src/video/SDL_stretch.c
author Ryan C. Gordon <icculus@icculus.org>
Fri, 06 Jan 2006 00:56:12 +0000
changeset 1233 228c94f0b5dc
parent 1228 f4a3a4129d04
child 1234 73676c1f56ee
permissions -rw-r--r--
Patched to compile.
slouken@0
     1
/*
slouken@0
     2
    SDL - Simple DirectMedia Layer
slouken@769
     3
    Copyright (C) 1997-2004 Sam Lantinga
slouken@0
     4
slouken@0
     5
    This library is free software; you can redistribute it and/or
slouken@0
     6
    modify it under the terms of the GNU Library General Public
slouken@0
     7
    License as published by the Free Software Foundation; either
slouken@0
     8
    version 2 of the License, or (at your option) any later version.
slouken@0
     9
slouken@0
    10
    This library is distributed in the hope that it will be useful,
slouken@0
    11
    but WITHOUT ANY WARRANTY; without even the implied warranty of
slouken@0
    12
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
slouken@0
    13
    Library General Public License for more details.
slouken@0
    14
slouken@0
    15
    You should have received a copy of the GNU Library General Public
slouken@0
    16
    License along with this library; if not, write to the Free
slouken@0
    17
    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
slouken@0
    18
slouken@0
    19
    Sam Lantinga
slouken@252
    20
    slouken@libsdl.org
slouken@0
    21
*/
slouken@0
    22
slouken@0
    23
#ifdef SAVE_RCSID
slouken@0
    24
static char rcsid =
slouken@0
    25
 "@(#) $Id$";
slouken@0
    26
#endif
slouken@0
    27
slouken@0
    28
/* This a stretch blit implementation based on ideas given to me by
slouken@0
    29
   Tomasz Cejner - thanks! :)
slouken@0
    30
slouken@0
    31
   April 27, 2000 - Sam Lantinga
slouken@0
    32
*/
slouken@0
    33
slouken@0
    34
#include "SDL_error.h"
slouken@0
    35
#include "SDL_video.h"
slouken@0
    36
#include "SDL_blit.h"
slouken@0
    37
slouken@0
    38
/* This isn't ready for general consumption yet - it should be folded
slouken@0
    39
   into the general blitting mechanism.
slouken@0
    40
*/
slouken@0
    41
slouken@288
    42
#if (defined(WIN32) && !defined(_M_ALPHA) && !defined(_WIN32_WCE) && \
slouken@288
    43
     !defined(__WATCOMC__) && !defined(__LCC__) && !defined(__FREEBCC__)) || \
slouken@288
    44
    (defined(i386) && defined(__GNUC__) && defined(USE_ASMBLIT))
slouken@0
    45
#define USE_ASM_STRETCH
slouken@0
    46
#endif
slouken@0
    47
slouken@0
    48
#ifdef USE_ASM_STRETCH
slouken@0
    49
slouken@0
    50
#if defined(WIN32) || defined(i386)
slouken@0
    51
#define PREFIX16	0x66
slouken@0
    52
#define STORE_BYTE	0xAA
slouken@0
    53
#define STORE_WORD	0xAB
slouken@0
    54
#define LOAD_BYTE	0xAC
slouken@0
    55
#define LOAD_WORD	0xAD
slouken@0
    56
#define RETURN		0xC3
slouken@0
    57
#else
slouken@0
    58
#error Need assembly opcodes for this architecture
slouken@0
    59
#endif
slouken@0
    60
slouken@0
    61
static unsigned char copy_row[4096];
slouken@0
    62
slouken@0
    63
static int generate_rowbytes(int src_w, int dst_w, int bpp)
slouken@0
    64
{
slouken@0
    65
	static struct {
slouken@0
    66
		int bpp;
slouken@0
    67
		int src_w;
slouken@0
    68
		int dst_w;
slouken@0
    69
	} last;
slouken@0
    70
slouken@0
    71
	int i;
slouken@0
    72
	int pos, inc;
slouken@0
    73
	unsigned char *eip;
slouken@0
    74
	unsigned char load, store;
slouken@0
    75
slouken@0
    76
	/* See if we need to regenerate the copy buffer */
slouken@0
    77
	if ( (src_w == last.src_w) &&
icculus@1164
