Commit | Line | Data |
---|---|---|
ef79bbde P |
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 | ||
528ad661 | 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 | */ | |
6d9f0559 | 35 | |
36 | /* Was set to 2 before but it would cause issues in R-types and Vandal Hearts videos. | |
37 | * Setting it to 6 as dmitrysmagin did fix those... except for Galerians. | |
38 | * Galerians needs this to be set to 10 (!!) before it looks properly. | |
39 | * I've tried this with a few other games (including R-Types) and so far, this | |
40 | * has not backfired. | |
41 | * */ | |
42 | #define MDEC_BIAS 10 | |
528ad661 | 43 | |
ef79bbde P |
44 | #define DSIZE 8 |
45 | #define DSIZE2 (DSIZE * DSIZE) | |
46 | ||
47 | #define SCALE(x, n) ((x) >> (n)) | |
48 | #define SCALER(x, n) (((x) + ((1 << (n)) >> 1)) >> (n)) | |
49 | ||
50 | #define AAN_CONST_BITS 12 | |
51 | #define AAN_PRESCALE_BITS 16 | |
52 | ||
53 | #define AAN_CONST_SIZE 24 | |
54 | #define AAN_CONST_SCALE (AAN_CONST_SIZE - AAN_CONST_BITS) | |
55 | ||
56 | #define AAN_PRESCALE_SIZE 20 | |
57 | #define AAN_PRESCALE_SCALE (AAN_PRESCALE_SIZE-AAN_PRESCALE_BITS) | |
58 | #define AAN_EXTRA 12 | |
59 | ||
60 | #define FIX_1_082392200 SCALER(18159528, AAN_CONST_SCALE) // B6 | |
61 | #define FIX_1_414213562 SCALER(23726566, AAN_CONST_SCALE) // A4 | |
62 | #define FIX_1_847759065 SCALER(31000253, AAN_CONST_SCALE) // A2 | |
63 | #define FIX_2_613125930 SCALER(43840978, AAN_CONST_SCALE) // B2 | |
64 | ||
65 | #define MULS(var, const) (SCALE((var) * (const), AAN_CONST_BITS)) | |
66 | ||
67 | #define RLE_RUN(a) ((a) >> 10) | |
68 | #define RLE_VAL(a) (((int)(a) << (sizeof(int) * 8 - 10)) >> (sizeof(int) * 8 - 10)) | |
69 | ||
70 | #if 0 | |
71 | static void printmatrixu8(u8 *m) { | |
72 | int i; | |
73 | for(i = 0; i < DSIZE2; i++) { | |
74 | printf("%3d ",m[i]); | |
75 | if((i+1) % 8 == 0) printf("\n"); | |
76 | } | |
77 | } | |
78 | #endif | |
79 | ||
80 | static inline void fillcol(int *blk, int val) { | |
81 | blk[0 * DSIZE] = blk[1 * DSIZE] = blk[2 * DSIZE] = blk[3 * DSIZE] | |
82 | = blk[4 * DSIZE] = blk[5 * DSIZE] = blk[6 * DSIZE] = blk[7 * DSIZE] = val; | |
83 | } | |
84 | ||
85 | static inline void fillrow(int *blk, int val) { | |
86 | blk[0] = blk[1] = blk[2] = blk[3] | |
87 | = blk[4] = blk[5] = blk[6] = blk[7] = val; | |
88 | } | |
89 | ||
528ad661 | 90 | static void idct(int *block,int used_col) { |
ef79bbde P |
91 | int tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7; |
92 | int z5, z10, z11, z12, z13; | |
93 | int *ptr; | |
94 | int i; | |
95 | ||
96 | // the block has only the DC coefficient | |
97 | if (used_col == -1) { | |
98 | int v = block[0]; | |
99 | for (i = 0; i < DSIZE2; i++) block[i] = v; | |
100 | return; | |
101 | } | |
102 | ||
103 | // last_col keeps track of the highest column with non zero coefficients | |
104 | ptr = block; | |
105 | for (i = 0; i < DSIZE; i++, ptr++) { | |
106 | if ((used_col & (1 << i)) == 0) { | |
107 | // the column is empty or has only the DC coefficient | |
