I422ToYUY2Row_C, 1, 2, 15)
michael@0: YANY(I422ToUYVYRow_Any_NEON, I422ToUYVYRow_NEON, I422ToUYVYRow_C, 1, 2, 15)
michael@0: #endif  // HAS_I422TOARGBROW_NEON
michael@0: #undef YANY
michael@0: 
michael@0: // Wrappers to handle odd width
michael@0: #define NV2NY(NAMEANY, NV12TORGB_SIMD, NV12TORGB_C, UV_SHIFT, BPP)             \
michael@0:     void NAMEANY(const uint8* y_buf,                                           \
michael@0:                  const uint8* uv_buf,                                          \
michael@0:                  uint8* rgb_buf,                                               \
michael@0:                  int width) {                                                  \
michael@0:       int n = width & ~7;                                                      \
michael@0:       NV12TORGB_SIMD(y_buf, uv_buf, rgb_buf, n);                               \
michael@0:       NV12TORGB_C(y_buf + n,                                                   \
michael@0:                   uv_buf + (n >> UV_SHIFT),                                    \
michael@0:                   rgb_buf + n * BPP, width & 7);                               \
michael@0:     }
michael@0: 
michael@0: #ifdef HAS_NV12TOARGBROW_SSSE3
michael@0: NV2NY(NV12ToARGBRow_Any_SSSE3, NV12ToARGBRow_Unaligned_SSSE3, NV12ToARGBRow_C,
michael@0:       0, 4)
michael@0: NV2NY(NV21ToARGBRow_Any_SSSE3, NV21ToARGBRow_Unaligned_SSSE3, NV21ToARGBRow_C,
michael@0:       0, 4)
michael@0: #endif  // HAS_NV12TOARGBROW_SSSE3
michael@0: #ifdef HAS_NV12TOARGBROW_NEON
michael@0: NV2NY(NV12ToARGBRow_Any_NEON, NV12ToARGBRow_NEON, NV12ToARGBRow_C, 0, 4)
michael@0: NV2NY(NV21ToARGBRow_Any_NEON, NV21ToARGBRow_NEON, NV21ToARGBRow_C, 0, 4)
michael@0: #endif  // HAS_NV12TOARGBROW_NEON
michael@0: #ifdef HAS_NV12TORGB565ROW_SSSE3
michael@0: NV2NY(NV12ToRGB565Row_Any_SSSE3, NV12ToRGB565Row_SSSE3, NV12ToRGB565Row_C,
michael@0:       0, 2)
michael@0: NV2NY(NV21ToRGB565Row_Any_SSSE3, NV21ToRGB565Row_SSSE3, NV21ToRGB565Row_C,
michael@0:       0, 2)
michael@0: #endif  // HAS_NV12TORGB565ROW_SSSE3
michael@0: #ifdef HAS_NV12TORGB565ROW_NEON
michael@0: NV2NY(NV12ToRGB565Row_Any_NEON, NV12ToRGB565Row_NEON, NV12ToRGB565Row_C, 0, 2)
michael@0: NV2NY(NV21ToRGB565Row_Any_NEON, NV21ToRGB565Row_NEON, NV21ToRGB565Row_C, 0, 2)
michael@0: #endif  // HAS_NV12TORGB565ROW_NEON
michael@0: #undef NVANY
michael@0: 
michael@0: #define RGBANY(NAMEANY, ARGBTORGB_SIMD, ARGBTORGB_C, MASK, SBPP, BPP)          \
michael@0:     void NAMEANY(const uint8* src,                                             \
michael@0:                  uint8* dst,                                                   \
michael@0:                  int width) {                                                  \
michael@0:       int n = width & ~MASK;                                                   \
michael@0:       ARGBTORGB_SIMD(src, dst, n);                                             \
michael@0:       ARGBTORGB_C(src + n * SBPP, dst + n * BPP, width & MASK);                \
michael@0:     }
michael@0: 
michael@0: #if defined(HAS_ARGBTORGB24ROW_SSSE3)
michael@0: RGBANY(ARGBToRGB24Row_Any_SSSE3, ARGBToRGB24Row_SSSE3, ARGBToRGB24Row_C,
michael@0:        15, 4, 3)
michael@0: RGBANY(ARGBToRAWRow_Any_SSSE3, ARGBToRAWRow_SSSE3, ARGBToRAWRow_C,
michael@0:        15, 4, 3)
michael@0: RGBANY(ARGBToRGB565Row_Any_SSE2, ARGBToRGB565Row_SSE2, ARGBToRGB565Row_C,
michael@0:        3, 4, 2)
michael@0: RGBANY(ARGBToARGB1555Row_Any_SSE2, ARGBToARGB1555Row_SSE2, ARGBToARGB1555Row_C,
michael@0:        3, 4, 2)
michael@0: RGBANY(ARGBToARGB4444Row_Any_SSE2, ARGBToARGB4444Row_SSE2, ARGBToARGB4444Row_C,
michael@0:        3, 4, 2)
michael@0: #endif
michael@0: #if defined(HAS_I400TOARGBROW_SSE2)
michael@0: RGBANY(I400ToARGBRow_Any_SSE2, I400ToARGBRow_Unaligned_SSE2, I400ToARGBRow_C,
michael@0:        7, 1, 4)
michael@0: #endif
michael@0: #if defined(HAS_YTOARGBROW_SSE2)
michael@0: RGBANY(YToARGBRow_Any_SSE2, YToARGBRow_SSE2, YToARGBRow_C,
michael@0:        7, 1, 4)
michael@0: RGBANY(YUY2ToARGBRow_Any_SSSE3, YUY2ToARGBRow_Unaligned_SSSE3, YUY2ToARGBRow_C,
michael@0:        15, 2, 4)
michael@0: RGBANY(UYVYToARGBRow_Any_SSSE3, UYVYToARGBRow_Unaligned_SSSE3, UYVYToARGBRow_C,
michael@0:        15, 2, 4)
michael@0: // These require alignment on ARGB, so C is used for remainder.
