draw.c

这是一个同样来自贝尔实验室的和UNIX有着渊源的操作系统, 其简洁的设计和实现易于我们学习和理解

		off = pt.x%npack;
		val = p[0] >> bpp*(npack-1-off);
		val &= (1<<bpp)-1;
		break;
	case 8:
		val = p[0];
		break;
	case 16:
		val = p[0]|(p[1]<<8);
		break;
	case 24:
		val = p[0]|(p[1]<<8)|(p[2]<<16);
		break;
	case 32:
		val = p[0]|(p[1]<<8)|(p[2]<<16)|(p[3]<<24);
		break;
	}
	while(bpp<32){
		val |= val<<bpp;
		bpp *= 2;
	}
	return val;
}

static Calcfn*
boolcopyfn(Memimage *img, Memimage *mask)
{
	if(mask->flags&Frepl && Dx(mask->r)==1 && Dy(mask->r)==1 && pixelbits(mask, mask->r.min)==~0)
		return boolmemmove;

	switch(img->depth){
	case 8:
		return boolcopy8;
	case 16:
		return boolcopy16;
	case 24:
		return boolcopy24;
	case 32:
		return boolcopy32;
	default:
		assert(0 /* boolcopyfn */);
	}
	return nil;
}

/*
 * Optimized draw for filling and scrolling; uses memset and memmove.
 */
static void
memsetb(void *vp, uchar val, int n)
{
	uchar *p, *ep;

	p = vp;
	ep = p+n;
	while(p<ep)
		*p++ = val;
}

static void
memsets(void *vp, ushort val, int n)
{
	ushort *p, *ep;

	p = vp;
	ep = p+n;
	while(p<ep)
		*p++ = val;
}

static void
memsetl(void *vp, ulong val, int n)
{
	ulong *p, *ep;

	p = vp;
	ep = p+n;
	while(p<ep)
		*p++ = val;
}

static void
memset24(void *vp, ulong val, int n)
{
	uchar *p, *ep;
	uchar a,b,c;

	p = vp;
	ep = p+3*n;
	a = val;
	b = val>>8;
	c = val>>16;
	while(p<ep){
		*p++ = a;
		*p++ = b;
		*p++ = c;
	}
}

static ulong
imgtorgba(Memimage *img, ulong val)
{
	uchar r, g, b, a;
	int nb, ov, v;
	ulong chan;
	uchar *p;

	a = 0xFF;
	r = g = b = 0xAA;	/* garbage */
	for(chan=img->chan; chan; chan>>=8){
		nb = NBITS(chan);
		ov = v = val&((1<<nb)-1);
		val >>= nb;

		while(nb < 8){
			v |= v<<nb;
			nb *= 2;
		}
		v >>= (nb-8);

		switch(TYPE(chan)){
		case CRed:
			r = v;
			break;
		case CGreen:
			g = v;
			break;
		case CBlue:
			b = v;
			break;
		case CAlpha:
			a = v;
			break;
		case CGrey:
			r = g = b = v;
			break;
		case CMap:
			p = img->cmap->cmap2rgb+3*ov;
			r = *p++;
			g = *p++;	
			b = *p;
			break;
		}
	}
	return (r<<24)|(g<<16)|(b<<8)|a;	
}

static ulong
rgbatoimg(Memimage *img, ulong rgba)
{
	ulong chan;
	int d, nb;
	ulong v;
	uchar *p, r, g, b, a, m;

	v = 0;
	r = rgba>>24;
	g = rgba>>16;
	b = rgba>>8;
	a = rgba;
	d = 0;
	for(chan=img->chan; chan; chan>>=8){
		nb = NBITS(chan);
		switch(TYPE(chan)){
		case CRed:
			v |= (r>>(8-nb))<<d;
			break;
		case CGreen:
			v |= (g>>(8-nb))<<d;
			break;
		case CBlue:
			v |= (b>>(8-nb))<<d;
			break;
		case CAlpha:
			v |= (a>>(8-nb))<<d;
			break;
		case CMap:
			p = img->cmap->rgb2cmap;
			m = p[(r>>4)*256+(g>>4)*16+(b>>4)];
			v |= (m>>(8-nb))<<d;
			break;
		case CGrey:
			m = RGB2K(r,g,b);
			v |= (m>>(8-nb))<<d;
			break;
		}
		d += nb;
	}
//	print("rgba2img %.8lux = %.*lux\n", rgba, 2*d/8, v);
	return v;
}

#define DBG if(0)
static int
memoptdraw(Memdrawparam *par)
{
	int m, y, dy, dx, op;
	ulong v;
	Memimage *src;
	Memimage *dst;

	dx = Dx(par->r);
	dy = Dy(par->r);
	src = par->src;
	dst = par->dst;
	op = par->op;