108 | if (ptr[DSIZE * 0]) { | |
109 | fillcol(ptr, ptr[0]); | |
110 | used_col |= (1 << i); | |
111 | } | |
112 | continue; | |
113 | } | |
114 | ||
115 | // further optimization could be made by keeping track of | |
116 | // last_row in rl2blk | |
117 | z10 = ptr[DSIZE * 0] + ptr[DSIZE * 4]; // s04 | |
118 | z11 = ptr[DSIZE * 0] - ptr[DSIZE * 4]; // d04 | |
119 | z13 = ptr[DSIZE * 2] + ptr[DSIZE * 6]; // s26 | |
120 | z12 = MULS(ptr[DSIZE * 2] - ptr[DSIZE * 6], FIX_1_414213562) - z13; | |
121 | //^^^^ d26=d26*2*A4-s26 | |
122 | ||
123 | tmp0 = z10 + z13; // os07 = s04 + s26 | |
124 | tmp3 = z10 - z13; // os34 = s04 - s26 | |
125 | tmp1 = z11 + z12; // os16 = d04 + d26 | |
126 | tmp2 = z11 - z12; // os25 = d04 - d26 | |
127 | ||
128 | z13 = ptr[DSIZE * 3] + ptr[DSIZE * 5]; //s53 | |
129 | z10 = ptr[DSIZE * 3] - ptr[DSIZE * 5]; //-d53 | |
130 | z11 = ptr[DSIZE * 1] + ptr[DSIZE * 7]; //s17 | |
131 | z12 = ptr[DSIZE * 1] - ptr[DSIZE * 7]; //d17 | |
132 | ||
133 | tmp7 = z11 + z13; // od07 = s17 + s53 | |
134 | ||
135 | z5 = (z12 - z10) * (FIX_1_847759065); | |
136 | tmp6 = SCALE(z10*(FIX_2_613125930) + z5, AAN_CONST_BITS) - tmp7; | |
137 | tmp5 = MULS(z11 - z13, FIX_1_414213562) - tmp6; | |
138 | tmp4 = SCALE(z12*(FIX_1_082392200) - z5, AAN_CONST_BITS) + tmp5; | |
139 | ||
140 | // path #1 | |
141 | //z5 = (z12 - z10)* FIX_1_847759065; | |
142 | // tmp0 = (d17 + d53) * 2*A2 | |
143 | ||
144 | //tmp6 = DESCALE(z10*FIX_2_613125930 + z5, CONST_BITS) - tmp7; | |
145 | // od16 = (d53*-2*B2 + tmp0) - od07 | |
146 | ||
147 | //tmp4 = DESCALE(z12*FIX_1_082392200 - z5, CONST_BITS) + tmp5; | |
148 | // od34 = (d17*2*B6 - tmp0) + od25 | |
149 | ||
150 | // path #2 | |
151 | ||
152 | // od34 = d17*2*(B6-A2) - d53*2*A2 | |
153 | // od16 = d53*2*(A2-B2) + d17*2*A2 | |
154 | ||
155 | // end | |
156 | ||
157 | // tmp5 = MULS(z11 - z13, FIX_1_414213562) - tmp6; | |
158 | // od25 = (s17 - s53)*2*A4 - od16 | |
159 | ||
160 | ptr[DSIZE * 0] = (tmp0 + tmp7); // os07 + od07 | |
161 | ptr[DSIZE * 7] = (tmp0 - tmp7); // os07 - od07 | |
162 | ptr[DSIZE * 1] = (tmp1 + tmp6); // os16 + od16 | |
163 | ptr[DSIZE * 6] = (tmp1 - tmp6); // os16 - od16 | |
164 | ptr[DSIZE * 2] = (tmp2 + tmp5); // os25 + od25 | |
165 | ptr[DSIZE * 5] = (tmp2 - tmp5); // os25 - od25 | |
166 | ptr[DSIZE * 4] = (tmp3 + tmp4); // os34 + od34 | |
167 | ptr[DSIZE * 3] = (tmp3 - tmp4); // os34 - od34 | |
168 | } | |
169 | ||
170 | ptr = block; | |
171 | if (used_col == 1) { | |
172 | for (i = 0; i < DSIZE; i++) | |
173 | fillrow(block + DSIZE * i, block[DSIZE * i]); | |
174 | } else { | |
175 | for (i = 0; i < DSIZE; i++, ptr += DSIZE) { | |
176 | z10 = ptr[0] + ptr[4]; | |
177 | z11 = ptr[0] - ptr[4]; | |
178 | z13 = ptr[2] + ptr[6]; | |
179 | z12 = MULS(ptr[2] - ptr[6], FIX_1_414213562) - z13; | |
180 | ||
181 | tmp0 = z10 + z13; | |
182 | tmp3 = z10 - z13; | |
183 | tmp1 = z11 + z12; | |
184 | tmp2 = z11 - z12; | |
185 | ||
186 | z13 = ptr[3] + ptr[5]; | |
187 | z10 = ptr[3] - ptr[5]; | |
188 | z11 = ptr[1] + ptr[7]; | |
