d_polysa.s

Quake 2 Source code for students by Theerthan You can also download from idsoftwares.com

							//  (t1*p01_minus_p21 - t0*p11_minus_p21) *
							//  xstepdenominv |
							//  t1*p00_minus_p20 - t0*p10_minus_p20
	fmulp	%st(0),%st(2)	// (t1*p01_minus_p21 - t0*p11_minus_p21) *
							//  xstepdenominv |
							//  (t1*p00_minus_p20 - t0*p10_minus_p20) *
							//  ystepdenominv
	fistpl	C(r_zistepx)	// (t1*p00_minus_p20 - t0*p10_minus_p20) *
							//  ystepdenominv
	fistpl	C(r_zistepy)

//	a_sstepxfrac = r_sstepx << 16;
//	a_tstepxfrac = r_tstepx << 16;
//
//	a_ststepxwhole = r_affinetridesc.skinwidth * (r_tstepx >> 16) +
//			(r_sstepx >> 16);

	movl	C(r_sstepx),%eax
	movl	C(r_tstepx),%edx
	shll	$16,%eax
	shll	$16,%edx
	movl	%eax,C(a_sstepxfrac)
	movl	%edx,C(a_tstepxfrac)

	movl	C(r_sstepx),%ecx
	movl	C(r_tstepx),%eax
	sarl	$16,%ecx
	sarl	$16,%eax
	imull	skinwidth(%esp)
	addl	%ecx,%eax
	movl	%eax,C(a_ststepxwhole)

	ret

#endif

//----------------------------------------------------------------------
// recursive subdivision affine triangle drawing code
//
// not C-callable because of stdcall return
//----------------------------------------------------------------------

#define lp1	4+16
#define lp2	8+16
#define lp3	12+16

.globl C(D_PolysetRecursiveTriangle)
C(D_PolysetRecursiveTriangle):
	pushl	%ebp				// preserve caller stack frame pointer
	pushl	%esi				// preserve register variables
	pushl	%edi
	pushl	%ebx

//	int		*temp;
//	int		d;
//	int		new[6];
//	int		i;
//	int		z;
//	short	*zbuf;
	movl	lp2(%esp),%esi
	movl	lp1(%esp),%ebx
	movl	lp3(%esp),%edi

//	d = lp2[0] - lp1[0];
//	if (d < -1 || d > 1)
//		goto split;
	movl	0(%esi),%eax

	movl	0(%ebx),%edx
	movl	4(%esi),%ebp

	subl	%edx,%eax
	movl	4(%ebx),%ecx

	subl	%ecx,%ebp
	incl	%eax

	cmpl	$2,%eax
	ja		LSplit

//	d = lp2[1] - lp1[1];
//	if (d < -1 || d > 1)
//		goto split;
	movl	0(%edi),%eax
	incl	%ebp

	cmpl	$2,%ebp
	ja		LSplit

//	d = lp3[0] - lp2[0];
//	if (d < -1 || d > 1)
//		goto split2;
	movl	0(%esi),%edx
	movl	4(%edi),%ebp

	subl	%edx,%eax
	movl	4(%esi),%ecx

	subl	%ecx,%ebp
	incl	%eax

	cmpl	$2,%eax
	ja		LSplit2

//	d = lp3[1] - lp2[1];
//	if (d < -1 || d > 1)
//		goto split2;
	movl	0(%ebx),%eax
	incl	%ebp

	cmpl	$2,%ebp
	ja		LSplit2

//	d = lp1[0] - lp3[0];
//	if (d < -1 || d > 1)
//		goto split3;
	movl	0(%edi),%edx
	movl	4(%ebx),%ebp

	subl	%edx,%eax
	movl	4(%edi),%ecx

	subl	%ecx,%ebp
	incl	%eax

	incl	%ebp
	movl	%ebx,%edx

	cmpl	$2,%eax
	ja		LSplit3

//	d = lp1[1] - lp3[1];
//	if (d < -1 || d > 1)
//	{
//split3:
//		temp = lp1;
//		lp3 = lp2;
//		lp1 = lp3;
//		lp2 = temp;
//		goto split;
//	}
//
//	return;			// entire tri is filled
//
	cmpl	$2,%ebp
	jna		LDone

