src/video/SDL_surface.c
author Sam Lantinga <slouken@libsdl.org>
Sun, 02 Sep 2012 16:05:29 -0700
branchSDL-1.2
changeset 6411 ec7ccf5ee3dd
parent 4159 a1b03ba2fcd0
permissions -rw-r--r--
Use fast path for RGB 565 -> 32-bit XRGB 8888

Hello Sam,
while profiling ScummVM I noticed it was making use of the generic
BlitNToN blitter, which struck me as odd because it should be a very
classical codepath.
After investigating, I saw that in the blit op chooser:

{ 0x0000F800,0x000007E0,0x0000001F, 4, 0x00FF0000,0x0000FF00,0x000000FF,
0, NULL, Blit_RGB565_ARGB8888, SET_ALPHA },
{ 0x0000F800,0x000007E0,0x0000001F, 4, 0x000000FF,0x0000FF00,0x00FF0000,
0, NULL, Blit_RGB565_ABGR8888, SET_ALPHA },
{ 0x0000F800,0x000007E0,0x0000001F, 4, 0xFF000000,0x00FF0000,0x0000FF00,
0, NULL, Blit_RGB565_RGBA8888, SET_ALPHA },
{ 0x0000F800,0x000007E0,0x0000001F, 4, 0x0000FF00,0x00FF0000,0xFF000000,
0, NULL, Blit_RGB565_BGRA8888, SET_ALPHA },

Couldn't the optimized versions be used for NO_ALPHA too? I take it
that the resulting alpha component can be undefined as it should never
be used.
I tried this (see attached patch) and it worked perfectly (and
therefore faster) on ScummVM but there might be a trick (I'm not
expert at the semantics of SDL, ie NO_ALPHA, SET_ALPHA and COPY_ALPHA
there).
What do you think?

Cheers,
Bertrand
     1 /*
     2     SDL - Simple DirectMedia Layer
     3     Copyright (C) 1997-2012 Sam Lantinga
     4 
     5     This library is free software; you can redistribute it and/or
     6     modify it under the terms of the GNU Lesser General Public
     7     License as published by the Free Software Foundation; either
     8     version 2.1 of the License, or (at your option) any later version.
     9 
    10     This library is distributed in the hope that it will be useful,
    11     but WITHOUT ANY WARRANTY; without even the implied warranty of
    12     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
    13     Lesser General Public License for more details.
    14 
    15     You should have received a copy of the GNU Lesser General Public
    16     License along with this library; if not, write to the Free Software
    17     Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
    18 
    19     Sam Lantinga
    20     slouken@libsdl.org
    21 */
    22 #include "SDL_config.h"
    23 
    24 #include "SDL_video.h"
    25 #include "SDL_sysvideo.h"
    26 #include "SDL_cursor_c.h"
    27 #include "SDL_blit.h"
    28 #include "SDL_RLEaccel_c.h"
    29 #include "SDL_pixels_c.h"
    30 #include "SDL_leaks.h"
    31 
    32 
    33 /* Public routines */
    34 /*
    35  * Create an empty RGB surface of the appropriate depth
    36  */
    37 SDL_Surface * SDL_CreateRGBSurface (Uint32 flags,
    38 			int width, int height, int depth,
    39 			Uint32 Rmask, Uint32 Gmask, Uint32 Bmask, Uint32 Amask)
    40 {
    41 	SDL_VideoDevice *video = current_video;
    42 	SDL_VideoDevice *this  = current_video;
    43 	SDL_Surface *screen;
    44 	SDL_Surface *surface;
    45 
    46 	/* Make sure the size requested doesn't overflow our datatypes */
    47 	/* Next time I write a library like SDL, I'll use int for size. :) */
    48 	if ( width >= 16384 || height >= 65536 ) {
    49 		SDL_SetError("Width or height is too large");
    50 		return(NULL);
    51 	}
    52 
    53 	/* Check to see if we desire the surface in video memory */
    54 	if ( video ) {
    55 		screen = SDL_PublicSurface;
    56 	} else {
    57 		screen = NULL;
    58 	}
    59 	if ( screen && ((screen->flags&SDL_HWSURFACE) == SDL_HWSURFACE) ) {
    60 		if ( (flags&(SDL_SRCCOLORKEY|SDL_SRCALPHA)) != 0 ) {