189 | z12 = ptr[1] - ptr[7]; | |
190 | ||
191 | tmp7 = z11 + z13; | |
192 | z5 = (z12 - z10) * FIX_1_847759065; | |
193 | tmp6 = SCALE(z10 * FIX_2_613125930 + z5, AAN_CONST_BITS) - tmp7; | |
194 | tmp5 = MULS(z11 - z13, FIX_1_414213562) - tmp6; | |
195 | tmp4 = SCALE(z12 * FIX_1_082392200 - z5, AAN_CONST_BITS) + tmp5; | |
196 | ||
197 | ptr[0] = tmp0 + tmp7; | |
198 | ||
199 | ptr[7] = tmp0 - tmp7; | |
200 | ptr[1] = tmp1 + tmp6; | |
201 | ptr[6] = tmp1 - tmp6; | |
202 | ptr[2] = tmp2 + tmp5; | |
203 | ptr[5] = tmp2 - tmp5; | |
204 | ptr[4] = tmp3 + tmp4; | |
205 | ptr[3] = tmp3 - tmp4; | |
206 | } | |
207 | } | |
208 | } | |
209 | ||
210 | // mdec0: command register | |
211 | #define MDEC0_STP 0x02000000 | |
212 | #define MDEC0_RGB24 0x08000000 | |
528ad661 | 213 | #define MDEC0_SIZE_MASK 0x0000FFFF |
ef79bbde P |
214 | |
215 | // mdec1: status register | |
216 | #define MDEC1_BUSY 0x20000000 | |
217 | #define MDEC1_DREQ 0x18000000 | |
218 | #define MDEC1_FIFO 0xc0000000 | |
219 | #define MDEC1_RGB24 0x02000000 | |
220 | #define MDEC1_STP 0x00800000 | |
221 | #define MDEC1_RESET 0x80000000 | |
222 | ||
528ad661 | 223 | struct _pending_dma1 { |
224 | u32 adr; | |
225 | u32 bcr; | |
226 | u32 chcr; | |
227 | }; | |
228 | ||
c62b43c9 | 229 | static struct { |
528ad661 | 230 | u32 reg0; |
231 | u32 reg1; | |
232 | u16 * rl; | |
233 | u16 * rl_end; | |
234 | u8 * block_buffer_pos; | |
235 | u8 block_buffer[16*16*3]; | |
236 | struct _pending_dma1 pending_dma1; | |
ef79bbde P |
237 | } mdec; |
238 | ||
239 | static int iq_y[DSIZE2], iq_uv[DSIZE2]; | |
240 | ||
241 | static int zscan[DSIZE2] = { | |
242 | 0 , 1 , 8 , 16, 9 , 2 , 3 , 10, | |
243 | 17, 24, 32, 25, 18, 11, 4 , 5 , | |
244 | 12, 19, 26, 33, 40, 48, 41, 34, | |
245 | 27, 20, 13, 6 , 7 , 14, 21, 28, | |
246 | 35, 42, 49, 56, 57, 50, 43, 36, | |
247 | 29, 22, 15, 23, 30, 37, 44, 51, | |
248 | 58, 59, 52, 45, 38, 31, 39, 46, | |
249 | 53, 60, 61, 54, 47, 55, 62, 63 | |
250 | }; | |
251 | ||
252 | static int aanscales[DSIZE2] = { | |
253 | 1048576, 1454417, 1370031, 1232995, 1048576, 823861, 567485, 289301, | |
254 | 1454417, 2017334, 1900287, 1710213, 1454417, 1142728, 787125, 401273, | |
255 | 1370031, 1900287, 1790031, 1610986, 1370031, 1076426, 741455, 377991, | |
256 | 1232995, 1710213, 1610986, 1449849, 1232995, 968758, 667292, 340183, | |
257 | 1048576, 1454417, 1370031, 1232995, 1048576, 823861, 567485, 289301, | |
258 | 823861, 1142728, 1076426, 968758, 823861, 647303, 445870, 227303, | |
259 | 567485, 787125, 741455, 667292, 567485, 445870, 307121, 156569, | |
260 | 289301, 401273, 377991, 340183, 289301, 227303, 156569, 79818 | |
261 | }; | |
262 | ||
263 | static void iqtab_init(int *iqtab, unsigned char *iq_y) { | |
264 | int i; | |
265 | ||
266 | for (i = 0; i < DSIZE2; i++) { | |
267 | iqtab[i] = (iq_y[i] * SCALER(aanscales[zscan[i]], AAN_PRESCALE_SCALE)); | |
268 | } | |
269 | } | |
270 | ||
271 | #define MDEC_END_OF_DATA 0xfe00 | |
272 | ||
