rleaccel.c

linux下的图形界面开发minigui最新源代码

		int ofs = 0;						      \
		for(;;) {						      \
		    unsigned run;					      \
		    ofs += *(Type *)srcbuf;				      \
		    run = ((Type *)srcbuf)[1];				      \
		    srcbuf += 2 * sizeof(Type);				      \
		    if(run) {						      \
			do_blit(dstbuf + ofs * bpp, srcbuf, run, bpp, alpha); \
			srcbuf += run * bpp;				      \
			ofs += run;					      \
		    } else if(!ofs)					      \
			break;						      \
		    if(ofs == w) {					      \
			ofs = 0;					      \
			dstbuf += dst->pitch;				      \
			if(!--linecount)				      \
			    break;					      \
		    }							      \
		}							      \
	    } while(0)

	    CHOOSE_BLIT(RLEBLIT, alpha, fmt);

#undef RLEBLIT
	}

done:
	/* Unlock the destination if necessary */
	if ( dst->flags & (GAL_HWSURFACE|GAL_ASYNCBLIT) ) {
		GAL_VideoDevice *video = current_video;
		GAL_VideoDevice *this  = current_video;
#if 0
		video->UnlockHWSurface(this, dst);
#endif
	}
	return(0);
}

#undef OPAQUE_BLIT

/*
 * Per-pixel blitting macros for translucent pixels:
 * These use the same techniques as the per-surface blitting macros
 */

/*
 * For 32bpp pixels, we have made sure the alpha is stored in the top
 * 8 bits, so proceed as usual
 */
#define BLIT_TRANSL_888(src, dst)				\
    do {							\
        Uint32 s = src;						\
	Uint32 d = dst;						\
	unsigned alpha = s >> 24;				\
	Uint32 s1 = s & 0xff00ff;				\
	Uint32 d1 = d & 0xff00ff;				\
	d1 = (d1 + ((s1 - d1) * alpha >> 8)) & 0xff00ff;	\
	s &= 0xff00;						\
	d &= 0xff00;						\
	d = (d + ((s - d) * alpha >> 8)) & 0xff00;		\
	dst = d1 | d;						\
    } while(0)

/*
 * For 16bpp pixels, we have stored the 5 most significant alpha bits in
 * bits 5-10. As before, we can process all 3 RGB components at the same time.
 */
#define BLIT_TRANSL_565(src, dst)		\
    do {					\
        Uint32 s = src;				\
	Uint32 d = dst;				\
	unsigned alpha = (s & 0x3e0) >> 5;	\
	s &= 0x07e0f81f;			\
	d = (d | d << 16) & 0x07e0f81f;		\
	d += (s - d) * alpha >> 5;		\
	d &= 0x07e0f81f;			\
	dst = d | d >> 16;			\
    } while(0)

#define BLIT_TRANSL_555(src, dst)		\
    do {					\
        Uint32 s = src;				\
	Uint32 d = dst;				\
	unsigned alpha = (s & 0x3e0) >> 5;	\
	s &= 0x03e07c1f;			\
	d = (d | d << 16) & 0x03e07c1f;		\
	d += (s - d) * alpha >> 5;		\
	d &= 0x03e07c1f;			\
	dst = d | d >> 16;			\
    } while(0)

/* used to save the destination format in the encoding. Designed to be
   macro-compatible with GAL_PixelFormat but without the unneeded fields */
typedef struct {
    	Uint8  BytesPerPixel;
	Uint8  Rloss;
	Uint8  Gloss;
	Uint8  Bloss;
	Uint8  Rshift;
	Uint8  Gshift;
	Uint8  Bshift;
	Uint8  Ashift;
	Uint32 Rmask;
	Uint32 Gmask;
	Uint32 Bmask;
	Uint32 Amask;
} RLEDestFormat;