    61 			flags |= SDL_HWSURFACE;
    62 		}
    63 		if ( (flags & SDL_SRCCOLORKEY) == SDL_SRCCOLORKEY ) {
    64 			if ( ! current_video->info.blit_hw_CC ) {
    65 				flags &= ~SDL_HWSURFACE;
    66 			}
    67 		}
    68 		if ( (flags & SDL_SRCALPHA) == SDL_SRCALPHA ) {
    69 			if ( ! current_video->info.blit_hw_A ) {
    70 				flags &= ~SDL_HWSURFACE;
    71 			}
    72 		}
    73 	} else {
    74 		flags &= ~SDL_HWSURFACE;
    75 	}
    76 
    77 	/* Allocate the surface */
    78 	surface = (SDL_Surface *)SDL_malloc(sizeof(*surface));
    79 	if ( surface == NULL ) {
    80 		SDL_OutOfMemory();
    81 		return(NULL);
    82 	}
    83 	surface->flags = SDL_SWSURFACE;
    84 	if ( (flags & SDL_HWSURFACE) == SDL_HWSURFACE ) {
    85 		if ((Amask) && (video->displayformatalphapixel))
    86 		{
    87 			depth = video->displayformatalphapixel->BitsPerPixel;
    88 			Rmask = video->displayformatalphapixel->Rmask;
    89 			Gmask = video->displayformatalphapixel->Gmask;
    90 			Bmask = video->displayformatalphapixel->Bmask;
    91 			Amask = video->displayformatalphapixel->Amask;
    92 		}
    93 		else
    94 		{
    95 			depth = screen->format->BitsPerPixel;
    96 			Rmask = screen->format->Rmask;
    97 			Gmask = screen->format->Gmask;
    98 			Bmask = screen->format->Bmask;
    99 			Amask = screen->format->Amask;
   100 		}
   101 	}
   102 	surface->format = SDL_AllocFormat(depth, Rmask, Gmask, Bmask, Amask);
   103 	if ( surface->format == NULL ) {
   104 		SDL_free(surface);
   105 		return(NULL);
   106 	}
   107 	if ( Amask ) {
   108 		surface->flags |= SDL_SRCALPHA;
   109 	}
   110 	surface->w = width;
   111 	surface->h = height;
   112 	surface->pitch = SDL_CalculatePitch(surface);
   113 	surface->pixels = NULL;
   114 	surface->offset = 0;
   115 	surface->hwdata = NULL;
   116 	surface->locked = 0;
   117 	surface->map = NULL;
   118 	surface->unused1 = 0;
   119 	SDL_SetClipRect(surface, NULL);
   120 	SDL_FormatChanged(surface);
   121 
   122 	/* Get the pixels */
   123 	if ( ((flags&SDL_HWSURFACE) == SDL_SWSURFACE) || 
   124 				(video->AllocHWSurface(this, surface) < 0) ) {
   125 		if ( surface->w && surface->h ) {
   126 			surface->pixels = SDL_malloc(surface->h*surface->pitch);
   127 			if ( surface->pixels == NULL ) {
   128 				SDL_FreeSurface(surface);
   129 				SDL_OutOfMemory();
   130 				return(NULL);
   131 			}
   132 			/* This is important for bitmaps */
   133 			SDL_memset(surface->pixels, 0, surface->h*surface->pitch);
   134 		}
   135 	}
   136 
   137 	/* Allocate an empty mapping */
   138 	surface->map = SDL_AllocBlitMap();
   139 	if ( surface->map == NULL ) {
   140 		SDL_FreeSurface(surface);
   141 		return(NULL);
   142 	}
   143 
   144 	/* The surface is ready to go */
   145 	surface->refcount = 1;
   146 #ifdef CHECK_LEAKS
   147 	++surfaces_allocated;
   148 #endif
   149 	return(surface);
   150 }
   151 /*
   152  * Create an RGB surface from an existing memory buffer
   153  */
   154 SDL_Surface * SDL_CreateRGBSurfaceFrom (void *pixels,
   155 			int width, int height, int depth, int pitch,
   156 			Uint32 Rmask, Uint32 Gmask, Uint32 Bmask, Uint32 Amask)
   157 {
   158 	SDL_Surface *surface;
   159 
   160 	surface = SDL_CreateRGBSurface(SDL_SWSURFACE, 0, 0, depth,
   161 	                               Rmask, Gmask, Bmask, Amask);
   162 	if ( surface != NULL ) {
   163 		surface->flags |= SDL_PREALLOC;
   164 		surface->pixels = pixels;
   165 		surface->w = width;
   166 		surface->h = height;
   167 		surface->pitch = pitch;
   168 		SDL_SetClipRect(surface, NULL);
   169 	}
