pixel-sse2.asm

从FFMPEG转换而来的H264解码程序,VC下编译..

;*****************************************************************************
;* pixel-sse2.asm: h264 encoder library
;*****************************************************************************
;* Copyright (C) 2005 x264 project
;*
;* Authors: Alex Izvorski <[e-mail]>
;*          Loren Merritt <[e-mail]>
;*
;* This program is free software; you can redistribute it and/or modify
;* it under the terms of the GNU General Public License as published by
;* the Free Software Foundation; either version 2 of the License, or
;* (at your option) any later version.
;*
;* This program is distributed in the hope that it will be useful,
;* but WITHOUT ANY WARRANTY; without even the implied warranty of
;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
;* GNU General Public License for more details.
;*
;* You should have received a copy of the GNU General Public License
;* along with this program; if not, write to the Free Software
;* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111, USA.
;*****************************************************************************

BITS 32

;=============================================================================
; Macros and other preprocessor constants
;=============================================================================

%include "i386inc.asm"

SECTION_RODATA

pw_1:    times 8 dw 1
ssim_c1: times 4 dd 416    ; .01*.01*255*255*64
ssim_c2: times 4 dd 235963 ; .03*.03*255*255*64*63
mask_ff: times 16 db 0xff
         times 16 db 0


SECTION .text

%macro HADDW 2    ; sum junk
    ; ebx is no longer used at this point, so no push needed
    picgetgot ebx
    pmaddwd %1, [pw_1 GOT_ebx]
    movhlps %2, %1
    paddd   %1, %2
    pshuflw %2, %1, 0xE 
    paddd   %1, %2
%endmacro

%macro SAD_INC_4x16P_SSE2 0
    movdqu  xmm1,   [ecx]
    movdqu  xmm2,   [ecx+edx]
    lea     ecx,    [ecx+2*edx]
    movdqu  xmm3,   [ecx]
    movdqu  xmm4,   [ecx+edx]
    psadbw  xmm1,   [eax]
    psadbw  xmm2,   [eax+ebx]
    lea     eax,    [eax+2*ebx]
    psadbw  xmm3,   [eax]
    psadbw  xmm4,   [eax+ebx]
    lea     eax,    [eax+2*ebx]
    lea     ecx,    [ecx+2*edx]
    paddw   xmm1,   xmm2
    paddw   xmm3,   xmm4
    paddw   xmm0,   xmm1
    paddw   xmm0,   xmm3
%endmacro

%macro SAD_START_SSE2 0
    push    ebx

    mov     eax,    [esp+ 8]    ; pix1
    mov     ebx,    [esp+12]    ; stride1
    mov     ecx,    [esp+16]    ; pix2
    mov     edx,    [esp+20]    ; stride2
%endmacro

%macro SAD_END_SSE2 0
    movdqa  xmm1, xmm0
    psrldq  xmm0,  8
    paddw   xmm0, xmm1
    movd    eax,  xmm0

