src/video/SDL_blit.h
author Sam Lantinga <slouken@libsdl.org>
Wed, 01 Feb 2006 06:32:25 +0000
changeset 1312 c9b51268668f
parent 1162 2651158f59b8
child 1402 d910939febfa
permissions -rw-r--r--
Updated copyright information and removed rcs id lines (problematic in branch merges)
I batch edited these files, so please let me know if I've accidentally removed anybody's
credit here.
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
#ifndef _SDL_blit_h
slouken@0
    24
#define _SDL_blit_h
slouken@0
    25
slouken@0
    26
#include "SDL_endian.h"
slouken@0
    27
slouken@0
    28
/* The structure passed to the low level blit functions */
slouken@0
    29
typedef struct {
slouken@0
    30
	Uint8 *s_pixels;
slouken@0
    31
	int s_width;
slouken@0
    32
	int s_height;
slouken@0
    33
	int s_skip;
slouken@0
    34
	Uint8 *d_pixels;
slouken@0
    35
	int d_width;
slouken@0
    36
	int d_height;
slouken@0
    37
	int d_skip;
slouken@0
    38
	void *aux_data;
slouken@0
    39
	SDL_PixelFormat *src;
slouken@0
    40
	Uint8 *table;
slouken@0
    41
	SDL_PixelFormat *dst;
slouken@0
    42
} SDL_BlitInfo;
slouken@0
    43
slouken@0
    44
/* The type definition for the low level blit functions */
slouken@0
    45
typedef void (*SDL_loblit)(SDL_BlitInfo *info);
slouken@0
    46
slouken@0
    47
/* This is the private info structure for software accelerated blits */
slouken@0
    48
struct private_swaccel {
slouken@0
    49
	SDL_loblit blit;
slouken@0
    50
	void *aux_data;
slouken@0
    51
};
slouken@0
    52
slouken@0
    53
/* Blit mapping definition */
slouken@0
    54
typedef struct SDL_BlitMap {
slouken@0
    55
	SDL_Surface *dst;
slouken@0
    56
	int identity;
slouken@0
    57
	Uint8 *table;
slouken@0
    58
	SDL_blit hw_blit;
slouken@0
    59
	SDL_blit sw_blit;
slouken@0
    60
	struct private_hwaccel *hw_data;
slouken@0
    61
	struct private_swaccel *sw_data;
slouken@0
    62
slouken@0
    63
	/* the version count matches the destination; mismatch indicates
slouken@0
    64
	   an invalid mapping */
slouken@0
    65
        unsigned int format_version;
slouken@0
    66
} SDL_BlitMap;
slouken@0
    67
slouken@0
    68
slouken@0
    69
/* Functions found in SDL_blit.c */
slouken@0
    70
extern int SDL_CalculateBlit(SDL_Surface *surface);
slouken@0
    71
slouken@0
    72
/* Functions found in SDL_blit_{0,1,N,A}.c */
slouken@0
    73
extern SDL_loblit SDL_CalculateBlit0(SDL_Surface *surface, int complex);
slouken@0
    74
extern SDL_loblit SDL_CalculateBlit1(SDL_Surface *surface, int complex);
slouken@0
    75
extern SDL_loblit SDL_CalculateBlitN(SDL_Surface *surface, int complex);
slouken@0
    76
extern SDL_loblit SDL_CalculateAlphaBlit(SDL_Surface *surface, int complex);
slouken@0
    77
slouken@0
    78
/*
slouken@0
    79
 * Useful macros for blitting routines
slouken@0
    80
 */
slouken@0
    81
slouken@0
    82
#define FORMAT_EQUAL(A, B)						\