   170 	return(surface);
   171 }
   172 /*
   173  * Set the color key in a blittable surface
   174  */
   175 int SDL_SetColorKey (SDL_Surface *surface, Uint32 flag, Uint32 key)
   176 {
   177 	/* Sanity check the flag as it gets passed in */
   178 	if ( flag & SDL_SRCCOLORKEY ) {
   179 		if ( flag & (SDL_RLEACCEL|SDL_RLEACCELOK) ) {
   180 			flag = (SDL_SRCCOLORKEY | SDL_RLEACCELOK);
   181 		} else {
   182 			flag = SDL_SRCCOLORKEY;
   183 		}
   184 	} else {
   185 		flag = 0;
   186 	}
   187 
   188 	/* Optimize away operations that don't change anything */
   189 	if ( (flag == (surface->flags & (SDL_SRCCOLORKEY|SDL_RLEACCELOK))) &&
   190 	     (key == surface->format->colorkey) ) {
   191 		return(0);
   192 	}
   193 
   194 	/* UnRLE surfaces before we change the colorkey */
   195 	if ( surface->flags & SDL_RLEACCEL ) {
   196 	        SDL_UnRLESurface(surface, 1);
   197 	}
   198 
   199 	if ( flag ) {
   200 		SDL_VideoDevice *video = current_video;
   201 		SDL_VideoDevice *this  = current_video;
   202 
   203 
   204 		surface->flags |= SDL_SRCCOLORKEY;
   205 		surface->format->colorkey = key;
   206 		if ( (surface->flags & SDL_HWACCEL) == SDL_HWACCEL ) {
   207 			if ( (video->SetHWColorKey == NULL) ||
   208 			     (video->SetHWColorKey(this, surface, key) < 0) ) {
   209 				surface->flags &= ~SDL_HWACCEL;
   210 			}
   211 		}
   212 		if ( flag & SDL_RLEACCELOK ) {
   213 			surface->flags |= SDL_RLEACCELOK;
   214 		} else {
   215 			surface->flags &= ~SDL_RLEACCELOK;
   216 		}
   217 	} else {
   218 		surface->flags &= ~(SDL_SRCCOLORKEY|SDL_RLEACCELOK);
   219 		surface->format->colorkey = 0;
   220 	}
   221 	SDL_InvalidateMap(surface->map);
   222 	return(0);
   223 }
   224 /* This function sets the alpha channel of a surface */
   225 int SDL_SetAlpha (SDL_Surface *surface, Uint32 flag, Uint8 value)
   226 {
   227 	Uint32 oldflags = surface->flags;
   228 	Uint32 oldalpha = surface->format->alpha;
   229 
   230 	/* Sanity check the flag as it gets passed in */
   231 	if ( flag & SDL_SRCALPHA ) {
   232 		if ( flag & (SDL_RLEACCEL|SDL_RLEACCELOK) ) {
   233 			flag = (SDL_SRCALPHA | SDL_RLEACCELOK);
   234 		} else {
   235 			flag = SDL_SRCALPHA;
   236 		}
   237 	} else {
   238 		flag = 0;
   239 	}
   240 
   241 	/* Optimize away operations that don't change anything */
   242 	if ( (flag == (surface->flags & (SDL_SRCALPHA|SDL_RLEACCELOK))) &&
   243 	     (!flag || value == oldalpha) ) {
   244 		return(0);
   245 	}
   246 
   247 	if(!(flag & SDL_RLEACCELOK) && (surface->flags & SDL_RLEACCEL))
   248 		SDL_UnRLESurface(surface, 1);
   249 
   250 	if ( flag ) {
   251 		SDL_VideoDevice *video = current_video;
   252 		SDL_VideoDevice *this  = current_video;
   253 
   254 		surface->flags |= SDL_SRCALPHA;
   255 		surface->format->alpha = value;
   256 		if ( (surface->flags & SDL_HWACCEL) == SDL_HWACCEL ) {
   257 			if ( (video->SetHWAlpha == NULL) ||
   258 			     (video->SetHWAlpha(this, surface, value) < 0) ) {
   259 				surface->flags &= ~SDL_HWACCEL;
   260 			}
   261 		}
   262 		if ( flag & SDL_RLEACCELOK ) {
   263 		        surface->flags |= SDL_RLEACCELOK;
   264 		} else {
   265 		        surface->flags &= ~SDL_RLEACCELOK;
   266 		}
   267 	} else {
   268 		surface->flags &= ~SDL_SRCALPHA;
   269 		surface->format->alpha = SDL_ALPHA_OPAQUE;
   270 	}
   271 	/*
   272 	 * The representation for software surfaces is independent of
   273 	 * per-surface alpha, so no need to invalidate the blit mapping
   274 	 * if just the alpha value was changed. (If either is 255, we still
   275 	 * need to invalidate.)
