| 1 | /*************************************************************************** |
| 2 | * Copyright (C) 2010 Gabriele Gorla * |
| 3 | * Copyright (C) 2007 Ryan Schultz, PCSX-df Team, PCSX team * |
| 4 | * * |
| 5 | * This program is free software; you can redistribute it and/or modify * |
| 6 | * it under the terms of the GNU General Public License as published by * |
| 7 | * the Free Software Foundation; either version 2 of the License, or * |
| 8 | * (at your option) any later version. * |
| 9 | * * |
| 10 | * This program 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 * |
| 13 | * GNU General Public License for more details. * |
| 14 | * * |
| 15 | * You should have received a copy of the GNU General Public License * |
| 16 | * along with this program; if not, write to the * |
| 17 | * Free Software Foundation, Inc., * |
| 18 | * 51 Franklin Street, Fifth Floor, Boston, MA 02111-1307 USA. * |
| 19 | ***************************************************************************/ |
| 20 | |
| 21 | #include "mdec.h" |
| 22 | |
| 23 | /* memory speed is 1 byte per MDEC_BIAS psx clock |
| 24 | * That mean (PSXCLK / MDEC_BIAS) B/s |
| 25 | * MDEC_BIAS = 2.0 => ~16MB/s |
| 26 | * MDEC_BIAS = 3.0 => ~11MB/s |
| 27 | * and so on ... |
| 28 | * I guess I have 50 images in 50Hz ... (could be 25 images ?) |
| 29 | * 320x240x24@50Hz => 11.52 MB/s |
| 30 | * 320x240x24@60Hz => 13.824 MB/s |
| 31 | * 320x240x16@50Hz => 7.68 MB/s |
| 32 | * 320x240x16@60Hz => 9.216 MB/s |
| 33 | * so 2.0 to 4.0 should be fine. |
| 34 | */ |
| 35 | #define MDEC_BIAS 2 |
| 36 | |
| 37 | #define DSIZE 8 |
| 38 | #define DSIZE2 (DSIZE * DSIZE) |
| 39 | |
| 40 | #define SCALE(x, n) ((x) >> (n)) |
| 41 | #define SCALER(x, n) (((x) + ((1 << (n)) >> 1)) >> (n)) |
| 42 | |
| 43 | #define AAN_CONST_BITS 12 |
| 44 | #define AAN_PRESCALE_BITS 16 |
| 45 | |
| 46 | #define AAN_CONST_SIZE 24 |
| 47 | #define AAN_CONST_SCALE (AAN_CONST_SIZE - AAN_CONST_BITS) |
| 48 | |
| 49 | #define AAN_PRESCALE_SIZE 20 |
| 50 | #define AAN_PRESCALE_SCALE (AAN_PRESCALE_SIZE-AAN_PRESCALE_BITS) |
| 51 | #define AAN_EXTRA 12 |
| 52 | |
| 53 | #define FIX_1_082392200 SCALER(18159528, AAN_CONST_SCALE) // B6 |
| 54 | #define FIX_1_414213562 SCALER(23726566, AAN_CONST_SCALE) // A4 |
| 55 | #define FIX_1_847759065 SCALER(31000253, AAN_CONST_SCALE) // A2 |
| 56 | #define FIX_2_613125930 SCALER(43840978, AAN_CONST_SCALE) // B2 |
| 57 | |
| 58 | #define MULS(var, const) (SCALE((var) * (const), AAN_CONST_BITS)) |
| 59 | |
| 60 | #define RLE_RUN(a) ((a) >> 10) |
| 61 | #define RLE_VAL(a) (((int)(a) << (sizeof(int) * 8 - 10)) >> (sizeof(int) * 8 - 10)) |
| 62 | |
| 63 | #if 0 |
| 64 | static void printmatrixu8(u8 *m) { |
| 65 | int i; |
| 66 | for(i = 0; i < DSIZE2; i++) { |
| 67 | printf("%3d ",m[i]); |
| 68 | if((i+1) % 8 == 0) printf("\n"); |
| 69 | } |
| 70 | } |
| 71 | #endif |
| 72 | |
| 73 | static inline void fillcol(int *blk, int val) { |
| 74 | blk[0 * DSIZE] = blk[1 * DSIZE] = blk[2 * DSIZE] = blk[3 * DSIZE] |