slouken@0
    83
    ((A)->BitsPerPixel == (B)->BitsPerPixel				\
slouken@0
    84
     && ((A)->Rmask == (B)->Rmask) && ((A)->Amask == (B)->Amask))
slouken@0
    85
slouken@0
    86
/* Load pixel of the specified format from a buffer and get its R-G-B values */
slouken@0
    87
/* FIXME: rescale values to 0..255 here? */
icculus@1162
    88
#define RGB_FROM_PIXEL(Pixel, fmt, r, g, b)				\
slouken@0
    89
{									\
icculus@1162
    90
	r = (((Pixel&fmt->Rmask)>>fmt->Rshift)<<fmt->Rloss); 		\
icculus@1162
    91
	g = (((Pixel&fmt->Gmask)>>fmt->Gshift)<<fmt->Gloss); 		\
icculus@1162
    92
	b = (((Pixel&fmt->Bmask)>>fmt->Bshift)<<fmt->Bloss); 		\
slouken@0
    93
}
icculus@1162
    94
#define RGB_FROM_RGB565(Pixel, r, g, b)					\
slouken@0
    95
{									\
icculus@1162
    96
	r = (((Pixel&0xF800)>>11)<<3);		 			\
icculus@1162
    97
	g = (((Pixel&0x07E0)>>5)<<2); 					\
icculus@1162
    98
	b = ((Pixel&0x001F)<<3); 					\
slouken@0
    99
}
icculus@1162
   100
#define RGB_FROM_RGB555(Pixel, r, g, b)					\
slouken@0
   101
{									\
icculus@1162
   102
	r = (((Pixel&0x7C00)>>10)<<3);		 			\
icculus@1162
   103
	g = (((Pixel&0x03E0)>>5)<<3); 					\
icculus@1162
   104
	b = ((Pixel&0x001F)<<3); 					\
slouken@0
   105
}
icculus@1162
   106
#define RGB_FROM_RGB888(Pixel, r, g, b)					\
slouken@0
   107
{									\
icculus@1162
   108
	r = ((Pixel&0xFF0000)>>16);		 			\
icculus@1162
   109
	g = ((Pixel&0xFF00)>>8);		 			\
icculus@1162
   110
	b = (Pixel&0xFF);			 			\
slouken@0
   111
}
icculus@1162
   112
#define RETRIEVE_RGB_PIXEL(buf, bpp, Pixel)				   \
slouken@0
   113
do {									   \
slouken@0
   114
	switch (bpp) {							   \
slouken@0
   115
		case 2:							   \
icculus@1162
   116
			Pixel = *((Uint16 *)(buf));			   \
slouken@0
   117
		break;							   \
slouken@0
   118
									   \
slouken@0
   119
		case 3: {						   \
slouken@0
   120
		        Uint8 *B = (Uint8 *)(buf);			   \
slouken@0
   121
			if(SDL_BYTEORDER == SDL_LIL_ENDIAN) {		   \
icculus@1162
   122
			        Pixel = B[0] + (B[1] << 8) + (B[2] << 16); \
slouken@0
   123
			} else {					   \
icculus@1162
   124
			        Pixel = (B[0] << 16) + (B[1] << 8) + B[2]; \
slouken@0
   125
			}						   \
slouken@0
   126
		}							   \
slouken@0
   127
		break;							   \
slouken@0
   128
									   \
slouken@0
   129
		case 4:							   \
icculus@1162
   130
			Pixel = *((Uint32 *)(buf));			   \
slouken@0
   131
		break;							   \
slouken@0
   132
									   \
slouken@0
   133
		default:						   \
icculus@1162
   134
			Pixel = 0; /* appease gcc */			   \
slouken@0
   135
		break;							   \
slouken@0
   136
	}								   \
slouken@0
   137
} while(0)
slouken@0
   138
icculus@1162
   139
#define DISEMBLE_RGB(buf, bpp, fmt, Pixel, r, g, b)			   \
slouken@0
   140
do {									   \
slouken@0
   141
	switch (bpp) {							   \
slouken@0
   142
		case 2:							   \
