colorspace_mmx.inc

wince下的xvidcore开发库,可用于MP4等视频播放开发

 ;------------------------------------------------------------------------------
 ;
 ; MAKE_COLORSPACE(NAME,STACK, BYTES,PIXELS,ROWS, FUNC, ARG1)
 ;
 ; This macro provides a assembler width/height scroll loop
 ; NAME          function name
 ; STACK         additional stack bytes required by FUNC
 ; BYTES         bytes-per-pixel for the given colorspace
 ; PIXELS        pixels (columns) operated on per FUNC call
 ; VPIXELS       vpixels (rows) operated on per FUNC call
 ; FUNC          conversion macro name; we expect to find FUNC_INIT and FUNC macros
 ; ARG1          argument passed to FUNC
 ; 
 ; throughout the FUNC the registers mean:
 ; eax           y_stride
 ; ebx           u_ptr
 ; ecx           v_ptr
 ; edx           x_stride
 ; esi           y_ptr
 ; edi           x_ptr
 ; ebp           width
 ;
 ;------------------------------------------------------------------------------
 %macro          MAKE_COLORSPACE                 8
 %define NAME            %1
 %define STACK           %2
 %define BYTES           %3
 %define PIXELS          %4
 %define VPIXELS         %5
 %define FUNC            %6
 %define ARG1            %7
 %define ARG2            %8
         ; --- define function global/symbol
 ALIGN 16
 cglobal NAME
 NAME:
         ; --- init stack ---
 
 %define pushsize        16
 %define localsize       20 + STACK
 
 %define vflip           esp + localsize + pushsize + 40
 %define height          esp + localsize + pushsize + 36
 %define width           esp + localsize + pushsize + 32
 %define uv_stride       esp + localsize + pushsize + 28
 %define y_stride        esp + localsize + pushsize + 24
 %define v_ptr           esp + localsize + pushsize + 20
 %define u_ptr           esp + localsize + pushsize + 16
 %define y_ptr           esp + localsize + pushsize + 12
 %define x_stride        esp + localsize + pushsize + 8
 %define x_ptr           esp + localsize + pushsize + 4
 %define _ip             esp + localsize + pushsize + 0
 
   push ebx    ;   esp + localsize + 16
   push esi    ;   esp + localsize + 8
   push edi    ;   esp + localsize + 4
   push ebp    ;   esp + localsize + 0
 
 %define x_dif           esp + localsize - 4
 %define y_dif           esp + localsize - 8
 %define uv_dif          esp + localsize - 12
 %define fixed_width     esp + localsize - 16
 %define tmp_height      esp + localsize - 20
 
     sub esp, localsize
 
     ; --- init varibles ---
     
   mov eax, [width]          ; fixed width
   add eax, 15               ;
   and eax, ~15              ;
   mov [fixed_width],eax     ;
 
   mov ebx, [x_stride]       ;
 %rep BYTES
   sub ebx, eax              ;
 %endrep
   mov [x_dif], ebx          ; x_dif = x_stride - BYTES*fixed_width
 
   mov ebx, [y_stride]       ;
   sub ebx, eax              ;
   mov [y_dif], ebx          ; y_dif = y_stride - fixed_width
 
   mov ebx, [uv_stride]      ;
   mov ecx, eax              ;
   shr ecx, 1                ;
   sub ebx, ecx              ;
   mov [uv_dif], ebx         ; uv_dif = uv_stride - fixed_width/2
 
   mov esi, [y_ptr]          ; $esi$ = y_ptr
   mov edi, [x_ptr]          ; $edi$ = x_ptr
   mov edx, [x_stride]       ; $edx$ = x_stride
   mov ebp, [height]         ; $ebp$ = height
 
 
   mov ebx, [vflip]
   or ebx, ebx
   jz .dont_flip
 
     ; --- do flipping ---
 
   xor ebx,ebx
 %rep BYTES
   sub ebx, eax
 %endrep
   sub ebx, edx
   mov [x_dif], ebx          ; x_dif = -BYTES*fixed_width - x_stride
 
   mov eax, ebp
   sub eax, 1
   push edx                
   mul edx
   pop edx
   add edi, eax              ; $edi$ += (height-1) * x_stride
 
   neg edx                   ; x_stride = -x_stride
 
 .dont_flip
 
     ; --- begin loop ---
 
   mov eax, [y_stride]       ; $eax$ = y_stride
   mov ebx, [u_ptr]          ; $ebx$ = u_ptr
   mov ecx, [v_ptr]          ; $ecx$ = v_ptr
 
   FUNC %+ _INIT ARG1, ARG2  ; call FUNC_INIT
 
 .y_loop
   mov [tmp_height], ebp
   mov ebp, [fixed_width]
 
 .x_loop
   FUNC ARG1, ARG2           ; call FUNC
 
   add edi, BYTES*PIXELS     ; x_ptr += BYTES*PIXELS
   add esi, PIXELS           ; y_ptr += PIXELS
   add ebx, PIXELS/2         ; u_ptr += PIXELS/2
   add ecx, PIXELS/2         ; v_ptr += PIXELS/2
         
   sub ebp, PIXELS           ; $ebp$ -= PIXELS
   jg .x_loop                ; if ($ebp$ > 0) goto .x_loop
 
   mov ebp, [tmp_height]
   add edi, [x_dif]          ; x_ptr += x_dif + (VPIXELS-1)*x_stride
   add esi, [y_dif]          ; y_ptr += y_dif + (VPIXELS-1)*y_stride
 %rep VPIXELS-1
   add edi, edx            
   add esi, eax            
 %endrep
 
   add ebx, [uv_dif]         ; u_ptr += uv_dif + ((VPIXELS/2)-1)*uv_stride
   add ecx, [uv_dif]         ; v_ptr += uv_dif + ((VPIXELS/2)-1)*uv_stride
 %rep (VPIXELS/2)-1
   add ebx, [uv_stride]
   add ecx, [uv_stride]
 %endrep
 
   sub ebp, VPIXELS          ; $ebp$ -= VPIXELS
   jg .y_loop                ; if ($ebp$ > 0) goto .y_loop
 
   ; cleanup stack & undef everything
 
   add esp, localsize
   pop ebp
   pop edi
   pop esi
   pop ebx
 
 %undef vflip
 %undef height
 %undef width
 %undef uv_stride
 %undef y_stride
 %undef v_ptr
 %undef u_ptr
 %undef y_ptr
 %undef x_stride
 %undef x_ptr
 %undef _ip
 %undef x_dif
 %undef y_dif
 %undef uv_dif
 %undef fixed_width
 %undef tmp_height
         ret
 %undef NAME
 %undef STACK
 %undef BYTES
 %undef PIXELS
 %undef VPIXELS
 %undef FUNC
 %undef ARG1
 %endmacro
 ;------------------------------------------------------------------------------