| 75 | = blk[4 * DSIZE] = blk[5 * DSIZE] = blk[6 * DSIZE] = blk[7 * DSIZE] = val; |
| 76 | } |
| 77 | |
| 78 | static inline void fillrow(int *blk, int val) { |
| 79 | blk[0] = blk[1] = blk[2] = blk[3] |
| 80 | = blk[4] = blk[5] = blk[6] = blk[7] = val; |
| 81 | } |
| 82 | |
| 83 | static void idct(int *block,int used_col) { |
| 84 | int tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7; |
| 85 | int z5, z10, z11, z12, z13; |
| 86 | int *ptr; |
| 87 | int i; |
| 88 | |
| 89 | // the block has only the DC coefficient |
| 90 | if (used_col == -1) { |
| 91 | int v = block[0]; |
| 92 | for (i = 0; i < DSIZE2; i++) block[i] = v; |
| 93 | return; |
| 94 | } |
| 95 | |
| 96 | // last_col keeps track of the highest column with non zero coefficients |
| 97 | ptr = block; |
| 98 | for (i = 0; i < DSIZE; i++, ptr++) { |
| 99 | if ((used_col & (1 << i)) == 0) { |
| 100 | // the column is empty or has only the DC coefficient |
| 101 | if (ptr[DSIZE * 0]) { |
| 102 | fillcol(ptr, ptr[0]); |
| 103 | used_col |= (1 << i); |
| 104 | } |
| 105 | continue; |
| 106 | } |
| 107 | |
| 108 | // further optimization could be made by keeping track of |
| 109 | // last_row in rl2blk |
| 110 | z10 = ptr[DSIZE * 0] + ptr[DSIZE * 4]; // s04 |
| 111 | z11 = ptr[DSIZE * 0] - ptr[DSIZE * 4]; // d04 |
| 112 | z13 = ptr[DSIZE * 2] + ptr[DSIZE * 6]; // s26 |
| 113 | z12 = MULS(ptr[DSIZE * 2] - ptr[DSIZE * 6], FIX_1_414213562) - z13; |
| 114 | //^^^^ d26=d26*2*A4-s26 |
| 115 | |
| 116 | tmp0 = z10 + z13; // os07 = s04 + s26 |
| 117 | tmp3 = z10 - z13; // os34 = s04 - s26 |
| 118 | tmp1 = z11 + z12; // os16 = d04 + d26 |
| 119 | tmp2 = z11 - z12; // os25 = d04 - d26 |
| 120 | |
| 121 | z13 = ptr[DSIZE * 3] + ptr[DSIZE * 5]; //s53 |
| 122 | z10 = ptr[DSIZE * 3] - ptr[DSIZE * 5]; //-d53 |
| 123 | z11 = ptr[DSIZE * 1] + ptr[DSIZE * 7]; //s17 |
| 124 | z12 = ptr[DSIZE * 1] - ptr[DSIZE * 7]; //d17 |
| 125 | |
| 126 | tmp7 = z11 + z13; // od07 = s17 + s53 |
| 127 | |
| 128 | z5 = (z12 - z10) * (FIX_1_847759065); |
| 129 | tmp6 = SCALE(z10*(FIX_2_613125930) + z5, AAN_CONST_BITS) - tmp7; |
| 130 | tmp5 = MULS(z11 - z13, FIX_1_414213562) - tmp6; |
| 131 | tmp4 = SCALE(z12*(FIX_1_082392200) - z5, AAN_CONST_BITS) + tmp5; |
| 132 | |
| 133 | // path #1 |
| 134 | //z5 = (z12 - z10)* FIX_1_847759065; |
| 135 | // tmp0 = (d17 + d53) * 2*A2 |
| 136 | |
| 137 | //tmp6 = DESCALE(z10*FIX_2_613125930 + z5, CONST_BITS) - tmp7; |
| 138 | // od16 = (d53*-2*B2 + tmp0) - od07 |
| 139 | |
| 140 | //tmp4 = DESCALE(z12*FIX_1_082392200 - z5, CONST_BITS) + tmp5; |
| 141 | // od34 = (d17*2*B6 - tmp0) + od25 |
| 142 | |
| 143 | // path #2 |
| 144 | |
| 145 | // od34 = d17*2*(B6-A2) - d53*2*A2 |
| 146 | // od16 = d53*2*(A2-B2) + d17*2*A2 |
| 147 | |
| 148 | // end |
| 149 | |
| 150 | // tmp5 = MULS(z11 - z13, FIX_1_414213562) - tmp6; |
| 151 | // od25 = (s17 - s53)*2*A4 - od16 |
| 152 | |
| 153 | ptr[DSIZE * 0] = (tmp0 + tmp7); // os07 + od07 |
| 154 | ptr[DSIZE * 7] = (tmp0 - tmp7); // os07 - od07 |
| 155 | ptr[DSIZE * 1] = (tmp1 + tmp6); // os16 + od16 |
| 156 | ptr[DSIZE * 6] = (tmp1 - tmp6); // os16 - od16 |
