WIP: 16x16 idct/recon merge

This patch eliminates the intermediate diff buffer usage by
combining the short idct and the add residual into one function.
The encoder can use the same code as well.

Change-Id: Iea7976b22b1927d24b8004d2a3fddae7ecca3ba1

1 files changed, 0 insertions, 59 deletions

diff --git a/vp9/decoder/x86/vp9_dequantize_sse2.c b/vp9/decoder/x86/vp9_dequantize_sse2.c
index 38fd5aaa4..796fc123c 100644
--- a/vp9/decoder/x86/vp9_dequantize_sse2.c
+++ b/vp9/decoder/x86/vp9_dequantize_sse2.c
@@ -122,65 +122,6 @@ void vp9_add_residual_8x8_sse2(const int16_t *diff, uint8_t *dest, int stride) {
   _mm_storel_epi64((__m128i *)(dest + 7 * stride), p6);
 }
 
-void vp9_add_residual_16x16_sse2(const int16_t *diff, uint8_t *dest,
-                                 int stride) {
-  const int width = 16;
-  int i = 4;
-  const __m128i zero = _mm_setzero_si128();
-
-  // Diff data
-  __m128i d0, d1, d2, d3, d4, d5, d6, d7;
-  __m128i p0, p1, p2, p3, p4, p5, p6, p7;
-
-  do {
-    d0 = _mm_load_si128((const __m128i *)(diff + 0 * width));
-    d1 = _mm_load_si128((const __m128i *)(diff + 0 * width + 8));
-    d2 = _mm_load_si128((const __m128i *)(diff + 1 * width));
-    d3 = _mm_load_si128((const __m128i *)(diff + 1 * width + 8));
-    d4 = _mm_load_si128((const __m128i *)(diff + 2 * width));
-    d5 = _mm_load_si128((const __m128i *)(diff + 2 * width + 8));
-    d6 = _mm_load_si128((const __m128i *)(diff + 3 * width));
-    d7 = _mm_load_si128((const __m128i *)(diff + 3 * width + 8));
-
-    // Prediction data.
-    p1 = _mm_load_si128((const __m128i *)(dest + 0 * stride));
-    p3 = _mm_load_si128((const __m128i *)(dest + 1 * stride));
-    p5 = _mm_load_si128((const __m128i *)(dest + 2 * stride));
-    p7 = _mm_load_si128((const __m128i *)(dest + 3 * stride));
-
-    p0 = _mm_unpacklo_epi8(p1, zero);
-    p1 = _mm_unpackhi_epi8(p1, zero);
-    p2 = _mm_unpacklo_epi8(p3, zero);
-    p3 = _mm_unpackhi_epi8(p3, zero);
-    p4 = _mm_unpacklo_epi8(p5, zero);
-    p5 = _mm_unpackhi_epi8(p5, zero);
-    p6 = _mm_unpacklo_epi8(p7, zero);
-    p7 = _mm_unpackhi_epi8(p7, zero);
-
-    p0 = _mm_add_epi16(p0, d0);
-    p1 = _mm_add_epi16(p1, d1);
-    p2 = _mm_add_epi16(p2, d2);
-    p3 = _mm_add_epi16(p3, d3);
-    p4 = _mm_add_epi16(p4, d4);
-    p5 = _mm_add_epi16(p5, d5);
-    p6 = _mm_add_epi16(p6, d6);
-    p7 = _mm_add_epi16(p7, d7);
-
-    p0 = _mm_packus_epi16(p0, p1);
-    p1 = _mm_packus_epi16(p2, p3);
-    p2 = _mm_packus_epi16(p4, p5);
-    p3 = _mm_packus_epi16(p6, p7);
-
-    _mm_store_si128((__m128i *)(dest + 0 * stride), p0);
-    _mm_store_si128((__m128i *)(dest + 1 * stride), p1);
-    _mm_store_si128((__m128i *)(dest + 2 * stride), p2);
-    _mm_store_si128((__m128i *)(dest + 3 * stride), p3);
-
-    diff += 4 * width;
-    dest += 4 * stride;
-  } while (--i);
-}
-
 void vp9_add_constant_residual_8x8_sse2(const int16_t diff, uint8_t *dest,
                                         int stride) {
   uint8_t abs_diff;