path: root/lib/gdi/gpixmap.cpp

#include <cstdlib>
#include <cstring>
#include <lib/gdi/gpixmap.h>
#include <lib/gdi/region.h>
#include <lib/gdi/accel.h>
#include <byteswap.h>

#ifndef BYTE_ORDER
#error "no BYTE_ORDER defined!"
#endif

gLookup::gLookup()
	:size(0), lookup(0)
{
}

gLookup::gLookup(int size, const gPalette &pal, const gRGB &start, const gRGB &end)
	:size(0), lookup(0)
{
	build(size, pal, start, end);
}

void gLookup::build(int _size, const gPalette &pal, const gRGB &start, const gRGB &end)
{
	if (lookup)
	{
		delete [] lookup;
		lookup=0;
		size=0;
	}
	size=_size;
	if (!size)
		return;
	lookup=new gColor[size];
	
	for (int i=0; i<size; i++)
	{
		gRGB col;
		if (i)
		{
			int rdiff=-start.r+end.r;
			int gdiff=-start.g+end.g;
			int bdiff=-start.b+end.b;
			int adiff=-start.a+end.a;
			rdiff*=i; rdiff/=(size-1);
			gdiff*=i; gdiff/=(size-1);
			bdiff*=i; bdiff/=(size-1);
			adiff*=i; adiff/=(size-1);
			col.r=start.r+rdiff;
			col.g=start.g+gdiff;
			col.b=start.b+bdiff;
			col.a=start.a+adiff;
		} else
			col=start;
		lookup[i]=pal.findColor(col);
	}
}

gSurface::gSurface()
{
	x = 0;
	y = 0;
	bpp = 0;
	bypp = 0;
	stride = 0;
	data = 0;
	data_phys = 0;
	clut.colors = 0;
	clut.data = 0;
	type = 0;
}

gSurface::gSurface(eSize size, int _bpp, int accel)
{
	x = size.width();
	y = size.height();
	bpp = _bpp;

	switch (bpp)
	{
	case 8:
		bypp = 1;
		break;
	case 15:
	case 16:
		bypp = 2;
		break;
	case 24:		// never use 24bit mode
	case 32:
		bypp = 4;
		break;
	default:
		bypp = (bpp+7)/8;
	}

	stride = x*bypp;
	
	data = 0;
	data_phys = 0;
	
	if (accel)
	{
		stride += 63;
		stride &=~63;
		
		int pal_size = 0;
		if (bpp == 8)
			pal_size = 256 * 4;
		
		if (gAccel::getInstance())
			eDebug("accel memory: %d", gAccel::getInstance()->accelAlloc(data, data_phys, y * stride + pal_size));
		else
			eDebug("no accel available");
	}
	
	clut.colors = 0;
	clut.data = 0;

	if (!data)
		data = new unsigned char [y * stride];
	
	type = 1;
}

gSurface::~gSurface()
{
	if (type)
	{
		if (data_phys)
			gAccel::getInstance()->accelFree(data_phys);
		else
			delete [] (unsigned char*)data;

		delete [] clut.data;
	}
}

gPixmap *gPixmap::lock()
{
	contentlock.lock(1);
	return this;
}

void gPixmap::unlock()
{
	contentlock.unlock(1);
}

void gPixmap::fill(const gRegion &region, const gColor &color)
{
	unsigned int i;
	for (i=0; i<region.rects.size(); ++i)
	{
		const eRect &area = region.rects[i];
		if (area.empty())
			continue;

		if (surface->bpp == 8)
		{
			for (int y=area.top(); y<area.bottom(); y++)
		 		memset(((__u8*)surface->data)+y*surface->stride+area.left(), color.color, area.width());
		} else if (surface->bpp == 16)
		{
			__u32 icol;

			if (surface->clut.data && color < surface->clut.colors)
				icol=(surface->clut.data[color].a<<24)|(surface->clut.data[color].r<<16)|(surface->clut.data[color].g<<8)|(surface->clut.data[color].b);
			else
				icol=0x10101*color;
#if BYTE_ORDER == LITTLE_ENDIAN
			__u16 col = bswap_16(((icol & 0xFF) >> 3) << 11 | ((icol & 0xFF00) >> 10) << 5 | (icol & 0xFF0000) >> 19);
#else
			__u16 col = ((icol & 0xFF) >> 3) << 11 | ((icol & 0xFF00) >> 10) << 5 | (icol & 0xFF0000) >> 19;
#endif
			for (int y=area.top(); y<area.bottom(); y++)
			{
				__u16 *dst=(__u16*)(((__u8*)surface->data)+y*surface->stride+area.left()*surface->bypp);
				int x=area.width();
				while (x--)
					*dst++=col;
			}
		} else if (surface->bpp == 32)
		{
			__u32 col;

