| 1 | /* |
| 2 | SDL - Simple DirectMedia Layer |
| 3 | Copyright (C) 1997-2009 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 | /* This a stretch blit implementation based on ideas given to me by |
| 25 | Tomasz Cejner - thanks! :) |
| 26 | |
| 27 | April 27, 2000 - Sam Lantinga |
| 28 | */ |
| 29 | |
| 30 | #include "SDL_video.h" |
| 31 | #include "SDL_blit.h" |
| 32 | |
| 33 | /* This isn't ready for general consumption yet - it should be folded |
| 34 | into the general blitting mechanism. |
| 35 | */ |
| 36 | |
| 37 | #if ((defined(_MFC_VER) && defined(_M_IX86)/* && !defined(_WIN32_WCE) still needed? */) || \ |
| 38 | defined(__WATCOMC__) || \ |
| 39 | (defined(__GNUC__) && defined(__i386__))) && SDL_ASSEMBLY_ROUTINES |
| 40 | /* There's a bug with gcc 4.4.1 and -O2 where srcp doesn't get the correct |
| 41 | * value after the first scanline. FIXME? */ |
| 42 | /*#define USE_ASM_STRETCH*/ |
| 43 | #endif |
| 44 | |
| 45 | #ifdef USE_ASM_STRETCH |
| 46 | |
| 47 | #ifdef HAVE_MPROTECT |
| 48 | #include <sys/types.h> |
| 49 | #include <sys/mman.h> |
| 50 | #endif |
| 51 | #ifdef __GNUC__ |
| 52 | #define PAGE_ALIGNED __attribute__((__aligned__(4096))) |
| 53 | #else |
| 54 | #define PAGE_ALIGNED |
| 55 | #endif |
| 56 | |
| 57 | #if defined(_M_IX86) || defined(i386) |
| 58 | #define PREFIX16 0x66 |
| 59 | #define STORE_BYTE 0xAA |
| 60 | #define STORE_WORD 0xAB |
| 61 | #define LOAD_BYTE 0xAC |
| 62 | #define LOAD_WORD 0xAD |
| 63 | #define RETURN 0xC3 |
| 64 | #else |
| 65 | #error Need assembly opcodes for this architecture |
| 66 | #endif |
| 67 | |
| 68 | static unsigned char copy_row[4096] PAGE_ALIGNED; |
| 69 | |
| 70 | static int generate_rowbytes(int src_w, int dst_w, int bpp) |
| 71 | { |
| 72 | static struct { |
| 73 | int bpp; |
| 74 | int src_w; |
| 75 | int dst_w; |
| 76 | int status; |
| 77 | } last; |
| 78 | |
| 79 | int i; |
| 80 | int pos, inc; |
| 81 | unsigned char *eip; |
| 82 | unsigned char load, store; |
| 83 | |
| 84 | /* See if we need to regenerate the copy buffer */ |
| 85 | if ( (src_w == last.src_w) && |
| 86 | (dst_w == last.dst_w) && (bpp == last.bpp) ) { |
| 87 | return(last.status); |
| 88 | } |
| 89 | last.bpp = bpp; |
| 90 | last.src_w = src_w; |
| 91 | last.dst_w = dst_w; |
| 92 | last.status = -1; |
| 93 | |
| 94 | switch (bpp) { |
| 95 | case 1: |
| 96 | load = LOAD_BYTE; |
| 97 | store = STORE_BYTE; |
| 98 | break; |
| 99 | case 2: |
| 100 | case 4: |
| 101 | load = LOAD_WORD; |
| 102 | store = STORE_WORD; |
| 103 | break; |
| 104 | default: |
| 105 | SDL_SetError("ASM stretch of %d bytes isn't supported\n", bpp); |
| 106 | return(-1); |
| 107 | } |
| 108 | #ifdef HAVE_MPROTECT |
| 109 | /* Make the code writeable */ |
| 110 | if ( mprotect(copy_row, sizeof(copy_row), PROT_READ|PROT_WRITE) < 0 ) { |
| 111 | SDL_SetError("Couldn't make copy buffer writeable"); |
| 112 | return(-1); |
| 113 | } |
| 114 | #endif |
| 115 | pos = 0x10000; |
| 116 | inc = (src_w << 16) / dst_w; |
| 117 | eip = copy_row; |
| 118 | for ( i=0; i<dst_w; ++i ) { |
| 119 | while ( pos >= 0x10000L ) { |
| 120 | if ( bpp == 2 ) { |
| 121 | *eip++ = PREFIX16; |
| 122 | } |
| 123 | *eip++ = load; |
| 124 | pos -= 0x10000L; |
| 125 | } |
| 126 | if ( bpp == 2 ) { |
| 127 | *eip++ = PREFIX16; |
| 128 | } |
| 129 | *eip++ = store; |
| 130 | pos += inc; |
| 131 | } |
| 132 | *eip++ = RETURN; |
| 133 | |