   276 	 */
   277 	if((surface->flags & SDL_HWACCEL) == SDL_HWACCEL
   278 	   || oldflags != surface->flags
   279 	   || (((oldalpha + 1) ^ (value + 1)) & 0x100))
   280 		SDL_InvalidateMap(surface->map);
   281 	return(0);
   282 }
   283 int SDL_SetAlphaChannel(SDL_Surface *surface, Uint8 value)
   284 {
   285 	int row, col;
   286 	int offset;
   287 	Uint8 *buf;
   288 
   289 	if ( (surface->format->Amask != 0xFF000000) &&
   290 	     (surface->format->Amask != 0x000000FF) ) {
   291 		SDL_SetError("Unsupported surface alpha mask format");
   292 		return -1;
   293 	}
   294 
   295 #if SDL_BYTEORDER == SDL_LIL_ENDIAN
   296 	if ( surface->format->Amask == 0xFF000000 ) {
   297 			offset = 3;
   298 	} else {
   299 			offset = 0;
   300 	}
   301 #else
   302 	if ( surface->format->Amask == 0xFF000000 ) {
   303 			offset = 0;
   304 	} else {
   305 			offset = 3;
   306 	}
   307 #endif /* Byte ordering */
   308 
   309 	/* Quickly set the alpha channel of an RGBA or ARGB surface */
   310 	if ( SDL_MUSTLOCK(surface) ) {
   311 		if ( SDL_LockSurface(surface) < 0 ) {
   312 			return -1;
   313 		}
   314 	}
   315 	row = surface->h;
   316 	while (row--) {
   317 		col = surface->w;
   318 		buf = (Uint8 *)surface->pixels + row * surface->pitch + offset;
   319 		while(col--) {
   320 			*buf = value;
   321 			buf += 4;
   322 		}
   323 	}
   324 	if ( SDL_MUSTLOCK(surface) ) {
   325 		SDL_UnlockSurface(surface);
   326 	}
   327 	return 0;
   328 }
   329 
   330 /*
   331  * A function to calculate the intersection of two rectangles:
   332  * return true if the rectangles intersect, false otherwise
   333  */
   334 static __inline__
   335 SDL_bool SDL_IntersectRect(const SDL_Rect *A, const SDL_Rect *B, SDL_Rect *intersection)
   336 {
   337 	int Amin, Amax, Bmin, Bmax;
   338 
   339 	/* Horizontal intersection */
   340 	Amin = A->x;
   341 	Amax = Amin + A->w;
   342 	Bmin = B->x;
   343 	Bmax = Bmin + B->w;
   344 	if(Bmin > Amin)
   345 	        Amin = Bmin;
   346 	intersection->x = Amin;
   347 	if(Bmax < Amax)
   348 	        Amax = Bmax;
   349 	intersection->w = Amax - Amin > 0 ? Amax - Amin : 0;
   350 
   351 	/* Vertical intersection */
   352 	Amin = A->y;
   353 	Amax = Amin + A->h;
   354 	Bmin = B->y;
   355 	Bmax = Bmin + B->h;
   356 	if(Bmin > Amin)
   357 	        Amin = Bmin;
   358 	intersection->y = Amin;
   359 	if(Bmax < Amax)
   360 	        Amax = Bmax;
   361 	intersection->h = Amax - Amin > 0 ? Amax - Amin : 0;
   362 
   363 	return (intersection->w && intersection->h);
   364 }
   365 /*
   366  * Set the clipping rectangle for a blittable surface
   367  */
   368 SDL_bool SDL_SetClipRect(SDL_Surface *surface, const SDL_Rect *rect)
   369 {
   370 	SDL_Rect full_rect;
   371 
   372 	/* Don't do anything if there's no surface to act on */
   373 	if ( ! surface ) {
   374 		return SDL_FALSE;
   375 	}
   376 
   377 	/* Set up the full surface rectangle */
   378 	full_rect.x = 0;
   379 	full_rect.y = 0;
   380 	full_rect.w = surface->w;
   381 	full_rect.h = surface->h;
   382 
   383 	/* Set the clipping rectangle */
   384 	if ( ! rect ) {
   385 		surface->clip_rect = full_rect;
   386 		return 1;
   387 	}
   388 	return SDL_IntersectRect(rect, &full_rect, &surface->clip_rect);
   389 }
   390 void SDL_GetClipRect(SDL_Surface *surface, SDL_Rect *rect)
   391 {
   392 	if ( surface && rect ) {
   393 		*rect = surface->clip_rect;
   394 	}
   395 }
   396 /* 
   397  * Set up a blit between two surfaces -- split into three parts:
   398  * The upper part, SDL_UpperBlit(), performs clipping and rectangle 
   399  * verification.  The lower part is a pointer to a low level
   400  * accelerated blitting function.
   401  *
   402  * These parts are separated out and each used internally by this 
   403  * library in the optimimum places.  They are exported so that if
   404  * you know exactly what you are doing, you can optimize your code
   405  * by calling the one(s) you need.