    78
	     (dst_w == last.dst_w) && (bpp == last.bpp) ) {
slouken@0
    79
		return(0);
slouken@0
    80
	}
slouken@0
    81
	last.bpp = bpp;
slouken@0
    82
	last.src_w = src_w;
slouken@0
    83
	last.dst_w = dst_w;
slouken@0
    84
slouken@0
    85
	switch (bpp) {
slouken@0
    86
	    case 1:
slouken@0
    87
		load = LOAD_BYTE;
slouken@0
    88
		store = STORE_BYTE;
slouken@0
    89
		break;
slouken@0
    90
	    case 2:
slouken@0
    91
	    case 4:
slouken@0
    92
		load = LOAD_WORD;
slouken@0
    93
		store = STORE_WORD;
slouken@0
    94
		break;
slouken@0
    95
	    default:
slouken@0
    96
		SDL_SetError("ASM stretch of %d bytes isn't supported\n", bpp);
slouken@0
    97
		return(-1);
slouken@0
    98
	}
slouken@0
    99
	pos = 0x10000;
slouken@0
   100
	inc = (src_w << 16) / dst_w;
slouken@0
   101
	eip = copy_row;
slouken@0
   102
	for ( i=0; i<dst_w; ++i ) {
slouken@0
   103
		while ( pos >= 0x10000L ) {
slouken@0
   104
			if ( bpp == 2 ) {
slouken@0
   105
				*eip++ = PREFIX16;
slouken@0
   106
			}
slouken@0
   107
			*eip++ = load;
slouken@0
   108
			pos -= 0x10000L;
slouken@0
   109
		}
slouken@0
   110
		if ( bpp == 2 ) {
slouken@0
   111
			*eip++ = PREFIX16;
slouken@0
   112
		}
slouken@0
   113
		*eip++ = store;
slouken@0
   114
		pos += inc;
slouken@0
   115
	}
slouken@0
   116
	*eip++ = RETURN;
slouken@0
   117
slouken@0
   118
	/* Verify that we didn't overflow (too late) */
slouken@0
   119
	if ( eip > (copy_row+sizeof(copy_row)) ) {
slouken@0
   120
		SDL_SetError("Copy buffer overflow");
slouken@0
   121
		return(-1);
slouken@0
   122
	}
slouken@0
   123
	return(0);
slouken@0
   124
}
slouken@0
   125
slouken@0
   126
#else
slouken@0
   127
slouken@0
   128
#define DEFINE_COPY_ROW(name, type)			\
slouken@0
   129
void name(type *src, int src_w, type *dst, int dst_w)	\
slouken@0
   130
{							\
slouken@0
   131
	int i;						\
slouken@0
   132
	int pos, inc;					\
slouken@0
   133
	type pixel = 0;					\
slouken@0
   134
							\
slouken@0
   135
	pos = 0x10000;					\
slouken@0
   136
	inc = (src_w << 16) / dst_w;			\
slouken@0
   137
	for ( i=dst_w; i>0; --i ) {			\
slouken@0
   138
		while ( pos >= 0x10000L ) {		\
slouken@0
   139
			pixel = *src++;			\
slouken@0
   140
			pos -= 0x10000L;		\
slouken@0
   141
		}					\
slouken@0
   142
		*dst++ = pixel;				\
slouken@0
   143
		pos += inc;				\
slouken@0
   144
	}						\
slouken@0
   145
}
slouken@0
   146
DEFINE_COPY_ROW(copy_row1, Uint8)
slouken@0
   147
DEFINE_COPY_ROW(copy_row2, Uint16)
slouken@0
   148
DEFINE_COPY_ROW(copy_row4, Uint32)
slouken@0
   149
slouken@0
   150
#endif /* USE_ASM_STRETCH */
slouken@0
   151
slouken@0
   152
/* The ASM code doesn't handle 24-bpp stretch blits */
slouken@0
   153
void copy_row3(Uint8 *src, int src_w, Uint8 *dst, int dst_w)
slouken@0
   154
{
slouken@0
   155
	int i;
slouken@0
   156
	int pos, inc;
slouken@0
   157
	Uint8 pixel[3];
slouken@0
   158
slouken@0
   159
	pos = 0x10000;
slouken@0
   160
	inc = (src_w << 16) / dst_w;
slouken@0
   161
	for ( i=dst_w; i>0; --i ) {
slouken@0
   162
		while ( pos >= 0x10000L ) {
slouken@0
   163
			pixel[0] = *src++;
slouken@0
   164
			pixel[1] = *src++;
slouken@0
   165
			pixel[2] = *src++;
slouken@0
   166
			pos -= 0x10000L;
slouken@0
   167
		}
slouken@0
   168
		*dst++ = pixel[0];
slouken@0
   169
		*dst++ = pixel[1];
slouken@0
   170
		*dst++ = pixel[2];
slouken@0
   171
		pos += inc;
slouken@0
   172
	}
slouken@0
   173
}
slouken@0
   174
slouken@0
   175
/* Perform a stretch blit between two surfaces of the same format.
