ref: 001baa5dd8bce66707a123de687a38c4da45cf9a
parent: 714a46a63c5434a4f18b52905e840516a5d85017
parent: 907b33cdc432773f04f53e06a7d2d4ac1c434c18
author: Johann Koenig <[email protected]>
date: Fri Jun 19 12:27:24 EDT 2015
Merge "Move vp8 variance files"
--- a/vp8/common/x86/variance_impl_mmx.asm
+++ /dev/null
@@ -1,353 +1,0 @@
-;
-; 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 "vpx_ports/x86_abi_support.asm"
-
-%define mmx_filter_shift 7
-
-;void vp8_filter_block2d_bil4x4_var_mmx
-;(
-; unsigned char *ref_ptr,
-; int ref_pixels_per_line,
-; unsigned char *src_ptr,
-; int src_pixels_per_line,
-; unsigned short *HFilter,
-; unsigned short *VFilter,
-; int *sum,
-; unsigned int *sumsquared
-;)
-global sym(vp8_filter_block2d_bil4x4_var_mmx) PRIVATE
-sym(vp8_filter_block2d_bil4x4_var_mmx):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 8
- GET_GOT rbx
- push rsi
- push rdi
- sub rsp, 16
- ; end prolog
-
-
- pxor mm6, mm6 ;
- pxor mm7, mm7 ;
-
- mov rax, arg(4) ;HFilter ;
- mov rdx, arg(5) ;VFilter ;
-
- mov rsi, arg(0) ;ref_ptr ;
- mov rdi, arg(2) ;src_ptr ;
-
- mov rcx, 4 ;
- pxor mm0, mm0 ;
-
- movd mm1, [rsi] ;
- movd mm3, [rsi+1] ;
-
- punpcklbw mm1, mm0 ;
- pmullw mm1, [rax] ;
-
- punpcklbw mm3, mm0 ;
- pmullw mm3, [rax+8] ;
-
- paddw mm1, mm3 ;
- paddw mm1, [GLOBAL(mmx_bi_rd)] ;
-
- psraw mm1, mmx_filter_shift ;
- movq mm5, mm1
-
-%if ABI_IS_32BIT
- add rsi, dword ptr arg(1) ;ref_pixels_per_line ;
-%else
- movsxd r8, dword ptr arg(1) ;ref_pixels_per_line ;
- add rsi, r8
-%endif
-
-.filter_block2d_bil4x4_var_mmx_loop:
-
- movd mm1, [rsi] ;
- movd mm3, [rsi+1] ;
-
- punpcklbw mm1, mm0 ;
- pmullw mm1, [rax] ;
-
- punpcklbw mm3, mm0 ;
- pmullw mm3, [rax+8] ;
-
- paddw mm1, mm3 ;
- paddw mm1, [GLOBAL(mmx_bi_rd)] ;
-
- psraw mm1, mmx_filter_shift ;
- movq mm3, mm5 ;
-
- movq mm5, mm1 ;
- pmullw mm3, [rdx] ;
-
- pmullw mm1, [rdx+8] ;
- paddw mm1, mm3 ;
-
-
- paddw mm1, [GLOBAL(mmx_bi_rd)] ;
- psraw mm1, mmx_filter_shift ;
-
- movd mm3, [rdi] ;
- punpcklbw mm3, mm0 ;
-
- psubw mm1, mm3 ;
- paddw mm6, mm1 ;
-
- pmaddwd mm1, mm1 ;
- paddd mm7, mm1 ;
-
-%if ABI_IS_32BIT
- add rsi, dword ptr arg(1) ;ref_pixels_per_line ;
- add rdi, dword ptr arg(3) ;src_pixels_per_line ;
-%else
- movsxd r8, dword ptr arg(1) ;ref_pixels_per_line
- movsxd r9, dword ptr arg(3) ;src_pixels_per_line
- add rsi, r8
- add rdi, r9
-%endif
- sub rcx, 1 ;
- jnz .filter_block2d_bil4x4_var_mmx_loop ;
-
-
- pxor mm3, mm3 ;
- pxor mm2, mm2 ;
-
- punpcklwd mm2, mm6 ;
- punpckhwd mm3, mm6 ;
-
- paddd mm2, mm3 ;
- movq mm6, mm2 ;
-
- psrlq mm6, 32 ;
- paddd mm2, mm6 ;
-
- psrad mm2, 16 ;
- movq mm4, mm7 ;
-
- psrlq mm4, 32 ;
- paddd mm4, mm7 ;
-
- mov rdi, arg(6) ;sum
- mov rsi, arg(7) ;sumsquared
-
- movd dword ptr [rdi], mm2 ;
- movd dword ptr [rsi], mm4 ;
-
-
-
- ; begin epilog
- add rsp, 16
- pop rdi
- pop rsi
- RESTORE_GOT
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-
-
-;void vp8_filter_block2d_bil_var_mmx
-;(
-; unsigned char *ref_ptr,
-; int ref_pixels_per_line,
-; unsigned char *src_ptr,
-; int src_pixels_per_line,
-; unsigned int Height,
-; unsigned short *HFilter,
-; unsigned short *VFilter,
-; int *sum,
-; unsigned int *sumsquared
-;)
-global sym(vp8_filter_block2d_bil_var_mmx) PRIVATE
-sym(vp8_filter_block2d_bil_var_mmx):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 9
- GET_GOT rbx
- push rsi
- push rdi
- sub rsp, 16
- ; end prolog
-
- pxor mm6, mm6 ;
- pxor mm7, mm7 ;
- mov rax, arg(5) ;HFilter ;
-
- mov rdx, arg(6) ;VFilter ;
- mov rsi, arg(0) ;ref_ptr ;
-
- mov rdi, arg(2) ;src_ptr ;
- movsxd rcx, dword ptr arg(4) ;Height ;
-
- pxor mm0, mm0 ;
- movq mm1, [rsi] ;
-
- movq mm3, [rsi+1] ;
- movq mm2, mm1 ;
-
- movq mm4, mm3 ;
- punpcklbw mm1, mm0 ;
-
- punpckhbw mm2, mm0 ;
- pmullw mm1, [rax] ;
-
- pmullw mm2, [rax] ;
- punpcklbw mm3, mm0 ;
-
- punpckhbw mm4, mm0 ;
- pmullw mm3, [rax+8] ;
-
- pmullw mm4, [rax+8] ;
- paddw mm1, mm3 ;
-
- paddw mm2, mm4 ;
- paddw mm1, [GLOBAL(mmx_bi_rd)] ;
-
- psraw mm1, mmx_filter_shift ;
- paddw mm2, [GLOBAL(mmx_bi_rd)] ;
-
- psraw mm2, mmx_filter_shift ;
- movq mm5, mm1
-
- packuswb mm5, mm2 ;
-%if ABI_IS_32BIT
- add rsi, dword ptr arg(1) ;ref_pixels_per_line
-%else
- movsxd r8, dword ptr arg(1) ;ref_pixels_per_line
- add rsi, r8
-%endif
-
-.filter_block2d_bil_var_mmx_loop:
-
- movq mm1, [rsi] ;
- movq mm3, [rsi+1] ;
-
- movq mm2, mm1 ;
- movq mm4, mm3 ;
-
- punpcklbw mm1, mm0 ;
- punpckhbw mm2, mm0 ;
-
- pmullw mm1, [rax] ;
- pmullw mm2, [rax] ;
-
- punpcklbw mm3, mm0 ;
- punpckhbw mm4, mm0 ;
-
- pmullw mm3, [rax+8] ;
- pmullw mm4, [rax+8] ;
-
- paddw mm1, mm3 ;
- paddw mm2, mm4 ;
-
- paddw mm1, [GLOBAL(mmx_bi_rd)] ;
- psraw mm1, mmx_filter_shift ;
-
- paddw mm2, [GLOBAL(mmx_bi_rd)] ;
- psraw mm2, mmx_filter_shift ;