/* blit a pixel-alpha RLE surface clipped at the right and/or left edges */
static void RLEAlphaClipBlit(int w, Uint8 *srcbuf, GAL_Surface *dst,
			     Uint8 *dstbuf, GAL_Rect *srcrect)
{
    GAL_PixelFormat *df = dst->format;
    /*
     * clipped blitter: Ptype is the destination pixel type,
     * Ctype the translucent count type, and do_blend the macro
     * to blend one pixel.
     */
#define RLEALPHACLIPBLIT(Ptype, Ctype, do_blend)			  \
    do {								  \
	int linecount = srcrect->h;					  \
	int left = srcrect->x;						  \
	int right = left + srcrect->w;					  \
	dstbuf -= left * sizeof(Ptype);					  \
	do {								  \
	    int ofs = 0;						  \
	    /* blit opaque pixels on one line */			  \
	    do {							  \
		unsigned run;						  \
		ofs += ((Ctype *)srcbuf)[0];				  \
		run = ((Ctype *)srcbuf)[1];				  \
		srcbuf += 2 * sizeof(Ctype);				  \
		if(run) {						  \
		    /* clip to left and right borders */		  \
		    int cofs = ofs;					  \
		    int crun = run;					  \
		    if(left - cofs > 0) {				  \
			crun -= left - cofs;				  \
			cofs = left;					  \
		    }							  \
		    if(crun > right - cofs)				  \
			crun = right - cofs;				  \
		    if(crun > 0)					  \
			PIXEL_COPY(dstbuf + cofs * sizeof(Ptype),	  \
				   srcbuf + (cofs - ofs) * sizeof(Ptype), \
				   (unsigned)crun, sizeof(Ptype));	  \
		    srcbuf += run * sizeof(Ptype);			  \
		    ofs += run;						  \
		} else if(!ofs)						  \
		    return;						  \
	    } while(ofs < w);						  \
	    /* skip padding if necessary */				  \
	    if(sizeof(Ptype) == 2)					  \
		srcbuf += (unsigned long)srcbuf & 2;			  \
	    /* blit translucent pixels on the same line */		  \
	    ofs = 0;							  \
	    do {							  \
		unsigned run;						  \
		ofs += ((Uint16 *)srcbuf)[0];				  \
		run = ((Uint16 *)srcbuf)[1];				  \
		srcbuf += 4;						  \
		if(run) {						  \
		    /* clip to left and right borders */		  \
		    int cofs = ofs;					  \
		    int crun = run;					  \
		    if(left - cofs > 0) {				  \
			crun -= left - cofs;				  \
			cofs = left;					  \
		    }							  \
		    if(crun > right - cofs)				  \
			crun = right - cofs;				  \
		    if(crun > 0) {					  \
			Ptype *dst = (Ptype *)dstbuf + cofs;		  \
			Uint32 *src = (Uint32 *)srcbuf + (cofs - ofs);	  \
			int i;						  \
			for(i = 0; i < crun; i++)			  \
			    do_blend(src[i], dst[i]);			  \
		    }							  \
		    srcbuf += run * 4;					  \
		    ofs += run;						  \
		}							  \
	    } while(ofs < w);						  \
	    dstbuf += dst->pitch;					  \
	} while(--linecount);						  \
    } while(0)

    switch(df->BytesPerPixel) {
    case 2:
	if(df->Gmask == 0x07e0 || df->Rmask == 0x07e0
	   || df->Bmask == 0x07e0)
	    RLEALPHACLIPBLIT(Uint16, Uint8, BLIT_TRANSL_565);
	else
	    RLEALPHACLIPBLIT(Uint16, Uint8, BLIT_TRANSL_555);
	break;
    case 4:
	RLEALPHACLIPBLIT(Uint32, Uint16, BLIT_TRANSL_888);
	break;
    }
}

/* blit a pixel-alpha RLE surface */
int GAL_RLEAlphaBlit(GAL_Surface *src, GAL_Rect *srcrect,
		     GAL_Surface *dst, GAL_Rect *dstrect)
{
    int x, y;
    int w = src->w;
    Uint8 *srcbuf, *dstbuf;
    GAL_PixelFormat *df = dst->format;

    /* Lock the destination if necessary */
    if(dst->flags & (GAL_HWSURFACE|GAL_ASYNCBLIT)) {
	GAL_VideoDevice *video = current_video;
	GAL_VideoDevice *this  = current_video;
#if 0
	if(video->LockHWSurface(this, dst) < 0) {
	    return -1;
	}
#endif
    }

    x = dstrect->x;
    y = dstrect->y;
    dstbuf = (Uint8 *)dst->pixels + dst->offset
	     + y * dst->pitch + x * df->BytesPerPixel;
    srcbuf = (Uint8 *)src->map->sw_data->aux_data + sizeof(RLEDestFormat);

    {
	/* skip lines at the top if necessary */
	int vskip = srcrect->y;
	if(vskip) {
	    int ofs;
	    if(df->BytesPerPixel == 2) {
		/* the 16/32 interleaved format */
		do {
		    /* skip opaque line */
		    ofs = 0;
		    do {
			int run;
			ofs += srcbuf[0];
			run = srcbuf[1];
			srcbuf += 2;
			if(run) {
			    srcbuf += 2 * run;
			    ofs += run;
			} else if(!ofs)
			    goto done;
		    } while(ofs < w);

		    /* skip padding */
		    srcbuf += (unsigned long)srcbuf & 2;

