| 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 is the Playstation 2 implementation of YUV video overlays */ |
| 25 | |
| 26 | #include <fcntl.h> |
| 27 | #include <unistd.h> |
| 28 | #include <sys/ioctl.h> |
| 29 | #include <sys/mman.h> |
| 30 | #include <asm/page.h> /* For definition of PAGE_SIZE */ |
| 31 | |
| 32 | #include "SDL_video.h" |
| 33 | #include "SDL_gsyuv_c.h" |
| 34 | #include "../SDL_yuvfuncs.h" |
| 35 | |
| 36 | /* The maximum number of 16x16 pixel block converted at once */ |
| 37 | #define MAX_MACROBLOCKS 1024 /* 2^10 macroblocks at once */ |
| 38 | |
| 39 | /* The functions used to manipulate video overlays */ |
| 40 | static struct private_yuvhwfuncs gs_yuvfuncs = { |
| 41 | GS_LockYUVOverlay, |
| 42 | GS_UnlockYUVOverlay, |
| 43 | GS_DisplayYUVOverlay, |
| 44 | GS_FreeYUVOverlay |
| 45 | }; |
| 46 | |
| 47 | struct private_yuvhwdata { |
| 48 | int ipu_fd; |
| 49 | Uint8 *pixels; |
| 50 | int macroblocks; |
| 51 | int dma_len; |
| 52 | caddr_t dma_mem; |
| 53 | caddr_t ipu_imem; |
| 54 | caddr_t ipu_omem; |
| 55 | caddr_t dma_tags; |
| 56 | unsigned long long *stretch_x1y1; |
| 57 | unsigned long long *stretch_x2y2; |
| 58 | struct ps2_plist plist; |
| 59 | |
| 60 | /* These are just so we don't have to allocate them separately */ |
| 61 | Uint16 pitches[3]; |
| 62 | Uint8 *planes[3]; |
| 63 | }; |
| 64 | |
| 65 | static int power_of_2(int value) |
| 66 | { |
| 67 | int shift; |
| 68 | |
| 69 | for ( shift = 0; (1<<shift) < value; ++shift ) { |
| 70 | /* Keep looking */ ; |
| 71 | } |
| 72 | return(shift); |
| 73 | } |
| 74 | |
| 75 | SDL_Overlay *GS_CreateYUVOverlay(_THIS, int width, int height, Uint32 format, SDL_Surface *display) |
| 76 | { |
| 77 | SDL_Overlay *overlay; |
| 78 | struct private_yuvhwdata *hwdata; |
| 79 | int map_offset; |
| 80 | unsigned long long *tags; |
| 81 | caddr_t base; |
| 82 | int bpp; |
| 83 | int fbp, fbw, psm; |
| 84 | int x, y, w, h; |
| 85 | int pnum; |
| 86 | struct ps2_packet *packet; |
| 87 | struct ps2_packet tex_packet; |
| 88 | |
| 89 | /* We can only decode blocks of 16x16 pixels */ |
| 90 | if ( (width & 15) || (height & 15) ) { |
| 91 | SDL_SetError("Overlay width/height must be multiples of 16"); |
| 92 | return(NULL); |
| 93 | } |
| 94 | /* Make sure the image isn't too large for a single DMA transfer */ |
| 95 | if ( ((width/16) * (height/16)) > MAX_MACROBLOCKS ) { |
| 96 | SDL_SetError("Overlay too large (maximum size: %d pixels)", |
| 97 | MAX_MACROBLOCKS * 16 * 16); |
| 98 | return(NULL); |
| 99 | } |
| 100 | |
| 101 | /* Double-check the requested format. For simplicity, we'll only |
| 102 | support planar YUV formats. |
| 103 | */ |
| 104 | switch (format) { |
| 105 | case SDL_YV12_OVERLAY: |
| 106 | case SDL_IYUV_OVERLAY: |
| 107 | /* Supported planar YUV format */ |
| 108 | break; |
| 109 | default: |
| 110 | SDL_SetError("Unsupported YUV format"); |
| 111 | return(NULL); |
| 112 | } |
| 113 | |
| 114 | /* Create the overlay structure */ |
| 115 | overlay = (SDL_Overlay *)SDL_malloc(sizeof *overlay); |
| 116 | if ( overlay == NULL ) { |
| 117 | SDL_OutOfMemory(); |