| 157 | ptr[DSIZE * 2] = (tmp2 + tmp5); // os25 + od25 |
| 158 | ptr[DSIZE * 5] = (tmp2 - tmp5); // os25 - od25 |
| 159 | ptr[DSIZE * 4] = (tmp3 + tmp4); // os34 + od34 |
| 160 | ptr[DSIZE * 3] = (tmp3 - tmp4); // os34 - od34 |
| 161 | } |
| 162 | |
| 163 | ptr = block; |
| 164 | if (used_col == 1) { |
| 165 | for (i = 0; i < DSIZE; i++) |
| 166 | fillrow(block + DSIZE * i, block[DSIZE * i]); |
| 167 | } else { |
| 168 | for (i = 0; i < DSIZE; i++, ptr += DSIZE) { |
| 169 | z10 = ptr[0] + ptr[4]; |
| 170 | z11 = ptr[0] - ptr[4]; |
| 171 | z13 = ptr[2] + ptr[6]; |
| 172 | z12 = MULS(ptr[2] - ptr[6], FIX_1_414213562) - z13; |
| 173 | |
| 174 | tmp0 = z10 + z13; |
| 175 | tmp3 = z10 - z13; |
| 176 | tmp1 = z11 + z12; |
| 177 | tmp2 = z11 - z12; |
| 178 | |
| 179 | z13 = ptr[3] + ptr[5]; |
| 180 | z10 = ptr[3] - ptr[5]; |
| 181 | z11 = ptr[1] + ptr[7]; |
| 182 | z12 = ptr[1] - ptr[7]; |
| 183 | |
| 184 | tmp7 = z11 + z13; |
| 185 | z5 = (z12 - z10) * FIX_1_847759065; |
| 186 | tmp6 = SCALE(z10 * FIX_2_613125930 + z5, AAN_CONST_BITS) - tmp7; |
| 187 | tmp5 = MULS(z11 - z13, FIX_1_414213562) - tmp6; |
| 188 | tmp4 = SCALE(z12 * FIX_1_082392200 - z5, AAN_CONST_BITS) + tmp5; |
| 189 | |
| 190 | ptr[0] = tmp0 + tmp7; |
| 191 | |
| 192 | ptr[7] = tmp0 - tmp7; |
| 193 | ptr[1] = tmp1 + tmp6; |
| 194 | ptr[6] = tmp1 - tmp6; |
| 195 | ptr[2] = tmp2 + tmp5; |
| 196 | ptr[5] = tmp2 - tmp5; |
| 197 | ptr[4] = tmp3 + tmp4; |
| 198 | ptr[3] = tmp3 - tmp4; |
| 199 | } |
| 200 | } |
| 201 | } |
| 202 | |
| 203 | // mdec0: command register |
| 204 | #define MDEC0_STP 0x02000000 |
| 205 | #define MDEC0_RGB24 0x08000000 |
| 206 | #define MDEC0_SIZE_MASK 0x0000FFFF |
| 207 | |
| 208 | // mdec1: status register |
| 209 | #define MDEC1_BUSY 0x20000000 |
| 210 | #define MDEC1_DREQ 0x18000000 |
| 211 | #define MDEC1_FIFO 0xc0000000 |
| 212 | #define MDEC1_RGB24 0x02000000 |
| 213 | #define MDEC1_STP 0x00800000 |
| 214 | #define MDEC1_RESET 0x80000000 |
| 215 | |
| 216 | struct _pending_dma1 { |
| 217 | u32 adr; |
| 218 | u32 bcr; |
| 219 | u32 chcr; |
| 220 | }; |
| 221 | |
| 222 | static struct { |
| 223 | u32 reg0; |
| 224 | u32 reg1; |
| 225 | u16 * rl; |
| 226 | u16 * rl_end; |
| 227 | u8 * block_buffer_pos; |
| 228 | u8 block_buffer[16*16*3]; |
| 229 | struct _pending_dma1 pending_dma1; |
| 230 | } mdec; |
| 231 | |
| 232 | static int iq_y[DSIZE2], iq_uv[DSIZE2]; |
| 233 | |
| 234 | static int zscan[DSIZE2] = { |
| 235 | 0 , 1 , 8 , 16, 9 , 2 , 3 , 10, |
| 236 | 17, 24, 32, 25, 18, 11, 4 , 5 , |
| 237 | 12, 19, 26, 33, 40, 48, 41, 34, |
| 238 | 27, 20, 13, 6 , 7 , 14, 21, 28, |
| 239 | 35, 42, 49, 56, 57, 50, 43, 36, |
| 240 | 29, 22, 15, 23, 30, 37, 44, 51, |
| 241 | 58, 59, 52, 45, 38, 31, 39, 46, |
| 242 | 53, 60, 61, 54, 47, 55, 62, 63 |
| 243 | }; |
| 244 | |
| 245 | static int aanscales[DSIZE2] = { |
| 246 | 1048576, 1454417, 1370031, 1232995, 1048576, 823861, 567485, 289301, |
| 247 | 1454417, 2017334, 1900287, 1710213, 1454417, 1142728, 787125, 401273, |
