Merge "SSE/SSE2 assembly for 4x4/8x8/16x16/32x32 DC intra prediction."

vp9/common/vp9_rtcd_defs.sh

@@ -52,44 +52,161 @@ vp9_copy_mem8x8_dspr2=vp9_copy_mem8x8_dspr2
 prototype void vp9_copy_mem8x4 "const uint8_t *src, int src_pitch, uint8_t *dst, int dst_pitch"
 specialize vp9_copy_mem8x4 mmx
 
-prototype void vp9_d27_predictor "uint8_t *ypred_ptr, int y_stride, int bs, uint8_t *yabove_row, uint8_t *yleft_col"
-specialize vp9_d27_predictor
+prototype void vp9_d27_predictor_4x4 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d27_predictor_4x4
 
-prototype void vp9_d45_predictor "uint8_t *ypred_ptr, int y_stride, int bs, uint8_t *yabove_row, uint8_t *yleft_col"
-specialize vp9_d45_predictor
+prototype void vp9_d45_predictor_4x4 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d45_predictor_4x4
 
-prototype void vp9_d63_predictor "uint8_t *ypred_ptr, int y_stride, int bs, uint8_t *yabove_row, uint8_t *yleft_col"
-specialize vp9_d63_predictor
+prototype void vp9_d63_predictor_4x4 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d63_predictor_4x4
 
-prototype void vp9_h_predictor "uint8_t *ypred_ptr, int y_stride, int bs, uint8_t *yabove_row, uint8_t *yleft_col"
-specialize vp9_h_predictor
+prototype void vp9_h_predictor_4x4 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_h_predictor_4x4
 
-prototype void vp9_d117_predictor "uint8_t *ypred_ptr, int y_stride, int bs, uint8_t *yabove_row, uint8_t *yleft_col"
-specialize vp9_d117_predictor
+prototype void vp9_d117_predictor_4x4 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d117_predictor_4x4
 
-prototype void vp9_d135_predictor "uint8_t *ypred_ptr, int y_stride, int bs, uint8_t *yabove_row, uint8_t *yleft_col"
-specialize vp9_d135_predictor
+prototype void vp9_d135_predictor_4x4 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d135_predictor_4x4
 
-prototype void vp9_d153_predictor "uint8_t *ypred_ptr, int y_stride, int bs, uint8_t *yabove_row, uint8_t *yleft_col"
-specialize vp9_d153_predictor
+prototype void vp9_d153_predictor_4x4 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d153_predictor_4x4
 
-prototype void vp9_v_predictor "uint8_t *ypred_ptr, int y_stride, int bs, uint8_t *yabove_row, uint8_t *yleft_col"
-specialize vp9_v_predictor
+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
 
-prototype void vp9_tm_predictor "uint8_t *ypred_ptr, int y_stride, int bs, uint8_t *yabove_row, uint8_t *yleft_col"
-specialize vp9_tm_predictor
+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
 
-prototype void vp9_dc_predictor "uint8_t *ypred_ptr, int y_stride, int bs, uint8_t *yabove_row, uint8_t *yleft_col"
-specialize vp9_dc_predictor
+prototype void vp9_dc_predictor_4x4 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_dc_predictor_4x4 sse
 
-prototype void vp9_dc_top_predictor "uint8_t *ypred_ptr, int y_stride, int bs, uint8_t *yabove_row, uint8_t *yleft_col"
-specialize vp9_dc_top_predictor
+prototype void vp9_dc_top_predictor_4x4 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_dc_top_predictor_4x4
 
-prototype void vp9_dc_left_predictor "uint8_t *ypred_ptr, int y_stride, int bs, uint8_t *yabove_row, uint8_t *yleft_col"
-specialize vp9_dc_left_predictor
+prototype void vp9_dc_left_predictor_4x4 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_dc_left_predictor_4x4
 
