kolibrios/programs/demos/3DS/TWO_TEX.INC

;SIZE_X equ 350
;SIZE_Y equ 350
;ROUND equ 8
;TEX_X equ 512
;TEX_Y equ 512
;TEXTURE_SIZE EQU (512*512)-1
;TEX_SHIFT EQU 9

;CATMULL_SHIFT equ 8
;TEXTURE_SIZE EQU (TEX_X * TEX_Y)-1
;Ext = SSE
;SSE = 3
;MMX = 1
;NON = 0
;use32
;------- Big thanks to Majuma (www.majuma.xt.pl) for absolutely great---
;------- DOS 13h mode demos --------------------------------------------
;------- Procedure draws triangle with two overlapped textures, I use --
;--------Catmull Z-buffer algorithm- (Z coordinate interpolation)-------
;--------I calc texture pixel by this way: col1*col2/256 ---------------
two_tex_triangle_z:
;------------------in - eax - x1 shl 16 + y1 -----------
;---------------------- ebx - x2 shl 16 + y2 -----------
;---------------------- ecx - x3 shl 16 + y3 -----------
;---------------------- edx - pointer to b. texture-----
;---------------------- esi - pointer to e. texture-----
;---------------------- edi - pointer to screen buffer--
;---------------------- stack : b. tex coordinates------
;----------------------         e. tex coordinates------
;----------------------         Z position coordinates--
;----------------------         pointer io Z buffer-----
;-- Z-buffer - filled with coordinates as dword --------
;-- (Z coor. as word) shl CATMULL_SHIFT ----------------
.b_x1   equ ebp+4   ; procedure don't save registers !!!
.b_y1   equ ebp+6   ; each coordinate as word
.b_x2   equ ebp+8
.b_y2   equ ebp+10       ; b - first texture
.b_x3   equ ebp+12
.b_y3   equ ebp+14       ; e - second texture
.e_x1   equ ebp+16
.e_y1   equ ebp+18
.e_x2   equ ebp+20
.e_y2   equ ebp+22
.e_x3   equ ebp+24
.e_y3   equ ebp+26
.z1     equ word[ebp+28]
.z2     equ word[ebp+30]
.z3     equ word[ebp+32]
.z_buff equ dword[ebp+34]       ; pointer to Z-buffer


.t_bmap equ dword[ebp-4]        ; pointer to b. texture
.t_emap equ dword[ebp-8]        ; pointer to e. texture
.x1     equ word[ebp-10]
.y1     equ word[ebp-12]
.x2     equ word[ebp-14]
.y2     equ word[ebp-16]
.x3     equ word[ebp-18]
.y3     equ word[ebp-20]

.dx12  equ dword[ebp-24]
.dbx12 equ dword[ebp-28]
.dby12 equ dword[ebp-32]
.dby12q equ     [ebp-32]
.dex12 equ dword[ebp-36]
.dey12 equ dword[ebp-40]
.dey12q equ     [ebp-40]
.dz12  equ dword[ebp-44]

.dx13  equ dword[ebp-48]
.dbx13 equ dword[ebp-52]
.dby13 equ dword[ebp-56]
.dby13q equ     [ebp-56]
.dex13 equ dword[ebp-60]
.dey13 equ dword[ebp-64]
.dey13q equ     [ebp-64]
.dz13  equ dword[ebp-68]

.dx23  equ dword[ebp-72]
.dbx23 equ dword[ebp-76]
.dby23 equ dword[ebp-80]
.dby23q equ     [ebp-80]
.dex23 equ dword[ebp-84]
.dey23 equ dword[ebp-88]
.dey23q equ     [ebp-88]
.dz23  equ dword[ebp-92]

.cx1   equ dword[ebp-96]   ; current variables
.cx2   equ dword[ebp-100]
.cbx1  equ dword[ebp-104]
.cby1  equ [ebp-108]
.cex1  equ dword[ebp-112]
.cey1  equ [ebp-116]
.cbx2  equ dword[ebp-120]
.cby2  equ [ebp-124]
.cex2  equ dword[ebp-128]
.cey2  equ [ebp-132]