| 248 | 1370031, 1900287, 1790031, 1610986, 1370031, 1076426, 741455, 377991, |
| 249 | 1232995, 1710213, 1610986, 1449849, 1232995, 968758, 667292, 340183, |
| 250 | 1048576, 1454417, 1370031, 1232995, 1048576, 823861, 567485, 289301, |
| 251 | 823861, 1142728, 1076426, 968758, 823861, 647303, 445870, 227303, |
| 252 | 567485, 787125, 741455, 667292, 567485, 445870, 307121, 156569, |
| 253 | 289301, 401273, 377991, 340183, 289301, 227303, 156569, 79818 |
| 254 | }; |
| 255 | |
| 256 | static void iqtab_init(int *iqtab, unsigned char *iq_y) { |
| 257 | int i; |
| 258 | |
| 259 | for (i = 0; i < DSIZE2; i++) { |
| 260 | iqtab[i] = (iq_y[i] * SCALER(aanscales[zscan[i]], AAN_PRESCALE_SCALE)); |
| 261 | } |
| 262 | } |
| 263 | |
| 264 | #define MDEC_END_OF_DATA 0xfe00 |
| 265 | |
| 266 | static unsigned short *rl2blk(int *blk, unsigned short *mdec_rl) { |
| 267 | int i, k, q_scale, rl, used_col; |
| 268 | int *iqtab; |
| 269 | |
| 270 | memset(blk, 0, 6 * DSIZE2 * sizeof(int)); |
| 271 | iqtab = iq_uv; |
| 272 | for (i = 0; i < 6; i++) { |
| 273 | // decode blocks (Cr,Cb,Y1,Y2,Y3,Y4) |
| 274 | if (i == 2) iqtab = iq_y; |
| 275 | |
| 276 | rl = SWAP16(*mdec_rl); mdec_rl++; |
| 277 | q_scale = RLE_RUN(rl); |
| 278 | blk[0] = SCALER(iqtab[0] * RLE_VAL(rl), AAN_EXTRA - 3); |
| 279 | for (k = 0, used_col = 0;;) { |
| 280 | rl = SWAP16(*mdec_rl); mdec_rl++; |
| 281 | if (rl == MDEC_END_OF_DATA) break; |
| 282 | k += RLE_RUN(rl) + 1; // skip zero-coefficients |
| 283 | |
| 284 | if (k > 63) { |
| 285 | // printf("run lenght exceeded 64 enties\n"); |
| 286 | break; |
| 287 | } |
| 288 | |
| 289 | // zigzag transformation |
| 290 | blk[zscan[k]] = SCALER(RLE_VAL(rl) * iqtab[k] * q_scale, AAN_EXTRA); |
| 291 | // keep track of used columns to speed up the idtc |
| 292 | used_col |= (zscan[k] > 7) ? 1 << (zscan[k] & 7) : 0; |
| 293 | } |
| 294 | |
| 295 | if (k == 0) used_col = -1; |
| 296 | // used_col is -1 for blocks with only the DC coefficient |
| 297 | // any other value is a bitmask of the columns that have |
| 298 | // at least one non zero cofficient in the rows 1-7 |
| 299 | // single coefficients in row 0 are treted specially |
| 300 | // in the idtc function |
| 301 | idct(blk, used_col); |
| 302 | blk += DSIZE2; |
| 303 | } |
| 304 | return mdec_rl; |
| 305 | } |
| 306 | |
| 307 | // full scale (JPEG) |
| 308 | // Y/Cb/Cr[0...255] -> R/G/B[0...255] |
| 309 | // R = 1.000 * (Y) + 1.400 * (Cr - 128) |
| 310 | // G = 1.000 * (Y) - 0.343 * (Cb - 128) - 0.711 (Cr - 128) |
| 311 | // B = 1.000 * (Y) + 1.765 * (Cb - 128) |
| 312 | #define MULR(a) ((1434 * (a))) |
| 313 | #define MULB(a) ((1807 * (a))) |
| 314 | #define MULG2(a, b) ((-351 * (a) - 728 * (b))) |
| 315 | #define MULY(a) ((a) << 10) |
| 316 | |
| 317 | #define MAKERGB15(r, g, b, a) (SWAP16(a | ((b) << 10) | ((g) << 5) | (r))) |
| 318 | #define SCALE8(c) SCALER(c, 20) |
| 319 | #define SCALE5(c) SCALER(c, 23) |
| 320 | |
| 321 | #define CLAMP5(c) ( ((c) < -16) ? 0 : (((c) > (31 - 16)) ? 31 : ((c) + 16)) ) |
| 322 | #define CLAMP8(c) ( ((c) < -128) ? 0 : (((c) > (255 - 128)) ? 255 : ((c) + 128)) ) |