icculus@1162
   143
			Pixel = *((Uint16 *)(buf));			   \
slouken@0
   144
		break;							   \
slouken@0
   145
									   \
slouken@0
   146
		case 3: {						   \
slouken@0
   147
		        Uint8 *B = (Uint8 *)buf;			   \
slouken@0
   148
			if(SDL_BYTEORDER == SDL_LIL_ENDIAN) {		   \
icculus@1162
   149
			        Pixel = B[0] + (B[1] << 8) + (B[2] << 16); \
slouken@0
   150
			} else {					   \
icculus@1162
   151
			        Pixel = (B[0] << 16) + (B[1] << 8) + B[2]; \
slouken@0
   152
			}						   \
slouken@0
   153
		}							   \
slouken@0
   154
		break;							   \
slouken@0
   155
									   \
slouken@0
   156
		case 4:							   \
icculus@1162
   157
			Pixel = *((Uint32 *)(buf));			   \
slouken@0
   158
		break;							   \
slouken@0
   159
									   \
slouken@0
   160
	        default:						   \
icculus@1162
   161
		        Pixel = 0;	/* prevent gcc from complaining */ \
slouken@0
   162
		break;							   \
slouken@0
   163
	}								   \
icculus@1162
   164
	RGB_FROM_PIXEL(Pixel, fmt, r, g, b);				   \
slouken@0
   165
} while(0)
slouken@0
   166
slouken@0
   167
/* Assemble R-G-B values into a specified pixel format and store them */
icculus@1162
   168
#define PIXEL_FROM_RGB(Pixel, fmt, r, g, b)				\
slouken@0
   169
{									\
icculus@1162
   170
	Pixel = ((r>>fmt->Rloss)<<fmt->Rshift)|				\
slouken@0
   171
		((g>>fmt->Gloss)<<fmt->Gshift)|				\
slouken@0
   172
		((b>>fmt->Bloss)<<fmt->Bshift);				\
slouken@0
   173
}
icculus@1162
   174
#define RGB565_FROM_RGB(Pixel, r, g, b)					\
slouken@0
   175
{									\
icculus@1162
   176
	Pixel = ((r>>3)<<11)|((g>>2)<<5)|(b>>3);			\
slouken@0
   177
}
icculus@1162
   178
#define RGB555_FROM_RGB(Pixel, r, g, b)					\
slouken@0
   179
{									\
icculus@1162
   180
	Pixel = ((r>>3)<<10)|((g>>3)<<5)|(b>>3);			\
slouken@0
   181
}
icculus@1162
   182
#define RGB888_FROM_RGB(Pixel, r, g, b)					\
slouken@0
   183
{									\
icculus@1162
   184
	Pixel = (r<<16)|(g<<8)|b;					\
slouken@0
   185
}
slouken@0
   186
#define ASSEMBLE_RGB(buf, bpp, fmt, r, g, b) 				\
slouken@0
   187
{									\
slouken@0
   188
	switch (bpp) {							\
slouken@0
   189
		case 2: {						\
icculus@1162
   190
			Uint16 Pixel;					\
slouken@0
   191
									\
icculus@1162
   192
			PIXEL_FROM_RGB(Pixel, fmt, r, g, b);		\
icculus@1162
   193
			*((Uint16 *)(buf)) = Pixel;			\
slouken@0
   194
		}							\
slouken@0
   195
		break;							\
slouken@0
   196
									\
slouken@0
   197
		case 3: {						\
slouken@0
   198
                        if(SDL_BYTEORDER == SDL_LIL_ENDIAN) {		\
slouken@0
   199
			        *((buf)+fmt->Rshift/8) = r;		\
slouken@0
   200
				*((buf)+fmt->Gshift/8) = g;		\
slouken@0
   201
				*((buf)+fmt->Bshift/8) = b;		\
slouken@0
   202
			} else {					\
slouken@0
   203
			        *((buf)+2-fmt->Rshift/8) = r;		\
slouken@0
   204
				*((buf)+2-fmt->Gshift/8) = g;		\