    pop ebx
    ret
%endmacro

;-----------------------------------------------------------------------------
;   int __cdecl x264_pixel_sad_16x16_sse2 (uint8_t *, int, uint8_t *, int )
;-----------------------------------------------------------------------------
cglobal x264_pixel_sad_16x16_sse2
    SAD_START_SSE2
    movdqu xmm0, [ecx]
    movdqu xmm1, [ecx+edx]
    lea    ecx,  [ecx+2*edx]
    movdqu xmm2, [ecx]
    movdqu xmm3, [ecx+edx]
    lea    ecx,  [ecx+2*edx]
    psadbw xmm0, [eax]
    psadbw xmm1, [eax+ebx]
    lea    eax,  [eax+2*ebx]
    movdqu xmm4, [ecx]
    paddw  xmm0, xmm1
    psadbw xmm2, [eax]
    psadbw xmm3, [eax+ebx]
    lea    eax,  [eax+2*ebx]
    movdqu xmm5, [ecx+edx]
    lea    ecx,  [ecx+2*edx]
    paddw  xmm2, xmm3
    movdqu xmm6, [ecx]
    movdqu xmm7, [ecx+edx]
    lea    ecx,  [ecx+2*edx]
    paddw  xmm0, xmm2
    psadbw xmm4, [eax]
    psadbw xmm5, [eax+ebx]
    lea    eax,  [eax+2*ebx]
    movdqu xmm1, [ecx]
    paddw  xmm4, xmm5
    psadbw xmm6, [eax]
    psadbw xmm7, [eax+ebx]
    lea    eax,  [eax+2*ebx]
    movdqu xmm2, [ecx+edx]
    lea    ecx,  [ecx+2*edx]
    paddw  xmm6, xmm7
    movdqu xmm3, [ecx]
    paddw  xmm0, xmm4
    movdqu xmm4, [ecx+edx]
    lea    ecx,  [ecx+2*edx]
    paddw  xmm0, xmm6
    psadbw xmm1, [eax]
    psadbw xmm2, [eax+ebx]
    lea    eax,  [eax+2*ebx]
    movdqu xmm5, [ecx]
    paddw  xmm1, xmm2
    psadbw xmm3, [eax]
    psadbw xmm4, [eax+ebx]
    lea    eax,  [eax+2*ebx]
    movdqu xmm6, [ecx+edx]
    lea    ecx,  [ecx+2*edx]
    paddw  xmm3, xmm4
    movdqu xmm7, [ecx]
    paddw  xmm0, xmm1
    movdqu xmm1, [ecx+edx]
    paddw  xmm0, xmm3
    psadbw xmm5, [eax]
    psadbw xmm6, [eax+ebx]
    lea    eax,  [eax+2*ebx]
    paddw  xmm5, xmm6
    psadbw xmm7, [eax]
    psadbw xmm1, [eax+ebx]
    paddw  xmm7, xmm1
    paddw  xmm0, xmm5
    paddw  xmm0, xmm7
    SAD_END_SSE2

;-----------------------------------------------------------------------------
;   int __cdecl x264_pixel_sad_16x8_sse2 (uint8_t *, int, uint8_t *, int )
;-----------------------------------------------------------------------------
cglobal x264_pixel_sad_16x8_sse2
    SAD_START_SSE2
    pxor    xmm0,   xmm0
    SAD_INC_4x16P_SSE2
    SAD_INC_4x16P_SSE2
    SAD_END_SSE2


%macro SAD_X3_START_1x16P 0
    push    edi
    push    esi
    mov     edi,    [esp+12]
    mov     eax,    [esp+16]
    mov     ecx,    [esp+20]
    mov     edx,    [esp+24]
    mov     esi,    [esp+28]
    movdqa  xmm3,   [edi]
    movdqu  xmm0,   [eax]
    movdqu  xmm1,   [ecx]
    movdqu  xmm2,   [edx]
    psadbw  xmm0,   xmm3
    psadbw  xmm1,   xmm3
    psadbw  xmm2,   xmm3
%endmacro

%macro SAD_X3_1x16P 2
    movdqa  xmm3,   [edi+%1]
    movdqu  xmm4,   [eax+%2]
    movdqu  xmm5,   [ecx+%2]
    movdqu  xmm6,   [edx+%2]
    psadbw  xmm4,   xmm3
    psadbw  xmm5,   xmm3
    psadbw  xmm6,   xmm3
    paddw   xmm0,   xmm4
    paddw   xmm1,   xmm5
    paddw   xmm2,   xmm6
%endmacro

%macro SAD_X3_2x16P 1
%if %1
    SAD_X3_START_1x16P
%else
    SAD_X3_1x16P 0, 0
%endif
    SAD_X3_1x16P FENC_STRIDE, esi
    add     edi, 2*FENC_STRIDE
    lea     eax, [eax+2*esi]
    lea     ecx, [ecx+2*esi]
    lea     edx, [edx+2*esi]
%endmacro

%macro SAD_X4_START_1x16P 0
    push    edi
    push    esi
    push    ebx
    mov     edi,    [esp+16]
    mov     eax,    [esp+20]
    mov     ebx,    [esp+24]
    mov     ecx,    [esp+28]
    mov     edx,    [esp+32]
    mov     esi,    [esp+36]
    movdqa  xmm7,   [edi]
    movdqu  xmm0,   [eax]
    movdqu  xmm1,   [ebx]
    movdqu  xmm2,   [ecx]
    movdqu  xmm3,   [edx]
    psadbw  xmm0,   xmm7
    psadbw  xmm1,   xmm7
    psadbw  xmm2,   xmm7
    psadbw  xmm3,   xmm7
%endmacro