-
- movq mm3, mm5 ;
- movq mm4, mm5 ;
-
- punpcklbw mm3, mm0 ;
- punpckhbw mm4, mm0 ;
-
- movq mm5, mm1 ;
- packuswb mm5, mm2 ;
-
- pmullw mm3, [rdx] ;
- pmullw mm4, [rdx] ;
-
- pmullw mm1, [rdx+8] ;
- pmullw mm2, [rdx+8] ;
-
- paddw mm1, mm3 ;
- paddw mm2, mm4 ;
-
- paddw mm1, [GLOBAL(mmx_bi_rd)] ;
- paddw mm2, [GLOBAL(mmx_bi_rd)] ;
-
- psraw mm1, mmx_filter_shift ;
- psraw mm2, mmx_filter_shift ;
-
- movq mm3, [rdi] ;
- movq mm4, mm3 ;
-
- punpcklbw mm3, mm0 ;
- punpckhbw mm4, mm0 ;
-
- psubw mm1, mm3 ;
- psubw mm2, mm4 ;
-
- paddw mm6, mm1 ;
- pmaddwd mm1, mm1 ;
-
- paddw mm6, mm2 ;
- pmaddwd mm2, mm2 ;
-
- paddd mm7, mm1 ;
- paddd mm7, mm2 ;
-
-%if ABI_IS_32BIT
- add rsi, dword ptr arg(1) ;ref_pixels_per_line ;
- add rdi, dword ptr arg(3) ;src_pixels_per_line ;
-%else
- movsxd r8, dword ptr arg(1) ;ref_pixels_per_line ;
- movsxd r9, dword ptr arg(3) ;src_pixels_per_line ;
- add rsi, r8
- add rdi, r9
-%endif
- sub rcx, 1 ;
- jnz .filter_block2d_bil_var_mmx_loop ;
-
-
- pxor mm3, mm3 ;
- pxor mm2, mm2 ;
-
- punpcklwd mm2, mm6 ;
- punpckhwd mm3, mm6 ;
-
- paddd mm2, mm3 ;
- movq mm6, mm2 ;
-
- psrlq mm6, 32 ;
- paddd mm2, mm6 ;
-
- psrad mm2, 16 ;
- movq mm4, mm7 ;
-
- psrlq mm4, 32 ;
- paddd mm4, mm7 ;
-
- mov rdi, arg(7) ;sum
- mov rsi, arg(8) ;sumsquared
-
- movd dword ptr [rdi], mm2 ;
- movd dword ptr [rsi], mm4 ;
-
- ; begin epilog
- add rsp, 16
- pop rdi
- pop rsi
- RESTORE_GOT
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-SECTION_RODATA
-;short mmx_bi_rd[4] = { 64, 64, 64, 64};-align 16
-mmx_bi_rd:
- times 4 dw 64
--- a/vp8/common/x86/variance_mmx.c
+++ /dev/null
@@ -1,244 +1,0 @@
-/*
- * 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 "./vp8_rtcd.h"
-#include "vpx_config.h"
-#include "vp8/common/variance.h"
-#include "vpx_ports/mem.h"
-#include "vp8/common/x86/filter_x86.h"
-
-extern void filter_block1d_h6_mmx
-(
- const unsigned char *src_ptr,
- unsigned short *output_ptr,
- unsigned int src_pixels_per_line,
- unsigned int pixel_step,
- unsigned int output_height,
- unsigned int output_width,
- short *filter
-);
-extern void filter_block1d_v6_mmx
-(
- const short *src_ptr,
- unsigned char *output_ptr,
- unsigned int pixels_per_line,
- unsigned int pixel_step,
- unsigned int output_height,
- unsigned int output_width,
- short *filter
-);
-
-extern void vp8_filter_block2d_bil4x4_var_mmx
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- const short *HFilter,
- const short *VFilter,
- int *sum,
- unsigned int *sumsquared
-);
-extern void vp8_filter_block2d_bil_var_mmx
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- unsigned int Height,
- const short *HFilter,
- const short *VFilter,
- int *sum,
- unsigned int *sumsquared
-);
-
-unsigned int vp8_sub_pixel_variance4x4_mmx
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse)
-
-{- int xsum;
- unsigned int xxsum;
- vp8_filter_block2d_bil4x4_var_mmx(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line,
- vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
- &xsum, &xxsum
- );
- *sse = xxsum;
- return (xxsum - (((unsigned int)xsum * xsum) >> 4));
-}
-
-
-unsigned int vp8_sub_pixel_variance8x8_mmx
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{-
- int xsum;
- unsigned int xxsum;
- vp8_filter_block2d_bil_var_mmx(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
- &xsum, &xxsum
- );
- *sse = xxsum;
- return (xxsum - (((unsigned int)xsum * xsum) >> 6));
-}
-
-unsigned int vp8_sub_pixel_variance16x16_mmx
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{-
- int xsum0, xsum1;
- unsigned int xxsum0, xxsum1;
-
-
- vp8_filter_block2d_bil_var_mmx(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
- &xsum0, &xxsum0
- );
-
-
- vp8_filter_block2d_bil_var_mmx(
- src_ptr + 8, src_pixels_per_line,
- dst_ptr + 8, dst_pixels_per_line, 16,
- vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
- &xsum1, &xxsum1
- );
-
- xsum0 += xsum1;
- xxsum0 += xxsum1;
-
- *sse = xxsum0;
- return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 8));
-
-
-}
-
-unsigned int vp8_sub_pixel_variance16x8_mmx
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- int xsum0, xsum1;
- unsigned int xxsum0, xxsum1;
-
-
- vp8_filter_block2d_bil_var_mmx(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
- &xsum0, &xxsum0
- );
-
-
- vp8_filter_block2d_bil_var_mmx(
- src_ptr + 8, src_pixels_per_line,
- dst_ptr + 8, dst_pixels_per_line, 8,
- vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
- &xsum1, &xxsum1
- );
-
- xsum0 += xsum1;
- xxsum0 += xxsum1;
-
- *sse = xxsum0;
- return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 7));
-}
-
-unsigned int vp8_sub_pixel_variance8x16_mmx
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- int xsum;
- unsigned int xxsum;
