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/convert_from_argb.h"
michael@0: 
michael@0: #include "libyuv/basic_types.h"
michael@0: #include "libyuv/cpu_id.h"
michael@0: #include "libyuv/format_conversion.h"
michael@0: #include "libyuv/planar_functions.h"
michael@0: #include "libyuv/row.h"
michael@0: 
michael@0: #ifdef __cplusplus
michael@0: namespace libyuv {
michael@0: extern "C" {
michael@0: #endif
michael@0: 
michael@0: // ARGB little endian (bgra in memory) to I444
michael@0: LIBYUV_API
michael@0: int ARGBToI444(const uint8* src_argb, int src_stride_argb,
michael@0:                uint8* dst_y, int dst_stride_y,
michael@0:                uint8* dst_u, int dst_stride_u,
michael@0:                uint8* dst_v, int dst_stride_v,
michael@0:                int width, int height) {
michael@0:   if (!src_argb || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) {
michael@0:     return -1;
michael@0:   }
michael@0:   if (height < 0) {
michael@0:     height = -height;
michael@0:     src_argb = src_argb + (height - 1) * src_stride_argb;
michael@0:     src_stride_argb = -src_stride_argb;
michael@0:   }
michael@0:   // Coalesce rows.
michael@0:   if (src_stride_argb == width * 4 &&
michael@0:       dst_stride_y == width &&
michael@0:       dst_stride_u == width &&
michael@0:       dst_stride_v == width) {
michael@0:     width *= height;
michael@0:     height = 1;
michael@0:     src_stride_argb = dst_stride_y = dst_stride_u = dst_stride_v = 0;
michael@0:   }
michael@0:   void (*ARGBToYRow)(const uint8* src_argb, uint8* dst_y, int pix) =
michael@0:       ARGBToYRow_C;
michael@0:   void (*ARGBToUV444Row)(const uint8* src_argb, uint8* dst_u, uint8* dst_v,
michael@0:                          int pix) = ARGBToUV444Row_C;
michael@0: #if defined(HAS_ARGBTOUV444ROW_SSSE3)
michael@0:     if (TestCpuFlag(kCpuHasSSSE3) && width >= 16) {
michael@0:       ARGBToUV444Row = ARGBToUV444Row_Any_SSSE3;
michael@0:       if (IS_ALIGNED(width, 16)) {
michael@0:         ARGBToUV444Row = ARGBToUV444Row_Unaligned_SSSE3;
michael@0:         if (IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16)) {
michael@0:           ARGBToUV444Row = ARGBToUV444Row_SSSE3;
michael@0:         }
michael@0:       }
michael@0:   }
michael@0: #endif
michael@0: #if defined(HAS_ARGBTOYROW_SSSE3)
michael@0:   if (TestCpuFlag(kCpuHasSSSE3) && width >= 16) {
michael@0:     ARGBToYRow = ARGBToYRow_Any_SSSE3;
michael@0:     if (IS_ALIGNED(width, 16)) {
michael@0:       ARGBToYRow = ARGBToYRow_Unaligned_SSSE3;
michael@0:       if (IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16) &&
michael@0:           IS_ALIGNED(dst_y, 16) && IS_ALIGNED(dst_stride_y, 16)) {
michael@0:         ARGBToYRow = ARGBToYRow_SSSE3;
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: 
michael@0: #elif defined(HAS_ARGBTOYROW_NEON)
michael@0:   if (TestCpuFlag(kCpuHasNEON) && width >= 8) {
michael@0:     ARGBToYRow = ARGBToYRow_Any_NEON;
michael@0:     ARGBToUV444Row = ARGBToUV444Row_Any_NEON;
michael@0:     if (IS_ALIGNED(width, 8)) {
michael@0:       ARGBToYRow = ARGBToYRow_NEON;
michael@0:       ARGBToUV444Row = ARGBToUV444Row_NEON;
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: 
michael@0:   for (int y = 0; y < height; ++y) {
michael@0:     ARGBToUV444Row(src_argb, dst_u, dst_v, width);
michael@0:     ARGBToYRow(src_argb, dst_y, width);
michael@0:     src_argb += src_stride_argb;
michael@0:     dst_y += dst_stride_y;
michael@0:     dst_u += dst_stride_u;
michael@0:     dst_v += dst_stride_v;
michael@0:   }
michael@0:   return 0;
michael@0: }
michael@0: 
michael@0: // ARGB little endian (bgra in memory) to I422
michael@0: LIBYUV_API
michael@0: int ARGBToI422(const uint8* src_argb, int src_stride_argb,
michael@0:                uint8* dst_y, int dst_stride_y,
michael@0:                uint8* dst_u, int dst_stride_u,
michael@0:                uint8* dst_v, int dst_stride_v,
michael@0:                int width, int height) {
michael@0:   if (!src_argb || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) {
michael@0:     return -1;
michael@0:   }
michael@0:   if (height < 0) {
michael@0:     height = -height;
michael@0:     src_argb = src_argb + (height - 1) * src_stride_argb;
michael@0:     src_stride_argb = -src_stride_argb;
michael@0:   }
michael@0:   // Coalesce rows.
michael@0:   if (src_stride_argb == width * 4 &&
michael@0:       dst_stride_y == width &&
michael@0:       dst_stride_u * 2 == width &&
michael@0:       dst_stride_v * 2 == width) {
michael@0:     width *= height;
michael@0:     height = 1;
michael@0:     src_stride_argb = dst_stride_y = dst_stride_u = dst_stride_v = 0;
michael@0:   }
michael@0:   void (*ARGBToUV422Row)(const uint8* src_argb, uint8* dst_u, uint8* dst_v,
michael@0:                          int pix) = ARGBToUV422Row_C;
michael@0: #if defined(HAS_ARGBTOUV422ROW_SSSE3)
michael@0:   if (TestCpuFlag(kCpuHasSSSE3) && width >= 16) {
michael@0:     ARGBToUV422Row = ARGBToUV422Row_Any_SSSE3;
michael@0:     if (IS_ALIGNED(width, 16)) {
michael@0:       ARGBToUV422Row = ARGBToUV422Row_Unaligned_SSSE3;
michael@0:       if (IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16)) {
michael@0:         ARGBToUV422Row = ARGBToUV422Row_SSSE3;
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: 
michael@0:   void (*ARGBToYRow)(const uint8* src_argb, uint8* dst_y, int pix) =
michael@0:       ARGBToYRow_C;
michael@0: #if defined(HAS_ARGBTOYROW_SSSE3)
michael@0:   if (TestCpuFlag(kCpuHasSSSE3) && width >= 16) {
michael@0:     ARGBToYRow = ARGBToYRow_Any_SSSE3;
michael@0:     if (IS_ALIGNED(width, 16)) {
michael@0:       ARGBToYRow = ARGBToYRow_Unaligned_SSSE3;
michael@0:       if (IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16) &&
michael@0:           IS_ALIGNED(dst_y, 16) && IS_ALIGNED(dst_stride_y, 16)) {
michael@0:         ARGBToYRow = ARGBToYRow_SSSE3;
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #elif defined(HAS_ARGBTOYROW_NEON)
michael@0:   if (TestCpuFlag(kCpuHasNEON) && width >= 8) {
michael@0:     ARGBToYRow = ARGBToYRow_Any_NEON;
michael@0:     if (IS_ALIGNED(width, 8)) {
michael@0:       ARGBToYRow = ARGBToYRow_NEON;
michael@0:     }
michael@0:     if (width >= 16) {
michael@0:       ARGBToUV422Row = ARGBToUV422Row_Any_NEON;
michael@0:       if (IS_ALIGNED(width, 16)) {
michael@0:         ARGBToUV422Row = ARGBToUV422Row_NEON;
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: 
michael@0:   for (int y = 0; y < height; ++y) {
michael@0:     ARGBToUV422Row(src_argb, dst_u, dst_v, width);
michael@0:     ARGBToYRow(src_argb, dst_y, width);
michael@0:     src_argb += src_stride_argb;
michael@0:     dst_y += dst_stride_y;
michael@0:     dst_u += dst_stride_u;
michael@0:     dst_v += dst_stride_v;
michael@0:   }
michael@0:   return 0;
michael@0: }
michael@0: 
michael@0: // ARGB little endian (bgra in memory) to I411
michael@0: LIBYUV_API
michael@0: int ARGBToI411(const uint8* src_argb, int src_stride_argb,
michael@0:                uint8* dst_y, int dst_stride_y,
michael@0:                uint8* dst_u, int dst_stride_u,
michael@0:                uint8* dst_v, int dst_stride_v,
michael@0:                int width, int height) {
michael@0:   if (!src_argb || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) {
michael@0:     return -1;
michael@0:   }
michael@0:   if (height < 0) {
michael@0:     height = -height;
michael@0:     src_argb = src_argb + (height - 1) * src_stride_argb;
michael@0:     src_stride_argb = -src_stride_argb;
michael@0:   }
michael@0:   // Coalesce rows.
