r_draw16.pas

delphi编的不错的贪吃蛇

 shl ecx,12	; left-justify tstep fractional part
 mov ds:dword ptr[advancetable],eax	; advance extra in t

 mov ds:dword ptr[tstep],ecx
 add edx,ecx	; advance tfrac fractional part by tstep frac

 sbb ecx,ecx	; turn tstep carry into -1 (0 if none)
 add ebx,ebp	; advance sfrac fractional part by sstep frac
 adc esi,ds:dword ptr[advancetable+4+ecx*4]	; point to next source texel

 add edx,ds:dword ptr[tstep]
 sbb ecx,ecx
 mov al,ds:byte ptr[esi]
 add ebx,ebp
 mov ds:byte ptr[1+edi],al
 adc esi,ds:dword ptr[advancetable+4+ecx*4]

 add edx,ds:dword ptr[tstep]
 sbb ecx,ecx
 add ebx,ebp
 mov al,ds:byte ptr[esi]
 adc esi,ds:dword ptr[advancetable+4+ecx*4]

 add edx,ds:dword ptr[tstep]
 sbb ecx,ecx
 mov ds:byte ptr[2+edi],al
 add ebx,ebp
 mov al,ds:byte ptr[esi]
 adc esi,ds:dword ptr[advancetable+4+ecx*4]

 add edx,ds:dword ptr[tstep]
 sbb ecx,ecx
 mov ds:byte ptr[3+edi],al
 add ebx,ebp
 mov al,ds:byte ptr[esi]
 adc esi,ds:dword ptr[advancetable+4+ecx*4]

 add edx,ds:dword ptr[tstep]
 sbb ecx,ecx
 mov ds:byte ptr[4+edi],al
 add ebx,ebp
 mov al,ds:byte ptr[esi]
 adc esi,ds:dword ptr[advancetable+4+ecx*4]

 add edx,ds:dword ptr[tstep]
 sbb ecx,ecx
 mov ds:byte ptr[5+edi],al
 add ebx,ebp
 mov al,ds:byte ptr[esi]
 adc esi,ds:dword ptr[advancetable+4+ecx*4]

 add edx,ds:dword ptr[tstep]
 sbb ecx,ecx
 mov ds:byte ptr[6+edi],al
 add ebx,ebp
 mov al,ds:byte ptr[esi]
 adc esi,ds:dword ptr[advancetable+4+ecx*4]

 add edx,ds:dword ptr[tstep]
 sbb ecx,ecx
 mov ds:byte ptr[7+edi],al
 add ebx,ebp
 mov al,ds:byte ptr[esi]
 adc esi,ds:dword ptr[advancetable+4+ecx*4]

;
; start FDIV for end of next segment in flight, so it can overlap
;
 mov ecx,ds:dword ptr[counttemp]
 cmp ecx,16	; more than one segment after this?
 ja LSetupNotLast2	; yes

 dec ecx
 jz LFDIVInFlight2	; if only one pixel, no need to start an FDIV
 mov ds:dword ptr[spancountminus1],ecx
 fild ds:dword ptr[spancountminus1]

 fld ds:dword ptr[_d_zistepu]	; C(d_zistepu) | spancountminus1
 fmul st(0),st(1)	; C(d_zistepu)*scm1 | scm1
 fld ds:dword ptr[_d_tdivzstepu]	; C(d_tdivzstepu) | C(d_zistepu)*scm1 | scm1
 fmul st(0),st(2)	; C(d_tdivzstepu)*scm1 | C(d_zistepu)*scm1 | scm1
 fxch st(1)	; C(d_zistepu)*scm1 | C(d_tdivzstepu)*scm1 | scm1
 faddp st(3),st(0)	; C(d_tdivzstepu)*scm1 | scm1
 fxch st(1)	; scm1 | C(d_tdivzstepu)*scm1
 fmul ds:dword ptr[_d_sdivzstepu]	; C(d_sdivzstepu)*scm1 | C(d_tdivzstepu)*scm1
 fxch st(1)	; C(d_tdivzstepu)*scm1 | C(d_sdivzstepu)*scm1
 faddp st(3),st(0)	; C(d_sdivzstepu)*scm1
 fld ds:dword ptr[fp_64k]	; 64k | C(d_sdivzstepu)*scm1
 fxch st(1)	; C(d_sdivzstepu)*scm1 | 64k
 faddp st(4),st(0)	; 64k