| 323 | |
| 324 | #define CLAMP_SCALE8(a) (CLAMP8(SCALE8(a))) |
| 325 | #define CLAMP_SCALE5(a) (CLAMP5(SCALE5(a))) |
| 326 | |
| 327 | static inline void putlinebw15(u16 *image, int *Yblk) { |
| 328 | int i; |
| 329 | int A = (mdec.reg0 & MDEC0_STP) ? 0x8000 : 0; |
| 330 | |
| 331 | for (i = 0; i < 8; i++, Yblk++) { |
| 332 | int Y = *Yblk; |
| 333 | // missing rounding |
| 334 | image[i] = SWAP16((CLAMP5(Y >> 3) * 0x421) | A); |
| 335 | } |
| 336 | } |
| 337 | |
| 338 | static inline void putquadrgb15(u16 *image, int *Yblk, int Cr, int Cb) { |
| 339 | int Y, R, G, B; |
| 340 | int A = (mdec.reg0 & MDEC0_STP) ? 0x8000 : 0; |
| 341 | R = MULR(Cr); |
| 342 | G = MULG2(Cb, Cr); |
| 343 | B = MULB(Cb); |
| 344 | |
| 345 | // added transparency |
| 346 | Y = MULY(Yblk[0]); |
| 347 | image[0] = MAKERGB15(CLAMP_SCALE5(Y + R), CLAMP_SCALE5(Y + G), CLAMP_SCALE5(Y + B), A); |
| 348 | Y = MULY(Yblk[1]); |
| 349 | image[1] = MAKERGB15(CLAMP_SCALE5(Y + R), CLAMP_SCALE5(Y + G), CLAMP_SCALE5(Y + B), A); |
| 350 | Y = MULY(Yblk[8]); |
| 351 | image[16] = MAKERGB15(CLAMP_SCALE5(Y + R), CLAMP_SCALE5(Y + G), CLAMP_SCALE5(Y + B), A); |
| 352 | Y = MULY(Yblk[9]); |
| 353 | image[17] = MAKERGB15(CLAMP_SCALE5(Y + R), CLAMP_SCALE5(Y + G), CLAMP_SCALE5(Y + B), A); |
| 354 | } |
| 355 | |
| 356 | static inline void yuv2rgb15(int *blk, unsigned short *image) { |
| 357 | int x, y; |
| 358 | int *Yblk = blk + DSIZE2 * 2; |
| 359 | int *Crblk = blk; |
| 360 | int *Cbblk = blk + DSIZE2; |
| 361 | |
| 362 | if (!Config.Mdec) { |
| 363 | for (y = 0; y < 16; y += 2, Crblk += 4, Cbblk += 4, Yblk += 8, image += 24) { |
| 364 | if (y == 8) Yblk += DSIZE2; |
| 365 | for (x = 0; x < 4; x++, image += 2, Crblk++, Cbblk++, Yblk += 2) { |
| 366 | putquadrgb15(image, Yblk, *Crblk, *Cbblk); |
| 367 | putquadrgb15(image + 8, Yblk + DSIZE2, *(Crblk + 4), *(Cbblk + 4)); |
| 368 | } |
| 369 | } |
| 370 | } else { |
| 371 | for (y = 0; y < 16; y++, Yblk += 8, image += 16) { |
| 372 | if (y == 8) Yblk += DSIZE2; |
| 373 | putlinebw15(image, Yblk); |
| 374 | putlinebw15(image + 8, Yblk + DSIZE2); |
| 375 | } |
| 376 | } |
| 377 | } |
| 378 | |
| 379 | static inline void putlinebw24(u8 * image, int *Yblk) { |
| 380 | int i; |
| 381 | unsigned char Y; |
| 382 | for (i = 0; i < 8 * 3; i += 3, Yblk++) { |
| 383 | Y = CLAMP8(*Yblk); |
| 384 | image[i + 0] = Y; |
| 385 | image[i + 1] = Y; |
| 386 | image[i + 2] = Y; |
| 387 | } |
| 388 | } |
| 389 | |
| 390 | static inline void putquadrgb24(u8 * image, int *Yblk, int Cr, int Cb) { |
| 391 | int Y, R, G, B; |
| 392 | |
| 393 | R = MULR(Cr); |
| 394 | G = MULG2(Cb,Cr); |
| 395 | B = MULB(Cb); |
| 396 | |
| 397 | Y = MULY(Yblk[0]); |
| 398 | image[0 * 3 + 0] = CLAMP_SCALE8(Y + R); |
| 399 | image[0 * 3 + 1] = CLAMP_SCALE8(Y + G); |
| 400 | image[0 * 3 + 2] = CLAMP_SCALE8(Y + B); |
| 401 | Y = MULY(Yblk[1]); |
| 402 | image[1 * 3 + 0] = CLAMP_SCALE8(Y + R); |
| 403 | image[1 * 3 + 1] = CLAMP_SCALE8(Y + G); |
| 404 | image[1 * 3 + 2] = CLAMP_SCALE8(Y + B); |