michael@0:   if (src_stride_argb == width * 4 &&
michael@0:       dst_stride_y == width &&
michael@0:       dst_stride_u * 4 == width &&
michael@0:       dst_stride_v * 4 == width) {
michael@0:     width *= height;
michael@0:     height = 1;
michael@0:     src_stride_argb = dst_stride_y = dst_stride_u = dst_stride_v = 0;
michael@0:   }
michael@0:   void (*ARGBToUV411Row)(const uint8* src_argb, uint8* dst_u, uint8* dst_v,
michael@0:                          int pix) = ARGBToUV411Row_C;
michael@0:   void (*ARGBToYRow)(const uint8* src_argb, uint8* dst_y, int pix) =
michael@0:       ARGBToYRow_C;
michael@0: #if defined(HAS_ARGBTOYROW_SSSE3)
michael@0:   if (TestCpuFlag(kCpuHasSSSE3) && width >= 16) {
michael@0:     ARGBToYRow = ARGBToYRow_Any_SSSE3;
michael@0:     if (IS_ALIGNED(width, 16)) {
michael@0:       ARGBToYRow = ARGBToYRow_Unaligned_SSSE3;
michael@0:       if (IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16) &&
michael@0:           IS_ALIGNED(dst_y, 16) && IS_ALIGNED(dst_stride_y, 16)) {
michael@0:         ARGBToYRow = ARGBToYRow_SSSE3;
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: #if defined(HAS_ARGBTOYROW_AVX2)
michael@0:   if (TestCpuFlag(kCpuHasAVX2) && width >= 32) {
michael@0:     ARGBToYRow = ARGBToYRow_Any_AVX2;
michael@0:     if (IS_ALIGNED(width, 32)) {
michael@0:       ARGBToYRow = ARGBToYRow_AVX2;
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: #if defined(HAS_ARGBTOYROW_NEON)
michael@0:   if (TestCpuFlag(kCpuHasNEON) && width >= 8) {
michael@0:     ARGBToYRow = ARGBToYRow_Any_NEON;
michael@0:     if (IS_ALIGNED(width, 8)) {
michael@0:       ARGBToYRow = ARGBToYRow_NEON;
michael@0:     }
michael@0:     if (width >= 32) {
michael@0:       ARGBToUV411Row = ARGBToUV411Row_Any_NEON;
michael@0:       if (IS_ALIGNED(width, 32)) {
michael@0:         ARGBToUV411Row = ARGBToUV411Row_NEON;
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: 
michael@0:   for (int y = 0; y < height; ++y) {
michael@0:     ARGBToUV411Row(src_argb, dst_u, dst_v, width);
michael@0:     ARGBToYRow(src_argb, dst_y, width);
michael@0:     src_argb += src_stride_argb;
michael@0:     dst_y += dst_stride_y;
michael@0:     dst_u += dst_stride_u;
michael@0:     dst_v += dst_stride_v;
michael@0:   }
michael@0:   return 0;
michael@0: }
michael@0: 
michael@0: LIBYUV_API
michael@0: int ARGBToNV12(const uint8* src_argb, int src_stride_argb,
michael@0:                uint8* dst_y, int dst_stride_y,
michael@0:                uint8* dst_uv, int dst_stride_uv,
michael@0:                int width, int height) {
michael@0:   if (!src_argb ||
michael@0:       !dst_y || !dst_uv ||
michael@0:       width <= 0 || height == 0) {
michael@0:     return -1;
michael@0:   }
michael@0:   // Negative height means invert the image.