- vp8_filter_block2d_bil_var_mmx(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
- &xsum, &xxsum
- );
- *sse = xxsum;
- return (xxsum - (((unsigned int)xsum * xsum) >> 7));
-}
-
-
-unsigned int vp8_variance_halfpixvar16x16_h_mmx(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- return vp8_sub_pixel_variance16x16_mmx(src_ptr, source_stride, 4, 0,
- ref_ptr, recon_stride, sse);
-}
-
-
-unsigned int vp8_variance_halfpixvar16x16_v_mmx(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- return vp8_sub_pixel_variance16x16_mmx(src_ptr, source_stride, 0, 4,
- ref_ptr, recon_stride, sse);
-}
-
-
-unsigned int vp8_variance_halfpixvar16x16_hv_mmx(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- return vp8_sub_pixel_variance16x16_mmx(src_ptr, source_stride, 4, 4,
- ref_ptr, recon_stride, sse);
-}
--- a/vp8/common/x86/variance_sse2.c
+++ /dev/null
@@ -1,403 +1,0 @@
-/*
- * 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 "./vp8_rtcd.h"
-#include "vpx_config.h"
-#include "vp8/common/variance.h"
-#include "vpx_ports/mem.h"
-#include "vp8/common/x86/filter_x86.h"
-
-extern void filter_block1d_h6_mmx(const unsigned char *src_ptr, unsigned short *output_ptr, unsigned int src_pixels_per_line, unsigned int pixel_step, unsigned int output_height, unsigned int output_width, short *filter);
-extern void filter_block1d_v6_mmx(const short *src_ptr, unsigned char *output_ptr, unsigned int pixels_per_line, unsigned int pixel_step, unsigned int output_height, unsigned int output_width, short *filter);
-extern void filter_block1d8_h6_sse2(const unsigned char *src_ptr, unsigned short *output_ptr, unsigned int src_pixels_per_line, unsigned int pixel_step, unsigned int output_height, unsigned int output_width, short *filter);
-extern void filter_block1d8_v6_sse2(const short *src_ptr, unsigned char *output_ptr, unsigned int pixels_per_line, unsigned int pixel_step, unsigned int output_height, unsigned int output_width, short *filter);
-
-extern void vp8_filter_block2d_bil4x4_var_mmx
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- const short *HFilter,
- const short *VFilter,
- int *sum,
- unsigned int *sumsquared
-);
-
-void vp8_filter_block2d_bil_var_sse2
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- unsigned int Height,
- int xoffset,
- int yoffset,
- int *sum,
- unsigned int *sumsquared
-);
-void vp8_half_horiz_vert_variance8x_h_sse2
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- unsigned int Height,
- int *sum,
- unsigned int *sumsquared
-);
-void vp8_half_horiz_vert_variance16x_h_sse2
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- unsigned int Height,
- int *sum,
- unsigned int *sumsquared
-);
-void vp8_half_horiz_variance8x_h_sse2
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- unsigned int Height,
- int *sum,
- unsigned int *sumsquared
-);
-void vp8_half_horiz_variance16x_h_sse2
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- unsigned int Height,
- int *sum,
- unsigned int *sumsquared
-);
-void vp8_half_vert_variance8x_h_sse2
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- unsigned int Height,
- int *sum,
- unsigned int *sumsquared
-);
-void vp8_half_vert_variance16x_h_sse2
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- unsigned int Height,
- int *sum,
- unsigned int *sumsquared
-);
-
-unsigned int vp8_sub_pixel_variance4x4_wmt
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- int xsum;
- unsigned int xxsum;
- vp8_filter_block2d_bil4x4_var_mmx(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line,
- vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
- &xsum, &xxsum
- );
- *sse = xxsum;
- return (xxsum - (((unsigned int)xsum * xsum) >> 4));
-}
-
-
-unsigned int vp8_sub_pixel_variance8x8_wmt
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- int xsum;
- unsigned int xxsum;
-
- if (xoffset == 4 && yoffset == 0)
- {- vp8_half_horiz_variance8x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- &xsum, &xxsum);
- }
- else if (xoffset == 0 && yoffset == 4)
- {- vp8_half_vert_variance8x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- &xsum, &xxsum);
- }
- else if (xoffset == 4 && yoffset == 4)
- {- vp8_half_horiz_vert_variance8x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- &xsum, &xxsum);
- }
- else
- {- vp8_filter_block2d_bil_var_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- xoffset, yoffset,
- &xsum, &xxsum);
- }
-
- *sse = xxsum;
- return (xxsum - (((unsigned int)xsum * xsum) >> 6));
-}
-
-unsigned int vp8_sub_pixel_variance16x16_wmt
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- int xsum0, xsum1;
- unsigned int xxsum0, xxsum1;
-
-
- /* note we could avoid these if statements if the calling function
- * just called the appropriate functions inside.