slouken@0
   205
				*((buf)+2-fmt->Bshift/8) = b;		\
slouken@0
   206
			}						\
slouken@0
   207
		}							\
slouken@0
   208
		break;							\
slouken@0
   209
									\
slouken@0
   210
		case 4: {						\
icculus@1162
   211
			Uint32 Pixel;					\
slouken@0
   212
									\
icculus@1162
   213
			PIXEL_FROM_RGB(Pixel, fmt, r, g, b);		\
icculus@1162
   214
			*((Uint32 *)(buf)) = Pixel;			\
slouken@0
   215
		}							\
slouken@0
   216
		break;							\
slouken@0
   217
	}								\
slouken@0
   218
}
slouken@0
   219
#define ASSEMBLE_RGB_AMASK(buf, bpp, fmt, r, g, b, Amask)		\
slouken@0
   220
{									\
slouken@0
   221
	switch (bpp) {							\
slouken@0
   222
		case 2: {						\
slouken@0
   223
			Uint16 *bufp;					\
icculus@1162
   224
			Uint16 Pixel;					\
slouken@0
   225
									\
slouken@0
   226
			bufp = (Uint16 *)buf;				\
icculus@1162
   227
			PIXEL_FROM_RGB(Pixel, fmt, r, g, b);		\
icculus@1162
   228
			*bufp = Pixel | (*bufp & Amask);		\
slouken@0
   229
		}							\
slouken@0
   230
		break;							\
slouken@0
   231
									\
slouken@0
   232
		case 3: {						\
slouken@0
   233
                        if(SDL_BYTEORDER == SDL_LIL_ENDIAN) {		\
slouken@0
   234
			        *((buf)+fmt->Rshift/8) = r;		\
slouken@0
   235
				*((buf)+fmt->Gshift/8) = g;		\
slouken@0
   236
				*((buf)+fmt->Bshift/8) = b;		\
slouken@0
   237
			} else {					\
slouken@0
   238
			        *((buf)+2-fmt->Rshift/8) = r;		\
slouken@0
   239
				*((buf)+2-fmt->Gshift/8) = g;		\
slouken@0
   240
				*((buf)+2-fmt->Bshift/8) = b;		\
slouken@0
   241
			}						\
slouken@0
   242
		}							\
slouken@0
   243
		break;							\
slouken@0
   244
									\
slouken@0
   245
		case 4: {						\
slouken@0
   246
			Uint32 *bufp;					\
icculus@1162
   247
			Uint32 Pixel;					\
slouken@0
   248
									\
slouken@0
   249
			bufp = (Uint32 *)buf;				\
icculus@1162
   250
			PIXEL_FROM_RGB(Pixel, fmt, r, g, b);		\
icculus@1162
   251
			*bufp = Pixel | (*bufp & Amask);		\
slouken@0
   252
		}							\
slouken@0
   253
		break;							\
slouken@0
   254
	}								\
slouken@0
   255
}
slouken@0
   256
slouken@0
   257
/* FIXME: Should we rescale alpha into 0..255 here? */
icculus@1162
   258
#define RGBA_FROM_PIXEL(Pixel, fmt, r, g, b, a)				\
slouken@0
   259
{									\
icculus@1162
   260
	r = ((Pixel&fmt->Rmask)>>fmt->Rshift)<<fmt->Rloss; 		\
icculus@1162
   261
	g = ((Pixel&fmt->Gmask)>>fmt->Gshift)<<fmt->Gloss; 		\
icculus@1162
   262
	b = ((Pixel&fmt->Bmask)>>fmt->Bshift)<<fmt->Bloss; 		\
icculus@1162
   263
	a = ((Pixel&fmt->Amask)>>fmt->Ashift)<<fmt->Aloss;	 	\
slouken@0
   264
}
icculus@1162
   265
#define RGBA_FROM_8888(Pixel, fmt, r, g, b, a)	\
slouken@0
   266
{						\
icculus@1162
   267
	r = (Pixel&fmt->Rmask)>>fmt->Rshift;	\
icculus@1162
   268
	g = (Pixel&fmt->Gmask)>>fmt->Gshift;	\
icculus@1162
   269
	b = (Pixel&fmt->Bmask)>>fmt->Bshift;	\
icculus@1162