%macro SAD_X4_1x16P 2
    movdqa  xmm7,   [edi+%1]
    movdqu  xmm4,   [eax+%2]
    movdqu  xmm5,   [ebx+%2]
    movdqu  xmm6,   [ecx+%2]
    psadbw  xmm4,   xmm7
    psadbw  xmm5,   xmm7
    paddw   xmm0,   xmm4
    psadbw  xmm6,   xmm7
    movdqu  xmm4,   [edx+%2]
    paddw   xmm1,   xmm5
    psadbw  xmm4,   xmm7
    paddw   xmm2,   xmm6
    paddw   xmm3,   xmm4
%endmacro

%macro SAD_X4_2x16P 1
%if %1
    SAD_X4_START_1x16P
%else
    SAD_X4_1x16P 0, 0
%endif
    SAD_X4_1x16P FENC_STRIDE, esi
    add     edi, 2*FENC_STRIDE
    lea     eax, [eax+2*esi]
    lea     ebx, [ebx+2*esi]
    lea     ecx, [ecx+2*esi]
    lea     edx, [edx+2*esi]
%endmacro

%macro SAD_X3_END 0
    mov     eax,  [esp+32]
    pshufd  xmm4, xmm0, 2
    pshufd  xmm5, xmm1, 2
    pshufd  xmm6, xmm2, 2
    paddw   xmm0, xmm4
    paddw   xmm1, xmm5
    paddw   xmm2, xmm6
    movd    [eax+0], xmm0
    movd    [eax+4], xmm1
    movd    [eax+8], xmm2
    pop     esi
    pop     edi
    ret
%endmacro

%macro SAD_X4_END 0
    mov     eax,  [esp+40]
    pshufd  xmm4, xmm0, 2
    pshufd  xmm5, xmm1, 2
    pshufd  xmm6, xmm2, 2
    pshufd  xmm7, xmm3, 2
    paddw   xmm0, xmm4
    paddw   xmm1, xmm5
    paddw   xmm2, xmm6
    paddw   xmm3, xmm7
    movd    [eax+0], xmm0
    movd    [eax+4], xmm1
    movd    [eax+8], xmm2
    movd    [eax+12], xmm3
    pop     ebx
    pop     esi
    pop     edi
    ret
%endmacro

;-----------------------------------------------------------------------------
;  void x264_pixel_sad_x3_16x16_sse2( uint8_t *fenc, uint8_t *pix0, uint8_t *pix1,
;                                     uint8_t *pix2, int i_stride, int scores[3] )
;-----------------------------------------------------------------------------
%macro SAD_X 3
cglobal x264_pixel_sad_x%1_%2x%3_sse2
    SAD_X%1_2x%2P 1
%rep %3/2-1
    SAD_X%1_2x%2P 0
%endrep
    SAD_X%1_END
%endmacro

SAD_X 3, 16, 16
SAD_X 3, 16,  8
SAD_X 4, 16, 16
SAD_X 4, 16,  8


%macro SSD_INC_2x16P_SSE2 0
    movdqu  xmm1,   [eax]
    movdqu  xmm2,   [ecx]
    movdqu  xmm3,   [eax+ebx]
    movdqu  xmm4,   [ecx+edx]

    movdqa  xmm5,   xmm1
    movdqa  xmm6,   xmm3
    psubusb xmm1,   xmm2
    psubusb xmm3,   xmm4
    psubusb xmm2,   xmm5
    psubusb xmm4,   xmm6
    por     xmm1,   xmm2
    por     xmm3,   xmm4

    movdqa  xmm2,   xmm1
    movdqa  xmm4,   xmm3
    punpcklbw xmm1, xmm7
    punpckhbw xmm2, xmm7
    punpcklbw xmm3, xmm7
    punpckhbw xmm4, xmm7
    pmaddwd xmm1,   xmm1
    pmaddwd xmm2,   xmm2
    pmaddwd xmm3,   xmm3
    pmaddwd xmm4,   xmm4

    lea     eax,    [eax+2*ebx]
    lea     ecx,    [ecx+2*edx]

    paddd   xmm1,   xmm2
    paddd   xmm3,   xmm4
    paddd   xmm0,   xmm1
    paddd   xmm0,   xmm3
%endmacro

%macro SSD_START_SSE2 0
    push    ebx

    mov     eax,    [esp+ 8]    ; pix1
    mov     ebx,    [esp+12]    ; stride1
    mov     ecx,    [esp+16]    ; pix2
    mov     edx,    [esp+20]    ; stride2

    pxor    xmm7,   xmm7         ; zero
    pxor    xmm0,   xmm0         ; mm0 holds the sum