| 118 | return(NULL); |
| 119 | } |
| 120 | SDL_memset(overlay, 0, (sizeof *overlay)); |
| 121 | |
| 122 | /* Fill in the basic members */ |
| 123 | overlay->format = format; |
| 124 | overlay->w = width; |
| 125 | overlay->h = height; |
| 126 | |
| 127 | /* Set up the YUV surface function structure */ |
| 128 | overlay->hwfuncs = &gs_yuvfuncs; |
| 129 | overlay->hw_overlay = 1; |
| 130 | |
| 131 | /* Create the pixel data */ |
| 132 | hwdata = (struct private_yuvhwdata *)SDL_malloc(sizeof *hwdata); |
| 133 | overlay->hwdata = hwdata; |
| 134 | if ( hwdata == NULL ) { |
| 135 | SDL_FreeYUVOverlay(overlay); |
| 136 | SDL_OutOfMemory(); |
| 137 | return(NULL); |
| 138 | } |
| 139 | hwdata->ipu_fd = -1; |
| 140 | hwdata->pixels = (Uint8 *)SDL_malloc(width*height*2); |
| 141 | if ( hwdata->pixels == NULL ) { |
| 142 | SDL_FreeYUVOverlay(overlay); |
| 143 | SDL_OutOfMemory(); |
| 144 | return(NULL); |
| 145 | } |
| 146 | hwdata->macroblocks = (width/16) * (height/16); |
| 147 | |
| 148 | /* Find the pitch and offset values for the overlay */ |
| 149 | overlay->pitches = hwdata->pitches; |
| 150 | overlay->pixels = hwdata->planes; |
| 151 | switch (format) { |
| 152 | case SDL_YV12_OVERLAY: |
| 153 | case SDL_IYUV_OVERLAY: |
| 154 | overlay->pitches[0] = overlay->w; |
| 155 | overlay->pitches[1] = overlay->pitches[0] / 2; |
| 156 | overlay->pitches[2] = overlay->pitches[0] / 2; |
| 157 | overlay->pixels[0] = hwdata->pixels; |
| 158 | overlay->pixels[1] = overlay->pixels[0] + |
| 159 | overlay->pitches[0] * overlay->h; |
| 160 | overlay->pixels[2] = overlay->pixels[1] + |
| 161 | overlay->pitches[1] * overlay->h / 2; |
| 162 | overlay->planes = 3; |
| 163 | break; |
| 164 | default: |
| 165 | /* We should never get here (caught above) */ |
| 166 | break; |
| 167 | } |
| 168 | |
| 169 | /* Theoretically we could support several concurrent decode |
| 170 | streams queueing up on the same file descriptor, but for |
| 171 | simplicity we'll support only one. Opening the IPU more |
| 172 | than once will fail with EBUSY. |
| 173 | */ |
| 174 | hwdata->ipu_fd = open("/dev/ps2ipu", O_RDWR); |
| 175 | if ( hwdata->ipu_fd < 0 ) { |
| 176 | SDL_FreeYUVOverlay(overlay); |
| 177 | SDL_SetError("Playstation 2 IPU busy"); |
| 178 | return(NULL); |
| 179 | } |
| 180 | |
| 181 | /* Allocate a DMA area for pixel conversion */ |
| 182 | bpp = this->screen->format->BytesPerPixel; |
| 183 | map_offset = (mapped_len + (sysconf(_SC_PAGESIZE) - 1)) & ~(sysconf(_SC_PAGESIZE) - 1); |
| 184 | hwdata->dma_len = hwdata->macroblocks * (16 * 16 + 8 * 8 + 8 * 8) + |
| 185 | width * height * bpp + |
| 186 | hwdata->macroblocks * (16 * sizeof(long long)) + |
| 187 | 12 * sizeof(long long); |
| 188 | hwdata->dma_mem = mmap(0, hwdata->dma_len, PROT_READ|PROT_WRITE, |
| 189 | MAP_SHARED, memory_fd, map_offset); |
| 190 | if ( hwdata->dma_mem == MAP_FAILED ) { |
| 191 | hwdata->ipu_imem = (caddr_t)0; |
| 192 | SDL_FreeYUVOverlay(overlay); |
| 193 | SDL_SetError("Unable to map %d bytes for DMA", hwdata->dma_len); |
| 194 | return(NULL); |
| 195 | } |
| 196 | hwdata->ipu_imem = hwdata->dma_mem; |