   270
	a = (Pixel&fmt->Amask)>>fmt->Ashift;	\
slouken@0
   271
}
icculus@1162
   272
#define RGBA_FROM_RGBA8888(Pixel, r, g, b, a)				\
slouken@0
   273
{									\
icculus@1162
   274
	r = (Pixel>>24);						\
icculus@1162
   275
	g = ((Pixel>>16)&0xFF);						\
icculus@1162
   276
	b = ((Pixel>>8)&0xFF);						\
icculus@1162
   277
	a = (Pixel&0xFF);						\
slouken@0
   278
}
icculus@1162
   279
#define RGBA_FROM_ARGB8888(Pixel, r, g, b, a)				\
slouken@0
   280
{									\
icculus@1162
   281
	r = ((Pixel>>16)&0xFF);						\
icculus@1162
   282
	g = ((Pixel>>8)&0xFF);						\
icculus@1162
   283
	b = (Pixel&0xFF);						\
icculus@1162
   284
	a = (Pixel>>24);						\
slouken@0
   285
}
icculus@1162
   286
#define RGBA_FROM_ABGR8888(Pixel, r, g, b, a)				\
slouken@0
   287
{									\
icculus@1162
   288
	r = (Pixel&0xFF);						\
icculus@1162
   289
	g = ((Pixel>>8)&0xFF);						\
icculus@1162
   290
	b = ((Pixel>>16)&0xFF);						\
icculus@1162
   291
	a = (Pixel>>24);						\
slouken@0
   292
}
icculus@1162
   293
#define DISEMBLE_RGBA(buf, bpp, fmt, Pixel, r, g, b, a)			   \
slouken@0
   294
do {									   \
slouken@0
   295
	switch (bpp) {							   \
slouken@0
   296
		case 2:							   \
icculus@1162
   297
			Pixel = *((Uint16 *)(buf));			   \
slouken@0
   298
		break;							   \
slouken@0
   299
									   \
slouken@0
   300
		case 3:	{/* FIXME: broken code (no alpha) */		   \
slouken@0
   301
		        Uint8 *b = (Uint8 *)buf;			   \
slouken@0
   302
			if(SDL_BYTEORDER == SDL_LIL_ENDIAN) {		   \
icculus@1162
   303
			        Pixel = b[0] + (b[1] << 8) + (b[2] << 16); \
slouken@0
   304
			} else {					   \
icculus@1162
   305
			        Pixel = (b[0] << 16) + (b[1] << 8) + b[2]; \
slouken@0
   306
			}						   \
slouken@0
   307
		}							   \
slouken@0
   308
		break;							   \
slouken@0
   309
									   \
slouken@0
   310
		case 4:							   \
icculus@1162
   311
			Pixel = *((Uint32 *)(buf));			   \
slouken@0
   312
		break;							   \
slouken@0
   313
									   \
slouken@0
   314
		default:						   \
icculus@1162
   315
		        Pixel = 0; /* stop gcc complaints */		   \
slouken@0
   316
		break;							   \
slouken@0
   317
	}								   \
icculus@1162
   318
	RGBA_FROM_PIXEL(Pixel, fmt, r, g, b, a);			   \
icculus@1162
   319
	Pixel &= ~fmt->Amask;						   \
slouken@0
   320
} while(0)
slouken@0
   321
slouken@0
   322
/* FIXME: this isn't correct, especially for Alpha (maximum != 255) */
icculus@1162
   323
#define PIXEL_FROM_RGBA(Pixel, fmt, r, g, b, a)				\
slouken@0
   324
{									\
icculus@1162
   325
	Pixel = ((r>>fmt->Rloss)<<fmt->Rshift)|				\
slouken@0
   326
		((g>>fmt->Gloss)<<fmt->Gshift)|				\
slouken@0
   327
		((b>>fmt->Bloss)<<fmt->Bshift)|				\
slouken@535
   328
		((a>>fmt->Aloss)<<fmt->Ashift);				\
slouken@0
   329
}
slouken@0
   330
#define ASSEMBLE_RGBA(buf, bpp, fmt, r, g, b, a)			\
slouken@0