slouken@0
   176
   NOTE:  This function is not safe to call from multiple threads!
slouken@0
   177
*/
slouken@0
   178
int SDL_SoftStretch(SDL_Surface *src, SDL_Rect *srcrect,
slouken@0
   179
                    SDL_Surface *dst, SDL_Rect *dstrect)
slouken@0
   180
{
slouken@894
   181
	int src_locked;
slouken@894
   182
	int dst_locked;
slouken@0
   183
	int pos, inc;
slouken@0
   184
	int dst_width;
slouken@0
   185
	int dst_maxrow;
slouken@0
   186
	int src_row, dst_row;
slouken@0
   187
	Uint8 *srcp = NULL;
slouken@0
   188
	Uint8 *dstp;
slouken@0
   189
	SDL_Rect full_src;
slouken@0
   190
	SDL_Rect full_dst;
slouken@0
   191
#if defined(USE_ASM_STRETCH) && defined(__GNUC__)
slouken@0
   192
	int u1, u2;
slouken@0
   193
#endif
slouken@0
   194
	const int bpp = dst->format->BytesPerPixel;
slouken@0
   195
slouken@0
   196
	if ( src->format->BitsPerPixel != dst->format->BitsPerPixel ) {
slouken@0
   197
		SDL_SetError("Only works with same format surfaces");
slouken@0
   198
		return(-1);
slouken@0
   199
	}
slouken@0
   200
slouken@0
   201
	/* Verify the blit rectangles */
slouken@0
   202
	if ( srcrect ) {
slouken@0
   203
		if ( (srcrect->x < 0) || (srcrect->y < 0) ||
slouken@0
   204
		     ((srcrect->x+srcrect->w) > src->w) ||
slouken@0
   205
		     ((srcrect->y+srcrect->h) > src->h) ) {
slouken@0
   206
			SDL_SetError("Invalid source blit rectangle");
slouken@0
   207
			return(-1);
slouken@0
   208
		}
slouken@0
   209
	} else {
slouken@0
   210
		full_src.x = 0;
slouken@0
   211
		full_src.y = 0;
slouken@0
   212
		full_src.w = src->w;
slouken@0
   213
		full_src.h = src->h;
slouken@0
   214
		srcrect = &full_src;
slouken@0
   215
	}
slouken@0
   216
	if ( dstrect ) {
slouken@0
   217
		if ( (dstrect->x < 0) || (dstrect->y < 0) ||
slouken@0
   218
		     ((dstrect->x+dstrect->w) > dst->w) ||
slouken@0
   219
		     ((dstrect->y+dstrect->h) > dst->h) ) {
slouken@0
   220
			SDL_SetError("Invalid destination blit rectangle");
slouken@0
   221
			return(-1);
slouken@0
   222
		}
slouken@0
   223
	} else {
slouken@0
   224
		full_dst.x = 0;
slouken@0
   225
		full_dst.y = 0;
slouken@0
   226
		full_dst.w = dst->w;
slouken@0
   227
		full_dst.h = dst->h;
slouken@0
   228
		dstrect = &full_dst;
slouken@0
   229
	}
slouken@0
   230
slouken@894
   231
	/* Lock the destination if it's in hardware */
slouken@894
   232
	dst_locked = 0;
slouken@894
   233
	if ( SDL_MUSTLOCK(dst) ) {
slouken@894
   234
		if ( SDL_LockSurface(dst) < 0 ) {
slouken@894
   235
			SDL_SetError("Unable to lock destination surface");
slouken@894
   236
			return(-1);
slouken@894
   237
		}
slouken@894
   238
		dst_locked = 1;
slouken@894
   239
	}
slouken@894
   240
	/* Lock the source if it's in hardware */
slouken@894
   241
	src_locked = 0;
slouken@894
   242
	if ( SDL_MUSTLOCK(src) ) {
slouken@894
   243
		if ( SDL_LockSurface(src) < 0 ) {
slouken@894
   244
			if ( dst_locked ) {
slouken@894
   245
				SDL_UnlockSurface(dst);
slouken@894
   246
			}
slouken@894
   247
			SDL_SetError("Unable to lock source surface");
slouken@894
   248
			return(-1);
slouken@894
   249
		}
slouken@894
   250
		src_locked = 1;
slouken@894
   251
	}
slouken@894
   252
slouken@0
   253
	/* Set up the data... */
slouken@0
   254
	pos = 0x10000;
slouken@0
   255