| 405 | Y = MULY(Yblk[8]); |
| 406 | image[16 * 3 + 0] = CLAMP_SCALE8(Y + R); |
| 407 | image[16 * 3 + 1] = CLAMP_SCALE8(Y + G); |
| 408 | image[16 * 3 + 2] = CLAMP_SCALE8(Y + B); |
| 409 | Y = MULY(Yblk[9]); |
| 410 | image[17 * 3 + 0] = CLAMP_SCALE8(Y + R); |
| 411 | image[17 * 3 + 1] = CLAMP_SCALE8(Y + G); |
| 412 | image[17 * 3 + 2] = CLAMP_SCALE8(Y + B); |
| 413 | } |
| 414 | |
| 415 | static void yuv2rgb24(int *blk, u8 *image) { |
| 416 | int x, y; |
| 417 | int *Yblk = blk + DSIZE2 * 2; |
| 418 | int *Crblk = blk; |
| 419 | int *Cbblk = blk + DSIZE2; |
| 420 | |
| 421 | if (!Config.Mdec) { |
| 422 | for (y = 0; y < 16; y += 2, Crblk += 4, Cbblk += 4, Yblk += 8, image += 8 * 3 * 3) { |
| 423 | if (y == 8) Yblk += DSIZE2; |
| 424 | for (x = 0; x < 4; x++, image += 6, Crblk++, Cbblk++, Yblk += 2) { |
| 425 | putquadrgb24(image, Yblk, *Crblk, *Cbblk); |
| 426 | putquadrgb24(image + 8 * 3, Yblk + DSIZE2, *(Crblk + 4), *(Cbblk + 4)); |
| 427 | } |
| 428 | } |
| 429 | } else { |
| 430 | for (y = 0; y < 16; y++, Yblk += 8, image += 16 * 3) { |
| 431 | if (y == 8) Yblk += DSIZE2; |
| 432 | putlinebw24(image, Yblk); |
| 433 | putlinebw24(image + 8 * 3, Yblk + DSIZE2); |
| 434 | } |
| 435 | } |
| 436 | } |
| 437 | |
| 438 | void mdecInit(void) { |
| 439 | memset(&mdec, 0, sizeof(mdec)); |
| 440 | memset(iq_y, 0, sizeof(iq_y)); |
| 441 | memset(iq_uv, 0, sizeof(iq_uv)); |
| 442 | mdec.rl = (u16 *)&psxM[0x100000]; |
| 443 | } |
| 444 | |
| 445 | // command register |
| 446 | void mdecWrite0(u32 data) { |
| 447 | mdec.reg0 = data; |
| 448 | } |
| 449 | |
| 450 | u32 mdecRead0(void) { |
| 451 | return mdec.reg0; |
| 452 | } |
| 453 | |
| 454 | // status register |
| 455 | void mdecWrite1(u32 data) { |
| 456 | if (data & MDEC1_RESET) { // mdec reset |
| 457 | mdec.reg0 = 0; |
| 458 | mdec.reg1 = 0; |
| 459 | mdec.pending_dma1.adr = 0; |
| 460 | mdec.block_buffer_pos = 0; |
| 461 | } |
| 462 | } |
| 463 | |
| 464 | u32 mdecRead1(void) { |
| 465 | u32 v = mdec.reg1; |
| 466 | return v; |
| 467 | } |
| 468 | |
| 469 | void psxDma0(u32 adr, u32 bcr, u32 chcr) { |
| 470 | int cmd = mdec.reg0; |
| 471 | int size; |
| 472 | |
| 473 | if (chcr != 0x01000201) { |
| 474 | return; |
| 475 | } |
| 476 | |
| 477 | /* mdec is STP till dma0 is released */ |
| 478 | mdec.reg1 |= MDEC1_STP; |
| 479 | |
| 480 | size = (bcr >> 16) * (bcr & 0xffff); |
| 481 | |
| 482 | switch (cmd >> 28) { |
| 483 | case 0x3: // decode |
| 484 | mdec.rl = (u16 *) PSXM(adr); |
| 485 | /* now the mdec is busy till all data are decoded */ |
| 486 | mdec.reg1 |= MDEC1_BUSY; |
| 487 | /* detect the end of decoding */ |
| 488 | mdec.rl_end = mdec.rl + (size * 2); |
| 489 | |
| 490 | /* sanity check */ |
| 491 | if(mdec.rl_end <= mdec.rl) { |
| 492 | MDECINDMA_INT( size / 4 ); |
| 493 | return; |
| 494 | } |
| 495 | |
| 496 | /* process the pending dma1 */ |
| 497 | if(mdec.pending_dma1.adr){ |
| 498 | psxDma1(mdec.pending_dma1.adr, mdec.pending_dma1.bcr, mdec.pending_dma1.chcr); |
| 499 | } |
| 500 | mdec.pending_dma1.adr = 0; |
| 501 | return; |