528ad661 | 273 | static unsigned short *rl2blk(int *blk, unsigned short *mdec_rl) { |
ef79bbde P |
274 | int i, k, q_scale, rl, used_col; |
275 | int *iqtab; | |
276 | ||
277 | memset(blk, 0, 6 * DSIZE2 * sizeof(int)); | |
278 | iqtab = iq_uv; | |
279 | for (i = 0; i < 6; i++) { | |
280 | // decode blocks (Cr,Cb,Y1,Y2,Y3,Y4) | |
281 | if (i == 2) iqtab = iq_y; | |
282 | ||
283 | rl = SWAP16(*mdec_rl); mdec_rl++; | |
284 | q_scale = RLE_RUN(rl); | |
285 | blk[0] = SCALER(iqtab[0] * RLE_VAL(rl), AAN_EXTRA - 3); | |
286 | for (k = 0, used_col = 0;;) { | |
287 | rl = SWAP16(*mdec_rl); mdec_rl++; | |
288 | if (rl == MDEC_END_OF_DATA) break; | |
289 | k += RLE_RUN(rl) + 1; // skip zero-coefficients | |
290 | ||
291 | if (k > 63) { | |
292 | // printf("run lenght exceeded 64 enties\n"); | |
293 | break; | |
294 | } | |
295 | ||
296 | // zigzag transformation | |
297 | blk[zscan[k]] = SCALER(RLE_VAL(rl) * iqtab[k] * q_scale, AAN_EXTRA); | |
298 | // keep track of used columns to speed up the idtc | |
299 | used_col |= (zscan[k] > 7) ? 1 << (zscan[k] & 7) : 0; | |
300 | } | |
301 | ||
302 | if (k == 0) used_col = -1; | |
303 | // used_col is -1 for blocks with only the DC coefficient | |
304 | // any other value is a bitmask of the columns that have | |
305 | // at least one non zero cofficient in the rows 1-7 | |
306 | // single coefficients in row 0 are treted specially | |
307 | // in the idtc function | |
308 | idct(blk, used_col); | |
309 | blk += DSIZE2; | |
310 | } | |
311 | return mdec_rl; | |
312 | } | |
313 | ||
314 | // full scale (JPEG) | |
315 | // Y/Cb/Cr[0...255] -> R/G/B[0...255] | |
316 | // R = 1.000 * (Y) + 1.400 * (Cr - 128) | |
317 | // G = 1.000 * (Y) - 0.343 * (Cb - 128) - 0.711 (Cr - 128) | |
318 | // B = 1.000 * (Y) + 1.765 * (Cb - 128) | |
319 | #define MULR(a) ((1434 * (a))) | |
320 | #define MULB(a) ((1807 * (a))) | |
321 | #define MULG2(a, b) ((-351 * (a) - 728 * (b))) | |
322 | #define MULY(a) ((a) << 10) | |
323 | ||
324 | #define MAKERGB15(r, g, b, a) (SWAP16(a | ((b) << 10) | ((g) << 5) | (r))) | |
325 | #define SCALE8(c) SCALER(c, 20) | |
326 | #define SCALE5(c) SCALER(c, 23) | |
327 | ||
328 | #define CLAMP5(c) ( ((c) < -16) ? 0 : (((c) > (31 - 16)) ? 31 : ((c) + 16)) ) | |
329 | #define CLAMP8(c) ( ((c) < -128) ? 0 : (((c) > (255 - 128)) ? 255 : ((c) + 128)) ) | |
330 | ||
331 | #define CLAMP_SCALE8(a) (CLAMP8(SCALE8(a))) | |
332 | #define CLAMP_SCALE5(a) (CLAMP5(SCALE5(a))) | |
333 | ||
528ad661 | 334 | static inline void putlinebw15(u16 *image, int *Yblk) { |
ef79bbde P |
335 | int i; |
336 | int A = (mdec.reg0 & MDEC0_STP) ? 0x8000 : 0; | |
337 | ||
338 | for (i = 0; i < 8; i++, Yblk++) { | |
339 | int Y = *Yblk; | |
340 | // missing rounding | |
341 | image[i] = SWAP16((CLAMP5(Y >> 3) * 0x421) | A); | |
342 | } | |
343 | } | |
344 | ||
528ad661 | 345 | static inline void putquadrgb15(u16 *image, int *Yblk, int Cr, int Cb) { |
ef79bbde P |
346 | int Y, R, G, B; |
347 | int A = (mdec.reg0 & MDEC0_STP) ? 0x8000 : 0; | |
348 | R = MULR(Cr); | |
349 | G = MULG2(Cb, Cr); | |