   406  */
   407 int SDL_LowerBlit (SDL_Surface *src, SDL_Rect *srcrect,
   408 				SDL_Surface *dst, SDL_Rect *dstrect)
   409 {
   410 	SDL_blit do_blit;
   411 	SDL_Rect hw_srcrect;
   412 	SDL_Rect hw_dstrect;
   413 
   414 	/* Check to make sure the blit mapping is valid */
   415 	if ( (src->map->dst != dst) ||
   416              (src->map->dst->format_version != src->map->format_version) ) {
   417 		if ( SDL_MapSurface(src, dst) < 0 ) {
   418 			return(-1);
   419 		}
   420 	}
   421 
   422 	/* Figure out which blitter to use */
   423 	if ( (src->flags & SDL_HWACCEL) == SDL_HWACCEL ) {
   424 		if ( src == SDL_VideoSurface ) {
   425 			hw_srcrect = *srcrect;
   426 			hw_srcrect.x += current_video->offset_x;
   427 			hw_srcrect.y += current_video->offset_y;
   428 			srcrect = &hw_srcrect;
   429 		}
   430 		if ( dst == SDL_VideoSurface ) {
   431 			hw_dstrect = *dstrect;
   432 			hw_dstrect.x += current_video->offset_x;
   433 			hw_dstrect.y += current_video->offset_y;
   434 			dstrect = &hw_dstrect;
   435 		}
   436 		do_blit = src->map->hw_blit;
   437 	} else {
   438 		do_blit = src->map->sw_blit;
   439 	}
   440 	return(do_blit(src, srcrect, dst, dstrect));
   441 }
   442 
   443 
   444 int SDL_UpperBlit (SDL_Surface *src, SDL_Rect *srcrect,
   445 		   SDL_Surface *dst, SDL_Rect *dstrect)
   446 {
   447         SDL_Rect fulldst;
   448 	int srcx, srcy, w, h;
   449 
   450 	/* Make sure the surfaces aren't locked */
   451 	if ( ! src || ! dst ) {
   452 		SDL_SetError("SDL_UpperBlit: passed a NULL surface");
   453 		return(-1);
   454 	}
   455 	if ( src->locked || dst->locked ) {
   456 		SDL_SetError("Surfaces must not be locked during blit");
   457 		return(-1);
   458 	}
   459 
   460 	/* If the destination rectangle is NULL, use the entire dest surface */
   461 	if ( dstrect == NULL ) {
   462 	        fulldst.x = fulldst.y = 0;
   463 		dstrect = &fulldst;
   464 	}
   465 
   466 	/* clip the source rectangle to the source surface */
   467 	if(srcrect) {
   468 	        int maxw, maxh;
   469 	
   470 		srcx = srcrect->x;
   471 		w = srcrect->w;
   472 		if(srcx < 0) {
   473 		        w += srcx;
   474 			dstrect->x -= srcx;
   475 			srcx = 0;
   476 		}
   477 		maxw = src->w - srcx;
   478 		if(maxw < w)
   479 			w = maxw;
   480 
   481 		srcy = srcrect->y;
   482 		h = srcrect->h;
   483 		if(srcy < 0) {
   484 		        h += srcy;
   485 			dstrect->y -= srcy;
   486 			srcy = 0;
   487 		}
   488 		maxh = src->h - srcy;
   489 		if(maxh < h)
   490 			h = maxh;
   491 	    
   492 	} else {
   493 	        srcx = srcy = 0;
   494 		w = src->w;
   495 		h = src->h;
   496 	}
   497 
   498 	/* clip the destination rectangle against the clip rectangle */
   499 	{
   500 	        SDL_Rect *clip = &dst->clip_rect;
   501 		int dx, dy;
   502 
   503 		dx = clip->x - dstrect->x;
   504 		if(dx > 0) {
   505 			w -= dx;
   506 			dstrect->x += dx;
   507 			srcx += dx;
   508 		}
   509 		dx = dstrect->x + w - clip->x - clip->w;
   510 		if(dx > 0)
   511 			w -= dx;
   512 
   513 		dy = clip->y - dstrect->y;
   514 		if(dy > 0) {
   515 			h -= dy;
   516 			dstrect->y += dy;
   517 			srcy += dy;
   518 		}
   519 		dy = dstrect->y + h - clip->y - clip->h;
   520 		if(dy > 0)
   521 			h -= dy;
   522 	}
   523 
   524 	if(w > 0 && h > 0) {
   525 	        SDL_Rect sr;
   526 	        sr.x = srcx;
   527 		sr.y = srcy;
   528 		sr.w = dstrect->w = w;
   529 		sr.h = dstrect->h = h;
   530 		return SDL_LowerBlit(src, &sr, dst, dstrect);
   531 	}
   532 	dstrect->w = dstrect->h = 0;
   533 	return 0;
   534 }
   535 
   536 static int SDL_FillRect1(SDL_Surface *dst, SDL_Rect *dstrect, Uint32 color)
   537 {
   538 	/* FIXME: We have to worry about packing order.. *sigh* */
   539 	SDL_SetError("1-bpp rect fill not yet implemented");
   540 	return -1;
   541 }
   542 
   543 static int SDL_FillRect4(SDL_Surface *dst, SDL_Rect *dstrect, Uint32 color)
   544 {
   545 	/* FIXME: We have to worry about packing order.. *sigh* */
   546 	SDL_SetError("4-bpp rect fill not yet implemented");
   547 	return -1;
   548 }
   549 
   550 /* 
   551  * This function performs a fast fill of the given rectangle with 'color'