- */
- if (xoffset == 4 && yoffset == 0)
- {- vp8_half_horiz_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum0, &xxsum0);
- }
- else if (xoffset == 0 && yoffset == 4)
- {- vp8_half_vert_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum0, &xxsum0);
- }
- else if (xoffset == 4 && yoffset == 4)
- {- vp8_half_horiz_vert_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum0, &xxsum0);
- }
- else
- {- vp8_filter_block2d_bil_var_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- xoffset, yoffset,
- &xsum0, &xxsum0
- );
-
- vp8_filter_block2d_bil_var_sse2(
- src_ptr + 8, src_pixels_per_line,
- dst_ptr + 8, dst_pixels_per_line, 16,
- xoffset, yoffset,
- &xsum1, &xxsum1
- );
- xsum0 += xsum1;
- xxsum0 += xxsum1;
- }
-
- *sse = xxsum0;
- return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 8));
-}
-
-unsigned int vp8_sub_pixel_variance16x8_wmt
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-
-)
-{- int xsum0, xsum1;
- unsigned int xxsum0, xxsum1;
-
- if (xoffset == 4 && yoffset == 0)
- {- vp8_half_horiz_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- &xsum0, &xxsum0);
- }
- else if (xoffset == 0 && yoffset == 4)
- {- vp8_half_vert_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- &xsum0, &xxsum0);
- }
- else if (xoffset == 4 && yoffset == 4)
- {- vp8_half_horiz_vert_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- &xsum0, &xxsum0);
- }
- else
- {- vp8_filter_block2d_bil_var_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- xoffset, yoffset,
- &xsum0, &xxsum0);
-
- vp8_filter_block2d_bil_var_sse2(
- src_ptr + 8, src_pixels_per_line,
- dst_ptr + 8, dst_pixels_per_line, 8,
- xoffset, yoffset,
- &xsum1, &xxsum1);
- xsum0 += xsum1;
- xxsum0 += xxsum1;
- }
-
- *sse = xxsum0;
- return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 7));
-}
-
-unsigned int vp8_sub_pixel_variance8x16_wmt
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- int xsum;
- unsigned int xxsum;
-
- if (xoffset == 4 && yoffset == 0)
- {- vp8_half_horiz_variance8x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum, &xxsum);
- }
- else if (xoffset == 0 && yoffset == 4)
- {- vp8_half_vert_variance8x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum, &xxsum);
- }
- else if (xoffset == 4 && yoffset == 4)
- {- vp8_half_horiz_vert_variance8x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum, &xxsum);
- }
- else
- {- vp8_filter_block2d_bil_var_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- xoffset, yoffset,
- &xsum, &xxsum);
- }
-
- *sse = xxsum;
- return (xxsum - (((unsigned int)xsum * xsum) >> 7));
-}
-
-
-unsigned int vp8_variance_halfpixvar16x16_h_wmt(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse)
-{- int xsum0;
- unsigned int xxsum0;
-
- vp8_half_horiz_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum0, &xxsum0);
-
- *sse = xxsum0;
- return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 8));
-}
-
-
-unsigned int vp8_variance_halfpixvar16x16_v_wmt(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse)
-{- int xsum0;
- unsigned int xxsum0;
- vp8_half_vert_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum0, &xxsum0);
-
- *sse = xxsum0;
- return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 8));
-}
-
-
-unsigned int vp8_variance_halfpixvar16x16_hv_wmt(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse)
-{- int xsum0;
- unsigned int xxsum0;
-
- vp8_half_horiz_vert_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum0, &xxsum0);
-
- *sse = xxsum0;
- return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 8));
-}
--- /dev/null
+++ b/vp8/common/x86/vp8_variance_impl_mmx.asm
@@ -1,0 +1,353 @@
+;
+; 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 "vpx_ports/x86_abi_support.asm"
+
+%define mmx_filter_shift 7
+
+;void vp8_filter_block2d_bil4x4_var_mmx
+;(
+; unsigned char *ref_ptr,
+; int ref_pixels_per_line,
+; unsigned char *src_ptr,
+; int src_pixels_per_line,
+; unsigned short *HFilter,
+; unsigned short *VFilter,
+; int *sum,
+; unsigned int *sumsquared
+;)
+global sym(vp8_filter_block2d_bil4x4_var_mmx) PRIVATE
+sym(vp8_filter_block2d_bil4x4_var_mmx):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 8
+ GET_GOT rbx
+ push rsi
+ push rdi
+ sub rsp, 16
+ ; end prolog
+
+
+ pxor mm6, mm6 ;
+ pxor mm7, mm7 ;
+
+ mov rax, arg(4) ;HFilter ;
+ mov rdx, arg(5) ;VFilter ;
+
+ mov rsi, arg(0) ;ref_ptr ;