			if (surface->clut.data && color < surface->clut.colors)
				col=(surface->clut.data[color].a<<24)|(surface->clut.data[color].r<<16)|(surface->clut.data[color].g<<8)|(surface->clut.data[color].b);
			else
				col=0x10101*color;
			
			col^=0xFF000000;
			
			if (surface->data_phys && gAccel::getInstance())
				if (!gAccel::getInstance()->fill(surface,  area, col))
					continue;

			for (int y=area.top(); y<area.bottom(); y++)
			{
				__u32 *dst=(__u32*)(((__u8*)surface->data)+y*surface->stride+area.left()*surface->bypp);
				int x=area.width();
				while (x--)
					*dst++=col;
			}
		}	else
			eWarning("couldn't fill %d bpp", surface->bpp);
	}
}

void gPixmap::fill(const gRegion &region, const gRGB &color)
{
	unsigned int i;
	for (i=0; i<region.rects.size(); ++i)
	{
		const eRect &area = region.rects[i];
		if (area.empty())
			continue;

		if (surface->bpp == 32)
		{
			__u32 col;

			col = color.argb();
			col^=0xFF000000;

			if (surface->data_phys && gAccel::getInstance())
				if (!gAccel::getInstance()->fill(surface,  area, col))
					continue;

			for (int y=area.top(); y<area.bottom(); y++)
			{
				__u32 *dst=(__u32*)(((__u8*)surface->data)+y*surface->stride+area.left()*surface->bypp);
				int x=area.width();
				while (x--)
					*dst++=col;
			}
		} else if (surface->bpp == 16)
		{
			__u32 icol = color.argb();
#if BYTE_ORDER == LITTLE_ENDIAN
			__u16 col = bswap_16(((icol & 0xFF) >> 3) << 11 | ((icol & 0xFF00) >> 10) << 5 | (icol & 0xFF0000) >> 19);
#else
			__u16 col = ((icol & 0xFF) >> 3) << 11 | ((icol & 0xFF00) >> 10) << 5 | (icol & 0xFF0000) >> 19;
#endif
			for (int y=area.top(); y<area.bottom(); y++)
			{
				__u16 *dst=(__u16*)(((__u8*)surface->data)+y*surface->stride+area.left()*surface->bypp);
				int x=area.width();
				while (x--)
					*dst++=col;
			}
		}	else
			eWarning("couldn't rgbfill %d bpp", surface->bpp);
	}
}

static inline void blit_8i_to_32(__u32 *dst, __u8 *src, __u32 *pal, int width)
{
	while (width--)
		*dst++=pal[*src++];
}

static inline void blit_8i_to_32_at(__u32 *dst, __u8 *src, __u32 *pal, int width)
{
	while (width--)
	{
		if (!(pal[*src]&0x80000000))
		{
			src++;
			dst++;
		} else
			*dst++=pal[*src++];
	}
}

static inline void blit_8i_to_16(__u16 *dst, __u8 *src, __u32 *pal, int width)
{
	while (width--)
		*dst++=pal[*src++] & 0xFFFF;
}

static inline void blit_8i_to_16_at(__u16 *dst, __u8 *src, __u32 *pal, int width)
{
	while (width--)
	{
		if (!(pal[*src]&0x80000000))
		{
			src++;
			dst++;
		} else
			*dst++=pal[*src++] & 0xFFFF;
	}
}