michael@0: RGBANY(RGB24ToARGBRow_Any_SSSE3, RGB24ToARGBRow_SSSE3, RGB24ToARGBRow_C,
michael@0:        15, 3, 4)
michael@0: RGBANY(RAWToARGBRow_Any_SSSE3, RAWToARGBRow_SSSE3, RAWToARGBRow_C,
michael@0:        15, 3, 4)
michael@0: RGBANY(RGB565ToARGBRow_Any_SSE2, RGB565ToARGBRow_SSE2, RGB565ToARGBRow_C,
michael@0:        7, 2, 4)
michael@0: RGBANY(ARGB1555ToARGBRow_Any_SSE2, ARGB1555ToARGBRow_SSE2, ARGB1555ToARGBRow_C,
michael@0:        7, 2, 4)
michael@0: RGBANY(ARGB4444ToARGBRow_Any_SSE2, ARGB4444ToARGBRow_SSE2, ARGB4444ToARGBRow_C,
michael@0:        7, 2, 4)
michael@0: #endif
michael@0: #if defined(HAS_ARGBTORGB24ROW_NEON)
michael@0: RGBANY(ARGBToRGB24Row_Any_NEON, ARGBToRGB24Row_NEON, ARGBToRGB24Row_C, 7, 4, 3)
michael@0: RGBANY(ARGBToRAWRow_Any_NEON, ARGBToRAWRow_NEON, ARGBToRAWRow_C, 7, 4, 3)
michael@0: RGBANY(ARGBToRGB565Row_Any_NEON, ARGBToRGB565Row_NEON, ARGBToRGB565Row_C,
michael@0:        7, 4, 2)
michael@0: RGBANY(ARGBToARGB1555Row_Any_NEON, ARGBToARGB1555Row_NEON, ARGBToARGB1555Row_C,
michael@0:        7, 4, 2)
michael@0: RGBANY(ARGBToARGB4444Row_Any_NEON, ARGBToARGB4444Row_NEON, ARGBToARGB4444Row_C,
michael@0:        7, 4, 2)
michael@0: RGBANY(I400ToARGBRow_Any_NEON, I400ToARGBRow_NEON, I400ToARGBRow_C,
michael@0:        7, 1, 4)
michael@0: RGBANY(YToARGBRow_Any_NEON, YToARGBRow_NEON, YToARGBRow_C,
michael@0:        7, 1, 4)
michael@0: RGBANY(YUY2ToARGBRow_Any_NEON, YUY2ToARGBRow_NEON, YUY2ToARGBRow_C,
michael@0:        7, 2, 4)
michael@0: RGBANY(UYVYToARGBRow_Any_NEON, UYVYToARGBRow_NEON, UYVYToARGBRow_C,
michael@0:        7, 2, 4)
michael@0: #endif
michael@0: #undef RGBANY
michael@0: 
michael@0: // ARGB to Bayer does multiple of 4 pixels, SSSE3 aligned src, unaligned dst.
