From 75b33c68c75f3d35e7b409aff87f12b307aa096b Mon Sep 17 00:00:00 2001
From: "Ronald S. Bultje" <rbultje@google.com>
Date: Tue, 9 Jul 2013 14:54:20 -0700
Subject: [PATCH] SSE/SSE2 assembly for 4x4/8x8/16x16/32x32 V intra prediction.

Change-Id: I55a6cfa2daba738cbc0c4a02f806893f7e556997
---
 vp9/common/vp9_rtcd_defs.sh           |  8 ++--
 vp9/common/x86/vp9_intrapred_sse2.asm | 63 +++++++++++++++++++++++++++
 2 files changed, 67 insertions(+), 4 deletions(-)

diff --git a/vp9/common/vp9_rtcd_defs.sh b/vp9/common/vp9_rtcd_defs.sh
index c5e108722..313cfd6b5 100644
--- a/vp9/common/vp9_rtcd_defs.sh
+++ b/vp9/common/vp9_rtcd_defs.sh
@@ -74,7 +74,7 @@ prototype void vp9_d153_predictor_4x4 "uint8_t *ypred_ptr, ptrdiff_t y_stride, u
 specialize vp9_d153_predictor_4x4
 
 prototype void vp9_v_predictor_4x4 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
-specialize vp9_v_predictor_4x4
+specialize vp9_v_predictor_4x4 sse
 
 prototype void vp9_tm_predictor_4x4 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
 specialize vp9_tm_predictor_4x4
@@ -113,7 +113,7 @@ prototype void vp9_d153_predictor_8x8 "uint8_t *ypred_ptr, ptrdiff_t y_stride, u
 specialize vp9_d153_predictor_8x8
 
 prototype void vp9_v_predictor_8x8 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
-specialize vp9_v_predictor_8x8
+specialize vp9_v_predictor_8x8 sse
 
 prototype void vp9_tm_predictor_8x8 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
 specialize vp9_tm_predictor_8x8
@@ -152,7 +152,7 @@ prototype void vp9_d153_predictor_16x16 "uint8_t *ypred_ptr, ptrdiff_t y_stride,
 specialize vp9_d153_predictor_16x16
 
 prototype void vp9_v_predictor_16x16 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
-specialize vp9_v_predictor_16x16
+specialize vp9_v_predictor_16x16 sse2
 
 prototype void vp9_tm_predictor_16x16 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
 specialize vp9_tm_predictor_16x16
@@ -191,7 +191,7 @@ prototype void vp9_d153_predictor_32x32 "uint8_t *ypred_ptr, ptrdiff_t y_stride,
 specialize vp9_d153_predictor_32x32
 
 prototype void vp9_v_predictor_32x32 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
-specialize vp9_v_predictor_32x32
+specialize vp9_v_predictor_32x32 sse2
 
 prototype void vp9_tm_predictor_32x32 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
 specialize vp9_tm_predictor_32x32
diff --git a/vp9/common/x86/vp9_intrapred_sse2.asm b/vp9/common/x86/vp9_intrapred_sse2.asm
index 550df2457..5aa3dba25 100644
--- a/vp9/common/x86/vp9_intrapred_sse2.asm
+++ b/vp9/common/x86/vp9_intrapred_sse2.asm
@@ -125,3 +125,66 @@ cglobal dc_predictor_32x32, 4, 4, 5, dst, stride, above, left
   dec              lines4d
   jnz .loop
   REP_RET
+
+INIT_MMX sse
+cglobal v_predictor_4x4, 3, 3, 1, dst, stride, above
+  movd                  m0, [aboveq]
+  movd      [dstq        ], m0
+  movd      [dstq+strideq], m0
+  lea                 dstq, [dstq+strideq*2]
+  movd      [dstq        ], m0
+  movd      [dstq+strideq], m0
+  RET
+
+INIT_MMX sse
+cglobal v_predictor_8x8, 3, 3, 1, dst, stride, above
+  movq                  m0, [aboveq]
+  DEFINE_ARGS dst, stride, stride3
+  lea             stride3q, [strideq*3]
+  movq    [dstq          ], m0
+  movq    [dstq+strideq  ], m0
+  movq    [dstq+strideq*2], m0
+  movq    [dstq+stride3q ], m0
+  lea                 dstq, [dstq+strideq*4]
+  movq    [dstq          ], m0
+  movq    [dstq+strideq  ], m0
+  movq    [dstq+strideq*2], m0
+  movq    [dstq+stride3q ], m0
+  RET
+
+INIT_XMM sse2
+cglobal v_predictor_16x16, 3, 4, 1, dst, stride, above
+  mova                  m0, [aboveq]
+  DEFINE_ARGS dst, stride, stride3, nlines4
+  lea             stride3q, [strideq*3]
+  mov              nlines4d, 4
+.loop:
+  mova    [dstq          ], m0
+  mova    [dstq+strideq  ], m0
+  mova    [dstq+strideq*2], m0
+  mova    [dstq+stride3q ], m0
+  lea                 dstq, [dstq+strideq*4]
+  dec             nlines4d
+  jnz .loop
+  REP_RET
+
+INIT_XMM sse2
+cglobal v_predictor_32x32, 3, 4, 2, dst, stride, above
+  mova                  m0, [aboveq]
+  mova                  m1, [aboveq+16]
+  DEFINE_ARGS dst, stride, stride3, nlines4
+  lea             stride3q, [strideq*3]
+  mov              nlines4d, 8
+.loop:
+  mova [dstq             ], m0
+  mova [dstq          +16], m1
+  mova [dstq+strideq     ], m0
+  mova [dstq+strideq  +16], m1
+  mova [dstq+strideq*2   ], m0
+  mova [dstq+strideq*2+16], m1
+  mova [dstq+stride3q    ], m0
+  mova [dstq+stride3q +16], m1
+  lea                 dstq, [dstq+strideq*4]
+  dec             nlines4d
+  jnz .loop
+  REP_RET