-prototype void vp9_dc_128_predictor "uint8_t *ypred_ptr, int y_stride, int bs, uint8_t *yabove_row, uint8_t *yleft_col"
-specialize vp9_dc_128_predictor
+prototype void vp9_dc_128_predictor_4x4 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_dc_128_predictor_4x4
+
+prototype void vp9_d27_predictor_8x8 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d27_predictor_8x8
+
+prototype void vp9_d45_predictor_8x8 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d45_predictor_8x8
+
+prototype void vp9_d63_predictor_8x8 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d63_predictor_8x8
+
+prototype void vp9_h_predictor_8x8 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_h_predictor_8x8
+
+prototype void vp9_d117_predictor_8x8 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d117_predictor_8x8
+
+prototype void vp9_d135_predictor_8x8 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d135_predictor_8x8
+
+prototype void vp9_d153_predictor_8x8 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+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
+
+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
+
+prototype void vp9_dc_predictor_8x8 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_dc_predictor_8x8 sse
+
+prototype void vp9_dc_top_predictor_8x8 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_dc_top_predictor_8x8
+
+prototype void vp9_dc_left_predictor_8x8 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_dc_left_predictor_8x8
+
+prototype void vp9_dc_128_predictor_8x8 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_dc_128_predictor_8x8
+
+prototype void vp9_d27_predictor_16x16 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d27_predictor_16x16
+
+prototype void vp9_d45_predictor_16x16 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d45_predictor_16x16
+
+prototype void vp9_d63_predictor_16x16 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d63_predictor_16x16
+
+prototype void vp9_h_predictor_16x16 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_h_predictor_16x16
+
+prototype void vp9_d117_predictor_16x16 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d117_predictor_16x16
+
+prototype void vp9_d135_predictor_16x16 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d135_predictor_16x16
+
+prototype void vp9_d153_predictor_16x16 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+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
+
+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
+
+prototype void vp9_dc_predictor_16x16 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_dc_predictor_16x16 sse2
+
+prototype void vp9_dc_top_predictor_16x16 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_dc_top_predictor_16x16
+
+prototype void vp9_dc_left_predictor_16x16 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_dc_left_predictor_16x16
+
+prototype void vp9_dc_128_predictor_16x16 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_dc_128_predictor_16x16
+
+prototype void vp9_d27_predictor_32x32 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d27_predictor_32x32
+
+prototype void vp9_d45_predictor_32x32 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d45_predictor_32x32
+
+prototype void vp9_d63_predictor_32x32 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d63_predictor_32x32
+
+prototype void vp9_h_predictor_32x32 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_h_predictor_32x32
+
+prototype void vp9_d117_predictor_32x32 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d117_predictor_32x32
+
+prototype void vp9_d135_predictor_32x32 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_d135_predictor_32x32
+
+prototype void vp9_d153_predictor_32x32 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+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
+
+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
+
+prototype void vp9_dc_predictor_32x32 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_dc_predictor_32x32 sse2
+
+prototype void vp9_dc_top_predictor_32x32 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_dc_top_predictor_32x32
+
+prototype void vp9_dc_left_predictor_32x32 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_dc_left_predictor_32x32
+
+prototype void vp9_dc_128_predictor_32x32 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
+specialize vp9_dc_128_predictor_32x32
 
 if [ "$CONFIG_VP9_DECODER" = "yes" ]; then
 prototype void vp9_add_constant_residual_8x8 "const int16_t diff, uint8_t *dest, int stride"