.cz1   equ dword[ebp-136]
.cz2   equ dword[ebp-140]

    if Ext >= MMX
       emms
    else
       cld
    end if
       mov     ebp,esp
       push    edx esi       ; store bump map
;       push    esi        ; store e. map
     ; sub     esp,120
 .sort3:                  ; sort triangle coordinates...
       cmp     ax,bx
       jle     .sort1
       xchg    eax,ebx
       mov     edx,dword[.b_x1]
       xchg    edx,dword[.b_x2]
       mov     dword[.b_x1],edx
       mov     edx,dword[.e_x1]
       xchg    edx,dword[.e_x2]
       mov     dword[.e_x1],edx
       mov     dx,.z1
       xchg    dx,.z2
       mov     .z1,dx
 .sort1:
       cmp      bx,cx
       jle      .sort2
       xchg     ebx,ecx
       mov      edx,dword[.b_x2]
       xchg     edx,dword[.b_x3]
       mov      dword[.b_x2],edx
       mov      edx,dword[.e_x2]
       xchg     edx,dword[.e_x3]
       mov      dword[.e_x2],edx
       mov     dx,.z2
       xchg    dx,.z3
       mov     .z2,dx
       jmp      .sort3
 .sort2:
       push     eax ebx ecx    ; store triangle coords in variables
;       push     ebx
;       push     ecx

         mov      edx,80008000h  ; eax,ebx,ecx are ANDd together into edx which means that
         and      edx,ebx        ; if *all* of them are negative a sign flag is raised
         and      edx,ecx
         and      edx,eax
         test     edx,80008000h  ; Check both X&Y at once
         jne      .loop23_done
    ;   mov     edx,eax         ; eax,ebx,ecx are ORd together into edx which means that
    ;   or      edx,ebx         ; if any *one* of them is negative a sign flag is raised
    ;   or      edx,ecx
    ;   test    edx,80000000h   ; Check only X
    ;   jne     .loop23_done

    ;   cmp     .x1,SIZE_X    ; {
    ;   jg      .loop23_done
    ;   cmp     .x2,SIZE_X     ; This can be optimized with effort
    ;   jg      .loop23_done
    ;   cmp     .x3,SIZE_X
    ;   jg      .loop23_done    ; {


       mov      bx,.y2       ; calc delta 12
       sub      bx,.y1
       jnz      .bt_dx12_make
       mov      ecx,6
       xor      edx,edx
     @@:
       push     edx   ;dword 0
       loop     @b
       jmp      .bt_dx12_done
 .bt_dx12_make:
       mov      ax,.x2
       sub      ax,.x1
       cwde
       movsx    ebx,bx
       shl      eax,ROUND
       cdq
       idiv     ebx
;      mov      .dx12,eax
       push      eax

if Ext=SSE

       sub       esp,16
       cvtsi2ss  xmm3,ebx            ;rcps
    ;   mov       eax,255
       cvtsi2ss  xmm4,[i255d] ;eax
       divss     xmm3,xmm4
       rcpss     xmm3,xmm3
    ;   mulss     xmm3,xmm4
       shufps    xmm3,xmm3,0

       movd      mm0,[.b_x1]
       movd      mm1,[.b_x2]
       movd      mm2,[.e_x1]
       movd      mm3,[.e_x2]
     ;  psubsw    mm3,mm2
     ;  psubsw    mm1,mm0
       pxor      mm4,mm4
       punpcklwd mm0,mm4
       punpcklwd mm1,mm4
       punpcklwd mm2,mm4
       punpcklwd mm3,mm4
   ;    pslld     mm0,ROUND
   ;    pslld     mm1,ROUND
   ;    pslld     mm2,ROUND
   ;    pslld     mm3,ROUND
       cvtpi2ps  xmm0,mm0
       movlhps   xmm0,xmm0
       cvtpi2ps  xmm0,mm2
       cvtpi2ps  xmm1,mm1
       movlhps   xmm1,xmm1
       cvtpi2ps  xmm1,mm3
       subps     xmm1,xmm0

     ;  pxor      mm4,mm4
     ;  movq      mm5,mm1
     ;  movq      mm6,mm1
     ;  pcmpeqb   mm5,mm4
;       psubd     mm1,mm0
;       psubd     mm3,mm2