		/* WARNING, this function is not endian safe! */
static void blit_8i_to_32_ab(__u32 *dst, __u8 *src, __u32 *pal, int width)
{
	while (width--)
	{
#define BLEND(x, y, a) (y + (((x-y) * a)>>8))
		__u32 srccol = pal[*src++];
		__u32 dstcol = *dst;
		unsigned char sb = srccol & 0xFF;
		unsigned char sg = (srccol >> 8) & 0xFF;
		unsigned char sr = (srccol >> 16) & 0xFF;
		unsigned char sa = (srccol >> 24) & 0xFF;

		unsigned char db = dstcol & 0xFF;
		unsigned char dg = (dstcol >> 8) & 0xFF;
		unsigned char dr = (dstcol >> 16) & 0xFF;
		unsigned char da = (dstcol >> 24) & 0xFF;

		da = BLEND(0xFF, da, sa) & 0xFF;
		dr = BLEND(sr, dr, sa) & 0xFF;
		dg = BLEND(sg, dg, sa) & 0xFF;
		db = BLEND(sb, db, sa) & 0xFF;

#undef BLEND
		*dst++ = db | (dg << 8) | (dr << 16) | (da << 24);
	}
}

#define FIX 0x10000

void gPixmap::blit(const gPixmap &src, const eRect &_pos, const gRegion &clip, int flag)
{
//	eDebug("blit: -> %d.%d %d:%d -> %d.%d %d:%d, flags=%d",
//		_pos.x(), _pos.y(), _pos.width(), _pos.height(),
//		clip.extends.x(), clip.extends.y(), clip.extends.width(), clip.extends.height(),
//		flag);
	eRect pos = _pos;
	
//	eDebug("source size: %d %d", src.size().width(), src.size().height());
	
	if (!(flag & blitScale)) /* pos' size is valid only when scaling */
		pos = eRect(pos.topLeft(), src.size());
	else if (pos.size() == src.size()) /* no scaling required */
		flag &= ~blitScale;

	int scale_x = FIX, scale_y = FIX;
	
	if (flag & blitScale)
	{
		ASSERT(src.size().width());
		ASSERT(src.size().height());
		scale_x = pos.size().width() * FIX / src.size().width();
		scale_y = pos.size().height() * FIX / src.size().height();
	}
	
//	eDebug("SCALE %x %x", scale_x, scale_y);

	for (unsigned int i=0; i<clip.rects.size(); ++i)
	{
//		eDebug("clip rect: %d %d %d %d", clip.rects[i].x(), clip.rects[i].y(), clip.rects[i].width(), clip.rects[i].height());
		eRect area = pos; /* pos is the virtual (pre-clipping) area on the dest, which can be larger/smaller than src if scaling is enabled */
		area&=clip.rects[i];
		area&=eRect(ePoint(0, 0), size());

		if (area.empty())
			continue;

		eRect srcarea = area;
		srcarea.moveBy(-pos.x(), -pos.y());

//		eDebug("srcarea before scale: %d %d %d %d",
//			srcarea.x(), srcarea.y(), srcarea.width(), srcarea.height());
		
		if (flag & blitScale)
			srcarea = eRect(srcarea.x() * FIX / scale_x, srcarea.y() * FIX / scale_y, srcarea.width() * FIX / scale_x, srcarea.height() * FIX / scale_y);

//		eDebug("srcarea after scale: %d %d %d %d",
//			srcarea.x(), srcarea.y(), srcarea.width(), srcarea.height());

		if ((surface->data_phys && src.surface->data_phys) && (gAccel::getInstance()))
			if (!gAccel::getInstance()->blit(surface, src.surface, area, srcarea, flag))
				continue;

		if (flag & blitScale)
		{
			eWarning("unimplemented: scale on non-accel surfaces");
			continue;
		}

		if ((surface->bpp == 8) && (src.surface->bpp==8))
		{
			__u8 *srcptr=(__u8*)src.surface->data;
			__u8 *dstptr=(__u8*)surface->data;

			srcptr+=srcarea.left()*src.surface->bypp+srcarea.top()*src.surface->stride;
			dstptr+=area.left()*surface->bypp+area.top()*surface->stride;
			for (int y=0; y<area.height(); y++)
			{
				if (flag & (blitAlphaTest|blitAlphaBlend))
				{
  		      // no real alphatest yet
					int width=area.width();
					unsigned char *src=(unsigned char*)srcptr;
					unsigned char *dst=(unsigned char*)dstptr;
						// use duff's device here!
					while (width--)
					{
						if (!*src)
						{
							src++;
							dst++;
						} else
							*dst++=*src++;
					}
				} else
					memcpy(dstptr, srcptr, area.width()*surface->bypp);
				srcptr+=src.surface->stride;
				dstptr+=surface->stride;
			}
		} else if ((surface->bpp == 32) && (src.surface->bpp==32))
		{
			__u32 *srcptr=(__u32*)src.surface->data;
			__u32 *dstptr=(__u32*)surface->data;