+ mov rdi, arg(2) ;src_ptr ;
+
+ mov rcx, 4 ;
+ pxor mm0, mm0 ;
+
+ movd mm1, [rsi] ;
+ movd mm3, [rsi+1] ;
+
+ punpcklbw mm1, mm0 ;
+ pmullw mm1, [rax] ;
+
+ punpcklbw mm3, mm0 ;
+ pmullw mm3, [rax+8] ;
+
+ paddw mm1, mm3 ;
+ paddw mm1, [GLOBAL(mmx_bi_rd)] ;
+
+ psraw mm1, mmx_filter_shift ;
+ movq mm5, mm1
+
+%if ABI_IS_32BIT
+ add rsi, dword ptr arg(1) ;ref_pixels_per_line ;
+%else
+ movsxd r8, dword ptr arg(1) ;ref_pixels_per_line ;
+ add rsi, r8
+%endif
+
+.filter_block2d_bil4x4_var_mmx_loop:
+
+ movd mm1, [rsi] ;
+ movd mm3, [rsi+1] ;
+
+ punpcklbw mm1, mm0 ;
+ pmullw mm1, [rax] ;
+
+ punpcklbw mm3, mm0 ;
+ pmullw mm3, [rax+8] ;
+
+ paddw mm1, mm3 ;
+ paddw mm1, [GLOBAL(mmx_bi_rd)] ;
+
+ psraw mm1, mmx_filter_shift ;
+ movq mm3, mm5 ;
+
+ movq mm5, mm1 ;
+ pmullw mm3, [rdx] ;
+
+ pmullw mm1, [rdx+8] ;
+ paddw mm1, mm3 ;
+
+
+ paddw mm1, [GLOBAL(mmx_bi_rd)] ;
+ psraw mm1, mmx_filter_shift ;
+
+ movd mm3, [rdi] ;
+ punpcklbw mm3, mm0 ;
+
+ psubw mm1, mm3 ;
+ paddw mm6, mm1 ;
+
+ pmaddwd mm1, mm1 ;
+ paddd mm7, mm1 ;
+
+%if ABI_IS_32BIT
+ add rsi, dword ptr arg(1) ;ref_pixels_per_line ;
+ add rdi, dword ptr arg(3) ;src_pixels_per_line ;
+%else
+ movsxd r8, dword ptr arg(1) ;ref_pixels_per_line
+ movsxd r9, dword ptr arg(3) ;src_pixels_per_line
+ add rsi, r8
+ add rdi, r9
+%endif
+ sub rcx, 1 ;
+ jnz .filter_block2d_bil4x4_var_mmx_loop ;
+
+
+ pxor mm3, mm3 ;
+ pxor mm2, mm2 ;
+
+ punpcklwd mm2, mm6 ;
+ punpckhwd mm3, mm6 ;
+
+ paddd mm2, mm3 ;
+ movq mm6, mm2 ;
+
+ psrlq mm6, 32 ;
+ paddd mm2, mm6 ;
+
+ psrad mm2, 16 ;
+ movq mm4, mm7 ;
+
+ psrlq mm4, 32 ;
+ paddd mm4, mm7 ;
+
+ mov rdi, arg(6) ;sum
+ mov rsi, arg(7) ;sumsquared
+
+ movd dword ptr [rdi], mm2 ;
+ movd dword ptr [rsi], mm4 ;
+
+
+
+ ; begin epilog
+ add rsp, 16
+ pop rdi
+ pop rsi
+ RESTORE_GOT
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+
+
+;void vp8_filter_block2d_bil_var_mmx
+;(
+; unsigned char *ref_ptr,
+; int ref_pixels_per_line,
+; unsigned char *src_ptr,
+; int src_pixels_per_line,
+; unsigned int Height,
+; unsigned short *HFilter,
+; unsigned short *VFilter,
+; int *sum,
+; unsigned int *sumsquared
+;)
+global sym(vp8_filter_block2d_bil_var_mmx) PRIVATE
+sym(vp8_filter_block2d_bil_var_mmx):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 9
+ GET_GOT rbx
+ push rsi
+ push rdi
+ sub rsp, 16
+ ; end prolog
+
+ pxor mm6, mm6 ;
+ pxor mm7, mm7 ;
+ mov rax, arg(5) ;HFilter ;
+
+ mov rdx, arg(6) ;VFilter ;
+ mov rsi, arg(0) ;ref_ptr ;
+
+ mov rdi, arg(2) ;src_ptr ;
+ movsxd rcx, dword ptr arg(4) ;Height ;
+
+ pxor mm0, mm0 ;
+ movq mm1, [rsi] ;
+
+ movq mm3, [rsi+1] ;
+ movq mm2, mm1 ;
+
+ movq mm4, mm3 ;
+ punpcklbw mm1, mm0 ;
+
+ punpckhbw mm2, mm0 ;
+ pmullw mm1, [rax] ;
+
+ pmullw mm2, [rax] ;
+ punpcklbw mm3, mm0 ;
+
+ punpckhbw mm4, mm0 ;
+ pmullw mm3, [rax+8] ;
+
+ pmullw mm4, [rax+8] ;
+ paddw mm1, mm3 ;
+
+ paddw mm2, mm4 ;
+ paddw mm1, [GLOBAL(mmx_bi_rd)] ;
+
+ psraw mm1, mmx_filter_shift ;
+ paddw mm2, [GLOBAL(mmx_bi_rd)] ;
+
+ psraw mm2, mmx_filter_shift ;
+ movq mm5, mm1
+
+ packuswb mm5, mm2 ;
+%if ABI_IS_32BIT
+ add rsi, dword ptr arg(1) ;ref_pixels_per_line
+%else
+ movsxd r8, dword ptr arg(1) ;ref_pixels_per_line
+ add rsi, r8
+%endif
+
+.filter_block2d_bil_var_mmx_loop:
+
+ movq mm1, [rsi] ;
+ movq mm3, [rsi+1] ;
+
+ movq mm2, mm1 ;
+ movq mm4, mm3 ;
+
+ punpcklbw mm1, mm0 ;
+ punpckhbw mm2, mm0 ;
+
+ pmullw mm1, [rax] ;
+ pmullw mm2, [rax] ;
+
+ punpcklbw mm3, mm0 ;
+ punpckhbw mm4, mm0 ;
+
+ pmullw mm3, [rax+8] ;
+ pmullw mm4, [rax+8] ;
+
+ paddw mm1, mm3 ;
+ paddw mm2, mm4 ;
+
+ paddw mm1, [GLOBAL(mmx_bi_rd)] ;
+ psraw mm1, mmx_filter_shift ;
+
+ paddw mm2, [GLOBAL(mmx_bi_rd)] ;
+ psraw mm2, mmx_filter_shift ;
+
+ movq mm3, mm5 ;
+ movq mm4, mm5 ;
+
+ punpcklbw mm3, mm0 ;
+ punpckhbw mm4, mm0 ;
+
+ movq mm5, mm1 ;
+ packuswb mm5, mm2 ;
+
+ pmullw mm3, [rdx] ;
+ pmullw mm4, [rdx] ;
+
+ pmullw mm1, [rdx+8] ;
+ pmullw mm2, [rdx+8] ;
+
+ paddw mm1, mm3 ;
+ paddw mm2, mm4 ;
+
+ paddw mm1, [GLOBAL(mmx_bi_rd)] ;
+ paddw mm2, [GLOBAL(mmx_bi_rd)] ;
+
+ psraw mm1, mmx_filter_shift ;
+ psraw mm2, mmx_filter_shift ;
+
+ movq mm3, [rdi] ;
+ movq mm4, mm3 ;
+
+ punpcklbw mm3, mm0 ;
+ punpckhbw mm4, mm0 ;
+
+ psubw mm1, mm3 ;
+ psubw mm2, mm4 ;
+
+ paddw mm6, mm1 ;
+ pmaddwd mm1, mm1 ;
+
+ paddw mm6, mm2 ;
+ pmaddwd mm2, mm2 ;
+
+ paddd mm7, mm1 ;
+ paddd mm7, mm2 ;
+
+%if ABI_IS_32BIT
+ add rsi, dword ptr arg(1) ;ref_pixels_per_line ;