DBG print("state %lux mval %lux dd %d\n", par->state, par->mval, dst->depth);
	/*
	 * If we have an opaque mask and source is one opaque pixel we can convert to the
	 * destination format and just replicate with memset.
	 */
	m = Simplesrc|Simplemask|Fullmask;
	if((par->state&m)==m && (par->srgba&0xFF) == 0xFF && (op ==S || op == SoverD)){
		uchar *dp, p[4];
		int d, dwid, ppb, np, nb;
		uchar lm, rm;

DBG print("memopt, dst %p, dst->data->bdata %p\n", dst, dst->data->bdata);
		dwid = dst->width*sizeof(ulong);
		dp = byteaddr(dst, par->r.min);
		v = par->sdval;
DBG print("sdval %lud, depth %d\n", v, dst->depth);
		switch(dst->depth){
		case 1:
		case 2:
		case 4:
			for(d=dst->depth; d<8; d*=2)
				v |= (v<<d);
			ppb = 8/dst->depth;	/* pixels per byte */
			m = ppb-1;
			/* left edge */
			np = par->r.min.x&m;		/* no. pixels unused on left side of word */
			dx -= (ppb-np);
			nb = 8 - np * dst->depth;		/* no. bits used on right side of word */
			lm = (1<<nb)-1;
DBG print("np %d x %d nb %d lm %ux ppb %d m %ux\n", np, par->r.min.x, nb, lm, ppb, m);	

			/* right edge */
			np = par->r.max.x&m;	/* no. pixels used on left side of word */
			dx -= np;
			nb = 8 - np * dst->depth;		/* no. bits unused on right side of word */
			rm = ~((1<<nb)-1);
DBG print("np %d x %d nb %d rm %ux ppb %d m %ux\n", np, par->r.max.x, nb, rm, ppb, m);	

DBG print("dx %d Dx %d\n", dx, Dx(par->r));
			/* lm, rm are masks that are 1 where we should touch the bits */
			if(dx < 0){	/* just one byte */
				lm &= rm;
				for(y=0; y<dy; y++, dp+=dwid)
					*dp ^= (v ^ *dp) & lm;
			}else if(dx == 0){	/* no full bytes */
				if(lm)
					dwid--;

				for(y=0; y<dy; y++, dp+=dwid){
					if(lm){
DBG print("dp %p v %lux lm %ux (v ^ *dp) & lm %lux\n", dp, v, lm, (v^*dp)&lm);
						*dp ^= (v ^ *dp) & lm;
						dp++;
					}
					*dp ^= (v ^ *dp) & rm;
				}
			}else{		/* full bytes in middle */
				dx /= ppb;
				if(lm)
					dwid--;
				dwid -= dx;

				for(y=0; y<dy; y++, dp+=dwid){
					if(lm){
						*dp ^= (v ^ *dp) & lm;
						dp++;
					}
					memset(dp, v, dx);
					dp += dx;
					*dp ^= (v ^ *dp) & rm;
				}
			}
			return 1;
		case 8:
			for(y=0; y<dy; y++, dp+=dwid)
				memset(dp, v, dx);
			return 1;
		case 16:
			p[0] = v;		/* make little endian */
			p[1] = v>>8;
			v = *(ushort*)p;
DBG print("dp=%p; dx=%d; for(y=0; y<%d; y++, dp+=%d)\nmemsets(dp, v, dx);\n",
	dp, dx, dy, dwid);
			for(y=0; y<dy; y++, dp+=dwid)
				memsets(dp, v, dx);
			return 1;
		case 24:
			for(y=0; y<dy; y++, dp+=dwid)
				memset24(dp, v, dx);
			return 1;
		case 32:
			p[0] = v;		/* make little endian */
			p[1] = v>>8;
			p[2] = v>>16;
			p[3] = v>>24;
			v = *(ulong*)p;
			for(y=0; y<dy; y++, dp+=dwid)
				memsetl(dp, v, dx);
			return 1;
		default:
			assert(0 /* bad dest depth in memoptdraw */);
		}
	}

	/*
	 * If no source alpha, an opaque mask, we can just copy the
	 * source onto the destination.  If the channels are the same and
	 * the source is not replicated, memmove suffices.
	 */
	m = Simplemask|Fullmask;
	if((par->state&(m|Replsrc))==m && src->depth >= 8 
	&& src->chan == dst->chan && !(src->flags&Falpha) && (op == S || op == SoverD)){
		uchar *sp, *dp;
		long swid, dwid, nb;
		int dir;

		if(src->data == dst->data && byteaddr(dst, par->r.min) > byteaddr(src, par->sr.min))
			dir = -1;
		else
			dir = 1;

		swid = src->width*sizeof(ulong);
		dwid = dst->width*sizeof(ulong);
		sp = byteaddr(src, par->sr.min);
		dp = byteaddr(dst, par->r.min);
		if(dir == -1){
			sp += (dy-1)*swid;
			dp += (dy-1)*dwid;
			swid = -swid;
			dwid = -dwid;
		}
		nb = (dx*src->depth)/8;
		for(y=0; y<dy; y++, sp+=swid, dp+=dwid)
			memmove(dp, sp, nb);
		return 1;
	}