   552  */
   553 int SDL_FillRect(SDL_Surface *dst, SDL_Rect *dstrect, Uint32 color)
   554 {
   555 	SDL_VideoDevice *video = current_video;
   556 	SDL_VideoDevice *this  = current_video;
   557 	int x, y;
   558 	Uint8 *row;
   559 
   560 	/* This function doesn't work on surfaces < 8 bpp */
   561 	if ( dst->format->BitsPerPixel < 8 ) {
   562 		switch(dst->format->BitsPerPixel) {
   563 		    case 1:
   564 			return SDL_FillRect1(dst, dstrect, color);
   565 			break;
   566 		    case 4:
   567 			return SDL_FillRect4(dst, dstrect, color);
   568 			break;
   569 		    default:
   570 			SDL_SetError("Fill rect on unsupported surface format");
   571 			return(-1);
   572 			break;
   573 		}
   574 	}
   575 
   576 	/* If 'dstrect' == NULL, then fill the whole surface */
   577 	if ( dstrect ) {
   578 		/* Perform clipping */
   579 		if ( !SDL_IntersectRect(dstrect, &dst->clip_rect, dstrect) ) {
   580 			return(0);
   581 		}
   582 	} else {
   583 		dstrect = &dst->clip_rect;
   584 	}
   585 
   586 	/* Check for hardware acceleration */
   587 	if ( ((dst->flags & SDL_HWSURFACE) == SDL_HWSURFACE) &&
   588 					video->info.blit_fill ) {
   589 		SDL_Rect hw_rect;
   590 		if ( dst == SDL_VideoSurface ) {
   591 			hw_rect = *dstrect;
   592 			hw_rect.x += current_video->offset_x;
   593 			hw_rect.y += current_video->offset_y;
   594 			dstrect = &hw_rect;
   595 		}
   596 		return(video->FillHWRect(this, dst, dstrect, color));
   597 	}
   598 
   599 	/* Perform software fill */
   600 	if ( SDL_LockSurface(dst) != 0 ) {
   601 		return(-1);
   602 	}
   603 	row = (Uint8 *)dst->pixels+dstrect->y*dst->pitch+
   604 			dstrect->x*dst->format->BytesPerPixel;
   605 	if ( dst->format->palette || (color == 0) ) {
   606 		x = dstrect->w*dst->format->BytesPerPixel;
   607 		if ( !color && !((uintptr_t)row&3) && !(x&3) && !(dst->pitch&3) ) {
   608 			int n = x >> 2;
   609 			for ( y=dstrect->h; y; --y ) {
   610 				SDL_memset4(row, 0, n);
   611 				row += dst->pitch;
   612 			}
   613 		} else {
   614 #ifdef __powerpc__
   615 			/*
   616 			 * SDL_memset() on PPC (both glibc and codewarrior) uses
   617 			 * the dcbz (Data Cache Block Zero) instruction, which
   618 			 * causes an alignment exception if the destination is
   619 			 * uncachable, so only use it on software surfaces
   620 			 */
   621 			if((dst->flags & SDL_HWSURFACE) == SDL_HWSURFACE) {
   622 				if(dstrect->w >= 8) {
   623 					/*
   624 					 * 64-bit stores are probably most
   625 					 * efficient to uncached video memory
   626 					 */
   627 					double fill;
   628 					SDL_memset(&fill, color, (sizeof fill));
   629 					for(y = dstrect->h; y; y--) {
   630 						Uint8 *d = row;
   631 						unsigned n = x;
   632 						unsigned nn;
   633 						Uint8 c = color;
   634 						double f = fill;
   635 						while((unsigned long)d
   636 						      & (sizeof(double) - 1)) {
   637 							*d++ = c;
   638 							n--;
   639 						}
   640 						nn = n / (sizeof(double) * 4);
   641 						while(nn) {
   642 							((double *)d)[0] = f;
   643 							((double *)d)[1] = f;
   644 							((double *)d)[2] = f;
   645 							((double *)d)[3] = f;
   646 							d += 4*sizeof(double);
   647 							nn--;
   648 						}
   649 						n &= ~(sizeof(double) * 4 - 1);
   650 						nn = n / sizeof(double);
   651 						while(nn) {
   652 							*(double *)d = f;
   653 							d += sizeof(double);
   654 							nn--;
   655 						}
   656 						n &= ~(sizeof(double) - 1);
   657 						while(n) {
   658 							*d++ = c;
   659 							n--;
   660 						}
   661 						row += dst->pitch;
   662 					}
   663 				} else {
   664 					/* narrow boxes */
   665 					for(y = dstrect->h; y; y--) {
   666 						Uint8 *d = row;
   667 						Uint8 c = color;
   668 						int n = x;
   669 						while(n) {
   670 							*d++ = c;
   671 							n--;
   672 						}
   673 						row += dst->pitch;
   674 					}
   675 				}
   676 			} else
   677 #endif /* __powerpc__ */
   678 			{
   679 				for(y = dstrect->h; y; y--) {
   680 					SDL_memset(row, color, x);
   681 					row += dst->pitch;
   682 				}
   683 			}
   684 		}
   685 	} else {