+ add rdi, dword ptr arg(3) ;src_pixels_per_line ;
+%else
+ movsxd r8, dword ptr arg(1) ;ref_pixels_per_line ;
+ movsxd r9, dword ptr arg(3) ;src_pixels_per_line ;
+ add rsi, r8
+ add rdi, r9
+%endif
+ sub rcx, 1 ;
+ jnz .filter_block2d_bil_var_mmx_loop ;
+
+
+ pxor mm3, mm3 ;
+ pxor mm2, mm2 ;
+
+ punpcklwd mm2, mm6 ;
+ punpckhwd mm3, mm6 ;
+
+ paddd mm2, mm3 ;
+ movq mm6, mm2 ;
+
+ psrlq mm6, 32 ;
+ paddd mm2, mm6 ;
+
+ psrad mm2, 16 ;
+ movq mm4, mm7 ;
+
+ psrlq mm4, 32 ;
+ paddd mm4, mm7 ;
+
+ mov rdi, arg(7) ;sum
+ mov rsi, arg(8) ;sumsquared
+
+ movd dword ptr [rdi], mm2 ;
+ movd dword ptr [rsi], mm4 ;
+
+ ; begin epilog
+ add rsp, 16
+ pop rdi
+ pop rsi
+ RESTORE_GOT
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+SECTION_RODATA
+;short mmx_bi_rd[4] = { 64, 64, 64, 64};+align 16
+mmx_bi_rd:
+ times 4 dw 64
--- /dev/null
+++ b/vp8/common/x86/vp8_variance_mmx.c
@@ -1,0 +1,244 @@
+/*
+ * 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 "./vp8_rtcd.h"
+#include "vpx_config.h"
+#include "vp8/common/variance.h"
+#include "vpx_ports/mem.h"
+#include "vp8/common/x86/filter_x86.h"
+
+extern void filter_block1d_h6_mmx
+(
+ const unsigned char *src_ptr,
+ unsigned short *output_ptr,
+ unsigned int src_pixels_per_line,
+ unsigned int pixel_step,
+ unsigned int output_height,
+ unsigned int output_width,
+ short *filter
+);
+extern void filter_block1d_v6_mmx
+(
+ const short *src_ptr,
+ unsigned char *output_ptr,
+ unsigned int pixels_per_line,
+ unsigned int pixel_step,
+ unsigned int output_height,
+ unsigned int output_width,
+ short *filter
+);
+
+extern void vp8_filter_block2d_bil4x4_var_mmx
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ const short *HFilter,
+ const short *VFilter,
+ int *sum,
+ unsigned int *sumsquared
+);
+extern void vp8_filter_block2d_bil_var_mmx
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ unsigned int Height,
+ const short *HFilter,
+ const short *VFilter,
+ int *sum,
+ unsigned int *sumsquared
+);
+
+unsigned int vp8_sub_pixel_variance4x4_mmx
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse)
+
+{+ int xsum;
+ unsigned int xxsum;
+ vp8_filter_block2d_bil4x4_var_mmx(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line,
+ vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
+ &xsum, &xxsum
+ );
+ *sse = xxsum;
+ return (xxsum - (((unsigned int)xsum * xsum) >> 4));
+}
+
+
+unsigned int vp8_sub_pixel_variance8x8_mmx
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+
+ int xsum;
+ unsigned int xxsum;
+ vp8_filter_block2d_bil_var_mmx(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
+ &xsum, &xxsum
+ );
+ *sse = xxsum;
+ return (xxsum - (((unsigned int)xsum * xsum) >> 6));
+}
+
+unsigned int vp8_sub_pixel_variance16x16_mmx
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+
+ int xsum0, xsum1;
+ unsigned int xxsum0, xxsum1;
+
+
+ vp8_filter_block2d_bil_var_mmx(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
+ &xsum0, &xxsum0
+ );
+
+
+ vp8_filter_block2d_bil_var_mmx(
+ src_ptr + 8, src_pixels_per_line,
+ dst_ptr + 8, dst_pixels_per_line, 16,
+ vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
+ &xsum1, &xxsum1
+ );
+
+ xsum0 += xsum1;
+ xxsum0 += xxsum1;
+
+ *sse = xxsum0;
+ return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 8));
+
+
+}
+
+unsigned int vp8_sub_pixel_variance16x8_mmx
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ int xsum0, xsum1;
+ unsigned int xxsum0, xxsum1;
+
+
+ vp8_filter_block2d_bil_var_mmx(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
+ &xsum0, &xxsum0
+ );
+
+
+ vp8_filter_block2d_bil_var_mmx(
+ src_ptr + 8, src_pixels_per_line,
+ dst_ptr + 8, dst_pixels_per_line, 8,
+ vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
+ &xsum1, &xxsum1
+ );
+
+ xsum0 += xsum1;
+ xxsum0 += xxsum1;
+
+ *sse = xxsum0;
+ return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 7));
+}
+
+unsigned int vp8_sub_pixel_variance8x16_mmx
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ int xsum;
+ unsigned int xxsum;
+ vp8_filter_block2d_bil_var_mmx(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
+ &xsum, &xxsum
+ );
+ *sse = xxsum;
+ return (xxsum - (((unsigned int)xsum * xsum) >> 7));
+}
+
+
+unsigned int vp8_variance_halfpixvar16x16_h_mmx(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ return vp8_sub_pixel_variance16x16_mmx(src_ptr, source_stride, 4, 0,