LSplit3:
	movl	%edi,%ebx
	movl	%esi,%edi
	movl	%edx,%esi
	jmp		LSplit

//split2:
LSplit2:

//	temp = lp1;
//	lp1 = lp2;
//	lp2 = lp3;
//	lp3 = temp;
	movl	%ebx,%eax
	movl	%esi,%ebx
	movl	%edi,%esi
	movl	%eax,%edi

//split:
LSplit:

	subl	$24,%esp		// allocate space for a new vertex

//// split this edge
//	new[0] = (lp1[0] + lp2[0]) >> 1;
//	new[1] = (lp1[1] + lp2[1]) >> 1;
//	new[2] = (lp1[2] + lp2[2]) >> 1;
//	new[3] = (lp1[3] + lp2[3]) >> 1;
//	new[5] = (lp1[5] + lp2[5]) >> 1;
	movl	8(%ebx),%eax

	movl	8(%esi),%edx
	movl	12(%ebx),%ecx

	addl	%edx,%eax
	movl	12(%esi),%edx

	sarl	$1,%eax
	addl	%edx,%ecx

	movl	%eax,8(%esp)
	movl	20(%ebx),%eax

	sarl	$1,%ecx
	movl	20(%esi),%edx

	movl	%ecx,12(%esp)
	addl	%edx,%eax

	movl	0(%ebx),%ecx
	movl	0(%esi),%edx

	sarl	$1,%eax
	addl	%ecx,%edx

	movl	%eax,20(%esp)
	movl	4(%ebx),%eax

	sarl	$1,%edx
	movl	4(%esi),%ebp

	movl	%edx,0(%esp)
	addl	%eax,%ebp

	sarl	$1,%ebp
	movl	%ebp,4(%esp)

//// draw the point if splitting a leading edge
//	if (lp2[1] > lp1[1])
//		goto nodraw;
	cmpl	%eax,4(%esi)
	jg		LNoDraw

//	if ((lp2[1] == lp1[1]) && (lp2[0] < lp1[0]))
//		goto nodraw;
	movl	0(%esi),%edx
	jnz		LDraw

	cmpl	%ecx,%edx
	jl		LNoDraw

LDraw:

// z = new[5] >> 16;
	movl	20(%esp),%edx
	movl	4(%esp),%ecx

	sarl	$16,%edx
	movl	0(%esp),%ebp

//	zbuf = zspantable[new[1]] + new[0];
	movl	C(zspantable)(,%ecx,4),%eax

//	if (z >= *zbuf)
//	{
	cmpw	(%eax,%ebp,2),%dx
	jnge	LNoDraw

//		int		pix;
//		
//		*zbuf = z;
	movw	%dx,(%eax,%ebp,2)

//		pix = d_pcolormap[skintable[new[3]>>16][new[2]>>16]];
	movl	12(%esp),%eax

	sarl	$16,%eax
	movl	8(%esp),%edx

	sarl	$16,%edx
	subl	%ecx,%ecx

	movl	C(skintable)(,%eax,4),%eax
	movl	4(%esp),%ebp

	movb	(%eax,%edx,),%cl
	movl	C(d_pcolormap),%edx

	movb	(%edx,%ecx,),%dl
	movl	0(%esp),%ecx

//		d_viewbuffer[d_scantable[new[1]] + new[0]] = pix;
	movl	C(d_scantable)(,%ebp,4),%eax
	addl	%eax,%ecx
	movl	C(d_viewbuffer),%eax
	movb	%dl,(%eax,%ecx,1)

//	}
//
//nodraw:
LNoDraw:

//// recursively continue
//	D_PolysetRecursiveTriangle (lp3, lp1, new);
	pushl	%esp
	pushl	%ebx
	pushl	%edi
	call	C(D_PolysetRecursiveTriangle)

