src/video/SDL_stretch.c
author Sam Lantinga <slouken@libsdl.org>
Thu, 16 Feb 2006 10:11:48 +0000
changeset 1361 19418e4422cb
parent 1358 c71e05b4dc2e
child 1402 d910939febfa
permissions -rw-r--r--
New configure-based build system. Still work in progress, but much improved
slouken@0
     1
/*
slouken@0
     2
    SDL - Simple DirectMedia Layer
slouken@1312
     3
    Copyright (C) 1997-2006 Sam Lantinga
slouken@0
     4
slouken@0
     5
    This library is free software; you can redistribute it and/or
slouken@1312
     6
    modify it under the terms of the GNU Lesser General Public
slouken@0
     7
    License as published by the Free Software Foundation; either
slouken@1312
     8
    version 2.1 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@1312
    13
    Lesser General Public License for more details.
slouken@0
    14
slouken@1312
    15
    You should have received a copy of the GNU Lesser General Public
slouken@1312
    16
    License along with this library; if not, write to the Free Software
slouken@1312
    17
    Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
slouken@0
    18
slouken@0
    19
    Sam Lantinga
slouken@252
    20
    slouken@libsdl.org
slouken@0
    21
*/
slouken@0
    22
slouken@0
    23
/* This a stretch blit implementation based on ideas given to me by
slouken@0
    24
   Tomasz Cejner - thanks! :)
slouken@0
    25
slouken@0
    26
   April 27, 2000 - Sam Lantinga
slouken@0
    27
*/
slouken@0
    28
slouken@0
    29
#include "SDL_video.h"
slouken@0
    30
#include "SDL_blit.h"
slouken@0
    31
slouken@0
    32
/* This isn't ready for general consumption yet - it should be folded
slouken@0
    33
   into the general blitting mechanism.
slouken@0
    34
*/
slouken@0
    35
slouken@1361
    36
#if ((defined(_MFC_VER) && defined(_M_IX86)/* && !defined(_WIN32_WCE) still needed? */) || \
slouken@1361
    37
     (defined(i386) && defined(__GNUC__))) && SDL_ASSEMBLY_BLITTERS
slouken@0
    38
#define USE_ASM_STRETCH
slouken@0
    39
#endif
slouken@0
    40
slouken@0
    41
#ifdef USE_ASM_STRETCH
slouken@0
    42
slouken@1361
    43
#if defined(_M_IX86) || defined(i386)
slouken@0
    44
#define PREFIX16	0x66
slouken@0
    45
#define STORE_BYTE	0xAA
slouken@0
    46
#define STORE_WORD	0xAB
slouken@0
    47
#define LOAD_BYTE	0xAC
slouken@0
    48
#define LOAD_WORD	0xAD
slouken@0
    49
#define RETURN		0xC3
slouken@0
    50
#else
slouken@0
    51
#error Need assembly opcodes for this architecture
slouken@0
    52
#endif
slouken@0
    53
slouken@0
    54
static unsigned char copy_row[4096];
slouken@0
    55
slouken@0
    56
static int generate_rowbytes(int src_w, int dst_w, int bpp)
slouken@0
    57
{
slouken@0
    58
	static struct {
slouken@0
    59
		int bpp;
slouken@0
    60
		int src_w;
slouken@0
    61
		int dst_w;
slouken@0
    62
	} last;
slouken@0
    63
slouken@0
    64
	int i;
slouken@0
    65
	int pos, inc;
slouken@0
    66
	unsigned char *eip;
slouken@0
    67
	unsigned char load, store;
slouken@0
    68
slouken@0
    69
	/* See if we need to regenerate the copy buffer */
slouken@0
    70
	if ( (src_w == last.src_w) &&
icculus@1164
    71
	     (dst_w == last.dst_w) && (bpp == last.bpp) ) {
slouken@0
    72
		return(0);
slouken@0
    73
	}
slouken@0
    74
	last.bpp = bpp;
slouken@0
    75
	last.src_w = src_w;
slouken@0
    76
	last.dst_w = dst_w;
slouken@0
    77
slouken@0
    78