michael@0:   if (height < 0) {
michael@0:     height = -height;
michael@0:     src_argb = src_argb + (height - 1) * src_stride_argb;
michael@0:     src_stride_argb = -src_stride_argb;
michael@0:   }
michael@0:   void (*ARGBToUVRow)(const uint8* src_argb0, int src_stride_argb,
michael@0:                       uint8* dst_u, uint8* dst_v, int width) = ARGBToUVRow_C;
michael@0:   void (*ARGBToYRow)(const uint8* src_argb, uint8* dst_y, int pix) =
michael@0:       ARGBToYRow_C;
michael@0: #if defined(HAS_ARGBTOYROW_SSSE3) && defined(HAS_ARGBTOUVROW_SSSE3)
michael@0:   if (TestCpuFlag(kCpuHasSSSE3) && width >= 16) {
michael@0:     ARGBToUVRow = ARGBToUVRow_Any_SSSE3;
michael@0:     ARGBToYRow = ARGBToYRow_Any_SSSE3;
michael@0:     if (IS_ALIGNED(width, 16)) {
michael@0:       ARGBToUVRow = ARGBToUVRow_Unaligned_SSSE3;
michael@0:       ARGBToYRow = ARGBToYRow_Unaligned_SSSE3;
michael@0:       if (IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16)) {
michael@0:         ARGBToUVRow = ARGBToUVRow_SSSE3;
michael@0:         if (IS_ALIGNED(dst_y, 16) && IS_ALIGNED(dst_stride_y, 16)) {
michael@0:           ARGBToYRow = ARGBToYRow_SSSE3;
michael@0:         }
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #elif defined(HAS_ARGBTOYROW_NEON)
michael@0:   if (TestCpuFlag(kCpuHasNEON) && width >= 8) {
michael@0:     ARGBToYRow = ARGBToYRow_Any_NEON;
michael@0:     if (IS_ALIGNED(width, 8)) {
michael@0:       ARGBToYRow = ARGBToYRow_NEON;
michael@0:     }
michael@0:     if (width >= 16) {
michael@0:       ARGBToUVRow = ARGBToUVRow_Any_NEON;
michael@0:       if (IS_ALIGNED(width, 16)) {
michael@0:         ARGBToUVRow = ARGBToUVRow_NEON;
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0:   int halfwidth = (width + 1) >> 1;
michael@0:   void (*MergeUVRow_)(const uint8* src_u, const uint8* src_v, uint8* dst_uv,
michael@0:                       int width) = MergeUVRow_C;
michael@0: #if defined(HAS_MERGEUVROW_SSE2)
michael@0:   if (TestCpuFlag(kCpuHasSSE2) && halfwidth >= 16) {
michael@0:     MergeUVRow_ = MergeUVRow_Any_SSE2;
michael@0:     if (IS_ALIGNED(halfwidth, 16)) {
michael@0:       MergeUVRow_ = MergeUVRow_Unaligned_SSE2;
michael@0:       if (IS_ALIGNED(dst_uv, 16) && IS_ALIGNED(dst_stride_uv, 16)) {
michael@0:         MergeUVRow_ = MergeUVRow_SSE2;
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: #if defined(HAS_MERGEUVROW_AVX2)
michael@0:   if (TestCpuFlag(kCpuHasAVX2) && halfwidth >= 32) {
michael@0:     MergeUVRow_ = MergeUVRow_Any_AVX2;
michael@0:     if (IS_ALIGNED(halfwidth, 32)) {
michael@0:       MergeUVRow_ = MergeUVRow_AVX2;
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: #if defined(HAS_MERGEUVROW_NEON)
michael@0:   if (TestCpuFlag(kCpuHasNEON) && halfwidth >= 16) {
michael@0:     MergeUVRow_ = MergeUVRow_Any_NEON;
michael@0:     if (IS_ALIGNED(halfwidth, 16)) {
michael@0:       MergeUVRow_ = MergeUVRow_NEON;
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: 
michael@0:   // Allocate a rows of uv.
michael@0:   align_buffer_64(row_u, ((halfwidth + 15) & ~15) * 2);
michael@0:   uint8* row_v = row_u + ((halfwidth + 15) & ~15);
michael@0: 
michael@0:   for (int y = 0; y < height - 1; y += 2) {
michael@0:     ARGBToUVRow(src_argb, src_stride_argb, row_u, row_v, width);
michael@0:     MergeUVRow_(row_u, row_v, dst_uv, halfwidth);
michael@0:     ARGBToYRow(src_argb, dst_y, width);
michael@0:     ARGBToYRow(src_argb + src_stride_argb, dst_y + dst_stride_y, width);
michael@0:     src_argb += src_stride_argb * 2;
michael@0:     dst_y += dst_stride_y * 2;
michael@0:     dst_uv += dst_stride_uv;
michael@0:   }
michael@0:   if (height & 1) {
michael@0:     ARGBToUVRow(src_argb, 0, row_u, row_v, width);
michael@0:     MergeUVRow_(row_u, row_v, dst_uv, halfwidth);
michael@0:     ARGBToYRow(src_argb, dst_y, width);
michael@0:   }
michael@0:   free_aligned_buffer_64(row_u);
michael@0:   return 0;
michael@0: }
michael@0: 
michael@0: // Same as NV12 but U and V swapped.
michael@0: LIBYUV_API
michael@0: int ARGBToNV21(const uint8* src_argb, int src_stride_argb,
michael@0:                uint8* dst_y, int dst_stride_y,
michael@0:                uint8* dst_uv, int dst_stride_uv,
michael@0:                int width, int height) {
michael@0:   if (!src_argb ||
michael@0:       !dst_y || !dst_uv ||
michael@0:       width <= 0 || height == 0) {
michael@0:     return -1;
michael@0:   }
michael@0:   // Negative height means invert the image.
michael@0:   if (height < 0) {
michael@0:     height = -height;
michael@0:     src_argb = src_argb + (height - 1) * src_stride_argb;
michael@0:     src_stride_argb = -src_stride_argb;
michael@0:   }
michael@0:   void (*ARGBToUVRow)(const uint8* src_argb0, int src_stride_argb,
michael@0:                       uint8* dst_u, uint8* dst_v, int width) = ARGBToUVRow_C;
michael@0:   void (*ARGBToYRow)(const uint8* src_argb, uint8* dst_y, int pix) =
michael@0:       ARGBToYRow_C;
michael@0: #if defined(HAS_ARGBTOYROW_SSSE3) && defined(HAS_ARGBTOUVROW_SSSE3)
michael@0:   if (TestCpuFlag(kCpuHasSSSE3) && width >= 16) {
michael@0:     ARGBToUVRow = ARGBToUVRow_Any_SSSE3;
michael@0:     ARGBToYRow = ARGBToYRow_Any_SSSE3;
michael@0:     if (IS_ALIGNED(width, 16)) {
michael@0:       ARGBToUVRow = ARGBToUVRow_Unaligned_SSSE3;
michael@0:       ARGBToYRow = ARGBToYRow_Unaligned_SSSE3;
michael@0:       if (IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16)) {
michael@0:         ARGBToUVRow = ARGBToUVRow_SSSE3;
michael@0:         if (IS_ALIGNED(dst_y, 16) && IS_ALIGNED(dst_stride_y, 16)) {
michael@0:           ARGBToYRow = ARGBToYRow_SSSE3;
michael@0:         }
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #elif defined(HAS_ARGBTOYROW_NEON)
michael@0:   if (TestCpuFlag(kCpuHasNEON) && width >= 8) {
michael@0:     ARGBToYRow = ARGBToYRow_Any_NEON;
michael@0:     if (IS_ALIGNED(width, 8)) {
michael@0:       ARGBToYRow = ARGBToYRow_NEON;
michael@0:     }
michael@0:     if (width >= 16) {
michael@0:       ARGBToUVRow = ARGBToUVRow_Any_NEON;
michael@0:       if (IS_ALIGNED(width, 16)) {
michael@0:         ARGBToUVRow = ARGBToUVRow_NEON;
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0:   int halfwidth = (width + 1) >> 1;
michael@0:   void (*MergeUVRow_)(const uint8* src_u, const uint8* src_v, uint8* dst_uv,
michael@0:                       int width) = MergeUVRow_C;
michael@0: #if defined(HAS_MERGEUVROW_SSE2)
michael@0:   if (TestCpuFlag(kCpuHasSSE2) && halfwidth >= 16) {
michael@0:     MergeUVRow_ = MergeUVRow_Any_SSE2;
michael@0:     if (IS_ALIGNED(halfwidth, 16)) {
michael@0:       MergeUVRow_ = MergeUVRow_Unaligned_SSE2;
michael@0:       if (IS_ALIGNED(dst_uv, 16) && IS_ALIGNED(dst_stride_uv, 16)) {
michael@0:         MergeUVRow_ = MergeUVRow_SSE2;
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: #if defined(HAS_MERGEUVROW_AVX2)
michael@0:   if (TestCpuFlag(kCpuHasAVX2) && halfwidth >= 32) {
michael@0:     MergeUVRow_ = MergeUVRow_Any_AVX2;
michael@0:     if (IS_ALIGNED(halfwidth, 32)) {
michael@0:       MergeUVRow_ = MergeUVRow_AVX2;
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: #if defined(HAS_MERGEUVROW_NEON)
michael@0:   if (TestCpuFlag(kCpuHasNEON) && halfwidth >= 16) {
michael@0:     MergeUVRow_ = MergeUVRow_Any_NEON;
michael@0:     if (IS_ALIGNED(halfwidth, 16)) {
michael@0:       MergeUVRow_ = MergeUVRow_NEON;
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: 
michael@0:   // Allocate a rows of uv.