+ ref_ptr, recon_stride, sse);
+}
+
+
+unsigned int vp8_variance_halfpixvar16x16_v_mmx(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ return vp8_sub_pixel_variance16x16_mmx(src_ptr, source_stride, 0, 4,
+ ref_ptr, recon_stride, sse);
+}
+
+
+unsigned int vp8_variance_halfpixvar16x16_hv_mmx(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ return vp8_sub_pixel_variance16x16_mmx(src_ptr, source_stride, 4, 4,
+ ref_ptr, recon_stride, sse);
+}
--- /dev/null
+++ b/vp8/common/x86/vp8_variance_sse2.c
@@ -1,0 +1,403 @@
+/*
+ * 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 "./vp8_rtcd.h"
+#include "vpx_config.h"
+#include "vp8/common/variance.h"
+#include "vpx_ports/mem.h"
+#include "vp8/common/x86/filter_x86.h"
+
+extern void filter_block1d_h6_mmx(const unsigned char *src_ptr, unsigned short *output_ptr, unsigned int src_pixels_per_line, unsigned int pixel_step, unsigned int output_height, unsigned int output_width, short *filter);
+extern void filter_block1d_v6_mmx(const short *src_ptr, unsigned char *output_ptr, unsigned int pixels_per_line, unsigned int pixel_step, unsigned int output_height, unsigned int output_width, short *filter);
+extern void filter_block1d8_h6_sse2(const unsigned char *src_ptr, unsigned short *output_ptr, unsigned int src_pixels_per_line, unsigned int pixel_step, unsigned int output_height, unsigned int output_width, short *filter);
+extern void filter_block1d8_v6_sse2(const short *src_ptr, unsigned char *output_ptr, unsigned int pixels_per_line, unsigned int pixel_step, unsigned int output_height, unsigned int output_width, short *filter);
+
+extern void vp8_filter_block2d_bil4x4_var_mmx
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ const short *HFilter,
+ const short *VFilter,
+ int *sum,
+ unsigned int *sumsquared
+);
+
+void vp8_filter_block2d_bil_var_sse2
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ unsigned int Height,
+ int xoffset,
+ int yoffset,
+ int *sum,
+ unsigned int *sumsquared
+);
+void vp8_half_horiz_vert_variance8x_h_sse2
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ unsigned int Height,
+ int *sum,
+ unsigned int *sumsquared
+);
+void vp8_half_horiz_vert_variance16x_h_sse2
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ unsigned int Height,
+ int *sum,
+ unsigned int *sumsquared
+);
+void vp8_half_horiz_variance8x_h_sse2
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ unsigned int Height,
+ int *sum,
+ unsigned int *sumsquared
+);
+void vp8_half_horiz_variance16x_h_sse2
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ unsigned int Height,
+ int *sum,
+ unsigned int *sumsquared
+);
+void vp8_half_vert_variance8x_h_sse2
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ unsigned int Height,
+ int *sum,
+ unsigned int *sumsquared
+);
+void vp8_half_vert_variance16x_h_sse2
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ unsigned int Height,
+ int *sum,
+ unsigned int *sumsquared
+);
+
+unsigned int vp8_sub_pixel_variance4x4_wmt
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ int xsum;
+ unsigned int xxsum;
+ vp8_filter_block2d_bil4x4_var_mmx(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line,
+ vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
+ &xsum, &xxsum
+ );
+ *sse = xxsum;
+ return (xxsum - (((unsigned int)xsum * xsum) >> 4));
+}
+
+
+unsigned int vp8_sub_pixel_variance8x8_wmt
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ int xsum;
+ unsigned int xxsum;
+
+ if (xoffset == 4 && yoffset == 0)
+ {+ vp8_half_horiz_variance8x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ &xsum, &xxsum);
+ }
+ else if (xoffset == 0 && yoffset == 4)
+ {+ vp8_half_vert_variance8x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ &xsum, &xxsum);
+ }
+ else if (xoffset == 4 && yoffset == 4)
+ {+ vp8_half_horiz_vert_variance8x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ &xsum, &xxsum);
+ }
+ else
+ {+ vp8_filter_block2d_bil_var_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ xoffset, yoffset,
+ &xsum, &xxsum);
+ }
+
+ *sse = xxsum;
+ return (xxsum - (((unsigned int)xsum * xsum) >> 6));
+}
+
+unsigned int vp8_sub_pixel_variance16x16_wmt
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ int xsum0, xsum1;
+ unsigned int xxsum0, xxsum1;
+
+
+ /* note we could avoid these if statements if the calling function
+ * just called the appropriate functions inside.
