gfx/ycbcr/yuv_convert_sse2.cpp

   1 // Copyright (c) 2010 The Chromium Authors. All rights reserved.
   2 // Use of this source code is governed by a BSD-style license that can be
   3 // found in the LICENSE file.
   4
   5 #include <emmintrin.h>
   6 #include "yuv_row.h"
   7
   8 namespace mozilla {
   9 namespace gfx {
  10
  11 // FilterRows combines two rows of the image using linear interpolation.
  12 // SSE2 version does 16 pixels at a time.
  13 void FilterRows_SSE2(uint8* ybuf, const uint8* y0_ptr, const uint8* y1_ptr,
  14                      int source_width, int source_y_fraction) {
  15   __m128i zero = _mm_setzero_si128();
  16   __m128i y1_fraction = _mm_set1_epi16(source_y_fraction);
  17   __m128i y0_fraction = _mm_set1_epi16(256 - source_y_fraction);
  18
  19   const __m128i* y0_ptr128 = reinterpret_cast<const __m128i*>(y0_ptr);
  20   const __m128i* y1_ptr128 = reinterpret_cast<const __m128i*>(y1_ptr);
  21   __m128i* dest128 = reinterpret_cast<__m128i*>(ybuf);
  22   __m128i* end128 = reinterpret_cast<__m128i*>(ybuf + source_width);
  23
  24   do {
  25     __m128i y0 = _mm_loadu_si128(y0_ptr128);
  26     __m128i y1 = _mm_loadu_si128(y1_ptr128);
  27     __m128i y2 = _mm_unpackhi_epi8(y0, zero);
  28     __m128i y3 = _mm_unpackhi_epi8(y1, zero);
  29     y0 = _mm_unpacklo_epi8(y0, zero);
  30     y1 = _mm_unpacklo_epi8(y1, zero);
  31     y0 = _mm_mullo_epi16(y0, y0_fraction);
  32     y1 = _mm_mullo_epi16(y1, y1_fraction);
  33     y2 = _mm_mullo_epi16(y2, y0_fraction);
  34     y3 = _mm_mullo_epi16(y3, y1_fraction);
  35     y0 = _mm_add_epi16(y0, y1);
  36     y2 = _mm_add_epi16(y2, y3);
  37     y0 = _mm_srli_epi16(y0, 8);
  38     y2 = _mm_srli_epi16(y2, 8);
  39     y0 = _mm_packus_epi16(y0, y2);
  40     *dest128++ = y0;
  41     ++y0_ptr128;
  42     ++y1_ptr128;
  43   } while (dest128 < end128);
  44 }
  45
  46 } // namespace gfx
  47 } // namespace mozilla