    ;   movq      mm0,[.b_x1]      ; bx1  by1   bx2    by2
    ;   movq      mm1,[.e_x1]      ; ex1  ey1   ex2    ey2
    ;   pxor
    ;   punpcklhd mm0,mm1   ; lwd  ;
    ;   psubw     mm1,mm0   ; mm1, mm0
    ;   pxor      mm2,mm2
  ;     pmovmaskb eax,mm1
  ;     and       eax,10101010b
    ;   pcmpgtw   mm2,mm1
    ;   punpcklwd mm1,mm2
 ;      psllw     mm0,ROUND
 ;      psllw     mm1,ROUND
 ;      movq      mm2,mm0
 ;      psrlq     mm0,32

;       cvtpi2ps  xmm0,mm1
;       movlhps   xmm0,xmm0
;       cvtpi2ps  xmm0,mm3
  ;     divps     xmm1,xmm3
       mulps     xmm1,xmm3
       shufps    xmm1,xmm1,10110001b
       cvtps2pi  mm0,xmm1          ; mm0 -> 2 delta dwords
       movhlps   xmm1,xmm1
       cvtps2pi  mm1,xmm1
       movq      .dey12q,mm0
       movq      .dby12q,mm1

;       movd      .dex12,mm0
;       psrlq     mm0,32
;       movd      .dey12,mm0
;       movhlps   xmm1,xmm1
;       cvtps2pi  mm0,xmm1
;       movd      .dbx12,mm0
;       psrlq     mm0,32
;       movd      .dby12,mm0

else
       mov      ax,word[.b_x2]
       sub      ax,word[.b_x1]
       cwde
       shl      eax,ROUND
       cdq
       idiv     ebx
 ;     mov      .dbx12,eax
       push      eax

       mov      ax,word[.b_y2]
       sub      ax,word[.b_y1]
       cwde
       shl      eax,ROUND
       cdq
       idiv     ebx
 ;     mov      .dby12,eax
       push      eax

 ;     mov       eax,.dbx12
 ;     mov       ebx,.dby12
 ;     int3

       mov      ax,word[.e_x2]
       sub      ax,word[.e_x1]
       cwde
       shl      eax,ROUND
       cdq
       idiv     ebx
 ;      mov      .dex12,eax
       push      eax

       mov      ax,word[.e_y2]
       sub      ax,word[.e_y1]
       cwde
       shl      eax,ROUND
       cdq
       idiv     ebx
 ;      mov      .dey12,eax
       push      eax

end if
        mov     ax,.z2
        sub     ax,.z1
        cwde
        shl     eax,CATMULL_SHIFT
        cdq
        idiv    ebx
        push    eax
   .bt_dx12_done:

       mov      bx,.y3       ; calc delta13
       sub      bx,.y1
       jnz      .bt_dx13_make
       mov      ecx,6
       xor      edx,edx
     @@:
       push     edx   ;dword 0
       loop     @b
       jmp      .bt_dx13_done
 .bt_dx13_make:
       mov      ax,.x3
       sub      ax,.x1
       cwde
       movsx    ebx,bx
       shl      eax,ROUND
       cdq
       idiv     ebx
 ;      mov      .dx13,eax
       push      eax

if Ext=SSE

       cvtsi2ss  xmm3,ebx
  ;     mov       eax,255
       cvtsi2ss  xmm4,[i255d]
       divss     xmm3,xmm4
       rcpss     xmm3,xmm3
;       mulss     xmm3,xmm4
       shufps    xmm3,xmm3,0
       sub       esp,16

       movd      mm0,[.b_x1]
       movd      mm1,[.b_x3]
       movd      mm2,[.e_x1]
       movd      mm3,[.e_x3]

       pxor      mm4,mm4
       punpcklwd mm0,mm4
       punpcklwd mm1,mm4
       punpcklwd mm2,mm4
       punpcklwd mm3,mm4

       cvtpi2ps  xmm0,mm0
       movlhps   xmm0,xmm0
       cvtpi2ps  xmm0,mm2
       cvtpi2ps  xmm1,mm1
       movlhps   xmm1,xmm1
       cvtpi2ps  xmm1,mm3
       subps     xmm1,xmm0