	/*
	 * If we have a 1-bit mask, 1-bit source, and 1-bit destination, and
	 * they're all bit aligned, we can just use bit operators.  This happens
	 * when we're manipulating boolean masks, e.g. in the arc code.
	 */
	if((par->state&(Simplemask|Simplesrc|Replmask|Replsrc))==0 
	&& dst->chan==GREY1 && src->chan==GREY1 && par->mask->chan==GREY1 
	&& (par->r.min.x&7)==(par->sr.min.x&7) && (par->r.min.x&7)==(par->mr.min.x&7)){
		uchar *sp, *dp, *mp;
		uchar lm, rm;
		long swid, dwid, mwid;
		int i, x, dir;

		sp = byteaddr(src, par->sr.min);
		dp = byteaddr(dst, par->r.min);
		mp = byteaddr(par->mask, par->mr.min);
		swid = src->width*sizeof(ulong);
		dwid = dst->width*sizeof(ulong);
		mwid = par->mask->width*sizeof(ulong);

		if(src->data == dst->data && byteaddr(dst, par->r.min) > byteaddr(src, par->sr.min)){
			dir = -1;
		}else
			dir = 1;

		lm = 0xFF>>(par->r.min.x&7);
		rm = 0xFF<<(8-(par->r.max.x&7));
		dx -= (8-(par->r.min.x&7)) + (par->r.max.x&7);

		if(dx < 0){	/* one byte wide */
			lm &= rm;
			if(dir == -1){
				dp += dwid*(dy-1);
				sp += swid*(dy-1);
				mp += mwid*(dy-1);
				dwid = -dwid;
				swid = -swid;
				mwid = -mwid;
			}
			for(y=0; y<dy; y++){
				*dp ^= (*dp ^ *sp) & *mp & lm;
				dp += dwid;
				sp += swid;
				mp += mwid;
			}
			return 1;
		}

		dx /= 8;
		if(dir == 1){
			i = (lm!=0)+dx+(rm!=0);
			mwid -= i;
			swid -= i;
			dwid -= i;
			for(y=0; y<dy; y++, dp+=dwid, sp+=swid, mp+=mwid){
				if(lm){
					*dp ^= (*dp ^ *sp++) & *mp++ & lm;
					dp++;
				}
				for(x=0; x<dx; x++){
					*dp ^= (*dp ^ *sp++) & *mp++;
					dp++;
				}
				if(rm){
					*dp ^= (*dp ^ *sp++) & *mp++ & rm;
					dp++;
				}
			}
			return 1;
		}else{
		/* dir == -1 */
			i = (lm!=0)+dx+(rm!=0);
			dp += dwid*(dy-1)+i-1;
			sp += swid*(dy-1)+i-1;
			mp += mwid*(dy-1)+i-1;
			dwid = -dwid+i;
			swid = -swid+i;
			mwid = -mwid+i;
			for(y=0; y<dy; y++, dp+=dwid, sp+=swid, mp+=mwid){
				if(rm){
					*dp ^= (*dp ^ *sp--) & *mp-- & rm;
					dp--;
				}
				for(x=0; x<dx; x++){
					*dp ^= (*dp ^ *sp--) & *mp--;
					dp--;
				}
				if(lm){
					*dp ^= (*dp ^ *sp--) & *mp-- & lm;
					dp--;
				}
			}
		}
		return 1;
	}
	return 0;	
}
#undef DBG

/*
 * Boolean character drawing.
 * Solid opaque color through a 1-bit greyscale mask.
 */
#define DBG if(0)
static int
chardraw(Memdrawparam *par)
{
	ulong bits;
	int i, ddepth, dy, dx, x, bx, ex, y, npack, bsh, depth, op;
	ulong v, maskwid, dstwid;
	uchar *wp, *rp, *q, *wc;
	ushort *ws;
	ulong *wl;
	uchar sp[4];
	Rectangle r, mr;
	Memimage *mask, *src, *dst;

if(0) if(drawdebug) iprint("chardraw? mf %lux md %d sf %lux dxs %d dys %d dd %d ddat %p sdat %p\n",
		par->mask->flags, par->mask->depth, par->src->flags, 
		Dx(par->src->r), Dy(par->src->r), par->dst->depth, par->dst->data, par->src->data);

	mask = par->mask;
	src = par->src;
	dst = par->dst;
	r = par->r;
	mr = par->mr;
	op = par->op;

	if((par->state&(Replsrc|Simplesrc|Replmask)) != (Replsrc|Simplesrc)
	|| mask->depth != 1 || src->flags&Falpha || dst->depth<8 || dst->data==src->data
	|| op != SoverD)
		return 0;

