pixel-a.asm

linux下编译已经通过

;*****************************************************************************
;* pixel.asm: h264 encoder library
;*****************************************************************************
;* Copyright (C) 2003 x264 project
;* $Id: pixel.asm,v 1.1 2004/06/03 19:27:07 fenrir Exp $
;*
;* Authors: Laurent Aimar <[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"

; sad

%macro SAD_INC_2x16P 0
    movq    mm1,    [eax]
    movq    mm2,    [eax+8]
    movq    mm3,    [eax+ebx]
    movq    mm4,    [eax+ebx+8]
    psadbw  mm1,    [ecx]
    psadbw  mm2,    [ecx+8]
    psadbw  mm3,    [ecx+edx]
    psadbw  mm4,    [ecx+edx+8]
    lea     eax,    [eax+2*ebx]
    paddw   mm1,    mm2
    paddw   mm3,    mm4
    lea     ecx,    [ecx+2*edx]
    paddw   mm0,    mm1
    paddw   mm0,    mm3
%endmacro

%macro SAD_INC_2x8P 0
    movq    mm1,    [eax]
    movq    mm2,    [eax+ebx]
    psadbw  mm1,    [ecx]
    psadbw  mm2,    [ecx+edx]
    lea     eax,    [eax+2*ebx]
    paddw   mm0,    mm1
    paddw   mm0,    mm2
    lea     ecx,    [ecx+2*edx]
%endmacro

%macro SAD_INC_2x4P 0
    movd    mm1,    [eax]
    movd    mm2,    [ecx]
    punpckldq mm1,  [eax+ebx]
    punpckldq mm2,  [ecx+edx]
    psadbw  mm1,    mm2
    paddw   mm0,    mm1
    lea     eax,    [eax+2*ebx]
    lea     ecx,    [ecx+2*edx]
%endmacro

; sad x3 / x4

%macro SAD_X3_START 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]
%endmacro

%macro SAD_X3_START_1x8P 0
    movq    mm3,    [edi]
    movq    mm0,    [eax]
    movq    mm1,    [ecx]
    movq    mm2,    [edx]
    psadbw  mm0,    mm3
    psadbw  mm1,    mm3
    psadbw  mm2,    mm3
%endmacro

%macro SAD_X3_1x8P 2
    movq    mm3,    [edi+%1]
    movq    mm4,    [eax+%2]
    movq    mm5,    [ecx+%2]
    movq    mm6,    [edx+%2]
    psadbw  mm4,    mm3
    psadbw  mm5,    mm3
    psadbw  mm6,    mm3
    paddw   mm0,    mm4
    paddw   mm1,    mm5
    paddw   mm2,    mm6
%endmacro

%macro SAD_X3_START_2x4P 3
    movd      mm3,  [edi]
    movd      %1,   [eax]
    movd      %2,   [ecx]
    movd      %3,   [edx]
    punpckldq mm3,  [edi+FENC_STRIDE]
    punpckldq %1,   [eax+esi]
    punpckldq %2,   [ecx+esi]
    punpckldq %3,   [edx+esi]
    psadbw    %1,   mm3
    psadbw    %2,   mm3
    psadbw    %3,   mm3
%endmacro

%macro SAD_X3_2x16P 1
%if %1
    SAD_X3_START
    SAD_X3_START_1x8P
%else
    SAD_X3_1x8P 0, 0
%endif
    SAD_X3_1x8P 8, 8
    SAD_X3_1x8P FENC_STRIDE, esi
    SAD_X3_1x8P FENC_STRIDE+8, esi+8
    add     edi, 2*FENC_STRIDE
    lea     eax, [eax+2*esi]
    lea     ecx, [ecx+2*esi]
    lea     edx, [edx+2*esi]
%endmacro

%macro SAD_X3_2x8P 1
%if %1
    SAD_X3_START
    SAD_X3_START_1x8P
%else
    SAD_X3_1x8P 0, 0
%endif
    SAD_X3_1x8P 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_X3_2x4P 1
%if %1
    SAD_X3_START
    SAD_X3_START_2x4P mm0, mm1, mm2
%else
    SAD_X3_START_2x4P mm4, mm5, mm6
    paddw   mm0, mm4
    paddw   mm1, mm5
    paddw   mm2, mm6
%endif
    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 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]
%endmacro

%macro SAD_X4_START_1x8P 0
    movq    mm7,    [edi]
    movq    mm0,    [eax]
    movq    mm1,    [ebx]
    movq    mm2,    [ecx]
    movq    mm3,    [edx]
    psadbw  mm0,    mm7
    psadbw  mm1,    mm7
    psadbw  mm2,    mm7
    psadbw  mm3,    mm7
%endmacro

%macro SAD_X4_1x8P 2
    movq    mm7,    [edi+%1]
    movq    mm4,    [eax+%2]
    movq    mm5,    [ebx+%2]
    movq    mm6,    [ecx+%2]
    psadbw  mm4,    mm7
    psadbw  mm5,    mm7
    psadbw  mm6,    mm7
    psadbw  mm7,    [edx+%2]
    paddw   mm0,    mm4
    paddw   mm1,    mm5
    paddw   mm2,    mm6
    paddw   mm3,    mm7
%endmacro