			srcptr+=srcarea.left()+srcarea.top()*src.surface->stride/4;
			dstptr+=area.left()+area.top()*surface->stride/4;
			for (int y=0; y<area.height(); y++)
			{
				if (flag & blitAlphaTest)
				{
					int width=area.width();
					unsigned long *src=(unsigned long*)srcptr;
					unsigned long *dst=(unsigned long*)dstptr;

					while (width--)
					{
						if (!((*src)&0xFF000000))
						{
							src++;
							dst++;
						} else
							*dst++=*src++;
					}
				} else if (flag & blitAlphaBlend)
				{
					// uh oh. this is only until hardware accel is working.

					int width=area.width();
							// ARGB color space!
					unsigned char *src=(unsigned char*)srcptr;
					unsigned char *dst=(unsigned char*)dstptr;

#define BLEND(x, y, a) (y + ((x-y) * a)/256)
					while (width--)
					{
						unsigned char sa = src[3];
						unsigned char sr = src[2];
						unsigned char sg = src[1];
						unsigned char sb = src[0];

						unsigned char da = dst[3];
						unsigned char dr = dst[2];
						unsigned char dg = dst[1];
						unsigned char db = dst[0];

						dst[3] = BLEND(0xFF, da, sa);
						dst[2] = BLEND(sr, dr, sa);
						dst[1] = BLEND(sg, dg, sa);
						dst[0] = BLEND(sb, db, sa);
#undef BLEND

						src += 4; dst += 4;
					}
				} else
					memcpy(dstptr, srcptr, area.width()*surface->bypp);
				srcptr+=src.surface->stride/4;
				dstptr+=surface->stride/4;
			}
		} else if ((surface->bpp == 32) && (src.surface->bpp==8))
		{	
			__u8 *srcptr=(__u8*)src.surface->data;
			__u8 *dstptr=(__u8*)surface->data; // !!
			__u32 pal[256];

			for (int i=0; i<256; ++i)
			{
				if (src.surface->clut.data && (i<src.surface->clut.colors))
					pal[i]=(src.surface->clut.data[i].a<<24)|(src.surface->clut.data[i].r<<16)|(src.surface->clut.data[i].g<<8)|(src.surface->clut.data[i].b);
				else
					pal[i]=0x010101*i;
				pal[i]^=0xFF000000;
			}

			srcptr+=srcarea.left()*src.surface->bypp+srcarea.top()*src.surface->stride;
			dstptr+=area.left()*surface->bypp+area.top()*surface->stride;
			for (int y=0; y<area.height(); y++)
			{
				int width=area.width();
				unsigned char *psrc=(unsigned char*)srcptr;
				__u32 *dst=(__u32*)dstptr;
				if (flag & blitAlphaTest)
					blit_8i_to_32_at(dst, psrc, pal, width);
				else if (flag & blitAlphaBlend)
					blit_8i_to_32_ab(dst, psrc, pal, width);
				else
					blit_8i_to_32(dst, psrc, pal, width);
				srcptr+=src.surface->stride;
				dstptr+=surface->stride;
			}
		} else if ((surface->bpp == 16) && (src.surface->bpp==8))
		{
			__u8 *srcptr=(__u8*)src.surface->data;
			__u8 *dstptr=(__u8*)surface->data; // !!
			__u32 pal[256];

			for (int i=0; i<256; ++i)
			{
				__u32 icol;
				if (src.surface->clut.data && (i<src.surface->clut.colors))
					icol=(src.surface->clut.data[i].a<<24)|(src.surface->clut.data[i].r<<16)|(src.surface->clut.data[i].g<<8)|(src.surface->clut.data[i].b);
				else
					icol=0x010101*i;
#if BYTE_ORDER == LITTLE_ENDIAN
				pal[i] = bswap_16(((icol & 0xFF) >> 3) << 11 | ((icol & 0xFF00) >> 10) << 5 | (icol & 0xFF0000) >> 19);
#else
				pal[i] = ((icol & 0xFF) >> 3) << 11 | ((icol & 0xFF00) >> 10) << 5 | (icol & 0xFF0000) >> 19;
#endif
				pal[i]^=0xFF000000;
			}