| 197 | hwdata->ipu_omem = hwdata->ipu_imem + |
| 198 | hwdata->macroblocks * (16 * 16 + 8 * 8 + 8 * 8); |
| 199 | hwdata->dma_tags = hwdata->ipu_omem + width * height * bpp; |
| 200 | |
| 201 | /* Allocate memory for the DMA packets */ |
| 202 | hwdata->plist.num = hwdata->macroblocks * 4 + 1; |
| 203 | hwdata->plist.packet = (struct ps2_packet *)SDL_malloc( |
| 204 | hwdata->plist.num*sizeof(struct ps2_packet)); |
| 205 | if ( ! hwdata->plist.packet ) { |
| 206 | SDL_FreeYUVOverlay(overlay); |
| 207 | SDL_OutOfMemory(); |
| 208 | return(NULL); |
| 209 | } |
| 210 | pnum = 0; |
| 211 | packet = hwdata->plist.packet; |
| 212 | |
| 213 | /* Set up the tags to send the image to the screen */ |
| 214 | tags = (unsigned long long *)hwdata->dma_tags; |
| 215 | base = hwdata->ipu_omem; |
| 216 | fbp = screen_image.fbp; |
| 217 | fbw = screen_image.fbw; |
| 218 | psm = screen_image.psm; |
| 219 | y = screen_image.y + screen_image.h; /* Offscreen video memory */ |
| 220 | for ( h=height/16; h; --h ) { |
| 221 | x = 0; /* Visible video memory */ |
| 222 | for ( w=width/16; w; --w ) { |
| 223 | /* The head tag */ |
| 224 | packet[pnum].ptr = &tags[0]; |
| 225 | packet[pnum].len = 10 * sizeof(*tags); |
| 226 | ++pnum; |
| 227 | tags[0] = 4 | (1LL << 60); /* GIFtag */ |
| 228 | tags[1] = 0x0e; /* A+D */ |
| 229 | tags[2] = ((unsigned long long)fbp << 32) | |
| 230 | ((unsigned long long)fbw << 48) | |
| 231 | ((unsigned long long)psm << 56); |
| 232 | tags[3] = PS2_GS_BITBLTBUF; |
| 233 | tags[4] = ((unsigned long long)x << 32) | |
| 234 | ((unsigned long long)y << 48); |
| 235 | tags[5] = PS2_GS_TRXPOS; |
| 236 | tags[6] = (unsigned long long)16 | |
| 237 | ((unsigned long long)16 << 32); |
| 238 | tags[7] = PS2_GS_TRXREG; |
| 239 | tags[8] = 0; |
| 240 | tags[9] = PS2_GS_TRXDIR; |
| 241 | /* Now the actual image data */ |
| 242 | packet[pnum].ptr = &tags[10]; |
| 243 | packet[pnum].len = 2 * sizeof(*tags); |
| 244 | ++pnum; |
| 245 | tags[10] = ((16*16*bpp) >> 4) | (2LL << 58); |
| 246 | tags[11] = 0; |
| 247 | packet[pnum].ptr = (void *)base; |
| 248 | packet[pnum].len = 16 * 16 * bpp; |
| 249 | ++pnum; |
| 250 | packet[pnum].ptr = &tags[12]; |
| 251 | packet[pnum].len = 2 * sizeof(*tags); |
| 252 | ++pnum; |
| 253 | tags[12] = (0 >> 4) | (1 << 15) | (2LL << 58); |
| 254 | tags[13] = 0; |
| 255 | |
| 256 | tags += 16; |
| 257 | base += 16 * 16 * bpp; |
| 258 | |
| 259 | x += 16; |
| 260 | } |
| 261 | y += 16; |
| 262 | } |
| 263 | |
| 264 | /* Set up the texture memory area for the video */ |
| 265 | tex_packet.ptr = tags; |
| 266 | tex_packet.len = 8 * sizeof(*tags); |
| 267 | tags[0] = 3 | (1LL << 60); /* GIFtag */ |
| 268 | tags[1] = 0x0e; /* A+D */ |
| 269 | tags[2] = ((screen_image.y + screen_image.h) * screen_image.w) / 64 + |
| 270 | ((unsigned long long)fbw << 14) + |
| 271 | ((unsigned long long)psm << 20) + |
| 272 | ((unsigned long long)power_of_2(width) << 26) + |
| 273 | ((unsigned long long)power_of_2(height) << 30) + |
| 274 | ((unsigned long long)1 << 34) + |
| 275 | ((unsigned long long)1 << 35); |
| 276 | tags[3] = PS2_GS_TEX0_1; |