%macro SAD_X4_START_2x4P 0
    movd      mm7,  [edi]
    movd      mm0,  [eax]
    movd      mm1,  [ebx]
    movd      mm2,  [ecx]
    movd      mm3,  [edx]
    punpckldq mm7,  [edi+FENC_STRIDE]
    punpckldq mm0,  [eax+esi]
    punpckldq mm1,  [ebx+esi]
    punpckldq mm2,  [ecx+esi]
    punpckldq mm3,  [edx+esi]
    psadbw    mm0,  mm7
    psadbw    mm1,  mm7
    psadbw    mm2,  mm7
    psadbw    mm3,  mm7
%endmacro   
    
%macro SAD_X4_INC_2x4P 0
    movd      mm7,  [edi]
    movd      mm4,  [eax]
    movd      mm5,  [ebx]
    punpckldq mm7,  [edi+FENC_STRIDE]
    punpckldq mm4,  [eax+esi]
    punpckldq mm5,  [ebx+esi]
    psadbw    mm4,  mm7
    psadbw    mm5,  mm7
    paddw     mm0,  mm4
    paddw     mm1,  mm5
    movd      mm4,  [ecx]
    movd      mm5,  [edx]
    punpckldq mm4,  [ecx+esi]
    punpckldq mm5,  [edx+esi]
    psadbw    mm4,  mm7
    psadbw    mm5,  mm7
    paddw     mm2,  mm4
    paddw     mm3,  mm5
%endmacro

%macro SAD_X4_2x16P 1
%if %1
    SAD_X4_START
    SAD_X4_START_1x8P
%else
    SAD_X4_1x8P 0, 0
%endif
    SAD_X4_1x8P 8, 8
    SAD_X4_1x8P FENC_STRIDE, esi
    SAD_X4_1x8P FENC_STRIDE+8, esi+8
    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_X4_2x8P 1
%if %1
    SAD_X4_START
    SAD_X4_START_1x8P
%else
    SAD_X4_1x8P 0, 0
%endif
    SAD_X4_1x8P 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_X4_2x4P 1
%if %1
    SAD_X4_START
    SAD_X4_START_2x4P
%else
    SAD_X4_INC_2x4P
%endif
    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]
    movd    [eax+0], mm0
    movd    [eax+4], mm1  
    movd    [eax+8], mm2
    pop     esi
    pop     edi
    ret
%endmacro

%macro SAD_X4_END 0
    mov     eax,  [esp+40]
    movd    [eax+0], mm0
    movd    [eax+4], mm1  
    movd    [eax+8], mm2
    movd    [eax+12], mm3
    pop     ebx
    pop     esi
    pop     edi
    ret
%endmacro

; ssd

%macro SSD_INC_1x16P 0
    movq    mm1,    [eax]
    movq    mm2,    [ecx]
    movq    mm3,    [eax+8]
    movq    mm4,    [ecx+8]

    movq    mm5,    mm2
    movq    mm6,    mm4
    psubusb mm2,    mm1
    psubusb mm4,    mm3
    psubusb mm1,    mm5
    psubusb mm3,    mm6
    por     mm1,    mm2
    por     mm3,    mm4

    movq    mm2,    mm1
    movq    mm4,    mm3
    punpcklbw mm1,  mm7
    punpcklbw mm3,  mm7
    punpckhbw mm2,  mm7
    punpckhbw mm4,  mm7
    pmaddwd mm1,    mm1
    pmaddwd mm2,    mm2
    pmaddwd mm3,    mm3
    pmaddwd mm4,    mm4

    add     eax,    ebx
    add     ecx,    edx
    paddd   mm0,    mm1
    paddd   mm0,    mm2
    paddd   mm0,    mm3
    paddd   mm0,    mm4
%endmacro

%macro SSD_INC_1x8P 0
    movq    mm1,    [eax]
    movq    mm2,    [ecx]

    movq    mm5,    mm2
    psubusb mm2,    mm1
    psubusb mm1,    mm5
    por     mm1,    mm2         ; mm1 = 8bit abs diff

    movq    mm2,    mm1
    punpcklbw mm1,  mm7
    punpckhbw mm2,  mm7         ; (mm1,mm2) = 16bit abs diff
    pmaddwd mm1,    mm1
    pmaddwd mm2,    mm2

    add     eax,    ebx
    add     ecx,    edx
    paddd   mm0,    mm1
    paddd   mm0,    mm2
%endmacro

%macro SSD_INC_1x4P 0
    movd    mm1,    [eax]
    movd    mm2,    [ecx]

    movq    mm5,    mm2
    psubusb mm2,    mm1
    psubusb mm1,    mm5
    por     mm1,    mm2
    punpcklbw mm1,  mm7
    pmaddwd mm1,    mm1

    add     eax,    ebx
    add     ecx,    edx
    paddd   mm0,    mm1
%endmacro

; satd

%macro SUMSUB_BADC 4
    paddw %1,   %2
    paddw %3,   %4
    paddw %2,   %2
    paddw %4,   %4
    psubw %2,   %1
    psubw %4,   %3
%endmacro