| 502 | |
| 503 | |
| 504 | case 0x4: // quantization table upload |
| 505 | { |
| 506 | u8 *p = (u8 *)PSXM(adr); |
| 507 | // printf("uploading new quantization table\n"); |
| 508 | // printmatrixu8(p); |
| 509 | // printmatrixu8(p + 64); |
| 510 | iqtab_init(iq_y, p); |
| 511 | iqtab_init(iq_uv, p + 64); |
| 512 | } |
| 513 | |
| 514 | MDECINDMA_INT( size / 4 ); |
| 515 | return; |
| 516 | |
| 517 | case 0x6: // cosine table |
| 518 | // printf("mdec cosine table\n"); |
| 519 | |
| 520 | MDECINDMA_INT( size / 4 ); |
| 521 | return; |
| 522 | |
| 523 | default: |
| 524 | // printf("mdec unknown command\n"); |
| 525 | break; |
| 526 | } |
| 527 | |
| 528 | HW_DMA0_CHCR &= SWAP32(~0x01000000); |
| 529 | DMA_INTERRUPT(0); |
| 530 | } |
| 531 | |
| 532 | void mdec0Interrupt() |
| 533 | { |
| 534 | if (HW_DMA0_CHCR & SWAP32(0x01000000)) |
| 535 | { |
| 536 | HW_DMA0_CHCR &= SWAP32(~0x01000000); |
| 537 | DMA_INTERRUPT(0); |
| 538 | } |
| 539 | } |
| 540 | |
| 541 | #define SIZE_OF_24B_BLOCK (16*16*3) |
| 542 | #define SIZE_OF_16B_BLOCK (16*16*2) |
| 543 | |
| 544 | void psxDma1(u32 adr, u32 bcr, u32 chcr) { |
| 545 | int blk[DSIZE2 * 6]; |
| 546 | u8 * image; |
| 547 | int size; |
| 548 | u32 words; |
| 549 | |
| 550 | if (chcr != 0x01000200) return; |
| 551 | |
| 552 | words = (bcr >> 16) * (bcr & 0xffff); |
| 553 | /* size in byte */ |
| 554 | size = words * 4; |
| 555 | |
| 556 | if (!(mdec.reg1 & MDEC1_BUSY)) { |
| 557 | /* add to pending */ |
| 558 | mdec.pending_dma1.adr = adr; |
| 559 | mdec.pending_dma1.bcr = bcr; |
| 560 | mdec.pending_dma1.chcr = chcr; |
| 561 | /* do not free the dma */ |
| 562 | } else { |
| 563 | |
| 564 | image = (u8 *)PSXM(adr); |
| 565 | |
| 566 | if (mdec.reg0 & MDEC0_RGB24) { |
| 567 | /* 16 bits decoding |
| 568 | * block are 16 px * 16 px, each px are 2 byte |
| 569 | */ |
| 570 | |
| 571 | /* there is some partial block pending ? */ |
| 572 | if(mdec.block_buffer_pos != 0) { |
| 573 | int n = mdec.block_buffer - mdec.block_buffer_pos + SIZE_OF_16B_BLOCK; |
| 574 | /* TODO: check if partial block do not larger than size */ |
| 575 | memcpy(image, mdec.block_buffer_pos, n); |
| 576 | image += n; |
| 577 | size -= n; |
| 578 | mdec.block_buffer_pos = 0; |
| 579 | } |
| 580 | |
| 581 | while(size >= SIZE_OF_16B_BLOCK) { |
| 582 | mdec.rl = rl2blk(blk, mdec.rl); |
| 583 | yuv2rgb15(blk, (u16 *)image); |
| 584 | image += SIZE_OF_16B_BLOCK; |
| 585 | size -= SIZE_OF_16B_BLOCK; |
| 586 | } |
| 587 | |
| 588 | if(size != 0) { |
| 589 | mdec.rl = rl2blk(blk, mdec.rl); |
| 590 | yuv2rgb15(blk, (u16 *)mdec.block_buffer); |
| 591 | memcpy(image, mdec.block_buffer, size); |
| 592 | mdec.block_buffer_pos = mdec.block_buffer + size; |
| 593 | } |
| 594 | |
| 595 | } else { |
| 596 | /* 24 bits decoding |
| 597 | * block are 16 px * 16 px, each px are 3 byte |
| 598 | */ |
| 599 | |
| 600 | /* there is some partial block pending ? */ |
| 601 | if(mdec.block_buffer_pos != 0) { |
| 602 | int n = mdec.block_buffer - mdec.block_buffer_pos + SIZE_OF_24B_BLOCK; |