michael@0: #define BAYERANY(NAMEANY, ARGBTORGB_SIMD, ARGBTORGB_C, MASK, SBPP, BPP)        \
michael@0:     void NAMEANY(const uint8* src,                                             \
michael@0:                  uint8* dst, uint32 selector,                                  \
michael@0:                  int width) {                                                  \
michael@0:       int n = width & ~MASK;                                                   \
michael@0:       ARGBTORGB_SIMD(src, dst, selector, n);                                   \
michael@0:       ARGBTORGB_C(src + n * SBPP, dst + n * BPP, selector, width & MASK);      \
michael@0:     }
michael@0: 
michael@0: #if defined(HAS_ARGBTOBAYERROW_SSSE3)
michael@0: BAYERANY(ARGBToBayerRow_Any_SSSE3, ARGBToBayerRow_SSSE3, ARGBToBayerRow_C,
michael@0:          7, 4, 1)
michael@0: #endif
michael@0: #if defined(HAS_ARGBTOBAYERROW_NEON)
michael@0: BAYERANY(ARGBToBayerRow_Any_NEON, ARGBToBayerRow_NEON, ARGBToBayerRow_C,
michael@0:          7, 4, 1)
michael@0: #endif
michael@0: #if defined(HAS_ARGBTOBAYERGGROW_SSE2)
michael@0: BAYERANY(ARGBToBayerGGRow_Any_SSE2, ARGBToBayerGGRow_SSE2, ARGBToBayerGGRow_C,
michael@0:          7, 4, 1)
michael@0: #endif
michael@0: #if defined(HAS_ARGBTOBAYERGGROW_NEON)
michael@0: BAYERANY(ARGBToBayerGGRow_Any_NEON, ARGBToBayerGGRow_NEON, ARGBToBayerGGRow_C,
michael@0:          7, 4, 1)
michael@0: #endif
michael@0: 
michael@0: #undef BAYERANY
michael@0: 
michael@0: // RGB/YUV to Y does multiple of 16 with SIMD and last 16 with SIMD.
michael@0: #define YANY(NAMEANY, ARGBTOY_SIMD, SBPP, BPP, NUM)                            \
michael@0:     void NAMEANY(const uint8* src_argb, uint8* dst_y, int width) {             \
michael@0:       ARGBTOY_SIMD(src_argb, dst_y, width - NUM);                              \
michael@0:       ARGBTOY_SIMD(src_argb + (width - NUM) * SBPP,                            \
michael@0:                    dst_y + (width - NUM) * BPP, NUM);                          \
michael@0:     }
michael@0: 
michael@0: #ifdef HAS_ARGBTOYROW_AVX2
michael@0: YANY(ARGBToYRow_Any_AVX2, ARGBToYRow_AVX2, 4, 1, 32)
michael@0: YANY(ARGBToYJRow_Any_AVX2, ARGBToYJRow_AVX2, 4, 1, 32)
michael@0: YANY(YUY2ToYRow_Any_AVX2, YUY2ToYRow_AVX2, 2, 1, 32)
michael@0: YANY(UYVYToYRow_Any_AVX2, UYVYToYRow_AVX2, 2, 1, 32)
michael@0: #endif
michael@0: #ifdef HAS_ARGBTOYROW_SSSE3
michael@0: YANY(ARGBToYRow_Any_SSSE3, ARGBToYRow_Unaligned_SSSE3, 4, 1, 16)
michael@0: #endif
michael@0: #ifdef HAS_BGRATOYROW_SSSE3
michael@0: YANY(BGRAToYRow_Any_SSSE3, BGRAToYRow_Unaligned_SSSE3, 4, 1, 16)
michael@0: YANY(ABGRToYRow_Any_SSSE3, ABGRToYRow_Unaligned_SSSE3, 4, 1, 16)
michael@0: YANY(RGBAToYRow_Any_SSSE3, RGBAToYRow_Unaligned_SSSE3, 4, 1, 16)
michael@0: YANY(YUY2ToYRow_Any_SSE2, YUY2ToYRow_Unaligned_SSE2, 2, 1, 16)
michael@0: YANY(UYVYToYRow_Any_SSE2, UYVYToYRow_Unaligned_SSE2, 2, 1, 16)
michael@0: #endif
michael@0: #ifdef HAS_ARGBTOYJROW_SSSE3
michael@0: YANY(ARGBToYJRow_Any_SSSE3, ARGBToYJRow_Unaligned_SSSE3, 4, 1, 16)
michael@0: #endif