%macro HADAMARD4x4 4
    SUMSUB_BADC %1, %2, %3, %4
    SUMSUB_BADC %1, %3, %2, %4
%endmacro

%macro SBUTTERFLYwd 3
    movq        %3, %1
    punpcklwd   %1, %2
    punpckhwd   %3, %2
%endmacro

%macro SBUTTERFLYdq 3
    movq        %3, %1
    punpckldq   %1, %2
    punpckhdq   %3, %2
%endmacro

%macro TRANSPOSE4x4 5   ; abcd-t -> adtc
    SBUTTERFLYwd %1, %2, %5
    SBUTTERFLYwd %3, %4, %2
    SBUTTERFLYdq %1, %3, %4
    SBUTTERFLYdq %5, %2, %3
%endmacro

%macro MMX_ABS 2        ; mma, tmp
    pxor    %2, %2
    psubw   %2, %1
    pmaxsw  %1, %2
%endmacro

%macro MMX_ABS_TWO 4    ; mma, mmb, tmp0, tmp1
    pxor    %3, %3
    pxor    %4, %4
    psubw   %3, %1
    psubw   %4, %2
    pmaxsw  %1, %3
    pmaxsw  %2, %4
%endmacro

%macro HADAMARD4x4_SUM 1    ; %1 - dest (row sum of one block)
    HADAMARD4x4 mm4, mm5, mm6, mm7
    TRANSPOSE4x4 mm4, mm5, mm6, mm7, %1
    HADAMARD4x4 mm4, mm7, %1, mm6
    MMX_ABS_TWO mm4, mm7, mm3, mm5
    MMX_ABS_TWO %1,  mm6, mm3, mm5
    paddw       %1,  mm4
    paddw       mm6, mm7
    pavgw       %1,  mm6
%endmacro

%macro LOAD_DIFF_4P 4  ; mmp, mmt, dx, dy
    movd        %1, [eax+ebx*%4+%3]
    movd        %2, [ecx+edx*%4+%3]
    punpcklbw   %1, %2
    punpcklbw   %2, %2
    psubw       %1, %2
%endmacro

; in: %2 = horizontal offset
; in: %3 = whether we need to increment pix1 and pix2
; clobber: mm3..mm7
; out: %1 = satd
%macro LOAD_DIFF_HADAMARD_SUM 3
%if %3
    LOAD_DIFF_4P mm4, mm3, %2, 0
    LOAD_DIFF_4P mm5, mm3, %2, 1
    lea  eax, [eax+2*ebx]
    lea  ecx, [ecx+2*edx]
    LOAD_DIFF_4P mm6, mm3, %2, 0
    LOAD_DIFF_4P mm7, mm3, %2, 1
    lea  eax, [eax+2*ebx]
    lea  ecx, [ecx+2*edx]
%else
    LOAD_DIFF_4P mm4, mm3, %2, 0
    LOAD_DIFF_4P mm6, mm3, %2, 2
    add  eax, ebx
    add  ecx, edx
    LOAD_DIFF_4P mm5, mm3, %2, 0
    LOAD_DIFF_4P mm7, mm3, %2, 2
%endif
    HADAMARD4x4_SUM %1
%endmacro

;=============================================================================
; Code
;=============================================================================

SECTION .text

%macro SAD_START 0
    push    ebx

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

    pxor    mm0,    mm0
%endmacro
%macro SAD_END 0
    movd eax,    mm0

    pop ebx
    ret
%endmacro

;-----------------------------------------------------------------------------
;   int __cdecl x264_pixel_sad_16x16_mmxext (uint8_t *, int, uint8_t *, int )
;-----------------------------------------------------------------------------
%macro SAD 2
cglobal x264_pixel_sad_%1x%2_mmxext
    SAD_START
%rep %2/2
    SAD_INC_2x%1P
%endrep
    SAD_END
%endmacro

SAD 16, 16
SAD 16,  8
SAD  8, 16
SAD  8,  8
SAD  8,  4
SAD  4,  8
SAD  4,  4

;-----------------------------------------------------------------------------
;  void x264_pixel_sad_x3_16x16_mmxext( 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_mmxext
    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 3,  8, 16
SAD_X 3,  8,  8
SAD_X 3,  8,  4
SAD_X 3,  4,  8
SAD_X 3,  4,  4
SAD_X 4, 16, 16
SAD_X 4, 16,  8
SAD_X 4,  8, 16
SAD_X 4,  8,  8
SAD_X 4,  8,  4
SAD_X 4,  4,  8
SAD_X 4,  4,  4


;-----------------------------------------------------------------------------
;   int __cdecl x264_pixel_sad_pde_16x16_mmxext (uint8_t *, int, uint8_t *, int, int )
;-----------------------------------------------------------------------------
%macro SAD_PDE 2
cglobal x264_pixel_sad_pde_%1x%2_mmxext
    SAD_START
%rep %2/4
    SAD_INC_2x%1P
%endrep

    movd ebx, mm0
    cmp  ebx, [esp+24] ; prev_score
    jl   .continue
    pop  ebx
    mov  eax, 0xffff
    ret
ALIGN 4