350 | B = MULB(Cb); | |
351 | ||
352 | // added transparency | |
353 | Y = MULY(Yblk[0]); | |
354 | image[0] = MAKERGB15(CLAMP_SCALE5(Y + R), CLAMP_SCALE5(Y + G), CLAMP_SCALE5(Y + B), A); | |
355 | Y = MULY(Yblk[1]); | |
356 | image[1] = MAKERGB15(CLAMP_SCALE5(Y + R), CLAMP_SCALE5(Y + G), CLAMP_SCALE5(Y + B), A); | |
357 | Y = MULY(Yblk[8]); | |
358 | image[16] = MAKERGB15(CLAMP_SCALE5(Y + R), CLAMP_SCALE5(Y + G), CLAMP_SCALE5(Y + B), A); | |
359 | Y = MULY(Yblk[9]); | |
360 | image[17] = MAKERGB15(CLAMP_SCALE5(Y + R), CLAMP_SCALE5(Y + G), CLAMP_SCALE5(Y + B), A); | |
361 | } | |
362 | ||
528ad661 | 363 | static inline void yuv2rgb15(int *blk, unsigned short *image) { |
ef79bbde P |
364 | int x, y; |
365 | int *Yblk = blk + DSIZE2 * 2; | |
366 | int *Crblk = blk; | |
367 | int *Cbblk = blk + DSIZE2; | |
368 | ||
369 | if (!Config.Mdec) { | |
370 | for (y = 0; y < 16; y += 2, Crblk += 4, Cbblk += 4, Yblk += 8, image += 24) { | |
371 | if (y == 8) Yblk += DSIZE2; | |
372 | for (x = 0; x < 4; x++, image += 2, Crblk++, Cbblk++, Yblk += 2) { | |
373 | putquadrgb15(image, Yblk, *Crblk, *Cbblk); | |
374 | putquadrgb15(image + 8, Yblk + DSIZE2, *(Crblk + 4), *(Cbblk + 4)); | |
375 | } | |
376 | } | |
377 | } else { | |
378 | for (y = 0; y < 16; y++, Yblk += 8, image += 16) { | |
379 | if (y == 8) Yblk += DSIZE2; | |
380 | putlinebw15(image, Yblk); | |
381 | putlinebw15(image + 8, Yblk + DSIZE2); | |
382 | } | |
383 | } | |
384 | } | |
385 | ||
528ad661 | 386 | static inline void putlinebw24(u8 * image, int *Yblk) { |
ef79bbde P |
387 | int i; |
388 | unsigned char Y; | |
389 | for (i = 0; i < 8 * 3; i += 3, Yblk++) { | |
390 | Y = CLAMP8(*Yblk); | |
391 | image[i + 0] = Y; | |
392 | image[i + 1] = Y; | |
393 | image[i + 2] = Y; | |
394 | } | |
395 | } | |
396 | ||
528ad661 | 397 | static inline void putquadrgb24(u8 * image, int *Yblk, int Cr, int Cb) { |
ef79bbde P |
398 | int Y, R, G, B; |
399 | ||
400 | R = MULR(Cr); | |
401 | G = MULG2(Cb,Cr); | |
402 | B = MULB(Cb); | |
403 | ||
404 | Y = MULY(Yblk[0]); | |
405 | image[0 * 3 + 0] = CLAMP_SCALE8(Y + R); | |
406 | image[0 * 3 + 1] = CLAMP_SCALE8(Y + G); | |
407 | image[0 * 3 + 2] = CLAMP_SCALE8(Y + B); | |
408 | Y = MULY(Yblk[1]); | |
409 | image[1 * 3 + 0] = CLAMP_SCALE8(Y + R); | |
410 | image[1 * 3 + 1] = CLAMP_SCALE8(Y + G); | |
411 | image[1 * 3 + 2] = CLAMP_SCALE8(Y + B); | |
412 | Y = MULY(Yblk[8]); | |
413 | image[16 * 3 + 0] = CLAMP_SCALE8(Y + R); | |
414 | image[16 * 3 + 1] = CLAMP_SCALE8(Y + G); | |
415 | image[16 * 3 + 2] = CLAMP_SCALE8(Y + B); | |
416 | Y = MULY(Yblk[9]); | |
417 | image[17 * 3 + 0] = CLAMP_SCALE8(Y + R); | |
418 | image[17 * 3 + 1] = CLAMP_SCALE8(Y + G); | |
419 | image[17 * 3 + 2] = CLAMP_SCALE8(Y + B); | |
420 | } | |
421 | ||
528ad661 | 422 | static void yuv2rgb24(int *blk, u8 *image) { |
ef79bbde P |
423 | int x, y; |
424 | int *Yblk = blk + DSIZE2 * 2; | |
425 | int *Crblk = blk; | |
426 | int *Cbblk = blk + DSIZE2; | |
427 | ||
428 | if (!Config.Mdec) { | |