			srcptr+=srcarea.left()*src.surface->bypp+srcarea.top()*src.surface->stride;
			dstptr+=area.left()*surface->bypp+area.top()*surface->stride;

			if (flag & blitAlphaBlend)
				eWarning("ignore unsupported 8bpp -> 16bpp alphablend!");

			for (int y=0; y<area.height(); y++)
			{
				int width=area.width();
				unsigned char *psrc=(unsigned char*)srcptr;
				__u16 *dst=(__u16*)dstptr;
				if (flag & blitAlphaTest)
					blit_8i_to_16_at(dst, psrc, pal, width);
				else
					blit_8i_to_16(dst, psrc, pal, width);
				srcptr+=src.surface->stride;
				dstptr+=surface->stride;
			}
		} else if ((surface->bpp == 16) && (src.surface->bpp==32))
		{
			__u8 *srcptr=(__u8*)src.surface->data;
			__u8 *dstptr=(__u8*)surface->data;

			srcptr+=srcarea.left()+srcarea.top()*src.surface->stride;
			dstptr+=area.left()+area.top()*surface->stride;

			if (flag & blitAlphaBlend)
				eWarning("ignore unsupported 32bpp -> 16bpp alphablend!");

			for (int y=0; y<area.height(); y++)
			{
				int width=area.width();
				__u32 *srcp=(__u32*)srcptr;
				__u16 *dstp=(__u16*)dstptr;

				if (flag & blitAlphaTest)
				{
					while (width--)
					{
						if (!((*srcp)&0xFF000000))
						{
							srcp++;
							dstp++;
						} else
						{
							__u32 icol = *srcp++;
#if BYTE_ORDER == LITTLE_ENDIAN
							*dstp++ = bswap_16(((icol & 0xFF) >> 3) << 11 | ((icol & 0xFF00) >> 10) << 5 | (icol & 0xFF0000) >> 19);
#else
							*dstp++ = ((icol & 0xFF) >> 3) << 11 | ((icol & 0xFF00) >> 10) << 5 | (icol & 0xFF0000) >> 19;
#endif
						}
					}
				} else
				{
					while (width--)
					{
						__u32 icol = *srcp++;
#if BYTE_ORDER == LITTLE_ENDIAN
						*dstp++ = bswap_16(((icol & 0xFF) >> 3) << 11 | ((icol & 0xFF00) >> 10) << 5 | (icol & 0xFF0000) >> 19);
#else
						*dstp++ = ((icol & 0xFF) >> 3) << 11 | ((icol & 0xFF00) >> 10) << 5 | (icol & 0xFF0000) >> 19;
#endif
					}
				}
				srcptr+=src.surface->stride;
				dstptr+=surface->stride;
			}
		} else
			eWarning("cannot blit %dbpp from %dbpp", surface->bpp, src.surface->bpp);
	}
}

#undef FIX

void gPixmap::mergePalette(const gPixmap &target)
{
	if ((!surface->clut.colors) || (!target.surface->clut.colors))
		return;

	gColor *lookup=new gColor[surface->clut.colors];

	for (int i=0; i<surface->clut.colors; i++)
		lookup[i].color=target.surface->clut.findColor(surface->clut.data[i]);
	
	delete [] surface->clut.data;
	surface->clut.colors=target.surface->clut.colors;
	surface->clut.data=new gRGB[surface->clut.colors];
	memcpy(surface->clut.data, target.surface->clut.data, sizeof(gRGB)*surface->clut.colors);

	__u8 *dstptr=(__u8*)surface->data;

	for (int ay=0; ay<surface->y; ay++)
	{
		for (int ax=0; ax<surface->x; ax++)
			dstptr[ax]=lookup[dstptr[ax]];
		dstptr+=surface->stride;
	}
	
	delete [] lookup;
}

static inline int sgn(int a)
{
	if (a < 0)
		return -1;
	else if (!a)
		return 0;
	else
		return 1;
}

void gPixmap::line(const gRegion &clip, ePoint start, ePoint dst, gColor color)
{
	__u8 *srf8 = 0;
	__u16 *srf16 = 0;
	__u32 *srf32 = 0;
	int stride = surface->stride;

	if (clip.rects.empty())
		return;