	switch (bpp) {
slouken@0
    79
	    case 1:
slouken@0
    80
		load = LOAD_BYTE;
slouken@0
    81
		store = STORE_BYTE;
slouken@0
    82
		break;
slouken@0
    83
	    case 2:
slouken@0
    84
	    case 4:
slouken@0
    85
		load = LOAD_WORD;
slouken@0
    86
		store = STORE_WORD;
slouken@0
    87
		break;
slouken@0
    88
	    default:
slouken@0
    89
		SDL_SetError("ASM stretch of %d bytes isn't supported\n", bpp);
slouken@0
    90
		return(-1);
slouken@0
    91
	}
slouken@0
    92
	pos = 0x10000;
slouken@0
    93
	inc = (src_w << 16) / dst_w;
slouken@0
    94
	eip = copy_row;
slouken@0
    95
	for ( i=0; i<dst_w; ++i ) {
slouken@0
    96
		while ( pos >= 0x10000L ) {
slouken@0
    97
			if ( bpp == 2 ) {
slouken@0
    98
				*eip++ = PREFIX16;
slouken@0
    99
			}
slouken@0
   100
			*eip++ = load;
slouken@0
   101
			pos -= 0x10000L;
slouken@0
   102
		}
slouken@0
   103
		if ( bpp == 2 ) {
slouken@0
   104
			*eip++ = PREFIX16;
slouken@0
   105
		}
slouken@0
   106
		*eip++ = store;
slouken@0
   107
		pos += inc;
slouken@0
   108
	}
slouken@0
   109
	*eip++ = RETURN;
slouken@0
   110
slouken@0
   111
	/* Verify that we didn't overflow (too late) */
slouken@0
   112
	if ( eip > (copy_row+sizeof(copy_row)) ) {
slouken@0
   113
		SDL_SetError("Copy buffer overflow");
slouken@0
   114
		return(-1);
slouken@0
   115
	}
slouken@0
   116
	return(0);
slouken@0
   117
}
slouken@0
   118
slouken@0
   119
#else
slouken@0
   120
slouken@0
   121
#define DEFINE_COPY_ROW(name, type)			\
slouken@0
   122
void name(type *src, int src_w, type *dst, int dst_w)	\
slouken@0
   123
{							\
slouken@0
   124
	int i;						\
slouken@0
   125
	int pos, inc;					\
slouken@0
   126
	type pixel = 0;					\
slouken@0
   127
							\
slouken@0
   128
	pos = 0x10000;					\
slouken@0
   129
	inc = (src_w << 16) / dst_w;			\
slouken@0
   130
	for ( i=dst_w; i>0; --i ) {			\
slouken@0
   131
		while ( pos >= 0x10000L ) {		\
slouken@0
   132
			pixel = *src++;			\
slouken@0
   133
			pos -= 0x10000L;		\
slouken@0
   134
		}					\
slouken@0
   135
		*dst++ = pixel;				\
slouken@0
   136
		pos += inc;				\
slouken@0
   137
	}						\
slouken@0
   138
}
slouken@0
   139
DEFINE_COPY_ROW(copy_row1, Uint8)
slouken@0
   140
DEFINE_COPY_ROW(copy_row2, Uint16)
slouken@0
   141
DEFINE_COPY_ROW(copy_row4, Uint32)
slouken@0
   142
slouken@0
   143
#endif /* USE_ASM_STRETCH */
slouken@0
   144
slouken@0
   145
/* The ASM code doesn't handle 24-bpp stretch blits */
slouken@0
   146
void copy_row3(Uint8 *src, int src_w, Uint8 *dst, int dst_w)
slouken@0
   147
{
slouken@0
   148
	int i;
slouken@0
   149
	int pos, inc;
slouken@0
   150
	Uint8 pixel[3];
slouken@0
   151
slouken@0
   152
	pos = 0x10000;
slouken@0
   153
	inc = (src_w << 16) / dst_w;
slouken@0
   154
	for ( i=dst_w; i>0; --i ) {
slouken@0
   155
		while ( pos >= 0x10000L ) {
slouken@0
   156
			pixel[0] = *src++;
slouken@0
   157
			pixel[1] = *src++;
slouken@0
   158
			pixel[2] = *src++;
slouken@0
   159
			pos -= 0x10000L;
slouken@0
   160
		}
slouken@0
   161
		*dst++ = pixel[0];
slouken@0
   162
		*dst++ = pixel[1];
slouken@0
   163
		*dst++ = pixel[2];
slouken@0
   164
		pos += inc;
slouken@0
   165
	}
slouken@0
   166
}
slouken@0
   167
slouken@0
   168
/* Perform a stretch blit between two surfaces of the same format.
slouken@0
   169
   NOTE:  This function is not safe to call from multiple threads!