+ */
+ if (xoffset == 4 && yoffset == 0)
+ {+ vp8_half_horiz_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum0, &xxsum0);
+ }
+ else if (xoffset == 0 && yoffset == 4)
+ {+ vp8_half_vert_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum0, &xxsum0);
+ }
+ else if (xoffset == 4 && yoffset == 4)
+ {+ vp8_half_horiz_vert_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum0, &xxsum0);
+ }
+ else
+ {+ vp8_filter_block2d_bil_var_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ xoffset, yoffset,
+ &xsum0, &xxsum0
+ );
+
+ vp8_filter_block2d_bil_var_sse2(
+ src_ptr + 8, src_pixels_per_line,
+ dst_ptr + 8, dst_pixels_per_line, 16,
+ xoffset, yoffset,
+ &xsum1, &xxsum1
+ );
+ xsum0 += xsum1;
+ xxsum0 += xxsum1;
+ }
+
+ *sse = xxsum0;
+ return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 8));
+}
+
+unsigned int vp8_sub_pixel_variance16x8_wmt
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+
+)
+{+ int xsum0, xsum1;
+ unsigned int xxsum0, xxsum1;
+
+ if (xoffset == 4 && yoffset == 0)
+ {+ vp8_half_horiz_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ &xsum0, &xxsum0);
+ }
+ else if (xoffset == 0 && yoffset == 4)
+ {+ vp8_half_vert_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ &xsum0, &xxsum0);
+ }
+ else if (xoffset == 4 && yoffset == 4)
+ {+ vp8_half_horiz_vert_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ &xsum0, &xxsum0);
+ }
+ else
+ {+ vp8_filter_block2d_bil_var_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ xoffset, yoffset,
+ &xsum0, &xxsum0);
+
+ vp8_filter_block2d_bil_var_sse2(
+ src_ptr + 8, src_pixels_per_line,
+ dst_ptr + 8, dst_pixels_per_line, 8,
+ xoffset, yoffset,
+ &xsum1, &xxsum1);
+ xsum0 += xsum1;
+ xxsum0 += xxsum1;
+ }
+
+ *sse = xxsum0;
+ return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 7));
+}
+
+unsigned int vp8_sub_pixel_variance8x16_wmt
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ int xsum;
+ unsigned int xxsum;
+
+ if (xoffset == 4 && yoffset == 0)
+ {+ vp8_half_horiz_variance8x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum, &xxsum);
+ }
+ else if (xoffset == 0 && yoffset == 4)
+ {+ vp8_half_vert_variance8x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum, &xxsum);
+ }
+ else if (xoffset == 4 && yoffset == 4)
+ {+ vp8_half_horiz_vert_variance8x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum, &xxsum);
+ }
+ else
+ {+ vp8_filter_block2d_bil_var_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ xoffset, yoffset,
+ &xsum, &xxsum);
+ }
+
+ *sse = xxsum;
+ return (xxsum - (((unsigned int)xsum * xsum) >> 7));
+}
+
+
+unsigned int vp8_variance_halfpixvar16x16_h_wmt(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse)
+{+ int xsum0;
+ unsigned int xxsum0;
+
+ vp8_half_horiz_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum0, &xxsum0);
+
+ *sse = xxsum0;
+ return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 8));
+}
+
+
+unsigned int vp8_variance_halfpixvar16x16_v_wmt(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse)
+{+ int xsum0;
+ unsigned int xxsum0;
+ vp8_half_vert_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum0, &xxsum0);
+
+ *sse = xxsum0;
+ return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 8));
+}
+
+
+unsigned int vp8_variance_halfpixvar16x16_hv_wmt(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse)
+{+ int xsum0;
+ unsigned int xxsum0;
+
+ vp8_half_horiz_vert_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum0, &xxsum0);
+
+ *sse = xxsum0;
+ return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 8));
+}
--- a/vp8/vp8_common.mk
+++ b/vp8/vp8_common.mk
@@ -86,8 +86,8 @@
VP8_COMMON_SRCS-$(HAVE_MMX) += common/x86/loopfilter_mmx.asm
VP8_COMMON_SRCS-$(HAVE_MMX) += common/x86/recon_mmx.asm
VP8_COMMON_SRCS-$(HAVE_MMX) += common/x86/subpixel_mmx.asm
-VP8_COMMON_SRCS-$(HAVE_MMX) += common/x86/variance_mmx.c
-VP8_COMMON_SRCS-$(HAVE_MMX) += common/x86/variance_impl_mmx.asm
+VP8_COMMON_SRCS-$(HAVE_MMX) += common/x86/vp8_variance_mmx.c
+VP8_COMMON_SRCS-$(HAVE_MMX) += common/x86/vp8_variance_impl_mmx.asm
VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/copy_sse2.asm
VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/idct_blk_sse2.c
VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/idctllm_sse2.asm
@@ -96,7 +96,7 @@
VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/subpixel_sse2.asm
VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/loopfilter_sse2.asm
VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/iwalsh_sse2.asm
-VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/variance_sse2.c
+VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp8_variance_sse2.c
VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/variance_impl_sse2.asm
VP8_COMMON_SRCS-$(HAVE_SSE3) += common/x86/copy_sse3.asm
VP8_COMMON_SRCS-$(HAVE_SSSE3) += common/x86/subpixel_ssse3.asm