   ;    divps     xmm1,xmm3
       mulps     xmm1,xmm3
       shufps    xmm1,xmm1,10110001b
       cvtps2pi  mm0,xmm1          ; mm0 -> 2 delta dwords
       movhlps   xmm1,xmm1
       cvtps2pi  mm1,xmm1
       movq      .dey13q,mm0
       movq      .dby13q,mm1

else

       mov      ax,word[.b_x3]
       sub      ax,word[.b_x1]
       cwde
       shl      eax,ROUND
       cdq
       idiv     ebx
 ;      mov      .dbx13,eax
       push      eax

       mov      ax,word[.b_y3]
       sub      ax,word[.b_y1]
       cwde
       shl      eax,ROUND
       cdq
       idiv     ebx
 ;      mov      .dby13,eax
       push      eax

       mov      ax,word[.e_x3]
       sub      ax,word[.e_x1]
       cwde
       shl      eax,ROUND
       cdq
       idiv     ebx
 ;      mov      .dex13,eax
       push      eax

       mov      ax,word[.e_y3]
       sub      ax,word[.e_y1]
       cwde
       shl      eax,ROUND
       cdq
       idiv     ebx
 ;      mov      .dey13,eax
       push      eax

end if

       mov     ax,.z3
       sub     ax,.z1
       cwde
       shl     eax,CATMULL_SHIFT
       cdq
       idiv    ebx
  ;    mov    .dz13,eax
       push    eax
   .bt_dx13_done:

       mov      bx,.y3       ; calc delta23
       sub      bx,.y2
       jnz      .bt_dx23_make
       mov      ecx,6
       xor      edx,edx
     @@:
       push     edx   ;dword 0
       loop     @b
       jmp      .bt_dx23_done
 .bt_dx23_make:
       mov      ax,.x3
       sub      ax,.x2
       cwde
       movsx    ebx,bx
       shl      eax,ROUND
       cdq
       idiv     ebx
 ;      mov      .dx23,eax
       push      eax

if Ext=SSE

       cvtsi2ss  xmm3,ebx
     ;  mov       eax,255
       cvtsi2ss  xmm4,[i255d] ;eax
       divss     xmm3,xmm4
       shufps    xmm3,xmm3,0
       sub       esp,16

       movd      mm0,[.b_x2]
       movd      mm1,[.b_x3]
       movd      mm2,[.e_x2]
       movd      mm3,[.e_x3]

       pxor      mm4,mm4
       punpcklwd mm0,mm4
       punpcklwd mm1,mm4
       punpcklwd mm2,mm4
       punpcklwd mm3,mm4

       cvtpi2ps  xmm0,mm0
       movlhps   xmm0,xmm0
       cvtpi2ps  xmm0,mm2
       cvtpi2ps  xmm1,mm1
       movlhps   xmm1,xmm1
       cvtpi2ps  xmm1,mm3
       subps     xmm1,xmm0

       divps     xmm1,xmm3
       shufps    xmm1,xmm1,10110001b
       cvtps2pi  mm0,xmm1          ; mm0 -> 2 delta dwords
       movhlps   xmm1,xmm1
       cvtps2pi  mm1,xmm1
       movq      .dey23q,mm0
       movq      .dby23q,mm1

else

       mov      ax,word[.b_x3]
       sub      ax,word[.b_x2]
       cwde
       shl      eax,ROUND
       cdq
       idiv     ebx
 ;      mov      .dbx23,eax
       push      eax

       mov      ax,word[.b_y3]
       sub      ax,word[.b_y2]
       cwde
       shl      eax,ROUND
       cdq
       idiv     ebx
 ;      mov      .dby23,eax
       push      eax

       mov      ax,word[.e_x3]
       sub      ax,word[.e_x2]
       cwde
       shl      eax,ROUND
       cdq
       idiv     ebx
 ;      mov      .dex23,eax
       push      eax