   686 		switch (dst->format->BytesPerPixel) {
   687 		    case 2:
   688 			for ( y=dstrect->h; y; --y ) {
   689 				Uint16 *pixels = (Uint16 *)row;
   690 				Uint16 c = (Uint16)color;
   691 				Uint32 cc = (Uint32)c << 16 | c;
   692 				int n = dstrect->w;
   693 				if((uintptr_t)pixels & 3) {
   694 					*pixels++ = c;
   695 					n--;
   696 				}
   697 				if(n >> 1)
   698 					SDL_memset4(pixels, cc, n >> 1);
   699 				if(n & 1)
   700 					pixels[n - 1] = c;
   701 				row += dst->pitch;
   702 			}
   703 			break;
   704 
   705 		    case 3:
   706 			#if SDL_BYTEORDER == SDL_BIG_ENDIAN
   707 				color <<= 8;
   708 			#endif
   709 			for ( y=dstrect->h; y; --y ) {
   710 				Uint8 *pixels = row;
   711 				for ( x=dstrect->w; x; --x ) {
   712 					SDL_memcpy(pixels, &color, 3);
   713 					pixels += 3;
   714 				}
   715 				row += dst->pitch;
   716 			}
   717 			break;
   718 
   719 		    case 4:
   720 			for(y = dstrect->h; y; --y) {
   721 				SDL_memset4(row, color, dstrect->w);
   722 				row += dst->pitch;
   723 			}
   724 			break;
   725 		}
   726 	}
   727 	SDL_UnlockSurface(dst);
   728 
   729 	/* We're done! */
   730 	return(0);
   731 }
   732 
   733 /*
   734  * Lock a surface to directly access the pixels
   735  */
   736 int SDL_LockSurface (SDL_Surface *surface)
   737 {
   738 	if ( ! surface->locked ) {
   739 		/* Perform the lock */
   740 		if ( surface->flags & (SDL_HWSURFACE|SDL_ASYNCBLIT) ) {
   741 			SDL_VideoDevice *video = current_video;
   742 			SDL_VideoDevice *this  = current_video;
   743 			if ( video->LockHWSurface(this, surface) < 0 ) {
   744 				return(-1);
   745 			}
   746 		}
   747 		if ( surface->flags & SDL_RLEACCEL ) {
   748 			SDL_UnRLESurface(surface, 1);
   749 			surface->flags |= SDL_RLEACCEL;	/* save accel'd state */
   750 		}
   751 		/* This needs to be done here in case pixels changes value */
   752 		surface->pixels = (Uint8 *)surface->pixels + surface->offset;
   753 	}
   754 
   755 	/* Increment the surface lock count, for recursive locks */
   756 	++surface->locked;
   757 
   758 	/* Ready to go.. */
   759 	return(0);
   760 }
   761 /*
   762  * Unlock a previously locked surface
   763  */
   764 void SDL_UnlockSurface (SDL_Surface *surface)
   765 {
   766 	/* Only perform an unlock if we are locked */
   767 	if ( ! surface->locked || (--surface->locked > 0) ) {
   768 		return;
   769 	}
   770 
   771 	/* Perform the unlock */
   772 	surface->pixels = (Uint8 *)surface->pixels - surface->offset;
   773 
   774 	/* Unlock hardware or accelerated surfaces */
   775 	if ( surface->flags & (SDL_HWSURFACE|SDL_ASYNCBLIT) ) {
   776 		SDL_VideoDevice *video = current_video;
   777 		SDL_VideoDevice *this  = current_video;
   778 		video->UnlockHWSurface(this, surface);
   779 	} else {
   780 		/* Update RLE encoded surface with new data */
   781 		if ( (surface->flags & SDL_RLEACCEL) == SDL_RLEACCEL ) {
   782 		        surface->flags &= ~SDL_RLEACCEL; /* stop lying */
   783 			SDL_RLESurface(surface);
   784 		}
   785 	}
   786 }
   787 
   788 /* 
   789  * Convert a surface into the specified pixel format.
   790  */
   791 SDL_Surface * SDL_ConvertSurface (SDL_Surface *surface,
   792 					SDL_PixelFormat *format, Uint32 flags)
   793 {
   794 	SDL_Surface *convert;
   795 	Uint32 colorkey = 0;
   796 	Uint8 alpha = 0;
   797 	Uint32 surface_flags;
   798 	SDL_Rect bounds;
   799 
   800 	/* Check for empty destination palette! (results in empty image) */
   801 	if ( format->palette != NULL ) {
   802 		int i;
   803 		for ( i=0; i<format->palette->ncolors; ++i ) {
   804 			if ( (format->palette->colors[i].r != 0) ||
   805 			     (format->palette->colors[i].g != 0) ||
   806 			     (format->palette->colors[i].b != 0) )
   807 				break;
   808 		}
   809 		if ( i == format->palette->ncolors ) {
   810 			SDL_SetError("Empty destination palette");
   811 			return(NULL);
   812 		}
   813 	}
   814 
   815 	/* Only create hw surfaces with alpha channel if hw alpha blits
   816 	   are supported */
   817 	if(format->Amask != 0 && (flags & SDL_HWSURFACE)) {
   818 		const SDL_VideoInfo *vi = SDL_GetVideoInfo();
   819 		if(!vi || !vi->blit_hw_A)