michael@0: #ifdef HAS_ARGBTOYROW_NEON
michael@0: YANY(ARGBToYRow_Any_NEON, ARGBToYRow_NEON, 4, 1, 8)
michael@0: YANY(ARGBToYJRow_Any_NEON, ARGBToYJRow_NEON, 4, 1, 8)
michael@0: YANY(BGRAToYRow_Any_NEON, BGRAToYRow_NEON, 4, 1, 8)
michael@0: YANY(ABGRToYRow_Any_NEON, ABGRToYRow_NEON, 4, 1, 8)
michael@0: YANY(RGBAToYRow_Any_NEON, RGBAToYRow_NEON, 4, 1, 8)
michael@0: YANY(RGB24ToYRow_Any_NEON, RGB24ToYRow_NEON, 3, 1, 8)
michael@0: YANY(RAWToYRow_Any_NEON, RAWToYRow_NEON, 3, 1, 8)
michael@0: YANY(RGB565ToYRow_Any_NEON, RGB565ToYRow_NEON, 2, 1, 8)
michael@0: YANY(ARGB1555ToYRow_Any_NEON, ARGB1555ToYRow_NEON, 2, 1, 8)
michael@0: YANY(ARGB4444ToYRow_Any_NEON, ARGB4444ToYRow_NEON, 2, 1, 8)
michael@0: YANY(YUY2ToYRow_Any_NEON, YUY2ToYRow_NEON, 2, 1, 16)
michael@0: YANY(UYVYToYRow_Any_NEON, UYVYToYRow_NEON, 2, 1, 16)
michael@0: YANY(RGB24ToARGBRow_Any_NEON, RGB24ToARGBRow_NEON, 3, 4, 8)
michael@0: YANY(RAWToARGBRow_Any_NEON, RAWToARGBRow_NEON, 3, 4, 8)
michael@0: YANY(RGB565ToARGBRow_Any_NEON, RGB565ToARGBRow_NEON, 2, 4, 8)
michael@0: YANY(ARGB1555ToARGBRow_Any_NEON, ARGB1555ToARGBRow_NEON, 2, 4, 8)
michael@0: YANY(ARGB4444ToARGBRow_Any_NEON, ARGB4444ToARGBRow_NEON, 2, 4, 8)
michael@0: #endif
michael@0: #undef YANY
michael@0: 
michael@0: #define YANY(NAMEANY, ARGBTOY_SIMD, ARGBTOY_C, SBPP, BPP, MASK)                \
michael@0:     void NAMEANY(const uint8* src_argb, uint8* dst_y, int width) {             \
michael@0:       int n = width & ~MASK;                                                   \
michael@0:       ARGBTOY_SIMD(src_argb, dst_y, n);                                        \
michael@0:       ARGBTOY_C(src_argb + n * SBPP,                                           \
michael@0:                 dst_y  + n * BPP, width & MASK);                               \
michael@0:     }
michael@0: 
michael@0: // Attenuate is destructive so last16 method can not be used due to overlap.
michael@0: #ifdef HAS_ARGBATTENUATEROW_SSSE3
michael@0: YANY(ARGBAttenuateRow_Any_SSSE3, ARGBAttenuateRow_SSSE3, ARGBAttenuateRow_C,
michael@0:      4, 4, 3)
michael@0: #endif
michael@0: #ifdef HAS_ARGBATTENUATEROW_SSE2
michael@0: YANY(ARGBAttenuateRow_Any_SSE2, ARGBAttenuateRow_SSE2, ARGBAttenuateRow_C,
michael@0:      4, 4, 3)
michael@0: #endif
michael@0: #ifdef HAS_ARGBUNATTENUATEROW_SSE2
michael@0: YANY(ARGBUnattenuateRow_Any_SSE2, ARGBUnattenuateRow_SSE2, ARGBUnattenuateRow_C,
michael@0:      4, 4, 3)
michael@0: #endif
michael@0: #ifdef HAS_ARGBATTENUATEROW_AVX2
michael@0: YANY(ARGBAttenuateRow_Any_AVX2, ARGBAttenuateRow_AVX2, ARGBAttenuateRow_C,
michael@0:      4, 4, 7)
michael@0: #endif
michael@0: #ifdef HAS_ARGBUNATTENUATEROW_AVX2
michael@0: YANY(ARGBUnattenuateRow_Any_AVX2, ARGBUnattenuateRow_AVX2, ARGBUnattenuateRow_C,
michael@0:      4, 4, 7)
michael@0: #endif
michael@0: #ifdef HAS_ARGBATTENUATEROW_NEON
michael@0: YANY(ARGBAttenuateRow_Any_NEON, ARGBAttenuateRow_NEON, ARGBAttenuateRow_C,
michael@0:      4, 4, 7)
michael@0: #endif
michael@0: #undef YANY
michael@0: 
michael@0: // RGB/YUV to UV does multiple of 16 with SIMD and remainder with C.