slouken@0
   170
*/
slouken@0
   171
int SDL_SoftStretch(SDL_Surface *src, SDL_Rect *srcrect,
slouken@0
   172
                    SDL_Surface *dst, SDL_Rect *dstrect)
slouken@0
   173
{
slouken@894
   174
	int src_locked;
slouken@894
   175
	int dst_locked;
slouken@0
   176
	int pos, inc;
slouken@0
   177
	int dst_width;
slouken@0
   178
	int dst_maxrow;
slouken@0
   179
	int src_row, dst_row;
slouken@0
   180
	Uint8 *srcp = NULL;
slouken@0
   181
	Uint8 *dstp;
slouken@0
   182
	SDL_Rect full_src;
slouken@0
   183
	SDL_Rect full_dst;
slouken@0
   184
#if defined(USE_ASM_STRETCH) && defined(__GNUC__)
slouken@0
   185
	int u1, u2;
slouken@0
   186
#endif
slouken@0
   187
	const int bpp = dst->format->BytesPerPixel;
slouken@0
   188
slouken@0
   189
	if ( src->format->BitsPerPixel != dst->format->BitsPerPixel ) {
slouken@0
   190
		SDL_SetError("Only works with same format surfaces");
slouken@0
   191
		return(-1);
slouken@0
   192
	}
slouken@0
   193
slouken@0
   194
	/* Verify the blit rectangles */
slouken@0
   195
	if ( srcrect ) {
slouken@0
   196
		if ( (srcrect->x < 0) || (srcrect->y < 0) ||
slouken@0
   197
		     ((srcrect->x+srcrect->w) > src->w) ||
slouken@0
   198
		     ((srcrect->y+srcrect->h) > src->h) ) {
slouken@0
   199
			SDL_SetError("Invalid source blit rectangle");
slouken@0
   200
			return(-1);
slouken@0
   201
		}
slouken@0
   202
	} else {
slouken@0
   203
		full_src.x = 0;
slouken@0
   204
		full_src.y = 0;
slouken@0
   205
		full_src.w = src->w;
slouken@0
   206
		full_src.h = src->h;
slouken@0
   207
		srcrect = &full_src;
slouken@0
   208
	}
slouken@0
   209
	if ( dstrect ) {
slouken@0
   210
		if ( (dstrect->x < 0) || (dstrect->y < 0) ||
slouken@0
   211
		     ((dstrect->x+dstrect->w) > dst->w) ||
slouken@0
   212
		     ((dstrect->y+dstrect->h) > dst->h) ) {
slouken@0
   213
			SDL_SetError("Invalid destination blit rectangle");
slouken@0
   214
			return(-1);
slouken@0
   215
		}
slouken@0
   216
	} else {
slouken@0
   217
		full_dst.x = 0;
slouken@0
   218
		full_dst.y = 0;
slouken@0
   219
		full_dst.w = dst->w;
slouken@0
   220
		full_dst.h = dst->h;
slouken@0
   221
		dstrect = &full_dst;
slouken@0
   222
	}
slouken@0
   223
slouken@894
   224
	/* Lock the destination if it's in hardware */
slouken@894
   225
	dst_locked = 0;
slouken@894
   226
	if ( SDL_MUSTLOCK(dst) ) {
slouken@894
   227
		if ( SDL_LockSurface(dst) < 0 ) {
slouken@894
   228
			SDL_SetError("Unable to lock destination surface");
slouken@894
   229
			return(-1);
slouken@894
   230
		}
slouken@894
   231
		dst_locked = 1;
slouken@894
   232
	}
slouken@894
   233
	/* Lock the source if it's in hardware */
slouken@894
   234
	src_locked = 0;
slouken@894
   235
	if ( SDL_MUSTLOCK(src) ) {
slouken@894
   236
		if ( SDL_LockSurface(src) < 0 ) {
slouken@894
   237
			if ( dst_locked ) {
slouken@894
   238
				SDL_UnlockSurface(dst);
slouken@894
   239
			}
slouken@894
   240
			SDL_SetError("Unable to lock source surface");
slouken@894
   241
			return(-1);
slouken@894
   242
		}
slouken@894
   243
		src_locked = 1;
slouken@894
   244
	}
slouken@894
   245
slouken@0
   246
	/* Set up the data... */
slouken@0
   247
	pos = 0x10000;
slouken@0
   248
	inc = (srcrect->h << 16) / dstrect->h;