528ad661 | 429 | for (y = 0; y < 16; y += 2, Crblk += 4, Cbblk += 4, Yblk += 8, image += 8 * 3 * 3) { |
ef79bbde P |
430 | if (y == 8) Yblk += DSIZE2; |
431 | for (x = 0; x < 4; x++, image += 6, Crblk++, Cbblk++, Yblk += 2) { | |
432 | putquadrgb24(image, Yblk, *Crblk, *Cbblk); | |
433 | putquadrgb24(image + 8 * 3, Yblk + DSIZE2, *(Crblk + 4), *(Cbblk + 4)); | |
434 | } | |
435 | } | |
436 | } else { | |
437 | for (y = 0; y < 16; y++, Yblk += 8, image += 16 * 3) { | |
438 | if (y == 8) Yblk += DSIZE2; | |
439 | putlinebw24(image, Yblk); | |
440 | putlinebw24(image + 8 * 3, Yblk + DSIZE2); | |
441 | } | |
442 | } | |
443 | } | |
444 | ||
445 | void mdecInit(void) { | |
c62b43c9 | 446 | memset(&mdec, 0, sizeof(mdec)); |
447 | memset(iq_y, 0, sizeof(iq_y)); | |
448 | memset(iq_uv, 0, sizeof(iq_uv)); | |
ef79bbde | 449 | mdec.rl = (u16 *)&psxM[0x100000]; |
ef79bbde P |
450 | } |
451 | ||
452 | // command register | |
453 | void mdecWrite0(u32 data) { | |
ef79bbde P |
454 | mdec.reg0 = data; |
455 | } | |
456 | ||
457 | u32 mdecRead0(void) { | |
ef79bbde P |
458 | return mdec.reg0; |
459 | } | |
460 | ||
461 | // status register | |
462 | void mdecWrite1(u32 data) { | |
ef79bbde P |
463 | if (data & MDEC1_RESET) { // mdec reset |
464 | mdec.reg0 = 0; | |
465 | mdec.reg1 = 0; | |
528ad661 | 466 | mdec.pending_dma1.adr = 0; |
467 | mdec.block_buffer_pos = 0; | |
ef79bbde P |
468 | } |
469 | } | |
470 | ||
471 | u32 mdecRead1(void) { | |
472 | u32 v = mdec.reg1; | |
ef79bbde P |
473 | return v; |
474 | } | |
475 | ||
476 | void psxDma0(u32 adr, u32 bcr, u32 chcr) { | |
477 | int cmd = mdec.reg0; | |
478 | int size; | |
ef79bbde P |
479 | |
480 | if (chcr != 0x01000201) { | |
ef79bbde P |
481 | return; |
482 | } | |
483 | ||
528ad661 | 484 | /* mdec is STP till dma0 is released */ |
485 | mdec.reg1 |= MDEC1_STP; | |
486 | ||
ef79bbde P |
487 | size = (bcr >> 16) * (bcr & 0xffff); |
488 | ||
489 | switch (cmd >> 28) { | |
490 | case 0x3: // decode | |
528ad661 | 491 | mdec.rl = (u16 *) PSXM(adr); |
492 | /* now the mdec is busy till all data are decoded */ | |
493 | mdec.reg1 |= MDEC1_BUSY; | |
494 | /* detect the end of decoding */ | |
495 | mdec.rl_end = mdec.rl + (size * 2); | |
496 | ||
497 | /* sanity check */ | |
498 | if(mdec.rl_end <= mdec.rl) { | |
499 | MDECINDMA_INT( size / 4 ); | |
500 | return; | |
501 | } | |
502 | ||
503 | /* process the pending dma1 */ | |
504 | if(mdec.pending_dma1.adr){ | |
505 | psxDma1(mdec.pending_dma1.adr, mdec.pending_dma1.bcr, mdec.pending_dma1.chcr); | |
506 | } | |
507 | mdec.pending_dma1.adr = 0; | |
508 | return; | |
509 | ||
ef79bbde P |
510 | |
511 | case 0x4: // quantization table upload | |
512 | { | |
513 | u8 *p = (u8 *)PSXM(adr); | |
514 | // printf("uploading new quantization table\n"); | |
515 | // printmatrixu8(p); | |
516 | // printmatrixu8(p + 64); | |
517 | iqtab_init(iq_y, p); | |
518 | iqtab_init(iq_uv, p + 64); | |
519 | } | |
528ad661 | 520 | |
521 | MDECINDMA_INT( size / 4 ); | |
522 | return; | |
ef79bbde P |
523 | |