michael@0: #define UVANY(NAMEANY, ANYTOUV_SIMD, ANYTOUV_C, BPP, MASK)                     \
michael@0:     void NAMEANY(const uint8* src_argb, int src_stride_argb,                   \
michael@0:                  uint8* dst_u, uint8* dst_v, int width) {                      \
michael@0:       int n = width & ~MASK;                                                   \
michael@0:       ANYTOUV_SIMD(src_argb, src_stride_argb, dst_u, dst_v, n);                \
michael@0:       ANYTOUV_C(src_argb  + n * BPP, src_stride_argb,                          \
michael@0:                 dst_u + (n >> 1),                                              \
michael@0:                 dst_v + (n >> 1),                                              \
michael@0:                 width & MASK);                                                 \
michael@0:     }
michael@0: 
michael@0: #ifdef HAS_ARGBTOUVROW_AVX2
michael@0: UVANY(ARGBToUVRow_Any_AVX2, ARGBToUVRow_AVX2, ARGBToUVRow_C, 4, 31)
michael@0: UVANY(YUY2ToUVRow_Any_AVX2, YUY2ToUVRow_AVX2, YUY2ToUVRow_C, 2, 31)
michael@0: UVANY(UYVYToUVRow_Any_AVX2, UYVYToUVRow_AVX2, UYVYToUVRow_C, 2, 31)
michael@0: #endif
michael@0: #ifdef HAS_ARGBTOUVROW_SSSE3
michael@0: UVANY(ARGBToUVRow_Any_SSSE3, ARGBToUVRow_Unaligned_SSSE3, ARGBToUVRow_C, 4, 15)
michael@0: UVANY(ARGBToUVJRow_Any_SSSE3, ARGBToUVJRow_Unaligned_SSSE3, ARGBToUVJRow_C,
michael@0:       4, 15)
michael@0: UVANY(BGRAToUVRow_Any_SSSE3, BGRAToUVRow_Unaligned_SSSE3, BGRAToUVRow_C, 4, 15)
michael@0: UVANY(ABGRToUVRow_Any_SSSE3, ABGRToUVRow_Unaligned_SSSE3, ABGRToUVRow_C, 4, 15)
michael@0: UVANY(RGBAToUVRow_Any_SSSE3, RGBAToUVRow_Unaligned_SSSE3, RGBAToUVRow_C, 4, 15)
michael@0: UVANY(YUY2ToUVRow_Any_SSE2, YUY2ToUVRow_Unaligned_SSE2, YUY2ToUVRow_C, 2, 15)
michael@0: UVANY(UYVYToUVRow_Any_SSE2, UYVYToUVRow_Unaligned_SSE2, UYVYToUVRow_C, 2, 15)
michael@0: #endif
michael@0: #ifdef HAS_ARGBTOUVROW_NEON
michael@0: UVANY(ARGBToUVRow_Any_NEON, ARGBToUVRow_NEON, ARGBToUVRow_C, 4, 15)
michael@0: UVANY(ARGBToUVJRow_Any_NEON, ARGBToUVJRow_NEON, ARGBToUVJRow_C, 4, 15)
michael@0: UVANY(BGRAToUVRow_Any_NEON, BGRAToUVRow_NEON, BGRAToUVRow_C, 4, 15)
michael@0: UVANY(ABGRToUVRow_Any_NEON, ABGRToUVRow_NEON, ABGRToUVRow_C, 4, 15)
michael@0: UVANY(RGBAToUVRow_Any_NEON, RGBAToUVRow_NEON, RGBAToUVRow_C, 4, 15)
michael@0: UVANY(RGB24ToUVRow_Any_NEON, RGB24ToUVRow_NEON, RGB24ToUVRow_C, 3, 15)
michael@0: UVANY(RAWToUVRow_Any_NEON, RAWToUVRow_NEON, RAWToUVRow_C, 3, 15)
michael@0: UVANY(RGB565ToUVRow_Any_NEON, RGB565ToUVRow_NEON, RGB565ToUVRow_C, 2, 15)
michael@0: UVANY(ARGB1555ToUVRow_Any_NEON, ARGB1555ToUVRow_NEON, ARGB1555ToUVRow_C, 2, 15)
michael@0: UVANY(ARGB4444ToUVRow_Any_NEON, ARGB4444ToUVRow_NEON, ARGB4444ToUVRow_C, 2, 15)
michael@0: UVANY(YUY2ToUVRow_Any_NEON, YUY2ToUVRow_NEON, YUY2ToUVRow_C, 2, 15)
michael@0: UVANY(UYVYToUVRow_Any_NEON, UYVYToUVRow_NEON, UYVYToUVRow_C, 2, 15)
michael@0: #endif
michael@0: #undef UVANY
michael@0: 
michael@0: #define UV422ANY(NAMEANY, ANYTOUV_SIMD, ANYTOUV_C, BPP, MASK, SHIFT)           \
michael@0:     void NAMEANY(const uint8* src_uv,                                          \
michael@0:                  uint8* dst_u, uint8* dst_v, int width) {                      \
michael@0:       int n = width & ~MASK;                                                   \
michael@0:       ANYTOUV_SIMD(src_uv, dst_u, dst_v, n);                                   \
michael@0:       ANYTOUV_C(src_uv  + n * BPP,                                             \
michael@0:                 dst_u + (n >> SHIFT),                                          \