michael@0:   align_buffer_64(row_u, ((halfwidth + 15) & ~15) * 2);
michael@0:   uint8* row_v = row_u + ((halfwidth + 15) & ~15);
michael@0: 
michael@0:   for (int y = 0; y < height - 1; y += 2) {
michael@0:     ARGBToUVRow(src_argb, src_stride_argb, row_u, row_v, width);
michael@0:     MergeUVRow_(row_v, row_u, dst_uv, halfwidth);
michael@0:     ARGBToYRow(src_argb, dst_y, width);
michael@0:     ARGBToYRow(src_argb + src_stride_argb, dst_y + dst_stride_y, width);
michael@0:     src_argb += src_stride_argb * 2;
michael@0:     dst_y += dst_stride_y * 2;
michael@0:     dst_uv += dst_stride_uv;
michael@0:   }
michael@0:   if (height & 1) {
michael@0:     ARGBToUVRow(src_argb, 0, row_u, row_v, width);
michael@0:     MergeUVRow_(row_v, row_u, dst_uv, halfwidth);
michael@0:     ARGBToYRow(src_argb, dst_y, width);
michael@0:   }
michael@0:   free_aligned_buffer_64(row_u);
michael@0:   return 0;
michael@0: }
michael@0: 
michael@0: // Convert ARGB to YUY2.
michael@0: LIBYUV_API
michael@0: int ARGBToYUY2(const uint8* src_argb, int src_stride_argb,
michael@0:                uint8* dst_yuy2, int dst_stride_yuy2,
michael@0:                int width, int height) {
michael@0:   if (!src_argb || !dst_yuy2 ||
michael@0:       width <= 0 || height == 0) {
michael@0:     return -1;
michael@0:   }
michael@0:   // Negative height means invert the image.
michael@0:   if (height < 0) {
michael@0:     height = -height;
michael@0:     dst_yuy2 = dst_yuy2 + (height - 1) * dst_stride_yuy2;
michael@0:     dst_stride_yuy2 = -dst_stride_yuy2;
michael@0:   }
michael@0:   // Coalesce rows.
michael@0:   if (src_stride_argb == width * 4 &&
michael@0:       dst_stride_yuy2 == width * 2) {
michael@0:     width *= height;
michael@0:     height = 1;
michael@0:     src_stride_argb = dst_stride_yuy2 = 0;
michael@0:   }
michael@0:   void (*ARGBToUV422Row)(const uint8* src_argb, uint8* dst_u, uint8* dst_v,
michael@0:                          int pix) = ARGBToUV422Row_C;
michael@0: #if defined(HAS_ARGBTOUV422ROW_SSSE3)
michael@0:   if (TestCpuFlag(kCpuHasSSSE3) && width >= 16) {
michael@0:     ARGBToUV422Row = ARGBToUV422Row_Any_SSSE3;
michael@0:     if (IS_ALIGNED(width, 16)) {
michael@0:       ARGBToUV422Row = ARGBToUV422Row_Unaligned_SSSE3;
michael@0:       if (IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16)) {
michael@0:         ARGBToUV422Row = ARGBToUV422Row_SSSE3;
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0:   void (*ARGBToYRow)(const uint8* src_argb, uint8* dst_y, int pix) =
michael@0:       ARGBToYRow_C;
michael@0: #if defined(HAS_ARGBTOYROW_SSSE3)
michael@0:   if (TestCpuFlag(kCpuHasSSSE3) && width >= 16) {
michael@0:     ARGBToYRow = ARGBToYRow_Any_SSSE3;
michael@0:     if (IS_ALIGNED(width, 16)) {
michael@0:       ARGBToYRow = ARGBToYRow_Unaligned_SSSE3;
michael@0:       if (IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16)) {
michael@0:         ARGBToYRow = ARGBToYRow_SSSE3;
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #elif defined(HAS_ARGBTOYROW_NEON)
michael@0:   if (TestCpuFlag(kCpuHasNEON) && width >= 8) {
michael@0:     ARGBToYRow = ARGBToYRow_Any_NEON;
michael@0:     if (IS_ALIGNED(width, 8)) {
michael@0:       ARGBToYRow = ARGBToYRow_NEON;
michael@0:     }
michael@0:     if (width >= 16) {
michael@0:       ARGBToUV422Row = ARGBToUV422Row_Any_NEON;
michael@0:       if (IS_ALIGNED(width, 16)) {
michael@0:         ARGBToUV422Row = ARGBToUV422Row_NEON;
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: 
michael@0:   void (*I422ToYUY2Row)(const uint8* src_y, const uint8* src_u,
michael@0:                         const uint8* src_v, uint8* dst_yuy2, int width) =
michael@0:       I422ToYUY2Row_C;
michael@0: #if defined(HAS_I422TOYUY2ROW_SSE2)
michael@0:   if (TestCpuFlag(kCpuHasSSE2) && width >= 16) {
michael@0:     I422ToYUY2Row = I422ToYUY2Row_Any_SSE2;
michael@0:     if (IS_ALIGNED(width, 16)) {
michael@0:       I422ToYUY2Row = I422ToYUY2Row_SSE2;
michael@0:     }
michael@0:   }
michael@0: #elif defined(HAS_I422TOYUY2ROW_NEON)
michael@0:   if (TestCpuFlag(kCpuHasNEON) && width >= 16) {
michael@0:     I422ToYUY2Row = I422ToYUY2Row_Any_NEON;
michael@0:     if (IS_ALIGNED(width, 16)) {
michael@0:       I422ToYUY2Row = I422ToYUY2Row_NEON;
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: 
michael@0:   // Allocate a rows of yuv.
michael@0:   align_buffer_64(row_y, ((width + 63) & ~63) * 2);
michael@0:   uint8* row_u = row_y + ((width + 63) & ~63);
michael@0:   uint8* row_v = row_u + ((width + 63) & ~63) / 2;
michael@0: 
michael@0:   for (int y = 0; y < height; ++y) {
michael@0:     ARGBToUV422Row(src_argb, row_u, row_v, width);
michael@0:     ARGBToYRow(src_argb, row_y, width);
michael@0:     I422ToYUY2Row(row_y, row_u, row_v, dst_yuy2, width);
michael@0:     src_argb += src_stride_argb;
michael@0:     dst_yuy2 += dst_stride_yuy2;
michael@0:   }
michael@0: 
michael@0:   free_aligned_buffer_64(row_y);
michael@0:   return 0;
michael@0: }
michael@0: 
michael@0: // Convert ARGB to UYVY.