       mov      ax,word[.e_y3]
       sub      ax,word[.e_y2]
       cwde
       shl      eax,ROUND
       cdq
       idiv     ebx
 ;      mov      .dey23,eax
       push      eax
end if
       mov     ax,.z3
       sub     ax,.z2
       cwde
       shl     eax,CATMULL_SHIFT
       cdq
       idiv    ebx
      ; mov     .dz23,eax
       push    eax
      ;  sub     esp,40
   .bt_dx23_done:
       movsx    eax,.x1
       shl      eax,ROUND
     ;  mov      .cx1,eax
     ;  mov      .cx2,eax
       push     eax eax
     ; push     eax

       movsx    eax,word[.b_x1]
       shl      eax,ROUND
       mov      .cbx1,eax
       mov      .cbx2,eax
     ;  push     eax eax
     ; push     eax

       movsx    eax,word[.b_y1]
       shl      eax,ROUND
       mov      .cby1,eax
       mov      .cby2,eax
     ;  push     eax eax
     ; push     eax

       movsx    eax,word[.e_x1]
       shl      eax,ROUND
       mov      .cex1,eax
       mov      .cex2,eax
      ; push     eax eax
      ;push     eax

       movsx    eax,word[.e_y1]
       shl      eax,ROUND
       mov      .cey1,eax
       mov      .cey2,eax
       sub      esp,32
      ; push     eax eax
      ;push     eax

       movsx    eax,.z1
       shl      eax,CATMULL_SHIFT
      ; mov      .cz1,eax
      ; mov      .cz2,eax
      push     eax eax
      ;push     eax

       movsx    ecx,.y1
       cmp      cx,.y2
       jge      .loop12_done
  .loop12:
       call     .call_line

       mov      eax,.dx13
       add      .cx1,eax
       mov      ebx,.dx12
       add      .cx2,ebx
if Ext>= SSE2
       movups   xmm0,.cey1
       movups   xmm1,.cey2
       movups   xmm2,.dey12q
       movups   xmm3,.dey13q
       paddd    xmm0,xmm3
       paddd    xmm1,xmm2
       movups   .cey1,xmm0
       movups   .cey2,xmm1
else if (Ext = MMX) | (Ext=SSE)
       movq     mm0,.cby2       ; with this optimization object
       movq     mm1,.cby1       ; looks bit annoying
       movq     mm2,.cey2
       movq     mm3,.cey1
       paddd    mm0,.dby12q
       paddd    mm1,.dby13q
       paddd    mm2,.dey12q
       paddd    mm3,.dey13q
       movq     .cby2,mm0
       movq     .cby1,mm1
       movq     .cey1,mm3
       movq     .cey2,mm2
else
       mov      edx,.dbx13
       add      .cbx1,edx
       mov      eax,.dbx12
       add      .cbx2,eax
       mov      ebx,.dby13
       add      .cby1,ebx
       mov      edx,.dby12
       add      .cby2,edx

       mov      eax,.dex13
       add      .cex1,eax
       mov      ebx,.dex12
       add      .cex2,ebx
       mov      edx,.dey13
       add      .cey1,edx
       mov      eax,.dey12
       add      .cey2,eax

end if
       mov      ebx,.dz13
       add      .cz1,ebx
       mov      edx,.dz12
       add      .cz2,edx

       inc      ecx
       cmp      cx,.y2
       jl       .loop12
    .loop12_done:

       movsx    ecx,.y2
       cmp      cx,.y3
       jge      .loop23_done

       movsx    eax,.z2
       shl      eax,CATMULL_SHIFT
       mov      .cz2,eax

       movsx    eax,.x2
       shl      eax,ROUND
       mov      .cx2,eax

       movzx    eax,word[.b_x2]
       shl      eax,ROUND
       mov      .cbx2,eax

       movzx    eax,word[.b_y2]
       shl      eax,ROUND
       mov      .cby2,eax

       movzx    eax,word[.e_x2]
       shl      eax,ROUND
       mov      .cex2,eax

       movzx    eax,word[.e_y2]
       shl      eax,ROUND
       mov      .cey2,eax

     .loop23:
       call     .call_line
;if Ext = NON
       mov      eax,.dx13
       add      .cx1,eax
       mov      ebx,.dx23
       add      .cx2,ebx
if Ext>= SSE2
       movups   xmm0,.cey1
       movups   xmm1,.cey2
       movups   xmm2,.dey23q
       movups   xmm3,.dey13q
       paddd    xmm0,xmm3
       paddd    xmm1,xmm2
       movups   .cey1,xmm0
       movups   .cey2,xmm1
else if (Ext = MMX) | ( Ext = SSE)
       movq     mm0,.cby2        ;    with this mmx optimization object looks bit
       movq     mm1,.cby1        ;    annoying
       movq     mm2,.cey2
       movq     mm3,.cey1
       paddd    mm0,.dby23q
       paddd    mm1,.dby13q
       paddd    mm2,.dey23q
       paddd    mm3,.dey13q
       movq     .cby2,mm0
       movq     .cby1,mm1
       movq     .cey2,mm2
       movq     .cey1,mm3

else
       mov      edx,.dbx13
       add      .cbx1,edx
       mov      eax,.dbx23
       add      .cbx2,eax
       mov      ebx,.dby13
       add      .cby1,ebx
       mov      edx,.dby23
       add      .cby2,edx

       mov      eax,.dex13
       add      .cex1,eax
       mov      ebx,.dex23
       add      .cex2,ebx
       mov      edx,.dey13
       add      .cey1,edx
       mov      eax,.dey23
       add      .cey2,eax
end if

       mov      ebx,.dz13
       add      .cz1,ebx
       mov      edx,.dz23
       add      .cz2,edx
;else
;       movq     mm0,.db13q
;       movq     mm1,.cbx1q

       inc      ecx
       cmp      cx,.y3
       jl       .loop23
    .loop23_done:

       mov      esp,ebp
ret   34

.call_line:

       pushad

       push     .cz1
       push     .cz2
       push     .z_buff
       push     .t_bmap
       push     .t_emap
       push     dword .cey2
       push     .cex2
       push     dword .cey1
       push     .cex1
       push     dword .cby2
       push     .cbx2
       push     dword .cby1
       push     .cbx1
       push     ecx

       mov      eax,.cx1
       sar      eax,ROUND
       mov      ebx,.cx2
       sar      ebx,ROUND

       call     two_tex_line_z

       popad
ret
two_tex_line_z:
;--------------in: eax - x1
;--------------    ebx - x2
;--------------    edi - pointer to screen buffer
;stack - another parameters :
.y      equ dword [ebp+4]
.bx1    equ  [ebp+8]   ;   ---
.by1    equ  [ebp+12]  ;       |
.bx2    equ  [ebp+16]  ;       |
.by2    equ  [ebp+20]  ;       |>   b. texture and e. texture coords
.ex1    equ  [ebp+24]  ;       |>   shifted shl ROUND
.ey1    equ  [ebp+28]  ;       |
.ex2    equ  [ebp+32]  ;       |
.ey2    equ  [ebp+36]  ;   ---
.emap   equ  [ebp+40]  ; b texture offset
.bmap   equ  [ebp+44]  ; e texture offset
.z_buff equ dword [ebp+48]
.z2     equ dword [ebp+52]  ;   -- |>   z coords shifted
.z1     equ dword [ebp+56]  ;   --       shl  CATMULL_SHIFT

.x1     equ dword [ebp-4]
.x2     equ dword [ebp-8]
.dbx    equ [ebp-12]
.dex    equ [ebp-16]
.dby    equ [ebp-20]
.dey    equ [ebp-24]
.dz     equ dword [ebp-28]
.cbx    equ [ebp-32]
.cex    equ [ebp-36]
.cby    equ [ebp-40]
.cey    equ [ebp-44]
.cz     equ dword [ebp-48]
.czbuff equ dword [ebp-52]

        mov     ebp,esp

        mov     ecx,.y
        or      ecx,ecx
        jl      .bl_end
        cmp     ecx,SIZE_Y
        jge     .bl_end

        cmp     eax,ebx
        jl      @f
        je      .bl_end

        xchg    eax,ebx
if Ext=NON
        mov     edx,.bx1
        xchg    edx,.bx2
        mov     .bx1,edx
        mov     edx,.by1
        xchg    edx,.by2
        mov     .by1,edx