michael@0:                 dst_v + (n >> SHIFT),                                          \
michael@0:                 width & MASK);                                                 \
michael@0:     }
michael@0: 
michael@0: #ifdef HAS_ARGBTOUV444ROW_SSSE3
michael@0: UV422ANY(ARGBToUV444Row_Any_SSSE3, ARGBToUV444Row_Unaligned_SSSE3,
michael@0:          ARGBToUV444Row_C, 4, 15, 0)
michael@0: #endif
michael@0: #ifdef HAS_YUY2TOUV422ROW_AVX2
michael@0: UV422ANY(YUY2ToUV422Row_Any_AVX2, YUY2ToUV422Row_AVX2,
michael@0:          YUY2ToUV422Row_C, 2, 31, 1)
michael@0: UV422ANY(UYVYToUV422Row_Any_AVX2, UYVYToUV422Row_AVX2,
michael@0:          UYVYToUV422Row_C, 2, 31, 1)
michael@0: #endif
michael@0: #ifdef HAS_ARGBTOUVROW_SSSE3
michael@0: UV422ANY(ARGBToUV422Row_Any_SSSE3, ARGBToUV422Row_Unaligned_SSSE3,
michael@0:          ARGBToUV422Row_C, 4, 15, 1)
michael@0: UV422ANY(YUY2ToUV422Row_Any_SSE2, YUY2ToUV422Row_Unaligned_SSE2,
michael@0:          YUY2ToUV422Row_C, 2, 15, 1)
michael@0: UV422ANY(UYVYToUV422Row_Any_SSE2, UYVYToUV422Row_Unaligned_SSE2,
michael@0:          UYVYToUV422Row_C, 2, 15, 1)
michael@0: #endif
michael@0: #ifdef HAS_YUY2TOUV422ROW_NEON
michael@0: UV422ANY(ARGBToUV444Row_Any_NEON, ARGBToUV444Row_NEON,
michael@0:          ARGBToUV444Row_C, 4, 7, 0)
michael@0: UV422ANY(ARGBToUV422Row_Any_NEON, ARGBToUV422Row_NEON,
michael@0:          ARGBToUV422Row_C, 4, 15, 1)
michael@0: UV422ANY(ARGBToUV411Row_Any_NEON, ARGBToUV411Row_NEON,
michael@0:          ARGBToUV411Row_C, 4, 31, 2)
michael@0: UV422ANY(YUY2ToUV422Row_Any_NEON, YUY2ToUV422Row_NEON,
michael@0:          YUY2ToUV422Row_C, 2, 15, 1)
michael@0: UV422ANY(UYVYToUV422Row_Any_NEON, UYVYToUV422Row_NEON,
michael@0:          UYVYToUV422Row_C, 2, 15, 1)
michael@0: #endif
michael@0: #undef UV422ANY
michael@0: 
michael@0: #define SPLITUVROWANY(NAMEANY, ANYTOUV_SIMD, ANYTOUV_C, MASK)                  \
michael@0:     void NAMEANY(const uint8* src_uv,                                          \
michael@0:                  uint8* dst_u, uint8* dst_v, int width) {                      \
michael@0:       int n = width & ~MASK;                                                   \
michael@0:       ANYTOUV_SIMD(src_uv, dst_u, dst_v, n);                                   \
michael@0:       ANYTOUV_C(src_uv + n * 2,                                                \
michael@0:                 dst_u + n,                                                     \
michael@0:                 dst_v + n,                                                     \
michael@0:                 width & MASK);                                                 \
michael@0:     }
michael@0: 
michael@0: #ifdef HAS_SPLITUVROW_SSE2
michael@0: SPLITUVROWANY(SplitUVRow_Any_SSE2, SplitUVRow_Unaligned_SSE2, SplitUVRow_C, 15)
michael@0: #endif
michael@0: #ifdef HAS_SPLITUVROW_AVX2
michael@0: SPLITUVROWANY(SplitUVRow_Any_AVX2, SplitUVRow_AVX2, SplitUVRow_C, 31)
michael@0: #endif
michael@0: #ifdef HAS_SPLITUVROW_NEON
michael@0: SPLITUVROWANY(SplitUVRow_Any_NEON, SplitUVRow_NEON, SplitUVRow_C, 15)
michael@0: #endif
michael@0: #ifdef HAS_SPLITUVROW_MIPS_DSPR2
michael@0: SPLITUVROWANY(SplitUVRow_Any_MIPS_DSPR2, SplitUVRow_Unaligned_MIPS_DSPR2,
michael@0:               SplitUVRow_C, 15)
michael@0: #endif
michael@0: #undef SPLITUVROWANY
michael@0: 
michael@0: #define MERGEUVROW_ANY(NAMEANY, ANYTOUV_SIMD, ANYTOUV_C, MASK)                 \
michael@0:     void NAMEANY(const uint8* src_u, const uint8* src_v,                       \