michael@0: LIBYUV_API
michael@0: int ARGBToUYVY(const uint8* src_argb, int src_stride_argb,
michael@0:                uint8* dst_uyvy, int dst_stride_uyvy,
michael@0:                int width, int height) {
michael@0:   if (!src_argb || !dst_uyvy ||
michael@0:       width <= 0 || height == 0) {
michael@0:     return -1;
michael@0:   }
michael@0:   // Negative height means invert the image.
michael@0:   if (height < 0) {
michael@0:     height = -height;
michael@0:     dst_uyvy = dst_uyvy + (height - 1) * dst_stride_uyvy;
michael@0:     dst_stride_uyvy = -dst_stride_uyvy;
michael@0:   }
michael@0:   // Coalesce rows.
michael@0:   if (src_stride_argb == width * 4 &&
michael@0:       dst_stride_uyvy == width * 2) {
michael@0:     width *= height;
michael@0:     height = 1;
michael@0:     src_stride_argb = dst_stride_uyvy = 0;
michael@0:   }
michael@0:   void (*ARGBToUV422Row)(const uint8* src_argb, uint8* dst_u, uint8* dst_v,
michael@0:                          int pix) = ARGBToUV422Row_C;
michael@0: #if defined(HAS_ARGBTOUV422ROW_SSSE3)
michael@0:   if (TestCpuFlag(kCpuHasSSSE3) && width >= 16) {
michael@0:     ARGBToUV422Row = ARGBToUV422Row_Any_SSSE3;
michael@0:     if (IS_ALIGNED(width, 16)) {
michael@0:       ARGBToUV422Row = ARGBToUV422Row_Unaligned_SSSE3;
michael@0:       if (IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16)) {
michael@0:         ARGBToUV422Row = ARGBToUV422Row_SSSE3;
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0:   void (*ARGBToYRow)(const uint8* src_argb, uint8* dst_y, int pix) =
michael@0:       ARGBToYRow_C;
michael@0: #if defined(HAS_ARGBTOYROW_SSSE3)
michael@0:   if (TestCpuFlag(kCpuHasSSSE3) && width >= 16) {
michael@0:     ARGBToYRow = ARGBToYRow_Any_SSSE3;
michael@0:     if (IS_ALIGNED(width, 16)) {
michael@0:       ARGBToYRow = ARGBToYRow_Unaligned_SSSE3;
michael@0:       if (IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16)) {
michael@0:         ARGBToYRow = ARGBToYRow_SSSE3;
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #elif defined(HAS_ARGBTOYROW_NEON)
michael@0:   if (TestCpuFlag(kCpuHasNEON) && width >= 8) {
michael@0:     ARGBToYRow = ARGBToYRow_Any_NEON;
michael@0:     if (IS_ALIGNED(width, 8)) {
michael@0:       ARGBToYRow = ARGBToYRow_NEON;
michael@0:     }
michael@0:     if (width >= 16) {
michael@0:       ARGBToUV422Row = ARGBToUV422Row_Any_NEON;
michael@0:       if (IS_ALIGNED(width, 16)) {
michael@0:         ARGBToUV422Row = ARGBToUV422Row_NEON;
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: 
michael@0:   void (*I422ToUYVYRow)(const uint8* src_y, const uint8* src_u,
michael@0:                         const uint8* src_v, uint8* dst_uyvy, int width) =
michael@0:       I422ToUYVYRow_C;
michael@0: #if defined(HAS_I422TOUYVYROW_SSE2)
michael@0:   if (TestCpuFlag(kCpuHasSSE2) && width >= 16) {
michael@0:     I422ToUYVYRow = I422ToUYVYRow_Any_SSE2;
michael@0:     if (IS_ALIGNED(width, 16)) {
michael@0:       I422ToUYVYRow = I422ToUYVYRow_SSE2;
michael@0:     }
michael@0:   }
michael@0: #elif defined(HAS_I422TOUYVYROW_NEON)
michael@0:   if (TestCpuFlag(kCpuHasNEON) && width >= 16) {
michael@0:     I422ToUYVYRow = I422ToUYVYRow_Any_NEON;
michael@0:     if (IS_ALIGNED(width, 16)) {
michael@0:       I422ToUYVYRow = I422ToUYVYRow_NEON;
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: 
michael@0:   // Allocate a rows of yuv.
michael@0:   align_buffer_64(row_y, ((width + 63) & ~63) * 2);
michael@0:   uint8* row_u = row_y + ((width + 63) & ~63);
michael@0:   uint8* row_v = row_u + ((width + 63) & ~63) / 2;
michael@0: 
michael@0:   for (int y = 0; y < height; ++y) {
michael@0:     ARGBToUV422Row(src_argb, row_u, row_v, width);
michael@0:     ARGBToYRow(src_argb, row_y, width);
michael@0:     I422ToUYVYRow(row_y, row_u, row_v, dst_uyvy, width);
michael@0:     src_argb += src_stride_argb;
michael@0:     dst_uyvy += dst_stride_uyvy;
michael@0:   }
michael@0: 
michael@0:   free_aligned_buffer_64(row_y);
michael@0:   return 0;
michael@0: }
michael@0: 
michael@0: // Convert ARGB to I400.
michael@0: LIBYUV_API
michael@0: int ARGBToI400(const uint8* src_argb, int src_stride_argb,
michael@0:                uint8* dst_y, int dst_stride_y,
michael@0:                int width, int height) {
michael@0:   if (!src_argb || !dst_y || width <= 0 || height == 0) {
michael@0:     return -1;
michael@0:   }
michael@0:   if (height < 0) {
michael@0:     height = -height;
michael@0:     src_argb = src_argb + (height - 1) * src_stride_argb;
michael@0:     src_stride_argb = -src_stride_argb;
michael@0:   }
michael@0:   // Coalesce rows.
michael@0:   if (src_stride_argb == width * 4 &&
michael@0:       dst_stride_y == width) {
michael@0:     width *= height;
michael@0:     height = 1;
michael@0:     src_stride_argb = dst_stride_y = 0;
michael@0:   }
michael@0:   void (*ARGBToYRow)(const uint8* src_argb, uint8* dst_y, int pix) =
michael@0:       ARGBToYRow_C;
michael@0: #if defined(HAS_ARGBTOYROW_SSSE3)
michael@0:   if (TestCpuFlag(kCpuHasSSSE3) && width >= 16) {
michael@0:     ARGBToYRow = ARGBToYRow_Any_SSSE3;
michael@0:     if (IS_ALIGNED(width, 16)) {
michael@0:       ARGBToYRow = ARGBToYRow_Unaligned_SSSE3;
michael@0:       if (IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16) &&
michael@0:           IS_ALIGNED(dst_y, 16) && IS_ALIGNED(dst_stride_y, 16)) {
michael@0:         ARGBToYRow = ARGBToYRow_SSSE3;
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: #if defined(HAS_ARGBTOYROW_AVX2)
michael@0:   if (TestCpuFlag(kCpuHasAVX2) && width >= 32) {
michael@0:     ARGBToYRow = ARGBToYRow_Any_AVX2;
michael@0:     if (IS_ALIGNED(width, 32)) {
michael@0:       ARGBToYRow = ARGBToYRow_AVX2;
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: #if defined(HAS_ARGBTOYROW_NEON)
michael@0:   if (TestCpuFlag(kCpuHasNEON) && width >= 8) {
michael@0:     ARGBToYRow = ARGBToYRow_Any_NEON;
michael@0:     if (IS_ALIGNED(width, 8)) {
michael@0:       ARGBToYRow = ARGBToYRow_NEON;
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: 
michael@0:   for (int y = 0; y < height; ++y) {
michael@0:     ARGBToYRow(src_argb, dst_y, width);
michael@0:     src_argb += src_stride_argb;
michael@0:     dst_y += dst_stride_y;
michael@0:   }
michael@0:   return 0;
michael@0: }
michael@0: 
michael@0: // Shuffle table for converting ARGB to RGBA.
