ref: 75b33c68c75f3d35e7b409aff87f12b307aa096b
parent: 92c5d3665dabe8b55b9b32df9fac92171d5f7947
author: Ronald S. Bultje <rbultje@google.com>
date: Tue Jul 9 10:54:20 EDT 2013
SSE/SSE2 assembly for 4x4/8x8/16x16/32x32 V intra prediction. Change-Id: I55a6cfa2daba738cbc0c4a02f806893f7e556997
--- a/vp9/common/vp9_rtcd_defs.sh
+++ b/vp9/common/vp9_rtcd_defs.sh
@@ -74,7 +74,7 @@
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 @@
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 @@
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 @@
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
--- a/vp9/common/x86/vp9_intrapred_sse2.asm
+++ b/vp9/common/x86/vp9_intrapred_sse2.asm
@@ -125,3 +125,66 @@
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