| 134 | /* Verify that we didn't overflow (too late!!!) */ |
| 135 | if ( eip > (copy_row+sizeof(copy_row)) ) { |
| 136 | SDL_SetError("Copy buffer overflow"); |
| 137 | return(-1); |
| 138 | } |
| 139 | #ifdef HAVE_MPROTECT |
| 140 | /* Make the code executable but not writeable */ |
| 141 | if ( mprotect(copy_row, sizeof(copy_row), PROT_READ|PROT_EXEC) < 0 ) { |
| 142 | SDL_SetError("Couldn't make copy buffer executable"); |
| 143 | return(-1); |
| 144 | } |
| 145 | #endif |
| 146 | last.status = 0; |
| 147 | return(0); |
| 148 | } |
| 149 | |
| 150 | #endif /* USE_ASM_STRETCH */ |
| 151 | |
| 152 | #define DEFINE_COPY_ROW(name, type) \ |
| 153 | void name(type *src, int src_w, type *dst, int dst_w) \ |
| 154 | { \ |
| 155 | int i; \ |
| 156 | int pos, inc; \ |
| 157 | type pixel = 0; \ |
| 158 | \ |
| 159 | pos = 0x10000; \ |
| 160 | inc = (src_w << 16) / dst_w; \ |
| 161 | for ( i=dst_w; i>0; --i ) { \ |
| 162 | while ( pos >= 0x10000L ) { \ |
| 163 | pixel = *src++; \ |
| 164 | pos -= 0x10000L; \ |
| 165 | } \ |
| 166 | *dst++ = pixel; \ |
| 167 | pos += inc; \ |
| 168 | } \ |
| 169 | } |
| 170 | DEFINE_COPY_ROW(copy_row1, Uint8) |
| 171 | DEFINE_COPY_ROW(copy_row2, Uint16) |
| 172 | DEFINE_COPY_ROW(copy_row4, Uint32) |
| 173 | |
| 174 | /* The ASM code doesn't handle 24-bpp stretch blits */ |
| 175 | void copy_row3(Uint8 *src, int src_w, Uint8 *dst, int dst_w) |
| 176 | { |
| 177 | int i; |
| 178 | int pos, inc; |
| 179 | Uint8 pixel[3] = { 0, 0, 0 }; |
| 180 | |
| 181 | pos = 0x10000; |
| 182 | inc = (src_w << 16) / dst_w; |
| 183 | for ( i=dst_w; i>0; --i ) { |
| 184 | while ( pos >= 0x10000L ) { |
| 185 | pixel[0] = *src++; |
| 186 | pixel[1] = *src++; |
| 187 | pixel[2] = *src++; |
| 188 | pos -= 0x10000L; |
| 189 | } |
| 190 | *dst++ = pixel[0]; |
| 191 | *dst++ = pixel[1]; |
| 192 | *dst++ = pixel[2]; |
| 193 | pos += inc; |
| 194 | } |
| 195 | } |
| 196 | |
| 197 | /* Perform a stretch blit between two surfaces of the same format. |
| 198 | NOTE: This function is not safe to call from multiple threads! |
| 199 | */ |
| 200 | int SDL_SoftStretch(SDL_Surface *src, SDL_Rect *srcrect, |
| 201 | SDL_Surface *dst, SDL_Rect *dstrect) |
| 202 | { |
| 203 | int src_locked; |
| 204 | int dst_locked; |
| 205 | int pos, inc; |
| 206 | int dst_width; |
| 207 | int dst_maxrow; |
| 208 | int src_row, dst_row; |
| 209 | Uint8 *srcp = NULL; |
| 210 | Uint8 *dstp; |
| 211 | SDL_Rect full_src; |
| 212 | SDL_Rect full_dst; |
| 213 | #ifdef USE_ASM_STRETCH |
| 214 | SDL_bool use_asm = SDL_TRUE; |
| 215 | #ifdef __GNUC__ |
| 216 | int u1, u2; |
| 217 | #endif |
| 218 | #endif /* USE_ASM_STRETCH */ |
| 219 | const int bpp = dst->format->BytesPerPixel; |
| 220 | |
| 221 | if ( src->format->BitsPerPixel != dst->format->BitsPerPixel ) { |
| 222 | SDL_SetError("Only works with same format surfaces"); |
| 223 | return(-1); |
| 224 | } |
| 225 | |
| 226 | /* Verify the blit rectangles */ |
| 227 | if ( srcrect ) { |
| 228 | if ( (srcrect->x < 0) || (srcrect->y < 0) || |
| 229 | ((srcrect->x+srcrect->w) > src->w) || |
| 230 | ((srcrect->y+srcrect->h) > src->h) ) { |
| 231 | SDL_SetError("Invalid source blit rectangle"); |
| 232 | return(-1); |
| 233 | } |
| 234 | } else { |
| 235 | full_src.x = 0; |
| 236 | full_src.y = 0; |
| 237 | full_src.w = src->w; |
| 238 | full_src.h = src->h; |
| 239 | srcrect = &full_src; |
| 240 | } |
| 241 | if ( dstrect ) { |