michael@0: static uvec8 kShuffleMaskARGBToRGBA = {
michael@0:   3u, 0u, 1u, 2u, 7u, 4u, 5u, 6u, 11u, 8u, 9u, 10u, 15u, 12u, 13u, 14u
michael@0: };
michael@0: 
michael@0: // Convert ARGB to RGBA.
michael@0: LIBYUV_API
michael@0: int ARGBToRGBA(const uint8* src_argb, int src_stride_argb,
michael@0:                uint8* dst_rgba, int dst_stride_rgba,
michael@0:                int width, int height) {
michael@0:   return ARGBShuffle(src_argb, src_stride_argb,
michael@0:                      dst_rgba, dst_stride_rgba,
michael@0:                      (const uint8*)(&kShuffleMaskARGBToRGBA),
michael@0:                      width, height);
michael@0: }
michael@0: 
michael@0: // Convert ARGB To RGB24.
michael@0: LIBYUV_API
michael@0: int ARGBToRGB24(const uint8* src_argb, int src_stride_argb,
michael@0:                 uint8* dst_rgb24, int dst_stride_rgb24,
michael@0:                 int width, int height) {
michael@0:   if (!src_argb || !dst_rgb24 || width <= 0 || height == 0) {
michael@0:     return -1;
michael@0:   }
michael@0:   if (height < 0) {
michael@0:     height = -height;
michael@0:     src_argb = src_argb + (height - 1) * src_stride_argb;
michael@0:     src_stride_argb = -src_stride_argb;
michael@0:   }
michael@0:   // Coalesce rows.
michael@0:   if (src_stride_argb == width * 4 &&
michael@0:       dst_stride_rgb24 == width * 3) {
michael@0:     width *= height;
michael@0:     height = 1;
michael@0:     src_stride_argb = dst_stride_rgb24 = 0;
michael@0:   }
michael@0:   void (*ARGBToRGB24Row)(const uint8* src_argb, uint8* dst_rgb, int pix) =
michael@0:       ARGBToRGB24Row_C;
michael@0: #if defined(HAS_ARGBTORGB24ROW_SSSE3)
michael@0:   if (TestCpuFlag(kCpuHasSSSE3) && width >= 16) {
michael@0:     ARGBToRGB24Row = ARGBToRGB24Row_Any_SSSE3;
michael@0:     if (IS_ALIGNED(width, 16)) {
michael@0:       ARGBToRGB24Row = ARGBToRGB24Row_SSSE3;
michael@0:     }
michael@0:   }
michael@0: #elif defined(HAS_ARGBTORGB24ROW_NEON)
michael@0:   if (TestCpuFlag(kCpuHasNEON) && width >= 8) {
michael@0:     ARGBToRGB24Row = ARGBToRGB24Row_Any_NEON;
michael@0:     if (IS_ALIGNED(width, 8)) {
michael@0:       ARGBToRGB24Row = ARGBToRGB24Row_NEON;
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: 
michael@0:   for (int y = 0; y < height; ++y) {
michael@0:     ARGBToRGB24Row(src_argb, dst_rgb24, width);
michael@0:     src_argb += src_stride_argb;
michael@0:     dst_rgb24 += dst_stride_rgb24;
michael@0:   }
michael@0:   return 0;
michael@0: }
michael@0: 
michael@0: // Convert ARGB To RAW.
michael@0: LIBYUV_API
michael@0: int ARGBToRAW(const uint8* src_argb, int src_stride_argb,
michael@0:               uint8* dst_raw, int dst_stride_raw,
michael@0:               int width, int height) {
michael@0:   if (!src_argb || !dst_raw || width <= 0 || height == 0) {
michael@0:     return -1;
michael@0:   }
michael@0:   if (height < 0) {
michael@0:     height = -height;
michael@0:     src_argb = src_argb + (height - 1) * src_stride_argb;
michael@0:     src_stride_argb = -src_stride_argb;
michael@0:   }
michael@0:   // Coalesce rows.
michael@0:   if (src_stride_argb == width * 4 &&
michael@0:       dst_stride_raw == width * 3) {
michael@0:     width *= height;
michael@0:     height = 1;
michael@0:     src_stride_argb = dst_stride_raw = 0;
michael@0:   }
michael@0:   void (*ARGBToRAWRow)(const uint8* src_argb, uint8* dst_rgb, int pix) =
michael@0:       ARGBToRAWRow_C;
michael@0: #if defined(HAS_ARGBTORAWROW_SSSE3)
michael@0:   if (TestCpuFlag(kCpuHasSSSE3) && width >= 16) {
michael@0:     ARGBToRAWRow = ARGBToRAWRow_Any_SSSE3;
michael@0:     if (IS_ALIGNED(width, 16)) {
michael@0:       ARGBToRAWRow = ARGBToRAWRow_SSSE3;
michael@0:     }
michael@0:   }
michael@0: #elif defined(HAS_ARGBTORAWROW_NEON)
michael@0:   if (TestCpuFlag(kCpuHasNEON) && width >= 8) {
michael@0:     ARGBToRAWRow = ARGBToRAWRow_Any_NEON;
michael@0:     if (IS_ALIGNED(width, 8)) {
michael@0:       ARGBToRAWRow = ARGBToRAWRow_NEON;
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: 
michael@0:   for (int y = 0; y < height; ++y) {
michael@0:     ARGBToRAWRow(src_argb, dst_raw, width);
michael@0:     src_argb += src_stride_argb;
michael@0:     dst_raw += dst_stride_raw;
michael@0:   }
michael@0:   return 0;
michael@0: }
michael@0: 
michael@0: // Convert ARGB To RGB565.
michael@0: LIBYUV_API
michael@0: int ARGBToRGB565(const uint8* src_argb, int src_stride_argb,
michael@0:                  uint8* dst_rgb565, int dst_stride_rgb565,
michael@0:                  int width, int height) {
michael@0:   if (!src_argb || !dst_rgb565 || width <= 0 || height == 0) {
michael@0:     return -1;
michael@0:   }
michael@0:   if (height < 0) {
michael@0:     height = -height;
michael@0:     src_argb = src_argb + (height - 1) * src_stride_argb;
michael@0:     src_stride_argb = -src_stride_argb;
michael@0:   }
michael@0:   // Coalesce rows.