   331
{									\
slouken@0
   332
	switch (bpp) {							\
slouken@0
   333
		case 2: {						\
icculus@1162
   334
			Uint16 Pixel;					\
slouken@0
   335
									\
icculus@1162
   336
			PIXEL_FROM_RGBA(Pixel, fmt, r, g, b, a);	\
icculus@1162
   337
			*((Uint16 *)(buf)) = Pixel;			\
slouken@0
   338
		}							\
slouken@0
   339
		break;							\
slouken@0
   340
									\
slouken@0
   341
		case 3: { /* FIXME: broken code (no alpha) */		\
slouken@0
   342
                        if(SDL_BYTEORDER == SDL_LIL_ENDIAN) {		\
slouken@0
   343
			        *((buf)+fmt->Rshift/8) = r;		\
slouken@0
   344
				*((buf)+fmt->Gshift/8) = g;		\
slouken@0
   345
				*((buf)+fmt->Bshift/8) = b;		\
slouken@0
   346
			} else {					\
slouken@0
   347
			        *((buf)+2-fmt->Rshift/8) = r;		\
slouken@0
   348
				*((buf)+2-fmt->Gshift/8) = g;		\
slouken@0
   349
				*((buf)+2-fmt->Bshift/8) = b;		\
slouken@0
   350
			}						\
slouken@0
   351
		}							\
slouken@0
   352
		break;							\
slouken@0
   353
									\
slouken@0
   354
		case 4: {						\
icculus@1162
   355
			Uint32 Pixel;					\
slouken@0
   356
									\
icculus@1162
   357
			PIXEL_FROM_RGBA(Pixel, fmt, r, g, b, a);	\
icculus@1162
   358
			*((Uint32 *)(buf)) = Pixel;			\
slouken@0
   359
		}							\
slouken@0
   360
		break;							\
slouken@0
   361
	}								\
slouken@0
   362
}
slouken@0
   363
icculus@1162
   364
/* Blend the RGB values of two Pixels based on a source alpha value */
slouken@0
   365
#define ALPHA_BLEND(sR, sG, sB, A, dR, dG, dB)	\
slouken@0
   366
do {						\
slouken@0
   367
	dR = (((sR-dR)*(A))>>8)+dR;		\
slouken@0
   368
	dG = (((sG-dG)*(A))>>8)+dG;		\
slouken@0
   369
	dB = (((sB-dB)*(A))>>8)+dB;		\
slouken@0
   370
} while(0)
slouken@0
   371
icculus@1162
   372
/* Blend the RGB values of two Pixels based on a source alpha value */
icculus@1047
   373
#define ACCURATE_ALPHA_BLEND(sR, sG, sB, sA, dR, dG, dB)	\
icculus@1047
   374
do {						\
icculus@1047
   375
    unsigned tR, tG, tB, tA; \
icculus@1047
   376
    tA = 255 - sA; \
icculus@1047
   377
    tR = 1 + (sR * sA) + (dR * tA); \
icculus@1047
   378
    dR = (tR + (tR >> 8)) >> 8; \
icculus@1047
   379
    tG = 1 + (sG * sA) + (dG * tA); \
icculus@1047
   380
    dG = (tG + (tG >> 8)) >> 8; \
icculus@1047
   381
    tB = 1 + (sB * sA) + (dB * tA); \
icculus@1047
   382
    dB = (tB + (tB >> 8)) >> 8; \
icculus@1047
   383
} while(0)
icculus@1047
   384
icculus@1047
   385
slouken@0
   386
/* This is a very useful loop for optimizing blitters */
slouken@553
   387
#if defined(_MSC_VER) && (_MSC_VER == 1300)
slouken@553
   388
/* There's a bug in the Visual C++ 7 optimizer when compiling this code */
slouken@553
   389
#else
slouken@0
   390
#define USE_DUFFS_LOOP
slouken@553
   391
#endif
slouken@0
   392
#ifdef USE_DUFFS_LOOP
slouken@0
   393
slouken@0
   394
/* 8-times unrolled loop */
slouken@0
   395
