void draw_spans8(register int y __asm__("a4"), register int ey __asm__("d5")); __asm (" .equ len,a6 .equ dest,a5 .equ y,a4 .equ storev,a3 .equ storeu,a2 .equ buf,a0 | parameter ... .equ prew,fp7 .equ prev,fp6 .equ preu,fp5 .equ w0,fp4 .equ v0,fp3 .equ u0,fp2 .equ slen,d7 .equ sx,d6 .equ ey,d5 .equ count,d4 | parameter ... .equ dv,d3 | .equ du,d2 | .equ v,d1 | .equ u,d0 | .text .globl _draw_spans8 | void draw_spans8(register int y __asm__ (a4), register int ey __asm__ (d5)) _draw_spans8: moveml d2-d4/d6-d7/a2-a3/a5-a6,sp@- fmovem fp2-fp7,sp@- movel _localDim+26,dest | localDim.frameBuffer movel y,d0 fmovel d0,preu fmoves _tmap+32,prew | tmap[8] fsglmulx preu,prew | y * tmap[8] fadds _tmap+24,prew | tmap[6] + y * tmap[8] fmoves _tmap+20,prev fsglmulx preu,prev fadds _tmap+12,prev fsglmuls _tmap+8,preu fadds _tmap+0,preu subl y,ey | y < ey subql #1,ey | (y - 1) <= ey blt .bigEnd8 .bigLoop8: lea @(_scan, y:l:8), a0 | scan[y][] movel a0@+,sx | FIX_INT(scan[y][0]) |addl #65535,sx |swap sx movel a0@,d0 | FIX_INT(scan[y][1]) |addl #65535,d0 |swap d0 subw sx,d0 | scan[y][1] - sx ble .noSpan8 | len > 0 movew d0,len movel _multRows,buf movel buf@(y:l:4),buf | multRows[y] addl dest,buf | localDim.frameBuffer + multRows[y] addw sx,buf | localDim.frameBuffer + multRows[y] + sx fmovew sx,u0 fmoves #0r65536.0,w0 | 1.0 fmoves _tmap+28,fp0 | tmap[7] fsglmulx u0,fp0 | sx * tmap[7] faddx prew,fp0 | prew + sx * tmap[7] fsgldivx fp0,w0 | 1 / (prew + sx * tmap[7]) fmoves _tmap+16,v0 | tmap[4] fsglmulx u0,v0 | sx * tmap[4] fsglmuls _tmap+4,u0 faddx prev,v0 | prev + sx * tmap[4] faddx preu,u0 fsglmulx w0,v0 | (prev + sx * tmap[4]) * w0 fsglmulx w0,u0 fmovel v0,v fmovel u0,u movew len,slen lsrw #4,slen subqw #1,slen | dbge if not blts ??? blt .noSub8 .sub8: addw #16,sx fmovew sx,u0 fmoves #0r65536.0,w0 fmoves _tmap+28,fp0 fsglmulx u0,fp0 faddx prew,fp0 fsgldivx fp0,w0 fmoves _tmap+16,v0 fsglmulx u0,v0 fsglmuls _tmap+4,u0 faddx prev,v0 faddx preu,u0 fsglmulx w0,v0 fsglmulx w0,u0 fmovel v0,dv fmovel u0,du movel dv,storev movel du,storeu subl v,dv asrl #4,dv subl u,du asrl #4,du |moveq #16,count moveq #15,count bsr _draw_affine8 movel storev,v movel storeu,u dbra slen,.sub8 .noSub8: |moveq #0,count | sometimes something in upper word movew len,count andw #0x000F,count subqw #1,count blts .noSpan8 | nothing to do beqs .noRest8 | avoid division by 0 addw count,sx | count is near 15 and sx smaller than 32767 fmovew sx,fp1 fmoves #0r65536.0,w0 | v0 and u0 are scaled by 65536.0 on entry fmoves _tmap+28,fp0 fsglmulx fp1,fp0 faddx prew,fp0 fsgldivx fp0,w0 fmoves _tmap+16,fp0 | tmap[4] fsglmulx fp1,fp0 | sx * tmap[4] fsglmuls _tmap+4,fp1 faddx prev,fp0 | prev + sx * tmap[4] faddx preu,fp1 fsglmulx w0,fp0 | (prev + sx * tmap[4]) * w1 fsglmulx w0,fp1 fsubx v0,fp0 | v1 - v0 fsubx u0,fp1 fsgldivw count,fp0 fsgldivw count,fp1 fmovel fp0,dv fmovel fp1,du .noRest8: |addqw #1,count bsrs _draw_affine8 .noSpan8: fadds _tmap+32,prew fadds _tmap+20,prev fadds _tmap+8,preu addql #1,y | y++ dbra ey,.bigLoop8 .bigEnd8: fmovem sp@+,fp2-fp7 moveml sp@+,d2-d4/d6-d7/a2-a3/a5-a6 rts .even _draw_affine8: movew _textureType,a1 cmpw #1,a1 bnes .noWall8 moveml d5-d7,sp@- movel _texture,a1 movel #0x0000FF00,d7 | from tables.h ,MAX_MASKX asrl #8,v | inaccurate but fast ,16-MAX_LOGX asrl #8,dv | inaccurate but fast ,16-MAX_LOGX .wall8: movel u,d6 movel v,d5 clrw d6 swap d6 andl d7,d5 | nessecary? -> yep moveb _textureRow+3,d5 addl @(_multTMap, d5:l:4),d6 moveb a1@(d6:l:1),buf@+ addl du,u addl dv,v dbra count,.wall8 moveml sp@+,d5-d7 rts .noWall8: cmpw #5,a1 bnes .noSky8 moveml d5-d7,sp@- movel _texture,a1 movew #0x7F7F,d7 asrl #8,u | inaccurate but fast asrl #8,du | inaccurate but fast .sky8: movel u,d6 movel v,d5 swap d5 moveb d5,d6 movew d6,d5 addw _skyMovementX1+2,d6 addw _skyMovementX2+2,d5 addb _skyMovementY1+3,d6 addb _skyMovementY2+3,d5 andw d7,d6 andw d7,d5 moveb a1@(0x80, d6:w:1),d6 moveb a1@(d5:w:1),d5 beqs .skipSky8 moveb d5,d6 .skipSky8: moveb d6,buf@+ addl du,u addl dv,v dbra count,.sky8 moveml sp@+,d5-d7 rts .noSky8: moveml a2/d5-d7,sp@- movel _texture,a1 movel _preTransparency,a2 moveq #0,d7 | nessecary ? (asr is faster then) .liquid8: movew _textureShift2,d7 movel u,d6 | bfextu u{d7:8},d6 movel v,d5 | bfextu v{d7:8},d5 asrl d7,d6 | ... asrl d7,d5 | ... andw #0x00FF,d6 | ... andw #0x00FF,d5 | ... movel @(_swim_v, d6:w:4),d6 movel @(_swim_u, d5:w:4),d5 movew _textureMip,d7 asrl d7,d6 asrl d7,d5 addl v,d6 addl u,d5 movew _textureShift1,d7 swap d6 asrl d7,d5 andl _textureMask2,d6 andl _textureMask1,d5 orl d6,d5 |moveq #0,d6 | d6 is not greater than 64*64, so the upper bits are 0 movew buf@,d6 moveb a1@(d5:w:1),d6 moveb a2@(d6:l:1),buf@+ addl du,u addl dv,v dbra count,.liquid8 moveml sp@+,a2/d5-d7 rts ");