	inc = (srcrect->h << 16) / dstrect->h;
slouken@0
   256
	src_row = srcrect->y;
slouken@0
   257
	dst_row = dstrect->y;
slouken@0
   258
	dst_width = dstrect->w*bpp;
slouken@0
   259
slouken@0
   260
#ifdef USE_ASM_STRETCH
slouken@0
   261
	/* Write the opcodes for this stretch */
slouken@0
   262
	if ( (bpp != 3) &&
slouken@0
   263
	     (generate_rowbytes(srcrect->w, dstrect->w, bpp) < 0) ) {
slouken@0
   264
		return(-1);
slouken@0
   265
	}
slouken@0
   266
#endif
slouken@0
   267
slouken@0
   268
	/* Perform the stretch blit */
slouken@0
   269
	for ( dst_maxrow = dst_row+dstrect->h; dst_row<dst_maxrow; ++dst_row ) {
slouken@0
   270
		dstp = (Uint8 *)dst->pixels + (dst_row*dst->pitch)
slouken@0
   271
		                            + (dstrect->x*bpp);
slouken@0
   272
		while ( pos >= 0x10000L ) {
slouken@0
   273
			srcp = (Uint8 *)src->pixels + (src_row*src->pitch)
slouken@0
   274
			                            + (srcrect->x*bpp);
slouken@0
   275
			++src_row;
slouken@0
   276
			pos -= 0x10000L;
slouken@0
   277
		}
slouken@0
   278
#ifdef USE_ASM_STRETCH
slouken@0
   279
		switch (bpp) {
slouken@0
   280
		    case 3:
slouken@0
   281
			copy_row3(srcp, srcrect->w, dstp, dstrect->w);
slouken@0
   282
			break;
slouken@0
   283
		    default:
slouken@0
   284
#ifdef __GNUC__
slouken@627
   285
			__asm__ __volatile__ (
icculus@1228
   286
			"call *%4"
slouken@0
   287
			: "=&D" (u1), "=&S" (u2)
icculus@1233
   288
			: "0" (dstp), "1" (srcp), "r" (&copy_row)
slouken@0
   289
			: "memory" );
slouken@0
   290
#else
slouken@0
   291
#ifdef WIN32
slouken@0
   292
		{ void *code = &copy_row;
slouken@0
   293
			__asm {
slouken@0
   294
				push edi
slouken@0
   295
				push esi
slouken@0
   296
	
slouken@0
   297
				mov edi, dstp
slouken@0
   298
				mov esi, srcp
slouken@0
   299
				call dword ptr code
slouken@0
   300
slouken@0
   301
				pop esi
slouken@0
   302
				pop edi
slouken@0
   303
			}
slouken@0
   304
		}
slouken@0
   305
#else
slouken@0
   306
#error Need inline assembly for this compiler
slouken@0
   307
#endif
slouken@0
   308
#endif /* __GNUC__ */
slouken@0
   309
			break;
slouken@0
   310
		}
slouken@0
   311
#else
slouken@0
   312
		switch (bpp) {
slouken@0
   313
		    case 1:
slouken@0
   314
			copy_row1(srcp, srcrect->w, dstp, dstrect->w);
slouken@0
   315
			break;
slouken@0
   316
		    case 2:
slouken@0
   317
			copy_row2((Uint16 *)srcp, srcrect->w,
slouken@0
   318
			          (Uint16 *)dstp, dstrect->w);
slouken@0
   319
			break;
slouken@0
   320
		    case 3:
slouken@0
   321
			copy_row3(srcp, srcrect->w, dstp, dstrect->w);
slouken@0
   322
			break;
slouken@0
   323
		    case 4:
slouken@0
   324
			copy_row4((Uint32 *)srcp, srcrect->w,
slouken@0
   325
			          (Uint32 *)dstp, dstrect->w);
slouken@0
   326
			break;
slouken@0
   327
		}
slouken@0
   328
#endif
slouken@0
   329
		pos += inc;
slouken@0
   330
	}
slouken@894
   331
slouken@894
   332
	/* We need to unlock the surfaces if they're locked */
slouken@894
   333
	if ( dst_locked ) {
slouken@894
   334
		SDL_UnlockSurface(dst);
slouken@894
   335
	}
slouken@894
   336
	if ( src_locked ) {
slouken@894
   337
		SDL_UnlockSurface(src);
slouken@894
   338
	}
slouken@0
   339
	return(0);
slouken@0
   340
}
slouken@0
   341