		    /* skip translucent line */
		    ofs = 0;
		    do {
			int run;
			ofs += ((Uint16 *)srcbuf)[0];
			run = ((Uint16 *)srcbuf)[1];
			srcbuf += 4 * (run + 1);
			ofs += run;
		    } while(ofs < w);
		} while(--vskip);
	    } else {
		/* the 32/32 interleaved format */
		vskip <<= 1;	/* opaque and translucent have same format */
		do {
		    ofs = 0;
		    do {
			int run;
			ofs += ((Uint16 *)srcbuf)[0];
			run = ((Uint16 *)srcbuf)[1];
			srcbuf += 4;
			if(run) {
			    srcbuf += 4 * run;
			    ofs += run;
			} else if(!ofs)
			    goto done;
		    } while(ofs < w);
		} while(--vskip);
	    }
	}
    }

    /* if left or right edge clipping needed, call clip blit */
    if(srcrect->x || srcrect->w != src->w) {
	RLEAlphaClipBlit(w, srcbuf, dst, dstbuf, srcrect);
    } else {

	/*
	 * non-clipped blitter. Ptype is the destination pixel type,
	 * Ctype the translucent count type, and do_blend the
	 * macro to blend one pixel.
	 */
#define RLEALPHABLIT(Ptype, Ctype, do_blend)				 \
	do {								 \
	    int linecount = srcrect->h;					 \
	    do {							 \
		int ofs = 0;						 \
		/* blit opaque pixels on one line */			 \
		do {							 \
		    unsigned run;					 \
		    ofs += ((Ctype *)srcbuf)[0];			 \
		    run = ((Ctype *)srcbuf)[1];				 \
		    srcbuf += 2 * sizeof(Ctype);			 \
		    if(run) {						 \
			PIXEL_COPY(dstbuf + ofs * sizeof(Ptype), srcbuf, \
				   run, sizeof(Ptype));			 \
			srcbuf += run * sizeof(Ptype);			 \
			ofs += run;					 \
		    } else if(!ofs)					 \
			goto done;					 \
		} while(ofs < w);					 \
		/* skip padding if necessary */				 \
		if(sizeof(Ptype) == 2)					 \
		    srcbuf += (unsigned long)srcbuf & 2;		 \
		/* blit translucent pixels on the same line */		 \
		ofs = 0;						 \
		do {							 \
		    unsigned run;					 \
		    ofs += ((Uint16 *)srcbuf)[0];			 \
		    run = ((Uint16 *)srcbuf)[1];			 \
		    srcbuf += 4;					 \
		    if(run) {						 \
			Ptype *dst = (Ptype *)dstbuf + ofs;		 \
			unsigned i;					 \
			for(i = 0; i < run; i++) {			 \
			    Uint32 src = *(Uint32 *)srcbuf;		 \
			    do_blend(src, *dst);			 \
			    srcbuf += 4;				 \
			    dst++;					 \
			}						 \
			ofs += run;					 \
		    }							 \
		} while(ofs < w);					 \
		dstbuf += dst->pitch;					 \
	    } while(--linecount);					 \
	} while(0)

	switch(df->BytesPerPixel) {
	case 2:
	    if(df->Gmask == 0x07e0 || df->Rmask == 0x07e0
	       || df->Bmask == 0x07e0)
		RLEALPHABLIT(Uint16, Uint8, BLIT_TRANSL_565);
	    else
		RLEALPHABLIT(Uint16, Uint8, BLIT_TRANSL_555);
	    break;
	case 4:
	    RLEALPHABLIT(Uint32, Uint16, BLIT_TRANSL_888);
	    break;
	}
    }

 done:
    /* Unlock the destination if necessary */
    if(dst->flags & (GAL_HWSURFACE|GAL_ASYNCBLIT)) {
	GAL_VideoDevice *video = current_video;
	GAL_VideoDevice *this  = current_video;
#if 0
	video->UnlockHWSurface(this, dst);
#endif
    }
    return 0;
}

/*
 * Auxiliary functions:
 * The encoding functions take 32bpp rgb + a, and
 * return the number of bytes copied to the destination.
 * The decoding functions copy to 32bpp rgb + a, and
 * return the number of bytes copied from the source.
 * These are only used in the encoder and un-RLE code and are therefore not
 * highly optimised.
 */

/* encode 32bpp rgb + a into 16bpp rgb, losing alpha */
static int copy_opaque_16(void *dst, Uint32 *src, int n,
			  GAL_PixelFormat *sfmt, GAL_PixelFormat *dfmt)
{
    int i;
    Uint16 *d = dst;
    for(i = 0; i < n; i++) {
	unsigned r, g, b;
	RGB_FROM_PIXEL(*src, sfmt, r, g, b);
	PIXEL_FROM_RGB(*d, dfmt, r, g, b);
	src++;
	d++;
    }
    return n * 2;
}

/* decode opaque pixels from 16bpp to 32bpp rgb + a */
static int uncopy_opaque_16(Uint32 *dst, void *src, int n,
			    RLEDestFormat *sfmt, GAL_PixelFormat *dfmt)
{
    int i;
    Uint16 *s = src;
    unsigned alpha = dfmt->Amask ? 255 : 0;
    for(i = 0; i < n; i++) {
	unsigned r, g, b;
	RGB_FROM_PIXEL(*s, sfmt, r, g, b);
	PIXEL_FROM_RGBA(*dst, dfmt, r, g, b, alpha);
	s++;
	dst++;
    }
    return n * 2;
}