        mov     edx,.ex1
        xchg    edx,.ex2
        mov     .ex1,edx
        mov     edx,.ey1
        xchg    edx,.ey2
        mov     .ey1,edx
else
        movq    mm0,.bx1
        movq    mm1,.ex1
        movq    mm2,.bx2
        movq    mm3,.ex2
        movq    .bx2,mm0
        movq    .ex2,mm1
        movq    .bx1,mm2
        movq    .ex1,mm3
end if
        mov     edx,.z1
        xchg    edx,.z2
        mov     .z1,edx
    @@:
        push    eax ebx
;        push    ebx           ;store x1, x2

        cmp     .x1,SIZE_X
        jge     .bl_end
        cmp     .x2,0
        jle     .bl_end

        mov     ebx,.x2
        sub     ebx,.x1

if Ext>=SSE

       sub       esp,16
       cvtsi2ss  xmm3,ebx            ;rcps
       shufps    xmm3,xmm3,0

  ;     movq      mm0,.bx1q
  ;     movq      mm1,.bx2q
  ;     movq      mm2,.ex1q
  ;     movq      mm3,.ex2q
  ;     psubd     mm1,mm0
  ;     psubd     mm3,mm2
  ;     cvtpi2ps  xmm1,mm1
  ;     movlhps   xmm1,xmm1
  ;     cvtpi2ps  xmm1,mm3

       cvtpi2ps  xmm0,.bx1 ;mm0     ; bx1; by1
       movlhps   xmm0,xmm0
       cvtpi2ps  xmm0,.ex1 ;mm2     ; ex1; ey1
       cvtpi2ps  xmm1,.bx2 ;mm1     ; bx2; by2
       movlhps   xmm1,xmm1
       cvtpi2ps  xmm1,.ex2 ;mm3     ; ex2; ey2
       subps     xmm1,xmm0
                                    ; hi             lo
       divps     xmm1,xmm3 ; xmm1 -> dby; dbx; dey; dex

       shufps    xmm1,xmm1,11011000b
       cvtps2pi  mm0,xmm1          ; mm0 -> 2 delta dwords
       movhlps   xmm1,xmm1
       cvtps2pi  mm1,xmm1
       movq      .dex,mm0 ; hi - lo  ->  dbx, dex
       movq      .dey,mm1 ; hi - lo  ->  dby, dey

else

        mov     eax,.bx2       ; calc .dbx
        sub     eax,.bx1
        cdq
        idiv    ebx
        push    eax

        mov     eax,.ex2       ; calc .dby
        sub     eax,.ex1
        cdq
        idiv    ebx
        push    eax

        mov     eax,.by2       ; calc .dex
        sub     eax,.by1
        cdq
        idiv    ebx
        push    eax

        mov     eax,.ey2       ; calc .dey
        sub     eax,.ey1
        cdq
        idiv    ebx
        push    eax

end if

        mov     eax,.z2        ; calc .dz
        sub     eax,.z1
        cdq
        idiv    ebx
        push    eax

        cmp     .x1,0         ; set correctly begin variable
        jge     @f            ; CLIPPING ON FUNCTION
                              ; cutting triangle exceedes screen
        mov     ebx,.x1
        neg     ebx
        imul    ebx           ; eax = .dz * abs(.x1)
        add     .z1,eax
        mov     .x1,0

        mov     eax,.dbx
        imul    ebx
        add    .bx1,eax

        mov     eax,.dby
        imul    ebx
        add     .by1,eax

        mov     eax,.dex
        imul    ebx
        add     .ex1,eax

        mov     eax,.dey
        imul    ebx
        add     .ey1,eax
      @@:
        cmp     .x2,SIZE_X
        jl      @f
        mov     .x2,SIZE_X
      @@:
        mov     eax,SIZE_X       ;calc memory begin in buffers
        mov     ebx,.y
        mul     ebx
        mov     ebx,.x1
        add     eax,ebx
        mov     ebx,eax
        lea     eax,[eax*3]
        add     edi,eax           ; edi - screen
        mov     esi,.z_buff       ; z-buffer filled with dd variables
        shl     ebx,2
        add     esi,ebx           ; esi - Z buffer