#define DUFFS_LOOP8(pixel_copy_increment, width)			\
slouken@0
   396
{ int n = (width+7)/8;							\
slouken@91
   397
	switch (width & 7) {						\
slouken@0
   398
	case 0: do {	pixel_copy_increment;				\
slouken@0
   399
	case 7:		pixel_copy_increment;				\
slouken@0
   400
	case 6:		pixel_copy_increment;				\
slouken@0
   401
	case 5:		pixel_copy_increment;				\
slouken@0
   402
	case 4:		pixel_copy_increment;				\
slouken@0
   403
	case 3:		pixel_copy_increment;				\
slouken@0
   404
	case 2:		pixel_copy_increment;				\
slouken@0
   405
	case 1:		pixel_copy_increment;				\
slouken@0
   406
		} while ( --n > 0 );					\
slouken@0
   407
	}								\
slouken@0
   408
}
slouken@0
   409
slouken@0
   410
/* 4-times unrolled loop */
slouken@0
   411
#define DUFFS_LOOP4(pixel_copy_increment, width)			\
slouken@0
   412
{ int n = (width+3)/4;							\
slouken@91
   413
	switch (width & 3) {						\
slouken@0
   414
	case 0: do {	pixel_copy_increment;				\
slouken@0
   415
	case 3:		pixel_copy_increment;				\
slouken@0
   416
	case 2:		pixel_copy_increment;				\
slouken@0
   417
	case 1:		pixel_copy_increment;				\
slouken@0
   418
		} while ( --n > 0 );					\
slouken@0
   419
	}								\
slouken@0
   420
}
slouken@0
   421
slouken@689
   422
/* 2 - times unrolled loop */
slouken@689
   423
#define DUFFS_LOOP_DOUBLE2(pixel_copy_increment,			\
slouken@689
   424
				double_pixel_copy_increment, width)	\
slouken@689
   425
{ int n, w = width;							\
slouken@689
   426
	if( w & 1 ) {							\
slouken@689
   427
	    pixel_copy_increment;					\
slouken@689
   428
	    w--;							\
slouken@689
   429
	}								\
slouken@689
   430
	if ( w > 0 )	{						\
slouken@689
   431
	    n = ( w + 2) / 4;						\
slouken@689
   432
	    switch( w & 2 ) {						\
slouken@689
   433
	    case 0: do {	double_pixel_copy_increment;		\
slouken@689
   434
	    case 2:		double_pixel_copy_increment;		\
slouken@689
   435
		    } while ( --n > 0 );					\
slouken@689
   436
	    }								\
slouken@689
   437
	}								\
slouken@689
   438
}
slouken@689
   439
slouken@689
   440
/* 2 - times unrolled loop 4 pixels */
slouken@689
   441
#define DUFFS_LOOP_QUATRO2(pixel_copy_increment,			\
slouken@689
   442
				double_pixel_copy_increment,		\
slouken@689
   443
				quatro_pixel_copy_increment, width)	\
slouken@689
   444
{ int n, w = width;								\
slouken@689
   445
        if(w & 1) {							\
slouken@689
   446
	  pixel_copy_increment;						\
slouken@689
   447
	  w--;								\
slouken@689
   448
	}								\
slouken@689
   449
	if(w & 2) {							\
slouken@689
   450
	  double_pixel_copy_increment;					\
slouken@689
   451
	  w -= 2;							\
slouken@689
   452
	}								\
slouken@689
   453
	if ( w > 0 ) {							\
slouken@689
   454
	    n = ( w + 7 ) / 8;						\
slouken@689
   455
	    switch( w & 4 ) {						\
slouken@689
   456
	    case 0: do {	quatro_pixel_copy_increment;		\