| 277 | tags[4] = (1 << 5) + (1 << 6); |
| 278 | tags[5] = PS2_GS_TEX1_1; |
| 279 | tags[6] = 0; |
| 280 | tags[7] = PS2_GS_TEXFLUSH; |
| 281 | ioctl(console_fd, PS2IOC_SEND, &tex_packet); |
| 282 | |
| 283 | /* Set up the tags for scaling the image */ |
| 284 | packet[pnum].ptr = tags; |
| 285 | packet[pnum].len = 12 * sizeof(*tags); |
| 286 | ++pnum; |
| 287 | tags[0] = 5 | (1LL << 60); /* GIFtag */ |
| 288 | tags[1] = 0x0e; /* A+D */ |
| 289 | tags[2] = 6 + (1 << 4) + (1 << 8); |
| 290 | tags[3] = PS2_GS_PRIM; |
| 291 | tags[4] = ((unsigned long long)0 * 16) + |
| 292 | (((unsigned long long)0 * 16) << 16); |
| 293 | tags[5] = PS2_GS_UV; |
| 294 | tags[6] = 0; /* X1, Y1 */ |
| 295 | tags[7] = PS2_GS_XYZ2; |
| 296 | hwdata->stretch_x1y1 = &tags[6]; |
| 297 | tags[8] = ((unsigned long long)overlay->w * 16) + |
| 298 | (((unsigned long long)overlay->h * 16) << 16); |
| 299 | tags[9] = PS2_GS_UV; |
| 300 | tags[10] = 0; /* X2, Y2 */ |
| 301 | tags[11] = PS2_GS_XYZ2; |
| 302 | hwdata->stretch_x2y2 = &tags[10]; |
| 303 | |
| 304 | /* We're all done.. */ |
| 305 | return(overlay); |
| 306 | } |
| 307 | |
| 308 | int GS_LockYUVOverlay(_THIS, SDL_Overlay *overlay) |
| 309 | { |
| 310 | return(0); |
| 311 | } |
| 312 | |
| 313 | void GS_UnlockYUVOverlay(_THIS, SDL_Overlay *overlay) |
| 314 | { |
| 315 | return; |
| 316 | } |
| 317 | |
| 318 | int GS_DisplayYUVOverlay(_THIS, SDL_Overlay *overlay, SDL_Rect *src, SDL_Rect *dst) |
| 319 | { |
| 320 | struct private_yuvhwdata *hwdata; |
| 321 | __u32 cmd; |
| 322 | struct ps2_packet packet; |
| 323 | int h, w, i; |
| 324 | Uint32 *lum, *Cr, *Cb; |
| 325 | int lum_pitch; |
| 326 | int crb_pitch; |
| 327 | Uint32 *lum_src, *Cr_src, *Cb_src; |
| 328 | Uint32 *srcp, *dstp; |
| 329 | unsigned int x, y; |
| 330 | SDL_Surface *screen; |
| 331 | |
| 332 | /* Find out where the various portions of the image are */ |
| 333 | hwdata = overlay->hwdata; |
| 334 | switch (overlay->format) { |
| 335 | case SDL_YV12_OVERLAY: |
| 336 | lum = (Uint32 *)overlay->pixels[0]; |
| 337 | Cr = (Uint32 *)overlay->pixels[1]; |
| 338 | Cb = (Uint32 *)overlay->pixels[2]; |
| 339 | break; |
| 340 | case SDL_IYUV_OVERLAY: |
| 341 | lum = (Uint32 *)overlay->pixels[0]; |
| 342 | Cr = (Uint32 *)overlay->pixels[2]; |
| 343 | Cb = (Uint32 *)overlay->pixels[1]; |
| 344 | default: |
| 345 | SDL_SetError("Unsupported YUV format in blit (?)"); |
| 346 | return(-1); |
| 347 | } |
| 348 | dstp = (Uint32 *)hwdata->ipu_imem; |
| 349 | lum_pitch = overlay->w/4; |
| 350 | crb_pitch = (overlay->w/2)/4; |
| 351 | |
| 352 | /* Copy blocks of 16x16 pixels to the DMA area */ |
| 353 | for ( h=overlay->h/16; h; --h ) { |
| 354 | lum_src = lum; |
| 355 | Cr_src = Cr; |
| 356 | Cb_src = Cb; |
| 357 | for ( w=overlay->w/16; w; --w ) { |
| 358 | srcp = lum_src; |
| 359 | for ( i=0; i<16; ++i ) { |
| 360 | dstp[0] = srcp[0]; |
| 361 | dstp[1] = srcp[1]; |
| 362 | dstp[2] = srcp[2]; |
| 363 | dstp[3] = srcp[3]; |
| 364 | srcp += lum_pitch; |
| 365 | dstp += 4; |
| 366 | } |
| 367 | srcp = Cb_src; |
| 368 | for ( i=0; i<8; ++i ) { |
| 369 | dstp[0] = srcp[0]; |