slouken@0
   249
	src_row = srcrect->y;
slouken@0
   250
	dst_row = dstrect->y;
slouken@0
   251
	dst_width = dstrect->w*bpp;
slouken@0
   252
slouken@0
   253
#ifdef USE_ASM_STRETCH
slouken@0
   254
	/* Write the opcodes for this stretch */
slouken@0
   255
	if ( (bpp != 3) &&
slouken@0
   256
	     (generate_rowbytes(srcrect->w, dstrect->w, bpp) < 0) ) {
slouken@0
   257
		return(-1);
slouken@0
   258
	}
slouken@0
   259
#endif
slouken@0
   260
slouken@0
   261
	/* Perform the stretch blit */
slouken@0
   262
	for ( dst_maxrow = dst_row+dstrect->h; dst_row<dst_maxrow; ++dst_row ) {
slouken@0
   263
		dstp = (Uint8 *)dst->pixels + (dst_row*dst->pitch)
slouken@0
   264
		                            + (dstrect->x*bpp);
slouken@0
   265
		while ( pos >= 0x10000L ) {
slouken@0
   266
			srcp = (Uint8 *)src->pixels + (src_row*src->pitch)
slouken@0
   267
			                            + (srcrect->x*bpp);
slouken@0
   268
			++src_row;
slouken@0
   269
			pos -= 0x10000L;
slouken@0
   270
		}
slouken@0
   271
#ifdef USE_ASM_STRETCH
slouken@0
   272
		switch (bpp) {
slouken@0
   273
		    case 3:
slouken@0
   274
			copy_row3(srcp, srcrect->w, dstp, dstrect->w);
slouken@0
   275
			break;
slouken@0
   276
		    default:
slouken@0
   277
#ifdef __GNUC__
slouken@627
   278
			__asm__ __volatile__ (
icculus@1228
   279
			"call *%4"
slouken@0
   280
			: "=&D" (u1), "=&S" (u2)
icculus@1234
   281
			: "0" (dstp), "1" (srcp), "r" (copy_row)
slouken@0
   282
			: "memory" );
slouken@0
   283
#else
slouken@1361
   284
#ifdef _MSC_VER
icculus@1234
   285
		{ void *code = copy_row;
slouken@0
   286
			__asm {
slouken@0
   287
				push edi
slouken@0
   288
				push esi
slouken@0
   289
	
slouken@0
   290
				mov edi, dstp
slouken@0
   291
				mov esi, srcp
slouken@0
   292
				call dword ptr code
slouken@0
   293
slouken@0
   294
				pop esi
slouken@0
   295
				pop edi
slouken@0
   296
			}
slouken@0
   297
		}
slouken@0
   298
#else
slouken@0
   299
#error Need inline assembly for this compiler
slouken@0
   300
#endif
slouken@0
   301
#endif /* __GNUC__ */
slouken@0
   302
			break;
slouken@0
   303
		}
slouken@0
   304
#else
slouken@0
   305
		switch (bpp) {
slouken@0
   306
		    case 1:
slouken@0
   307
			copy_row1(srcp, srcrect->w, dstp, dstrect->w);
slouken@0
   308
			break;
slouken@0
   309
		    case 2:
slouken@0
   310
			copy_row2((Uint16 *)srcp, srcrect->w,
slouken@0
   311
			          (Uint16 *)dstp, dstrect->w);
slouken@0
   312
			break;
slouken@0
   313
		    case 3:
slouken@0
   314
			copy_row3(srcp, srcrect->w, dstp, dstrect->w);
slouken@0
   315
			break;
slouken@0
   316
		    case 4:
slouken@0
   317
			copy_row4((Uint32 *)srcp, srcrect->w,
slouken@0
   318
			          (Uint32 *)dstp, dstrect->w);
slouken@0
   319
			break;
slouken@0
   320
		}
slouken@0
   321
#endif
slouken@0
   322
		pos += inc;
slouken@0
   323
	}
slouken@894
   324
slouken@894
   325
	/* We need to unlock the surfaces if they're locked */
slouken@894
   326
	if ( dst_locked ) {
slouken@894
   327
		SDL_UnlockSurface(dst);
slouken@894
   328
	}
slouken@894
   329
	if ( src_locked ) {
slouken@894
   330
		SDL_UnlockSurface(src);
slouken@894
   331
	}
slouken@0
   332
	return(0);
slouken@0
   333
}
slouken@0
   334