524 | case 0x6: // cosine table | |
525 | // printf("mdec cosine table\n"); | |
528ad661 | 526 | |
527 | MDECINDMA_INT( size / 4 ); | |
528 | return; | |
ef79bbde P |
529 | |
530 | default: | |
531 | // printf("mdec unknown command\n"); | |
532 | break; | |
533 | } | |
534 | ||
535 | HW_DMA0_CHCR &= SWAP32(~0x01000000); | |
536 | DMA_INTERRUPT(0); | |
537 | } | |
538 | ||
528ad661 | 539 | void mdec0Interrupt() |
540 | { | |
ad418c19 | 541 | if (HW_DMA0_CHCR & SWAP32(0x01000000)) |
542 | { | |
543 | HW_DMA0_CHCR &= SWAP32(~0x01000000); | |
544 | DMA_INTERRUPT(0); | |
545 | } | |
528ad661 | 546 | } |
547 | ||
548 | #define SIZE_OF_24B_BLOCK (16*16*3) | |
549 | #define SIZE_OF_16B_BLOCK (16*16*2) | |
550 | ||
ef79bbde P |
551 | void psxDma1(u32 adr, u32 bcr, u32 chcr) { |
552 | int blk[DSIZE2 * 6]; | |
528ad661 | 553 | u8 * image; |
ef79bbde | 554 | int size; |
fc4803bd | 555 | u32 words; |
ef79bbde P |
556 | |
557 | if (chcr != 0x01000200) return; | |
558 | ||
fc4803bd | 559 | words = (bcr >> 16) * (bcr & 0xffff); |
528ad661 | 560 | /* size in byte */ |
fc4803bd | 561 | size = words * 4; |
528ad661 | 562 | |
563 | if (!(mdec.reg1 & MDEC1_BUSY)) { | |
564 | /* add to pending */ | |
565 | mdec.pending_dma1.adr = adr; | |
566 | mdec.pending_dma1.bcr = bcr; | |
567 | mdec.pending_dma1.chcr = chcr; | |
568 | /* do not free the dma */ | |
569 | } else { | |
ef79bbde | 570 | |
528ad661 | 571 | image = (u8 *)PSXM(adr); |
572 | ||
573 | if (mdec.reg0 & MDEC0_RGB24) { | |
574 | /* 16 bits decoding | |
575 | * block are 16 px * 16 px, each px are 2 byte | |
576 | */ | |
577 | ||
578 | /* there is some partial block pending ? */ | |
579 | if(mdec.block_buffer_pos != 0) { | |
580 | int n = mdec.block_buffer - mdec.block_buffer_pos + SIZE_OF_16B_BLOCK; | |
581 | /* TODO: check if partial block do not larger than size */ | |
582 | memcpy(image, mdec.block_buffer_pos, n); | |
583 | image += n; | |
584 | size -= n; | |
585 | mdec.block_buffer_pos = 0; | |
586 | } | |
ef79bbde | 587 | |
528ad661 | 588 | while(size >= SIZE_OF_16B_BLOCK) { |
ef79bbde | 589 | mdec.rl = rl2blk(blk, mdec.rl); |
528ad661 | 590 | yuv2rgb15(blk, (u16 *)image); |
591 | image += SIZE_OF_16B_BLOCK; | |
592 | size -= SIZE_OF_16B_BLOCK; | |
ef79bbde | 593 | } |
528ad661 | 594 | |
595 | if(size != 0) { | |
ef79bbde | 596 | mdec.rl = rl2blk(blk, mdec.rl); |
528ad661 | 597 | yuv2rgb15(blk, (u16 *)mdec.block_buffer); |
598 | memcpy(image, mdec.block_buffer, size); | |
599 | mdec.block_buffer_pos = mdec.block_buffer + size; | |
600 | } | |
601 | ||
602 | } else { | |
603 | /* 24 bits decoding | |
604 | * block are 16 px * 16 px, each px are 3 byte | |
605 | */ | |
606 | ||
607 | /* there is some partial block pending ? */ | |
608 | if(mdec.block_buffer_pos != 0) { | |
609 | int n = mdec.block_buffer - mdec.block_buffer_pos + SIZE_OF_24B_BLOCK; | |
610 | /* TODO: check if partial block do not larger than size */ | |
611 | memcpy(image, mdec.block_buffer_pos, n); | |
612 | image += n; | |
613 | size -= n; | |