michael@0:                  uint8* dst_uv, int width) {                                   \
michael@0:       int n = width & ~MASK;                                                   \
michael@0:       ANYTOUV_SIMD(src_u, src_v, dst_uv, n);                                   \
michael@0:       ANYTOUV_C(src_u + n,                                                     \
michael@0:                 src_v + n,                                                     \
michael@0:                 dst_uv + n * 2,                                                \
michael@0:                 width & MASK);                                                 \
michael@0:     }
michael@0: 
michael@0: #ifdef HAS_MERGEUVROW_SSE2
michael@0: MERGEUVROW_ANY(MergeUVRow_Any_SSE2, MergeUVRow_Unaligned_SSE2, MergeUVRow_C, 15)
michael@0: #endif
michael@0: #ifdef HAS_MERGEUVROW_AVX2
michael@0: MERGEUVROW_ANY(MergeUVRow_Any_AVX2, MergeUVRow_AVX2, MergeUVRow_C, 31)
michael@0: #endif
michael@0: #ifdef HAS_MERGEUVROW_NEON
michael@0: MERGEUVROW_ANY(MergeUVRow_Any_NEON, MergeUVRow_NEON, MergeUVRow_C, 15)
michael@0: #endif
michael@0: #undef MERGEUVROW_ANY
michael@0: 
michael@0: #define MATHROW_ANY(NAMEANY, ARGBMATH_SIMD, ARGBMATH_C, MASK)                  \
michael@0:     void NAMEANY(const uint8* src_argb0, const uint8* src_argb1,               \
michael@0:                  uint8* dst_argb, int width) {                                 \
michael@0:       int n = width & ~MASK;                                                   \
michael@0:       ARGBMATH_SIMD(src_argb0, src_argb1, dst_argb, n);                        \
michael@0:       ARGBMATH_C(src_argb0 + n * 4,                                            \
michael@0:                  src_argb1 + n * 4,                                            \
michael@0:                  dst_argb + n * 4,                                             \
michael@0:                  width & MASK);                                                \
michael@0:     }
michael@0: 
michael@0: #ifdef HAS_ARGBMULTIPLYROW_SSE2
michael@0: MATHROW_ANY(ARGBMultiplyRow_Any_SSE2, ARGBMultiplyRow_SSE2, ARGBMultiplyRow_C,
michael@0:             3)
michael@0: #endif
michael@0: #ifdef HAS_ARGBADDROW_SSE2
michael@0: MATHROW_ANY(ARGBAddRow_Any_SSE2, ARGBAddRow_SSE2, ARGBAddRow_C, 3)
michael@0: #endif
michael@0: #ifdef HAS_ARGBSUBTRACTROW_SSE2
michael@0: MATHROW_ANY(ARGBSubtractRow_Any_SSE2, ARGBSubtractRow_SSE2, ARGBSubtractRow_C,
michael@0:             3)
michael@0: #endif
michael@0: #ifdef HAS_ARGBMULTIPLYROW_AVX2
michael@0: MATHROW_ANY(ARGBMultiplyRow_Any_AVX2, ARGBMultiplyRow_AVX2, ARGBMultiplyRow_C,
michael@0:             7)
michael@0: #endif
michael@0: #ifdef HAS_ARGBADDROW_AVX2
michael@0: MATHROW_ANY(ARGBAddRow_Any_AVX2, ARGBAddRow_AVX2, ARGBAddRow_C, 7)
michael@0: #endif
michael@0: #ifdef HAS_ARGBSUBTRACTROW_AVX2
michael@0: MATHROW_ANY(ARGBSubtractRow_Any_AVX2, ARGBSubtractRow_AVX2, ARGBSubtractRow_C,
michael@0:             7)
michael@0: #endif
michael@0: #ifdef HAS_ARGBMULTIPLYROW_NEON
michael@0: MATHROW_ANY(ARGBMultiplyRow_Any_NEON, ARGBMultiplyRow_NEON, ARGBMultiplyRow_C,
michael@0:             7)
michael@0: #endif
michael@0: #ifdef HAS_ARGBADDROW_NEON
michael@0: MATHROW_ANY(ARGBAddRow_Any_NEON, ARGBAddRow_NEON, ARGBAddRow_C, 7)
michael@0: #endif
michael@0: #ifdef HAS_ARGBSUBTRACTROW_NEON
michael@0: MATHROW_ANY(ARGBSubtractRow_Any_NEON, ARGBSubtractRow_NEON, ARGBSubtractRow_C,
michael@0:             7)
michael@0: #endif
michael@0: #undef MATHROW_ANY
michael@0: 
michael@0: // Shuffle may want to work in place, so last16 method can not be used.