//if(drawdebug) iprint("chardraw...");

	depth = mask->depth;
	maskwid = mask->width*sizeof(ulong);
	rp = byteaddr(mask, mr.min);
	npack = 8/depth;
	bsh = (mr.min.x % npack) * depth;

	wp = byteaddr(dst, r.min);
	dstwid = dst->width*sizeof(ulong);
DBG print("bsh %d\n", bsh);
	dy = Dy(r);
	dx = Dx(r);

	ddepth = dst->depth;

	/*
	 * for loop counts from bsh to bsh+dx
	 *
	 * we want the bottom bits to be the amount
	 * to shift the pixels down, so for n≡0 (mod 8) we want 
	 * bottom bits 7.  for n≡1, 6, etc.
	 * the bits come from -n-1.
	 */

	bx = -bsh-1;
	ex = -bsh-1-dx;
	SET(bits);
	v = par->sdval;

	/* make little endian */
	sp[0] = v;
	sp[1] = v>>8;
	sp[2] = v>>16;
	sp[3] = v>>24;

//print("sp %x %x %x %x\n", sp[0], sp[1], sp[2], sp[3]);
	for(y=0; y<dy; y++, rp+=maskwid, wp+=dstwid){
		q = rp;
		if(bsh)
			bits = *q++;
		switch(ddepth){
		case 8:
//if(drawdebug) iprint("8loop...");
			wc = wp;
			for(x=bx; x>ex; x--, wc++){
				i = x&7;
				if(i == 8-1)
					bits = *q++;
DBG print("bits %lux sh %d...", bits, i);
				if((bits>>i)&1)
					*wc = v;
			}
			break;
		case 16:
			ws = (ushort*)wp;
			v = *(ushort*)sp;
			for(x=bx; x>ex; x--, ws++){
				i = x&7;
				if(i == 8-1)
					bits = *q++;
DBG print("bits %lux sh %d...", bits, i);
				if((bits>>i)&1)
					*ws = v;
			}
			break;
		case 24:
			wc = wp;
			for(x=bx; x>ex; x--, wc+=3){
				i = x&7;
				if(i == 8-1)
					bits = *q++;
DBG print("bits %lux sh %d...", bits, i);
				if((bits>>i)&1){
					wc[0] = sp[0];
					wc[1] = sp[1];
					wc[2] = sp[2];
				}
			}
			break;
		case 32:
			wl = (ulong*)wp;
			v = *(ulong*)sp;
			for(x=bx; x>ex; x--, wl++){
				i = x&7;
				if(i == 8-1)
					bits = *q++;
DBG iprint("bits %lux sh %d...", bits, i);
				if((bits>>i)&1)
					*wl = v;
			}
			break;
		}
	}

DBG print("\n");	
	return 1;	
}
#undef DBG


/*
 * Fill entire byte with replicated (if necessary) copy of source pixel,
 * assuming destination ldepth is >= source ldepth.
 *
 * This code is just plain wrong for >8bpp.
 *
ulong
membyteval(Memimage *src)
{
	int i, val, bpp;
	uchar uc;

	unloadmemimage(src, src->r, &uc, 1);
	bpp = src->depth;
	uc <<= (src->r.min.x&(7/src->depth))*src->depth;
	uc &= ~(0xFF>>bpp);
	/* pixel value is now in high part of byte. repeat throughout byte 
	val = uc;
	for(i=bpp; i<8; i<<=1)
		val |= val>>i;
	return val;
}
 * 
 */

void
memfillcolor(Memimage *i, ulong val)
{
	ulong bits;
	int d, y;

	if(val == DNofill)
		return;

	bits = rgbatoimg(i, val);
	switch(i->depth){
	case 24:	/* 24-bit images suck */
		for(y=i->r.min.y; y<i->r.max.y; y++)
			memset24(byteaddr(i, Pt(i->r.min.x, y)), bits, Dx(i->r));
		break;
	default:	/* 1, 2, 4, 8, 16, 32 */
		for(d=i->depth; d<32; d*=2)
			bits = (bits << d) | bits;
		memsetl(wordaddr(i, i->r.min), bits, i->width*Dy(i->r));
		break;
	}
}