| 370 | dstp[1] = srcp[1]; |
| 371 | srcp += crb_pitch; |
| 372 | dstp += 2; |
| 373 | } |
| 374 | srcp = Cr_src; |
| 375 | for ( i=0; i<8; ++i ) { |
| 376 | dstp[0] = srcp[0]; |
| 377 | dstp[1] = srcp[1]; |
| 378 | srcp += crb_pitch; |
| 379 | dstp += 2; |
| 380 | } |
| 381 | lum_src += 16 / 4; |
| 382 | Cb_src += 8 / 4; |
| 383 | Cr_src += 8 / 4; |
| 384 | } |
| 385 | lum += lum_pitch * 16; |
| 386 | Cr += crb_pitch * 8; |
| 387 | Cb += crb_pitch * 8; |
| 388 | } |
| 389 | |
| 390 | /* Send the macroblock data to the IPU */ |
| 391 | #ifdef DEBUG_YUV |
| 392 | fprintf(stderr, "Sending data to IPU..\n"); |
| 393 | #endif |
| 394 | packet.ptr = hwdata->ipu_imem; |
| 395 | packet.len = hwdata->macroblocks * (16 * 16 + 8 * 8 + 8 * 8); |
| 396 | ioctl(hwdata->ipu_fd, PS2IOC_SENDA, &packet); |
| 397 | |
| 398 | /* Trigger the DMA to the IPU for conversion */ |
| 399 | #ifdef DEBUG_YUV |
| 400 | fprintf(stderr, "Trigging conversion command\n"); |
| 401 | #endif |
| 402 | cmd = (7 << 28) + hwdata->macroblocks; |
| 403 | if ( screen_image.psm == PS2_GS_PSMCT16 ) { |
| 404 | cmd += (1 << 27) + /* Output RGB 555 */ |
| 405 | (1 << 26); /* Dither output */ |
| 406 | } |
| 407 | ioctl(hwdata->ipu_fd, PS2IOC_SIPUCMD, &cmd); |
| 408 | |
| 409 | /* Retrieve the converted image from the IPU */ |
| 410 | #ifdef DEBUG_YUV |
| 411 | fprintf(stderr, "Retrieving data from IPU..\n"); |
| 412 | #endif |
| 413 | packet.ptr = hwdata->ipu_omem; |
| 414 | packet.len = overlay->w * overlay->h * |
| 415 | this->screen->format->BytesPerPixel; |
| 416 | ioctl(hwdata->ipu_fd, PS2IOC_RECV, &packet); |
| 417 | |
| 418 | #ifdef DEBUG_YUV |
| 419 | fprintf(stderr, "Copying image to screen..\n"); |
| 420 | #endif |
| 421 | /* Wait for previous DMA to complete */ |
| 422 | ioctl(console_fd, PS2IOC_SENDQCT, 1); |
| 423 | |
| 424 | /* Send the current image to the screen and scale it */ |
| 425 | screen = this->screen; |
| 426 | x = (unsigned int)dst->x; |
| 427 | y = (unsigned int)dst->y; |
| 428 | if ( screen->offset ) { |
| 429 | x += (screen->offset % screen->pitch) / |
| 430 | screen->format->BytesPerPixel; |
| 431 | y += (screen->offset / screen->pitch); |
| 432 | } |
| 433 | y += screen_image.y; |
| 434 | *hwdata->stretch_x1y1 = (x * 16) + ((y * 16) << 16); |
| 435 | x += (unsigned int)dst->w; |
| 436 | y += (unsigned int)dst->h; |
| 437 | *hwdata->stretch_x2y2 = (x * 16) + ((y * 16) << 16); |
| 438 | return ioctl(console_fd, PS2IOC_SENDL, &hwdata->plist); |
| 439 | } |
| 440 | |
| 441 | void GS_FreeYUVOverlay(_THIS, SDL_Overlay *overlay) |
| 442 | { |
| 443 | struct private_yuvhwdata *hwdata; |
| 444 | |
| 445 | hwdata = overlay->hwdata; |
| 446 | if ( hwdata ) { |
| 447 | if ( hwdata->ipu_fd >= 0 ) { |
| 448 | close(hwdata->ipu_fd); |
| 449 | } |
| 450 | if ( hwdata->dma_mem ) { |
| 451 | munmap(hwdata->dma_mem, hwdata->dma_len); |
| 452 | } |
| 453 | if ( hwdata->plist.packet ) { |
| 454 | SDL_free(hwdata->plist.packet); |
| 455 | } |
| 456 | if ( hwdata->pixels ) { |
| 457 | SDL_free(hwdata->pixels); |
| 458 | } |
| 459 | SDL_free(hwdata); |
| 460 | } |
| 461 | } |