michael@0: #define YANY(NAMEANY, ARGBTOY_SIMD, ARGBTOY_C, SBPP, BPP, MASK)                \
michael@0:     void NAMEANY(const uint8* src_argb, uint8* dst_argb,                       \
michael@0:                  const uint8* shuffler, int width) {                           \
michael@0:       int n = width & ~MASK;                                                   \
michael@0:       ARGBTOY_SIMD(src_argb, dst_argb, shuffler, n);                           \
michael@0:       ARGBTOY_C(src_argb + n * SBPP,                                           \
michael@0:                 dst_argb  + n * BPP, shuffler, width & MASK);                  \
michael@0:     }
michael@0: 
michael@0: #ifdef HAS_ARGBSHUFFLEROW_SSE2
michael@0: YANY(ARGBShuffleRow_Any_SSE2, ARGBShuffleRow_SSE2,
michael@0:      ARGBShuffleRow_C, 4, 4, 3)
michael@0: #endif
michael@0: #ifdef HAS_ARGBSHUFFLEROW_SSSE3
michael@0: YANY(ARGBShuffleRow_Any_SSSE3, ARGBShuffleRow_Unaligned_SSSE3,
michael@0:      ARGBShuffleRow_C, 4, 4, 7)
michael@0: #endif
michael@0: #ifdef HAS_ARGBSHUFFLEROW_AVX2
michael@0: YANY(ARGBShuffleRow_Any_AVX2, ARGBShuffleRow_AVX2,
michael@0:      ARGBShuffleRow_C, 4, 4, 15)
michael@0: #endif
michael@0: #ifdef HAS_ARGBSHUFFLEROW_NEON
michael@0: YANY(ARGBShuffleRow_Any_NEON, ARGBShuffleRow_NEON,
michael@0:      ARGBShuffleRow_C, 4, 4, 3)
michael@0: #endif
michael@0: #undef YANY
michael@0: 
michael@0: // Interpolate may want to work in place, so last16 method can not be used.
michael@0: #define NANY(NAMEANY, TERP_SIMD, TERP_C, SBPP, BPP, MASK)                      \
michael@0:     void NAMEANY(uint8* dst_ptr, const uint8* src_ptr,                         \
michael@0:                  ptrdiff_t src_stride_ptr, int width,                          \
michael@0:                  int source_y_fraction) {                                      \
michael@0:       int n = width & ~MASK;                                                   \
michael@0:       TERP_SIMD(dst_ptr, src_ptr, src_stride_ptr,                              \
michael@0:                 n, source_y_fraction);                                         \
michael@0:       TERP_C(dst_ptr + n * BPP,                                                \
michael@0:              src_ptr + n * SBPP, src_stride_ptr,                               \
michael@0:              width & MASK, source_y_fraction);                                 \
michael@0:     }
michael@0: 
michael@0: #ifdef HAS_INTERPOLATEROW_AVX2
michael@0: NANY(InterpolateRow_Any_AVX2, InterpolateRow_AVX2,
michael@0:      InterpolateRow_C, 1, 1, 32)
michael@0: #endif
michael@0: #ifdef HAS_INTERPOLATEROW_SSSE3
michael@0: NANY(InterpolateRow_Any_SSSE3, InterpolateRow_Unaligned_SSSE3,
michael@0:      InterpolateRow_C, 1, 1, 15)
michael@0: #endif
michael@0: #ifdef HAS_INTERPOLATEROW_SSE2
michael@0: NANY(InterpolateRow_Any_SSE2, InterpolateRow_Unaligned_SSE2,
michael@0:      InterpolateRow_C, 1, 1, 15)
michael@0: #endif
michael@0: #ifdef HAS_INTERPOLATEROW_NEON
michael@0: NANY(InterpolateRow_Any_NEON, InterpolateRow_NEON,
michael@0:      InterpolateRow_C, 1, 1, 15)
michael@0: #endif
michael@0: #ifdef HAS_INTERPOLATEROW_MIPS_DSPR2
michael@0: NANY(InterpolateRow_Any_MIPS_DSPR2, InterpolateRow_MIPS_DSPR2,
michael@0:      InterpolateRow_C, 1, 1, 3)
michael@0: #endif
michael@0: #undef NANY
michael@0: 
michael@0: #ifdef __cplusplus
michael@0: }  // extern "C"
michael@0: }  // namespace libyuv
michael@0: #endif