//	D_PolysetRecursiveTriangle (lp3, new, lp2);
	movl	%esp,%ebx
	pushl	%esi
	pushl	%ebx
	pushl	%edi
	call	C(D_PolysetRecursiveTriangle)
	addl	$24,%esp

LDone:
	popl	%ebx				// restore register variables
	popl	%edi
	popl	%esi
	popl	%ebp				// restore caller stack frame pointer
	ret		$12


//----------------------------------------------------------------------
// 8-bpp horizontal span drawing code for affine polygons, with smooth
// shading and no transparency
//----------------------------------------------------------------------

#define pspans	4+8

.globl C(D_PolysetAff8Start)
C(D_PolysetAff8Start):

.globl C(R_PolysetDrawSpans8_Opaque)
C(R_PolysetDrawSpans8_Opaque):
	pushl	%esi				// preserve register variables
	pushl	%ebx

	movl	pspans(%esp),%esi	// point to the first span descriptor
	movl	C(r_zistepx),%ecx

	pushl	%ebp				// preserve caller's stack frame
	pushl	%edi

	rorl	$16,%ecx			// put high 16 bits of 1/z step in low word
	movl	spanpackage_t_count(%esi),%edx

	movl	%ecx,lzistepx

LSpanLoop:

//		lcount = d_aspancount - pspanpackage->count;
//
//		errorterm += erroradjustup;
//		if (errorterm >= 0)
//		{
//			d_aspancount += d_countextrastep;
//			errorterm -= erroradjustdown;
//		}
//		else
//		{
//			d_aspancount += ubasestep;
//		}
	movl	C(d_aspancount),%eax
	subl	%edx,%eax

	movl	C(erroradjustup),%edx
	movl	C(errorterm),%ebx
	addl	%edx,%ebx
	js		LNoTurnover

	movl	C(erroradjustdown),%edx
	movl	C(d_countextrastep),%edi
	subl	%edx,%ebx
	movl	C(d_aspancount),%ebp
	movl	%ebx,C(errorterm)
	addl	%edi,%ebp
	movl	%ebp,C(d_aspancount)
	jmp		LRightEdgeStepped

LNoTurnover:
	movl	C(d_aspancount),%edi
	movl	C(ubasestep),%edx
	movl	%ebx,C(errorterm)
	addl	%edx,%edi
	movl	%edi,C(d_aspancount)

LRightEdgeStepped:
	cmpl	$1,%eax

	jl		LNextSpan
	jz		LExactlyOneLong

//
// set up advancetable
//
	movl	C(a_ststepxwhole),%ecx
	movl	C(r_affinetridesc)+atd_skinwidth,%edx

	movl	%ecx,advancetable+4	// advance base in t
	addl	%edx,%ecx

	movl	%ecx,advancetable	// advance extra in t
	movl	C(a_tstepxfrac),%ecx

	movw	C(r_lstepx),%cx
	movl	%eax,%edx			// count

	movl	%ecx,tstep
	addl	$7,%edx

	shrl	$3,%edx				// count of full and partial loops
	movl	spanpackage_t_sfrac(%esi),%ebx

	movw	%dx,%bx
	movl	spanpackage_t_pz(%esi),%ecx

	negl	%eax

	movl	spanpackage_t_pdest(%esi),%edi
	andl	$7,%eax		// 0->0, 1->7, 2->6, ... , 7->1

	subl	%eax,%edi	// compensate for hardwired offsets
	subl	%eax,%ecx

	subl	%eax,%ecx
	movl	spanpackage_t_tfrac(%esi),%edx

	movw	spanpackage_t_light(%esi),%dx
	movl	spanpackage_t_zi(%esi),%ebp

	rorl	$16,%ebp	// put high 16 bits of 1/z in low word
	pushl	%esi

	movl	spanpackage_t_ptex(%esi),%esi
	jmp		aff8entryvec_table(,%eax,4)

// %bx = count of full and partial loops