   820 			flags &= ~SDL_HWSURFACE;
   821 	}
   822 
   823 	/* Create a new surface with the desired format */
   824 	convert = SDL_CreateRGBSurface(flags,
   825 				surface->w, surface->h, format->BitsPerPixel,
   826 		format->Rmask, format->Gmask, format->Bmask, format->Amask);
   827 	if ( convert == NULL ) {
   828 		return(NULL);
   829 	}
   830 
   831 	/* Copy the palette if any */
   832 	if ( format->palette && convert->format->palette ) {
   833 		SDL_memcpy(convert->format->palette->colors,
   834 				format->palette->colors,
   835 				format->palette->ncolors*sizeof(SDL_Color));
   836 		convert->format->palette->ncolors = format->palette->ncolors;
   837 	}
   838 
   839 	/* Save the original surface color key and alpha */
   840 	surface_flags = surface->flags;
   841 	if ( (surface_flags & SDL_SRCCOLORKEY) == SDL_SRCCOLORKEY ) {
   842 		/* Convert colourkeyed surfaces to RGBA if requested */
   843 		if((flags & SDL_SRCCOLORKEY) != SDL_SRCCOLORKEY
   844 		   && format->Amask) {
   845 			surface_flags &= ~SDL_SRCCOLORKEY;
   846 		} else {
   847 			colorkey = surface->format->colorkey;
   848 			SDL_SetColorKey(surface, 0, 0);
   849 		}
   850 	}
   851 	if ( (surface_flags & SDL_SRCALPHA) == SDL_SRCALPHA ) {
   852 		/* Copy over the alpha channel to RGBA if requested */
   853 		if ( format->Amask ) {
   854 			surface->flags &= ~SDL_SRCALPHA;
   855 		} else {
   856 			alpha = surface->format->alpha;
   857 			SDL_SetAlpha(surface, 0, 0);
   858 		}
   859 	}
   860 
   861 	/* Copy over the image data */
   862 	bounds.x = 0;
   863 	bounds.y = 0;
   864 	bounds.w = surface->w;
   865 	bounds.h = surface->h;
   866 	SDL_LowerBlit(surface, &bounds, convert, &bounds);
   867 
   868 	/* Clean up the original surface, and update converted surface */
   869 	if ( convert != NULL ) {
   870 		SDL_SetClipRect(convert, &surface->clip_rect);
   871 	}
   872 	if ( (surface_flags & SDL_SRCCOLORKEY) == SDL_SRCCOLORKEY ) {
   873 		Uint32 cflags = surface_flags&(SDL_SRCCOLORKEY|SDL_RLEACCELOK);
   874 		if ( convert != NULL ) {
   875 			Uint8 keyR, keyG, keyB;
   876 
   877 			SDL_GetRGB(colorkey,surface->format,&keyR,&keyG,&keyB);
   878 			SDL_SetColorKey(convert, cflags|(flags&SDL_RLEACCELOK),
   879 				SDL_MapRGB(convert->format, keyR, keyG, keyB));
   880 		}
   881 		SDL_SetColorKey(surface, cflags, colorkey);
   882 	}
   883 	if ( (surface_flags & SDL_SRCALPHA) == SDL_SRCALPHA ) {
   884 		Uint32 aflags = surface_flags&(SDL_SRCALPHA|SDL_RLEACCELOK);
   885 		if ( convert != NULL ) {
   886 		        SDL_SetAlpha(convert, aflags|(flags&SDL_RLEACCELOK),
   887 				alpha);
   888 		}
   889 		if ( format->Amask ) {
   890 			surface->flags |= SDL_SRCALPHA;
   891 		} else {
   892 			SDL_SetAlpha(surface, aflags, alpha);
   893 		}
   894 	}
   895 
   896 	/* We're ready to go! */
   897 	return(convert);
   898 }
   899 
   900 /*
   901  * Free a surface created by the above function.
   902  */
   903 void SDL_FreeSurface (SDL_Surface *surface)
   904 {
   905 	/* Free anything that's not NULL, and not the screen surface */
   906 	if ((surface == NULL) ||
   907 	    (current_video &&
   908 	    ((surface == SDL_ShadowSurface)||(surface == SDL_VideoSurface)))) {
   909 		return;
   910 	}
   911 	if ( --surface->refcount > 0 ) {
   912 		return;
   913 	}
   914 	while ( surface->locked > 0 ) {
   915 		SDL_UnlockSurface(surface);
   916 	}
   917 	if ( (surface->flags & SDL_RLEACCEL) == SDL_RLEACCEL ) {
   918 	        SDL_UnRLESurface(surface, 0);
   919 	}
   920 	if ( surface->format ) {
   921 		SDL_FreeFormat(surface->format);
   922 		surface->format = NULL;
   923 	}
   924 	if ( surface->map != NULL ) {
   925 		SDL_FreeBlitMap(surface->map);
   926 		surface->map = NULL;
   927 	}
   928 	if ( surface->hwdata ) {
   929 		SDL_VideoDevice *video = current_video;
   930 		SDL_VideoDevice *this  = current_video;
   931 		video->FreeHWSurface(this, surface);
   932 	}
   933 	if ( surface->pixels &&
   934 	     ((surface->flags & SDL_PREALLOC) != SDL_PREALLOC) ) {
   935 		SDL_free(surface->pixels);
   936 	}
   937 	SDL_free(surface);
   938 #ifdef CHECK_LEAKS
   939 	--surfaces_allocated;
   940 #endif
   941 }