        mov     ecx,.x2
        sub     ecx,.x1
        ; init current variables
        push    dword .bx1 ;.by1 .ex1 .ey1 .z1 esi
        push    dword .ex1
        push    dword .by1
        push    dword .ey1

        push    .z1              ; current z shl CATMULL_SHIFT
        push    esi

if Ext >= MMX
     pxor   mm0,mm0
     movq   mm3,.cex   ; hi - lo -> cbx; cex
     movq   mm4,.cey   ; hi - lo -> cby; cey
;     movq   mm5,mm3
;     movq   mm6,mm4
;     psrad  mm5,ROUND
;     psrad  mm6,ROUND
;     movq   .ceyq,mm5
;     movq   .cbyq,mm6
     mov    edx,.czbuff
else
     cld
end if
     .draw:
    ; if TEX = SHIFTING   ;bump drawing only in shifting mode
if Ext=NON
        mov     esi,.czbuff      ; .czbuff current address in buffer
        mov     ebx,.cz          ; .cz - cur z position
        cmp     ebx,dword[esi]
else
        mov     ebx,.cz
        cmp     ebx,dword[edx]
end if
        jge     .skip

if Ext=NON
        mov     eax,.cby
        mov     esi,.cbx
        sar     eax,ROUND
        sar     esi,ROUND
        shl     eax,TEX_SHIFT   ;-
        add     esi,eax
        lea     esi,[esi*3]         ;-  ; esi - current b. texture addres
        add     esi,.bmap

        mov     ebx,.cex       ;.cex - current env map X
        mov     eax,.cey       ;.cey - current  env map y
        sar     ebx,ROUND
        sar     eax,ROUND

        shl     eax,TEX_SHIFT
        add     ebx,eax
        lea     ebx,[ebx*3]
        add     ebx,.emap


else
        movq    mm5,mm4 ;.cey
        psrad   mm5,ROUND
        pslld   mm5,TEX_SHIFT
        movq    mm6,mm3 ;.cex
        psrad   mm6,ROUND
        paddd   mm5,mm6
        movq    mm6,mm5
        paddd   mm5,mm5
        paddd   mm5,mm6
        paddd   mm5,.emap
        movd    esi,mm5
        psrlq   mm5,32
        movd    ebx,mm5
end if
if Ext>=MMX
        movd      mm1,[esi]
        movd      mm2,[ebx]
        punpcklbw mm1,mm0
        punpcklbw mm2,mm0
        pmullw    mm1,mm2
        psrlw     mm1,8
        packuswb  mm1,mm0
        movd      [edi],mm1
        mov       ebx,.cz
        mov       dword[edx],ebx
else
        cld                     ; esi - tex e.
        lodsb                   ; ebx - tex b.
        mov     dl,[ebx]
        mul     dl
        shr     ax,8
        stosb
        inc     ebx
        lodsb
        mov     dl,[ebx]
        mul     dl
        shr     ax,8
        stosb
        inc     ebx
        lodsb
        mov     dl,[ebx]
        mul     dl
        shr     ax,8
        stosb
        mov     ebx,.cz
        mov     esi,.czbuff
        mov     dword[esi],ebx
        jmp     .no_skip
end if
     .skip:
        add     edi,3

   if Ext = NON
     .no_skip:
        add     .czbuff,4
        mov     eax,.dbx
        add     .cbx,eax
        mov     eax,.dby
        add     .cby,eax
        mov     eax,.dex
        add     .cex,eax
        mov     eax,.dey
        add     .cey,eax
    else
        add     edx,4
        paddd   mm3,.dex
        paddd   mm4,.dey
  ;      movq    mm5,mm3
  ;      movq    mm6,mm4
  ;      psrad   mm5,ROUND
  ;      psrad   mm6,ROUND
     ;   movq    .cex,mm3
     ;   movq    .cey,mm4
    end if
        mov     eax,.dz
        add     .cz,eax
    if Ext = NON
        dec     ecx
        jnz     .draw
    else
        loop    .draw
    end if

  .bl_end:
        mov     esp,ebp
ret 56