| 242 | if ( (dstrect->x < 0) || (dstrect->y < 0) || |
| 243 | ((dstrect->x+dstrect->w) > dst->w) || |
| 244 | ((dstrect->y+dstrect->h) > dst->h) ) { |
| 245 | SDL_SetError("Invalid destination blit rectangle"); |
| 246 | return(-1); |
| 247 | } |
| 248 | } else { |
| 249 | full_dst.x = 0; |
| 250 | full_dst.y = 0; |
| 251 | full_dst.w = dst->w; |
| 252 | full_dst.h = dst->h; |
| 253 | dstrect = &full_dst; |
| 254 | } |
| 255 | |
| 256 | /* Lock the destination if it's in hardware */ |
| 257 | dst_locked = 0; |
| 258 | if ( SDL_MUSTLOCK(dst) ) { |
| 259 | if ( SDL_LockSurface(dst) < 0 ) { |
| 260 | SDL_SetError("Unable to lock destination surface"); |
| 261 | return(-1); |
| 262 | } |
| 263 | dst_locked = 1; |
| 264 | } |
| 265 | /* Lock the source if it's in hardware */ |
| 266 | src_locked = 0; |
| 267 | if ( SDL_MUSTLOCK(src) ) { |
| 268 | if ( SDL_LockSurface(src) < 0 ) { |
| 269 | if ( dst_locked ) { |
| 270 | SDL_UnlockSurface(dst); |
| 271 | } |
| 272 | SDL_SetError("Unable to lock source surface"); |
| 273 | return(-1); |
| 274 | } |
| 275 | src_locked = 1; |
| 276 | } |
| 277 | |
| 278 | /* Set up the data... */ |
| 279 | pos = 0x10000; |
| 280 | inc = (srcrect->h << 16) / dstrect->h; |
| 281 | src_row = srcrect->y; |
| 282 | dst_row = dstrect->y; |
| 283 | dst_width = dstrect->w*bpp; |
| 284 | |
| 285 | #ifdef USE_ASM_STRETCH |
| 286 | /* Write the opcodes for this stretch */ |
| 287 | if ( (bpp == 3) || |
| 288 | (generate_rowbytes(srcrect->w, dstrect->w, bpp) < 0) ) { |
| 289 | use_asm = SDL_FALSE; |
| 290 | } |
| 291 | #endif |
| 292 | |
| 293 | /* Perform the stretch blit */ |
| 294 | for ( dst_maxrow = dst_row+dstrect->h; dst_row<dst_maxrow; ++dst_row ) { |
| 295 | dstp = (Uint8 *)dst->pixels + (dst_row*dst->pitch) |
| 296 | + (dstrect->x*bpp); |
| 297 | while ( pos >= 0x10000L ) { |
| 298 | srcp = (Uint8 *)src->pixels + (src_row*src->pitch) |
| 299 | + (srcrect->x*bpp); |
| 300 | ++src_row; |
| 301 | pos -= 0x10000L; |
| 302 | } |
| 303 | #ifdef USE_ASM_STRETCH |
| 304 | if (use_asm) { |
| 305 | #ifdef __GNUC__ |
| 306 | __asm__ __volatile__ ( |
| 307 | "call *%4" |
| 308 | : "=&D" (u1), "=&S" (u2) |
| 309 | : "0" (dstp), "1" (srcp), "r" (copy_row) |
| 310 | : "memory" ); |
| 311 | #elif defined(_MSC_VER) || defined(__WATCOMC__) |
| 312 | { void *code = copy_row; |
| 313 | __asm { |
| 314 | push edi |
| 315 | push esi |
| 316 | |
| 317 | mov edi, dstp |
| 318 | mov esi, srcp |
| 319 | call dword ptr code |
| 320 | |
| 321 | pop esi |
| 322 | pop edi |
| 323 | } |
| 324 | } |
| 325 | #else |
| 326 | #error Need inline assembly for this compiler |
| 327 | #endif |
| 328 | } else |
| 329 | #endif |
| 330 | switch (bpp) { |
| 331 | case 1: |
| 332 | copy_row1(srcp, srcrect->w, dstp, dstrect->w); |
| 333 | break; |
| 334 | case 2: |
| 335 | copy_row2((Uint16 *)srcp, srcrect->w, |
| 336 | (Uint16 *)dstp, dstrect->w); |
| 337 | break; |
| 338 | case 3: |
| 339 | copy_row3(srcp, srcrect->w, dstp, dstrect->w); |
| 340 | break; |
| 341 | case 4: |
| 342 | copy_row4((Uint32 *)srcp, srcrect->w, |
| 343 | (Uint32 *)dstp, dstrect->w); |
| 344 | break; |
| 345 | } |
| 346 | pos += inc; |
| 347 | } |
| 348 | |
| 349 | /* We need to unlock the surfaces if they're locked */ |
| 350 | if ( dst_locked ) { |
| 351 | SDL_UnlockSurface(dst); |
| 352 | } |
| 353 | if ( src_locked ) { |
| 354 | SDL_UnlockSurface(src); |
| 355 | } |
| 356 | return(0); |
| 357 | } |
| 358 | |