michael@0:   if (src_stride_argb == width * 4 &&
michael@0:       dst_stride_rgb565 == width * 2) {
michael@0:     width *= height;
michael@0:     height = 1;
michael@0:     src_stride_argb = dst_stride_rgb565 = 0;
michael@0:   }
michael@0:   void (*ARGBToRGB565Row)(const uint8* src_argb, uint8* dst_rgb, int pix) =
michael@0:       ARGBToRGB565Row_C;
michael@0: #if defined(HAS_ARGBTORGB565ROW_SSE2)
michael@0:   if (TestCpuFlag(kCpuHasSSE2) && width >= 4 &&
michael@0:       IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16)) {
michael@0:     ARGBToRGB565Row = ARGBToRGB565Row_Any_SSE2;
michael@0:     if (IS_ALIGNED(width, 4)) {
michael@0:       ARGBToRGB565Row = ARGBToRGB565Row_SSE2;
michael@0:     }
michael@0:   }
michael@0: #elif defined(HAS_ARGBTORGB565ROW_NEON)
michael@0:   if (TestCpuFlag(kCpuHasNEON) && width >= 8) {
michael@0:     ARGBToRGB565Row = ARGBToRGB565Row_Any_NEON;
michael@0:     if (IS_ALIGNED(width, 8)) {
michael@0:       ARGBToRGB565Row = ARGBToRGB565Row_NEON;
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: 
michael@0:   for (int y = 0; y < height; ++y) {
michael@0:     ARGBToRGB565Row(src_argb, dst_rgb565, width);
michael@0:     src_argb += src_stride_argb;
michael@0:     dst_rgb565 += dst_stride_rgb565;
michael@0:   }
michael@0:   return 0;
michael@0: }
michael@0: 
michael@0: // Convert ARGB To ARGB1555.
michael@0: LIBYUV_API
michael@0: int ARGBToARGB1555(const uint8* src_argb, int src_stride_argb,
michael@0:                    uint8* dst_argb1555, int dst_stride_argb1555,
michael@0:                    int width, int height) {
michael@0:   if (!src_argb || !dst_argb1555 || width <= 0 || height == 0) {
michael@0:     return -1;
michael@0:   }
michael@0:   if (height < 0) {
michael@0:     height = -height;
michael@0:     src_argb = src_argb + (height - 1) * src_stride_argb;
michael@0:     src_stride_argb = -src_stride_argb;
michael@0:   }
michael@0:   // Coalesce rows.
michael@0:   if (src_stride_argb == width * 4 &&
michael@0:       dst_stride_argb1555 == width * 2) {
michael@0:     width *= height;
michael@0:     height = 1;
michael@0:     src_stride_argb = dst_stride_argb1555 = 0;
michael@0:   }
michael@0:   void (*ARGBToARGB1555Row)(const uint8* src_argb, uint8* dst_rgb, int pix) =
michael@0:       ARGBToARGB1555Row_C;
michael@0: #if defined(HAS_ARGBTOARGB1555ROW_SSE2)
michael@0:   if (TestCpuFlag(kCpuHasSSE2) && width >= 4 &&
michael@0:       IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16)) {
michael@0:     ARGBToARGB1555Row = ARGBToARGB1555Row_Any_SSE2;
michael@0:     if (IS_ALIGNED(width, 4)) {
michael@0:       ARGBToARGB1555Row = ARGBToARGB1555Row_SSE2;
michael@0:     }
michael@0:   }
michael@0: #elif defined(HAS_ARGBTOARGB1555ROW_NEON)
michael@0:   if (TestCpuFlag(kCpuHasNEON) && width >= 8) {
michael@0:     ARGBToARGB1555Row = ARGBToARGB1555Row_Any_NEON;
michael@0:     if (IS_ALIGNED(width, 8)) {
michael@0:       ARGBToARGB1555Row = ARGBToARGB1555Row_NEON;
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: 
michael@0:   for (int y = 0; y < height; ++y) {
michael@0:     ARGBToARGB1555Row(src_argb, dst_argb1555, width);
michael@0:     src_argb += src_stride_argb;
michael@0:     dst_argb1555 += dst_stride_argb1555;
michael@0:   }
michael@0:   return 0;
michael@0: }
michael@0: 
michael@0: // Convert ARGB To ARGB4444.
michael@0: LIBYUV_API
michael@0: int ARGBToARGB4444(const uint8* src_argb, int src_stride_argb,
michael@0:                    uint8* dst_argb4444, int dst_stride_argb4444,
michael@0:                    int width, int height) {
michael@0:   if (!src_argb || !dst_argb4444 || width <= 0 || height == 0) {
michael@0:     return -1;
michael@0:   }
michael@0:   if (height < 0) {
michael@0:     height = -height;
michael@0:     src_argb = src_argb + (height - 1) * src_stride_argb;
michael@0:     src_stride_argb = -src_stride_argb;
michael@0:   }
michael@0:   // Coalesce rows.
michael@0:   if (src_stride_argb == width * 4 &&
michael@0:       dst_stride_argb4444 == width * 2) {
michael@0:     width *= height;
michael@0:     height = 1;
michael@0:     src_stride_argb = dst_stride_argb4444 = 0;
michael@0:   }
michael@0:   void (*ARGBToARGB4444Row)(const uint8* src_argb, uint8* dst_rgb, int pix) =
michael@0:       ARGBToARGB4444Row_C;
michael@0: #if defined(HAS_ARGBTOARGB4444ROW_SSE2)
michael@0:   if (TestCpuFlag(kCpuHasSSE2) && width >= 4 &&
michael@0:       IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16)) {
michael@0:     ARGBToARGB4444Row = ARGBToARGB4444Row_Any_SSE2;
michael@0:     if (IS_ALIGNED(width, 4)) {
michael@0:       ARGBToARGB4444Row = ARGBToARGB4444Row_SSE2;
michael@0:     }
michael@0:   }
michael@0: #elif defined(HAS_ARGBTOARGB4444ROW_NEON)
michael@0:   if (TestCpuFlag(kCpuHasNEON) && width >= 8) {
michael@0:     ARGBToARGB4444Row = ARGBToARGB4444Row_Any_NEON;
michael@0:     if (IS_ALIGNED(width, 8)) {
michael@0:       ARGBToARGB4444Row = ARGBToARGB4444Row_NEON;
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: 
michael@0:   for (int y = 0; y < height; ++y) {
michael@0:     ARGBToARGB4444Row(src_argb, dst_argb4444, width);
michael@0:     src_argb += src_stride_argb;
michael@0:     dst_argb4444 += dst_stride_argb4444;
michael@0:   }
michael@0:   return 0;
michael@0: }
michael@0: 
michael@0: // Convert ARGB to J420. (JPeg full range I420).