614 | mdec.block_buffer_pos = 0; | |
615 | } | |
616 | ||
617 | while(size >= SIZE_OF_24B_BLOCK) { | |
618 | mdec.rl = rl2blk(blk, mdec.rl); | |
619 | yuv2rgb24(blk, image); | |
620 | image += SIZE_OF_24B_BLOCK; | |
621 | size -= SIZE_OF_24B_BLOCK; | |
ef79bbde | 622 | } |
ef79bbde | 623 | |
528ad661 | 624 | if(size != 0) { |
625 | mdec.rl = rl2blk(blk, mdec.rl); | |
626 | yuv2rgb24(blk, mdec.block_buffer); | |
627 | memcpy(image, mdec.block_buffer, size); | |
628 | mdec.block_buffer_pos = mdec.block_buffer + size; | |
629 | } | |
630 | } | |
631 | ||
632 | /* define the power of mdec */ | |
fc4803bd | 633 | MDECOUTDMA_INT(words * MDEC_BIAS); |
528ad661 | 634 | } |
ef79bbde P |
635 | } |
636 | ||
637 | void mdec1Interrupt() { | |
528ad661 | 638 | /* Author : gschwind |
639 | * | |
640 | * in that case we have done all decoding stuff | |
641 | * Note that : each block end with 0xfe00 flags | |
642 | * the list of blocks end with the same 0xfe00 flags | |
643 | * data loock like : | |
644 | * | |
645 | * data block ... | |
646 | * 0xfe00 | |
647 | * data block ... | |
648 | * 0xfe00 | |
649 | * a lost of block .. | |
650 | * | |
651 | * 0xfe00 | |
652 | * the last block | |
653 | * 0xfe00 | |
654 | * 0xfe00 | |
655 | * | |
656 | * OR | |
657 | * | |
658 | * if the 0xfe00 is not present the data size is important. | |
659 | * | |
660 | */ | |
661 | ||
ad418c19 | 662 | /* MDEC_END_OF_DATA avoids read outside memory */ |
663 | if (mdec.rl >= mdec.rl_end || SWAP16(*(mdec.rl)) == MDEC_END_OF_DATA) { | |
664 | mdec.reg1 &= ~(MDEC1_STP|MDEC1_BUSY); | |
665 | if (HW_DMA0_CHCR & SWAP32(0x01000000)) | |
666 | { | |
667 | HW_DMA0_CHCR &= SWAP32(~0x01000000); | |
668 | DMA_INTERRUPT(0); | |
669 | } | |
ef79bbde | 670 | } |
528ad661 | 671 | |
ad418c19 | 672 | if (HW_DMA1_CHCR & SWAP32(0x01000000)) |
673 | { | |
674 | HW_DMA1_CHCR &= SWAP32(~0x01000000); | |
675 | DMA_INTERRUPT(1); | |
676 | } | |
ef79bbde P |
677 | } |
678 | ||
496d88d4 | 679 | int mdecFreeze(void *f, int Mode) { |
8e1040b6 | 680 | u8 *base = (u8 *)&psxM[0x100000]; |
681 | u32 v; | |
682 | ||
683 | gzfreeze(&mdec.reg0, sizeof(mdec.reg0)); | |
684 | gzfreeze(&mdec.reg1, sizeof(mdec.reg1)); | |
685 | ||
686 | // old code used to save raw pointers.. | |
687 | v = (u8 *)mdec.rl - base; | |
688 | gzfreeze(&v, sizeof(v)); | |
689 | mdec.rl = (u16 *)(base + (v & 0xffffe)); | |
690 | v = (u8 *)mdec.rl_end - base; | |
691 | gzfreeze(&v, sizeof(v)); | |
692 | mdec.rl_end = (u16 *)(base + (v & 0xffffe)); | |
693 | ||
694 | v = 0; | |
695 | if (mdec.block_buffer_pos) | |
696 | v = mdec.block_buffer_pos - base; | |
697 | gzfreeze(&v, sizeof(v)); | |
698 | mdec.block_buffer_pos = 0; | |
699 | if (v) | |
700 | mdec.block_buffer_pos = base + (v & 0xfffff); | |
701 | ||
702 | gzfreeze(&mdec.block_buffer, sizeof(mdec.block_buffer)); | |
703 | gzfreeze(&mdec.pending_dma1, sizeof(mdec.pending_dma1)); | |
ef79bbde P |
704 | gzfreeze(iq_y, sizeof(iq_y)); |
705 | gzfreeze(iq_uv, sizeof(iq_uv)); | |
706 | ||
707 | return 0; | |
708 | } |