vp9/common/x86/vp9_intrapred_sse2.asm

+;
+;  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+;
+;  Use of this source code is governed by a BSD-style license
+;  that can be found in the LICENSE file in the root of the source
+;  tree. An additional intellectual property rights grant can be found
+;  in the file PATENTS.  All contributing project authors may
+;  be found in the AUTHORS file in the root of the source tree.
+;
+
+%include "third_party/x86inc/x86inc.asm"
+
+SECTION_RODATA
+pw_4:  times 8 dw 4
+pw_8:  times 8 dw 8
+pw_16: times 8 dw 16
+pw_32: times 8 dw 32
+
+SECTION .text
+
+INIT_MMX sse
+cglobal dc_predictor_4x4, 4, 4, 2, dst, stride, above, left
+  pxor                  m1, m1
+  movd                  m0, [aboveq]
+  punpckldq             m0, [leftq]
+  psadbw                m0, m1
+  paddw                 m0, [pw_4]
+  psraw                 m0, 3
+  pshufw                m0, m0, 0x0
+  packuswb              m0, m0
+  movd      [dstq        ], m0
+  movd      [dstq+strideq], m0
+  lea                 dstq, [dstq+strideq*2]
+  movd      [dstq        ], m0
+  movd      [dstq+strideq], m0
+  RET
+
+INIT_MMX sse
+cglobal dc_predictor_8x8, 4, 4, 3, dst, stride, above, left
+  pxor                  m1, m1
+  movq                  m0, [aboveq]
+  movq                  m2, [leftq]
+  DEFINE_ARGS dst, stride, stride3
+  lea             stride3q, [strideq*3]
+  psadbw                m0, m1
+  psadbw                m2, m1
+  paddw                 m0, m2
+  paddw                 m0, [pw_8]
+  psraw                 m0, 4
+  pshufw                m0, m0, 0x0
+  packuswb              m0, m0
+  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 dc_predictor_16x16, 4, 4, 3, dst, stride, above, left
+  pxor                  m1, m1
+  mova                  m0, [aboveq]
+  mova                  m2, [leftq]
+  DEFINE_ARGS dst, stride, stride3, lines4
+  lea             stride3q, [strideq*3]
+  mov              lines4d, 4
+  psadbw                m0, m1
+  psadbw                m2, m1
+  paddw                 m0, m2
+  movhlps               m2, m0
+  paddw                 m0, m2
+  paddw                 m0, [pw_16]
+  psraw                 m0, 5
+  pshuflw               m0, m0, 0x0
+  punpcklqdq            m0, m0
+  packuswb              m0, m0
+.loop:
+  mova    [dstq          ], m0
+  mova    [dstq+strideq  ], m0
+  mova    [dstq+strideq*2], m0
+  mova    [dstq+stride3q ], m0
+  lea                 dstq, [dstq+strideq*4]
+  dec              lines4d
+  jnz .loop
+  REP_RET
+
+INIT_XMM sse2
+cglobal dc_predictor_32x32, 4, 4, 5, dst, stride, above, left
+  pxor                  m1, m1
+  mova                  m0, [aboveq]
+  mova                  m2, [aboveq+16]
+  mova                  m3, [leftq]
+  mova                  m4, [leftq+16]
+  DEFINE_ARGS dst, stride, stride3, lines4
+  lea             stride3q, [strideq*3]
+  mov              lines4d, 8
+  psadbw                m0, m1
+  psadbw                m2, m1
+  psadbw                m3, m1
+  psadbw                m4, m1
+  paddw                 m0, m2
+  paddw                 m0, m3
+  paddw                 m0, m4
+  movhlps               m2, m0
+  paddw                 m0, m2
+  paddw                 m0, [pw_32]
+  psraw                 m0, 6
+  pshuflw               m0, m0, 0x0
+  punpcklqdq            m0, m0
+  packuswb              m0, m0
+.loop:
+  mova [dstq             ], m0
+  mova [dstq          +16], m0
+  mova [dstq+strideq     ], m0
+  mova [dstq+strideq  +16], m0
+  mova [dstq+strideq*2   ], m0
+  mova [dstq+strideq*2+16], m0
+  mova [dstq+stride3q    ], m0
+  mova [dstq+stride3q +16], m0
+  lea                 dstq, [dstq+strideq*4]
+  dec              lines4d
+  jnz .loop
+  REP_RET

vp9/vp9_common.mk

@@ -81,6 +81,7 @@ VP9_COMMON_SRCS-$(HAVE_MMX) += common/x86/vp9_loopfilter_mmx.asm
 VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_iwalsh_sse2.asm
 VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_loopfilter_sse2.asm
 VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_recon_sse2.asm
+VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_intrapred_sse2.asm
 VP9_COMMON_SRCS-$(HAVE_SSSE3) += common/x86/vp9_subpixel_8t_ssse3.asm
 ifeq ($(CONFIG_POSTPROC),yes)
 VP9_COMMON_SRCS-$(HAVE_MMX) += common/x86/vp9_postproc_mmx.asm