| 603 | /* TODO: check if partial block do not larger than size */ |
| 604 | memcpy(image, mdec.block_buffer_pos, n); |
| 605 | image += n; |
| 606 | size -= n; |
| 607 | mdec.block_buffer_pos = 0; |
| 608 | } |
| 609 | |
| 610 | while(size >= SIZE_OF_24B_BLOCK) { |
| 611 | mdec.rl = rl2blk(blk, mdec.rl); |
| 612 | yuv2rgb24(blk, image); |
| 613 | image += SIZE_OF_24B_BLOCK; |
| 614 | size -= SIZE_OF_24B_BLOCK; |
| 615 | } |
| 616 | |
| 617 | if(size != 0) { |
| 618 | mdec.rl = rl2blk(blk, mdec.rl); |
| 619 | yuv2rgb24(blk, mdec.block_buffer); |
| 620 | memcpy(image, mdec.block_buffer, size); |
| 621 | mdec.block_buffer_pos = mdec.block_buffer + size; |
| 622 | } |
| 623 | } |
| 624 | |
| 625 | /* define the power of mdec */ |
| 626 | MDECOUTDMA_INT(words * MDEC_BIAS); |
| 627 | } |
| 628 | } |
| 629 | |
| 630 | void mdec1Interrupt() { |
| 631 | /* Author : gschwind |
| 632 | * |
| 633 | * in that case we have done all decoding stuff |
| 634 | * Note that : each block end with 0xfe00 flags |
| 635 | * the list of blocks end with the same 0xfe00 flags |
| 636 | * data loock like : |
| 637 | * |
| 638 | * data block ... |
| 639 | * 0xfe00 |
| 640 | * data block ... |
| 641 | * 0xfe00 |
| 642 | * a lost of block .. |
| 643 | * |
| 644 | * 0xfe00 |
| 645 | * the last block |
| 646 | * 0xfe00 |
| 647 | * 0xfe00 |
| 648 | * |
| 649 | * OR |
| 650 | * |
| 651 | * if the 0xfe00 is not present the data size is important. |
| 652 | * |
| 653 | */ |
| 654 | |
| 655 | /* MDEC_END_OF_DATA avoids read outside memory */ |
| 656 | if (mdec.rl >= mdec.rl_end || SWAP16(*(mdec.rl)) == MDEC_END_OF_DATA) { |
| 657 | mdec.reg1 &= ~(MDEC1_STP|MDEC1_BUSY); |
| 658 | if (HW_DMA0_CHCR & SWAP32(0x01000000)) |
| 659 | { |
| 660 | HW_DMA0_CHCR &= SWAP32(~0x01000000); |
| 661 | DMA_INTERRUPT(0); |
| 662 | } |
| 663 | } |
| 664 | |
| 665 | if (HW_DMA1_CHCR & SWAP32(0x01000000)) |
| 666 | { |
| 667 | HW_DMA1_CHCR &= SWAP32(~0x01000000); |
| 668 | DMA_INTERRUPT(1); |
| 669 | } |
| 670 | } |
| 671 | |
| 672 | int mdecFreeze(void *f, int Mode) { |
| 673 | u8 *base = (u8 *)&psxM[0x100000]; |
| 674 | u32 v; |
| 675 | |
| 676 | gzfreeze(&mdec.reg0, sizeof(mdec.reg0)); |
| 677 | gzfreeze(&mdec.reg1, sizeof(mdec.reg1)); |
| 678 | |
| 679 | // old code used to save raw pointers.. |
| 680 | v = (u8 *)mdec.rl - base; |
| 681 | gzfreeze(&v, sizeof(v)); |
| 682 | mdec.rl = (u16 *)(base + (v & 0xffffe)); |
| 683 | v = (u8 *)mdec.rl_end - base; |
| 684 | gzfreeze(&v, sizeof(v)); |
| 685 | mdec.rl_end = (u16 *)(base + (v & 0xffffe)); |
| 686 | |
| 687 | v = 0; |
| 688 | if (mdec.block_buffer_pos) |
| 689 | v = mdec.block_buffer_pos - base; |
| 690 | gzfreeze(&v, sizeof(v)); |
| 691 | mdec.block_buffer_pos = 0; |
| 692 | if (v) |
| 693 | mdec.block_buffer_pos = base + (v & 0xfffff); |
| 694 | |
| 695 | gzfreeze(&mdec.block_buffer, sizeof(mdec.block_buffer)); |
| 696 | gzfreeze(&mdec.pending_dma1, sizeof(mdec.pending_dma1)); |
| 697 | gzfreeze(iq_y, sizeof(iq_y)); |
| 698 | gzfreeze(iq_uv, sizeof(iq_uv)); |
| 699 | |
| 700 | return 0; |
| 701 | } |