 fdiv st(0),st(1)	; this is what we've gone to all this trouble to
;  overlap
 jmp LFDIVInFlight2

 align 4
LSetupNotLast2:
 fadd ds:dword ptr[zi16stepu]
 fxch st(2)
 fadd ds:dword ptr[sdivz16stepu]
 fxch st(2)
 fld ds:dword ptr[tdivz16stepu]
 faddp st(2),st(0)
 fld ds:dword ptr[fp_64k]
 fdiv st(0),st(1)	; z = 1/1/z
; this is what we've gone to all this trouble to
;  overlap
LFDIVInFlight2:
 mov ds:dword ptr[counttemp],ecx

 add edx,ds:dword ptr[tstep]
 sbb ecx,ecx
 mov ds:byte ptr[8+edi],al
 add ebx,ebp
 mov al,ds:byte ptr[esi]
 adc esi,ds:dword ptr[advancetable+4+ecx*4]

 add edx,ds:dword ptr[tstep]
 sbb ecx,ecx
 mov ds:byte ptr[9+edi],al
 add ebx,ebp
 mov al,ds:byte ptr[esi]
 adc esi,ds:dword ptr[advancetable+4+ecx*4]

 add edx,ds:dword ptr[tstep]
 sbb ecx,ecx
 mov ds:byte ptr[10+edi],al
 add ebx,ebp
 mov al,ds:byte ptr[esi]
 adc esi,ds:dword ptr[advancetable+4+ecx*4]

 add edx,ds:dword ptr[tstep]
 sbb ecx,ecx
 mov ds:byte ptr[11+edi],al
 add ebx,ebp
 mov al,ds:byte ptr[esi]
 adc esi,ds:dword ptr[advancetable+4+ecx*4]

 add edx,ds:dword ptr[tstep]
 sbb ecx,ecx
 mov ds:byte ptr[12+edi],al
 add ebx,ebp
 mov al,ds:byte ptr[esi]
 adc esi,ds:dword ptr[advancetable+4+ecx*4]

 add edx,ds:dword ptr[tstep]
 sbb ecx,ecx
 mov ds:byte ptr[13+edi],al
 add ebx,ebp
 mov al,ds:byte ptr[esi]
 adc esi,ds:dword ptr[advancetable+4+ecx*4]

 add edx,ds:dword ptr[tstep]
 sbb ecx,ecx
 mov ds:byte ptr[14+edi],al
 add ebx,ebp
 mov al,ds:byte ptr[esi]
 adc esi,ds:dword ptr[advancetable+4+ecx*4]

 add edi,16
 mov ds:dword ptr[tfracf],edx
 mov edx,ds:dword ptr[snext]
 mov ds:dword ptr[sfracf],ebx
 mov ebx,ds:dword ptr[tnext]
 mov ds:dword ptr[s],edx
 mov ds:dword ptr[t],ebx

 mov ecx,ds:dword ptr[counttemp]	; retrieve count

;
; determine whether last span or not
;
 cmp ecx,16	; are there multiple segments remaining?
 mov ds:byte ptr[-1+edi],al
 ja LNotLastSegment	; yes

;
; last segment of scan
;
LLastSegment:

;
; advance s/z, t/z, and 1/z, and calculate s & t at end of span and steps to
; get there. The number of pixels left is variable, and we want to land on the
; last pixel, not step one past it, so we can't run into arithmetic problems
;
 test ecx,ecx
 jz LNoSteps	; just draw the last pixel and we're done