slouken@689
   457
	    case 4:		quatro_pixel_copy_increment;		\
slouken@689
   458
		    } while ( --n > 0 );					\
slouken@689
   459
	    }								\
slouken@689
   460
	}								\
slouken@689
   461
}
slouken@689
   462
slouken@0
   463
/* Use the 8-times version of the loop by default */
slouken@0
   464
#define DUFFS_LOOP(pixel_copy_increment, width)				\
slouken@0
   465
	DUFFS_LOOP8(pixel_copy_increment, width)
slouken@0
   466
slouken@0
   467
#else
slouken@0
   468
slouken@0
   469
/* Don't use Duff's device to unroll loops */
slouken@689
   470
#define DUFFS_LOOP_DOUBLE2(pixel_copy_increment,			\
slouken@689
   471
			 double_pixel_copy_increment, width)		\
slouken@689
   472
{ int n = width;								\
slouken@689
   473
    if( n & 1 ) {							\
slouken@689
   474
	pixel_copy_increment;						\
slouken@689
   475
	n--;								\
slouken@689
   476
    }									\
slouken@689
   477
    n=n>>1;								\
slouken@689
   478
    for(; n > 0; --n) {   						\
slouken@689
   479
	double_pixel_copy_increment;					\
slouken@689
   480
    }									\
slouken@689
   481
}
slouken@689
   482
slouken@689
   483
/* Don't use Duff's device to unroll loops */
slouken@689
   484
#define DUFFS_LOOP_QUATRO2(pixel_copy_increment,			\
slouken@689
   485
				double_pixel_copy_increment,		\
slouken@689
   486
				quatro_pixel_copy_increment, width)	\
slouken@689
   487
{ int n = width;								\
slouken@689
   488
        if(n & 1) {							\
slouken@689
   489
	  pixel_copy_increment;						\
slouken@689
   490
	  n--;								\
slouken@689
   491
	}								\
slouken@689
   492
	if(n & 2) {							\
slouken@689
   493
	  double_pixel_copy_increment;					\
slouken@689
   494
	  n -= 2;							\
slouken@689
   495
	}								\
slouken@689
   496
	n=n>>2;								\
slouken@689
   497
	for(; n > 0; --n) {   						\
slouken@689
   498
	  quatro_pixel_copy_increment;					\
slouken@689
   499
        }								\
slouken@689
   500
}
slouken@689
   501
slouken@689
   502
/* Don't use Duff's device to unroll loops */
slouken@0
   503
#define DUFFS_LOOP(pixel_copy_increment, width)				\
slouken@0
   504
{ int n;								\
slouken@0
   505
	for ( n=width; n > 0; --n ) {					\
slouken@0
   506
		pixel_copy_increment;					\
slouken@0
   507
	}								\
slouken@0
   508
}
slouken@0
   509
#define DUFFS_LOOP8(pixel_copy_increment, width)			\
slouken@0
   510
	DUFFS_LOOP(pixel_copy_increment, width)
slouken@0
   511
#define DUFFS_LOOP4(pixel_copy_increment, width)			\
slouken@0
   512
	DUFFS_LOOP(pixel_copy_increment, width)
slouken@0
   513
slouken@0
   514
#endif /* USE_DUFFS_LOOP */
slouken@0
   515
slouken@0
   516
/* Prevent Visual C++ 6.0 from printing out stupid warnings */
slouken@0
   517
#if defined(_MSC_VER) && (_MSC_VER >= 600)
slouken@0
   518
#pragma warning(disable: 4550)
slouken@0
   519
#endif
slouken@0
   520
slouken@0
   521
#endif /* _SDL_blit_h */