michael@0: LIBYUV_API
michael@0: int ARGBToJ420(const uint8* src_argb, int src_stride_argb,
michael@0:                uint8* dst_yj, int dst_stride_yj,
michael@0:                uint8* dst_u, int dst_stride_u,
michael@0:                uint8* dst_v, int dst_stride_v,
michael@0:                int width, int height) {
michael@0:   if (!src_argb ||
michael@0:       !dst_yj || !dst_u || !dst_v ||
michael@0:       width <= 0 || height == 0) {
michael@0:     return -1;
michael@0:   }
michael@0:   // Negative height means invert the image.
michael@0:   if (height < 0) {
michael@0:     height = -height;
michael@0:     src_argb = src_argb + (height - 1) * src_stride_argb;
michael@0:     src_stride_argb = -src_stride_argb;
michael@0:   }
michael@0:   void (*ARGBToUVJRow)(const uint8* src_argb0, int src_stride_argb,
michael@0:                       uint8* dst_u, uint8* dst_v, int width) = ARGBToUVJRow_C;
michael@0:   void (*ARGBToYJRow)(const uint8* src_argb, uint8* dst_yj, int pix) =
michael@0:       ARGBToYJRow_C;
michael@0: #if defined(HAS_ARGBTOYJROW_SSSE3) && defined(HAS_ARGBTOUVJROW_SSSE3)
michael@0:   if (TestCpuFlag(kCpuHasSSSE3) && width >= 16) {
michael@0:     ARGBToUVJRow = ARGBToUVJRow_Any_SSSE3;
michael@0:     ARGBToYJRow = ARGBToYJRow_Any_SSSE3;
michael@0:     if (IS_ALIGNED(width, 16)) {
michael@0:       ARGBToUVJRow = ARGBToUVJRow_Unaligned_SSSE3;
michael@0:       ARGBToYJRow = ARGBToYJRow_Unaligned_SSSE3;
michael@0:       if (IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16)) {
michael@0:         ARGBToUVJRow = ARGBToUVJRow_SSSE3;
michael@0:         if (IS_ALIGNED(dst_yj, 16) && IS_ALIGNED(dst_stride_yj, 16)) {
michael@0:           ARGBToYJRow = ARGBToYJRow_SSSE3;
michael@0:         }
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: #if defined(HAS_ARGBTOYJROW_AVX2) && defined(HAS_ARGBTOUVJROW_AVX2)
michael@0:   if (TestCpuFlag(kCpuHasAVX2) && width >= 32) {
michael@0:     ARGBToYJRow = ARGBToYJRow_Any_AVX2;
michael@0:     if (IS_ALIGNED(width, 32)) {
michael@0:       ARGBToYJRow = ARGBToYJRow_AVX2;
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: #if defined(HAS_ARGBTOYJROW_NEON)
michael@0:   if (TestCpuFlag(kCpuHasNEON) && width >= 8) {
michael@0:     ARGBToYJRow = ARGBToYJRow_Any_NEON;
michael@0:     if (IS_ALIGNED(width, 8)) {
michael@0:       ARGBToYJRow = ARGBToYJRow_NEON;
michael@0:     }
michael@0:     if (width >= 16) {
michael@0:       ARGBToUVJRow = ARGBToUVJRow_Any_NEON;
michael@0:       if (IS_ALIGNED(width, 16)) {
michael@0:         ARGBToUVJRow = ARGBToUVJRow_NEON;
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: 
michael@0:   for (int y = 0; y < height - 1; y += 2) {
michael@0:     ARGBToUVJRow(src_argb, src_stride_argb, dst_u, dst_v, width);
michael@0:     ARGBToYJRow(src_argb, dst_yj, width);
michael@0:     ARGBToYJRow(src_argb + src_stride_argb, dst_yj + dst_stride_yj, width);
michael@0:     src_argb += src_stride_argb * 2;
michael@0:     dst_yj += dst_stride_yj * 2;
michael@0:     dst_u += dst_stride_u;
michael@0:     dst_v += dst_stride_v;
michael@0:   }
michael@0:   if (height & 1) {
michael@0:     ARGBToUVJRow(src_argb, 0, dst_u, dst_v, width);
michael@0:     ARGBToYJRow(src_argb, dst_yj, width);
michael@0:   }
michael@0:   return 0;
michael@0: }
michael@0: 
michael@0: // Convert ARGB to J400.
michael@0: LIBYUV_API
michael@0: int ARGBToJ400(const uint8* src_argb, int src_stride_argb,
michael@0:                uint8* dst_yj, int dst_stride_yj,
michael@0:                int width, int height) {
michael@0:   if (!src_argb || !dst_yj || width <= 0 || height == 0) {
michael@0:     return -1;
michael@0:   }
michael@0:   if (height < 0) {
michael@0:     height = -height;
michael@0:     src_argb = src_argb + (height - 1) * src_stride_argb;
michael@0:     src_stride_argb = -src_stride_argb;
michael@0:   }
michael@0:   // Coalesce rows.
michael@0:   if (src_stride_argb == width * 4 &&
michael@0:       dst_stride_yj == width) {
michael@0:     width *= height;
michael@0:     height = 1;
michael@0:     src_stride_argb = dst_stride_yj = 0;
michael@0:   }
michael@0:   void (*ARGBToYJRow)(const uint8* src_argb, uint8* dst_yj, int pix) =
michael@0:       ARGBToYJRow_C;
michael@0: #if defined(HAS_ARGBTOYJROW_SSSE3)
michael@0:   if (TestCpuFlag(kCpuHasSSSE3) && width >= 16) {
michael@0:     ARGBToYJRow = ARGBToYJRow_Any_SSSE3;
michael@0:     if (IS_ALIGNED(width, 16)) {
michael@0:       ARGBToYJRow = ARGBToYJRow_Unaligned_SSSE3;
michael@0:       if (IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16) &&
michael@0:           IS_ALIGNED(dst_yj, 16) && IS_ALIGNED(dst_stride_yj, 16)) {
michael@0:         ARGBToYJRow = ARGBToYJRow_SSSE3;
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: #if defined(HAS_ARGBTOYJROW_AVX2)
michael@0:   if (TestCpuFlag(kCpuHasAVX2) && width >= 32) {
michael@0:     ARGBToYJRow = ARGBToYJRow_Any_AVX2;
michael@0:     if (IS_ALIGNED(width, 32)) {
michael@0:       ARGBToYJRow = ARGBToYJRow_AVX2;
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: #if defined(HAS_ARGBTOYJROW_NEON)
michael@0:   if (TestCpuFlag(kCpuHasNEON) && width >= 8) {
michael@0:     ARGBToYJRow = ARGBToYJRow_Any_NEON;
michael@0:     if (IS_ALIGNED(width, 8)) {
michael@0:       ARGBToYJRow = ARGBToYJRow_NEON;
michael@0:     }
michael@0:   }
michael@0: #endif
michael@0: 
michael@0:   for (int y = 0; y < height; ++y) {
michael@0:     ARGBToYJRow(src_argb, dst_yj, width);
michael@0:     src_argb += src_stride_argb;
michael@0:     dst_yj += dst_stride_yj;
michael@0:   }
michael@0:   return 0;
michael@0: }
michael@0: 
michael@0: #ifdef __cplusplus
michael@0: }  // extern "C"
michael@0: }  // namespace libyuv
michael@0: #endif