	__u16 col16;
	__u32 col = 0;
	if (surface->bpp == 8)
		srf8 = (__u8*)surface->data;
	else
	{
		srf32 = (__u32*)surface->data;
		if (surface->clut.data && color < surface->clut.colors)
			col=(surface->clut.data[color].a<<24)|(surface->clut.data[color].r<<16)|(surface->clut.data[color].g<<8)|(surface->clut.data[color].b);
		else
			col=0x10101*color;
		col^=0xFF000000;
	}

	if (surface->bpp == 16)
#if BYTE_ORDER == LITTLE_ENDIAN
		col16=bswap_16(((col & 0xFF) >> 3) << 11 | ((col & 0xFF00) >> 10) << 5 | (col & 0xFF0000) >> 19);
#else
		col16=((col & 0xFF) >> 3) << 11 | ((col & 0xFF00) >> 10) << 5 | (col & 0xFF0000) >> 19;
#endif

	int xa = start.x(), ya = start.y(), xb = dst.x(), yb = dst.y();
	int dx, dy, x, y, s1, s2, e, temp, swap, i;
	dy=abs(yb-ya);
	dx=abs(xb-xa);
	s1=sgn(xb-xa);
	s2=sgn(yb-ya);
	x=xa;
	y=ya;
	if (dy>dx)
	{
		temp=dx;
		dx=dy;
		dy=temp;
		swap=1;
	} else
		swap=0;
	e = 2*dy-dx;

	int lasthit = 0;
	for(i=1; i<=dx; i++)
	{
				/* i don't like this clipping loop, but the only */
				/* other choice i see is to calculate the intersections */
				/* before iterating through the pixels. */
				
				/* one could optimize this because of the ordering */
				/* of the bands. */
				
		lasthit = 0;
		int a = lasthit;
		
			/* if last pixel was invisble, first check bounding box */
		if (a == -1)
		{
				/* check if we just got into the bbox again */
			if (clip.extends.contains(x, y))
				lasthit = a = 0;
			else
				goto fail;
		} else if (!clip.rects[a].contains(x, y))
		{
			do
			{
				++a;
				if ((unsigned int)a == clip.rects.size())
					a = 0;
				if (a == lasthit)
				{
					goto fail;
					lasthit = -1;
				}
			} while (!clip.rects[a].contains(x, y));
			lasthit = a;
		}

		if (srf8)
			srf8[y * stride + x] = color;
		else if (srf16)
			srf16[y * stride/2 + x] = col16;
		else
			srf32[y * stride/4 + x] = col;
fail:
		while (e>=0)
		{
			if (swap==1)
				x+=s1;
			else
				y+=s2;
			e-=2*dx;
		}

		if (swap==1)
			y+=s2;
		else
			x+=s1;
		e+=2*dy;
	}
}

gColor gPalette::findColor(const gRGB &rgb) const
{
		/* grayscale? */
	if (!data)
		return (rgb.r + rgb.g + rgb.b) / 3;
	
	int difference=1<<30, best_choice=0;
	for (int t=0; t<colors; t++)
	{
		int ttd;
		int td=(signed)(rgb.r-data[t].r); td*=td; td*=(255-data[t].a);
		ttd=td;
		if (ttd>=difference)
			continue;
		td=(signed)(rgb.g-data[t].g); td*=td; td*=(255-data[t].a);
		ttd+=td;
		if (ttd>=difference)
			continue;
		td=(signed)(rgb.b-data[t].b); td*=td; td*=(255-data[t].a);
		ttd+=td;
		if (ttd>=difference)
			continue;
		td=(signed)(rgb.a-data[t].a); td*=td; td*=255;
		ttd+=td;
		if (ttd>=difference)
			continue;
		if (!ttd)
			return t;
		difference=ttd;
		best_choice=t;
	}
	return best_choice;
}

DEFINE_REF(gPixmap);

gPixmap::~gPixmap()
{
	if (must_delete_surface)
		delete surface;
}

gPixmap::gPixmap(gSurface *surface)
	:surface(surface), must_delete_surface(false)
{
}

gPixmap::gPixmap(eSize size, int bpp, int accel)
	:must_delete_surface(true)
{
	surface = new gSurface(size, bpp, accel);
}