/* encode 32bpp rgb + a into 32bpp G0RAB format for blitting into 565 */
static int copy_transl_565(void *dst, Uint32 *src, int n,
			   GAL_PixelFormat *sfmt, GAL_PixelFormat *dfmt)
{
    int i;
    Uint32 *d = dst;
    for(i = 0; i < n; i++) {
	unsigned r, g, b, a;
	Uint16 pix;
	RGBA_FROM_8888(*src, sfmt, r, g, b, a);
	PIXEL_FROM_RGB(pix, dfmt, r, g, b);
	*d = ((pix & 0x7e0) << 16) | (pix & 0xf81f) | ((a << 2) & 0x7e0);
	src++;
	d++;
    }
    return n * 4;
}

/* encode 32bpp rgb + a into 32bpp G0RAB format for blitting into 555 */
static int copy_transl_555(void *dst, Uint32 *src, int n,
			   GAL_PixelFormat *sfmt, GAL_PixelFormat *dfmt)
{
    int i;
    Uint32 *d = dst;
    for(i = 0; i < n; i++) {
	unsigned r, g, b, a;
	Uint16 pix;
	RGBA_FROM_8888(*src, sfmt, r, g, b, a);
	PIXEL_FROM_RGB(pix, dfmt, r, g, b);
	*d = ((pix & 0x3e0) << 16) | (pix & 0xfc1f) | ((a << 2) & 0x3e0);
	src++;
	d++;
    }
    return n * 4;
}

/* decode translucent pixels from 32bpp GORAB to 32bpp rgb + a */
static int uncopy_transl_16(Uint32 *dst, void *src, int n,
			    RLEDestFormat *sfmt, GAL_PixelFormat *dfmt)
{
    int i;
    Uint32 *s = src;
    for(i = 0; i < n; i++) {
	unsigned r, g, b, a;
	Uint32 pix = *s++;
	a = (pix & 0x3e0) >> 2;
	pix = (pix & ~0x3e0) | pix >> 16;
	RGB_FROM_PIXEL(pix, sfmt, r, g, b);
	PIXEL_FROM_RGBA(*dst, dfmt, r, g, b, a);
	dst++;
    }
    return n * 4;
}

/* encode 32bpp rgba into 32bpp rgba, keeping alpha (dual purpose) */
static int copy_32(void *dst, Uint32 *src, int n,
		   GAL_PixelFormat *sfmt, GAL_PixelFormat *dfmt)
{
    int i;
    Uint32 *d = dst;
    for(i = 0; i < n; i++) {
	unsigned r, g, b, a;
	Uint32 pixel;
	RGBA_FROM_8888(*src, sfmt, r, g, b, a);
	PIXEL_FROM_RGB(pixel, dfmt, r, g, b);
	*d++ = pixel | a << 24;
	src++;
    }
    return n * 4;
}

/* decode 32bpp rgba into 32bpp rgba, keeping alpha (dual purpose) */
static int uncopy_32(Uint32 *dst, void *src, int n,
		     RLEDestFormat *sfmt, GAL_PixelFormat *dfmt)
{
    int i;
    Uint32 *s = src;
    for(i = 0; i < n; i++) {
	unsigned r, g, b, a;
	Uint32 pixel = *s++;
	RGB_FROM_PIXEL(pixel, sfmt, r, g, b);
	a = pixel >> 24;
	PIXEL_FROM_RGBA(*dst, dfmt, r, g, b, a);
	dst++;
    }
    return n * 4;
}

#define ISOPAQUE(pixel, fmt) ((((pixel) & fmt->Amask) >> fmt->Ashift) == 255)

#define ISTRANSL(pixel, fmt)	\
    ((unsigned)((((pixel) & fmt->Amask) >> fmt->Ashift) - 1U) < 254U)

/* convert surface to be quickly alpha-blittable onto dest, if possible */
static int RLEAlphaSurface(GAL_Surface *surface)
{
    GAL_Surface *dest;
    GAL_PixelFormat *df;
    int maxsize = 0;
    int max_opaque_run;
    int max_transl_run = 65535;
    unsigned masksum;
    Uint8 *rlebuf, *dst;
    int (*copy_opaque)(void *, Uint32 *, int,
		       GAL_PixelFormat *, GAL_PixelFormat *);
    int (*copy_transl)(void *, Uint32 *, int,
		       GAL_PixelFormat *, GAL_PixelFormat *);

    dest = surface->map->dst;
    if(!dest)
	return -1;
    df = dest->format;
    if(surface->format->BitsPerPixel != 32)