; pick up after the FDIV that was left in flight previously

 fld st(0)	; duplicate it
 fmul st(0),st(4)	; s = s/z * z
 fxch st(1)
 fmul st(0),st(3)	; t = t/z * z
 fxch st(1)
 fistp ds:dword ptr[snext]
 fistp ds:dword ptr[tnext]

 mov al,ds:byte ptr[esi]	; load first texel in segment
 mov ebx,ds:dword ptr[_tadjust]
 mov ds:byte ptr[edi],al	; store first pixel in segment
 mov eax,ds:dword ptr[_sadjust]

 add eax,ds:dword ptr[snext]
 add ebx,ds:dword ptr[tnext]

 mov ebp,ds:dword ptr[_bbextents]
 mov edx,ds:dword ptr[_bbextentt]

 cmp eax,4096
 jl LClampLow4
 cmp eax,ebp
 ja LClampHigh4
LClampReentry4:
 mov ds:dword ptr[snext],eax

 cmp ebx,4096
 jl LClampLow5
 cmp ebx,edx
 ja LClampHigh5
LClampReentry5:

 cmp ecx,1	; don't bother
 je LOnlyOneStep	; if two pixels in segment, there's only one step,
;  of the segment length
 sub eax,ds:dword ptr[s]
 sub ebx,ds:dword ptr[t]

 add eax,eax	; convert to 15.17 format so multiply by 1.31
 add ebx,ebx	;  reciprocal yields 16.48

 imul ds:dword ptr[reciprocal_table_16-8+ecx*4]	; sstep = (snext - s) /
;  (spancount-1)
 mov ebp,edx

 mov eax,ebx
 imul ds:dword ptr[reciprocal_table_16-8+ecx*4]	; tstep = (tnext - t) /
;  (spancount-1)
LSetEntryvec:
;
; set up advancetable
;
 mov ebx,ds:dword ptr[entryvec_table_16+ecx*4]
 mov eax,edx
 mov ds:dword ptr[jumptemp],ebx	; entry point into code for RET later
 mov ecx,ebp
 sar edx,16	; tstep >>= 16;
 mov ebx,ds:dword ptr[_cachewidth]
 sar ecx,16	; sstep >>= 16;
 imul edx,ebx

 add edx,ecx	; add in sstep
; (tstep >> 16) * cachewidth + (sstep >> 16);
 mov ecx,ds:dword ptr[tfracf]
 mov ds:dword ptr[advancetable+4],edx	; advance base in t
 add edx,ebx	; ((tstep >> 16) + 1) * cachewidth +
;  (sstep >> 16);
 shl ebp,16	; left-justify sstep fractional part
 mov ebx,ds:dword ptr[sfracf]
 shl eax,16	; left-justify tstep fractional part
 mov ds:dword ptr[advancetable],edx	; advance extra in t

 mov ds:dword ptr[tstep],eax
 mov edx,ecx
 add edx,eax
 sbb ecx,ecx
 add ebx,ebp
 adc esi,ds:dword ptr[advancetable+4+ecx*4]

 jmp dword ptr[jumptemp]	; jump to the number-of-pixels handler

;----------------------------------------

LNoSteps:
 mov al,ds:byte ptr[esi]	; load first texel in segment
 sub edi,15	; adjust for hardwired offset
 jmp LEndSpan

LOnlyOneStep:
 sub eax,ds:dword ptr[s]
 sub ebx,ds:dword ptr[t]
 mov ebp,eax
 mov edx,ebx
 jmp LSetEntryvec

;----------------------------------------

 public Entry2_16, Entry3_16, Entry4_16, Entry5_16
 public Entry6_16, Entry7_16, Entry8_16, Entry9_16
 public Entry10_16, Entry11_16, Entry12_16, Entry13_16
 public Entry14_16, Entry15_16, Entry16_16

Entry2_16:
 sub edi,14	; adjust for hardwired offsets
 mov al,ds:byte ptr[esi]
 jmp LEntry2_16

;----------------------------------------

Entry3_16:
 sub edi,13	; adjust for hardwired offsets
 add edx,eax
 mov al,ds:byte ptr[esi]
 sbb ecx,ecx
 add ebx,ebp
 adc esi,ds:dword ptr[advancetable+4+ecx*4]
 jmp LEntry3_16

;----------------------------------------

Entry4_16:
 sub edi,12	; adjust for hardwired offsets
 add edx,eax
 mov al,ds:byte ptr[esi]
 sbb ecx,ecx
 add ebx,ebp
 adc esi,ds:dword ptr[advancetable+4+ecx*4]
 add edx,ds:dword ptr[tstep]
 jmp LEntry4_16

;----------------------------------------

Entry5_16:
 sub edi,11	; adjust for hardwired offsets
 add edx,eax
 mov al,ds:byte ptr[esi]
 sbb ecx,ecx
 add ebx,ebp
 adc esi,ds:dword ptr[advancetable+4+ecx*4]
 add edx,ds:dword ptr[tstep]
 jmp LEntry5_16

;----------------------------------------

Entry6_16:
 sub edi,10	; adjust for hardwired offsets
 add edx,eax
 mov al,ds:byte ptr[esi]
 sbb ecx,ecx
 add ebx,ebp
 adc esi,ds:dword ptr[advancetable+4+ecx*4]
 add edx,ds:dword ptr[tstep]
 jmp LEntry6_16

;----------------------------------------

Entry7_16:
 sub edi,9	; adjust for hardwired offsets
 add edx,eax
 mov al,ds:byte ptr[esi]
 sbb ecx,ecx
 add ebx,ebp
 adc esi,ds:dword ptr[advancetable+4+ecx*4]
 add edx,ds:dword ptr[tstep]
 jmp LEntry7_16

;----------------------------------------

Entry8_16:
 sub edi,8	; adjust for hardwired offsets
 add edx,eax
 mov al,ds:byte ptr[esi]
 sbb ecx,ecx
 add ebx,ebp
 adc esi,ds:dword ptr[advancetable+4+ecx*4]
 add edx,ds:dword ptr[tstep]
 jmp LEntry8_16

;----------------------------------------

Entry9_16:
 sub edi,7	; adjust for hardwired offsets
 add edx,eax
 mov al,ds:byte ptr[esi]
 sbb ecx,ecx
 add ebx,ebp
 adc esi,ds:dword ptr[advancetable+4+ecx*4]
 add edx,ds:dword ptr[tstep]
 jmp LEntry9_16

;----------------------------------------

Entry10_16:
 sub edi,6	; adjust for hardwired offsets
 add edx,eax
 mov al,ds:byte ptr[esi]
 sbb ecx,ecx
 add ebx,ebp
 adc esi,ds:dword ptr[advancetable+4+ecx*4]
 add edx,ds:dword ptr[tstep]
 jmp LEntry10_16

;----------------------------------------

Entry11_16:
 sub edi,5	; adjust for hardwired offsets
 add edx,eax
 mov al,ds:byte ptr[esi]
 sbb ecx,ecx
 add ebx,ebp
 adc esi,ds:dword ptr[advancetable+4+ecx*4]
 add edx,ds:dword ptr[tstep]
 jmp LEntry11_16

;----------------------------------------

Entry12_16:
 sub edi,4	; adjust for hardwired offsets
 add edx,eax
 mov al,ds:byte ptr[esi]
 sbb ecx,ecx
 add ebx,ebp
 adc esi,ds:dword ptr[advancetable+4+ecx*4]
 add edx,ds:dword ptr[tstep]
 jmp LEntry12_16

;----------------------------------------

Entry13_16:
 sub edi,3	; adjust for hardwired offsets
 add edx,eax
 mov al,ds:byte ptr[esi]
 sbb ecx,ecx
 add ebx,ebp
 adc esi,ds:dword ptr[advancetable+4+ecx*4]
 add edx,ds:dword ptr[tstep]
 jmp LEntry13_16