src/gba/gba-bios.c (view raw)
1#include "gba-bios.h"
2
3#include "gba.h"
4#include "gba-io.h"
5#include "gba-memory.h"
6
7const uint32_t GBA_BIOS_CHECKSUM = 0xBAAE187F;
8const uint32_t GBA_DS_BIOS_CHECKSUM = 0xBAAE1880;
9
10static void _unLz77(struct GBA* gba, uint32_t source, uint8_t* dest);
11static void _unHuffman(struct GBA* gba, uint32_t source, uint32_t* dest);
12static void _unRl(struct GBA* gba, uint32_t source, uint8_t* dest);
13
14static void _RegisterRamReset(struct GBA* gba) {
15 uint32_t registers = gba->cpu->gprs[0];
16 (void)(registers);
17 GBALog(gba, GBA_LOG_STUB, "RegisterRamReset unimplemented");
18}
19
20static void _CpuSet(struct GBA* gba) {
21 struct ARMCore* cpu = gba->cpu;
22 uint32_t source = cpu->gprs[0];
23 uint32_t dest = cpu->gprs[1];
24 uint32_t mode = cpu->gprs[2];
25 int count = mode & 0x000FFFFF;
26 int fill = mode & 0x01000000;
27 int wordsize = (mode & 0x04000000) ? 4 : 2;
28 int i;
29 if (fill) {
30 if (wordsize == 4) {
31 source &= 0xFFFFFFFC;
32 dest &= 0xFFFFFFFC;
33 int32_t word = cpu->memory.load32(cpu, source, &cpu->cycles);
34 for (i = 0; i < count; ++i) {
35 cpu->memory.store32(cpu, dest + (i << 2), word, &cpu->cycles);
36 cpu->irqh.processEvents(cpu);
37 }
38 } else {
39 source &= 0xFFFFFFFE;
40 dest &= 0xFFFFFFFE;
41 uint16_t word = cpu->memory.load16(cpu, source, &cpu->cycles);
42 for (i = 0; i < count; ++i) {
43 cpu->memory.store16(cpu, dest + (i << 1), word, &cpu->cycles);
44 cpu->irqh.processEvents(cpu);
45 }
46 }
47 } else {
48 if (wordsize == 4) {
49 source &= 0xFFFFFFFC;
50 dest &= 0xFFFFFFFC;
51 for (i = 0; i < count; ++i) {
52 int32_t word = cpu->memory.load32(cpu, source + (i << 2), &cpu->cycles);
53 cpu->memory.store32(cpu, dest + (i << 2), word, &cpu->cycles);
54 cpu->irqh.processEvents(cpu);
55 }
56 } else {
57 source &= 0xFFFFFFFE;
58 dest &= 0xFFFFFFFE;
59 for (i = 0; i < count; ++i) {
60 uint16_t word = cpu->memory.load16(cpu, source + (i << 1), &cpu->cycles);
61 cpu->memory.store16(cpu, dest + (i << 1), word, &cpu->cycles);
62 cpu->irqh.processEvents(cpu);
63 }
64 }
65 }
66}
67
68static void _FastCpuSet(struct GBA* gba) {
69 struct ARMCore* cpu = gba->cpu;
70 uint32_t source = cpu->gprs[0] & 0xFFFFFFFC;
71 uint32_t dest = cpu->gprs[1] & 0xFFFFFFFC;
72 uint32_t mode = cpu->gprs[2];
73 int count = mode & 0x000FFFFF;
74 int storeCycles = cpu->memory.waitMultiple(cpu, dest, 4);
75 count = ((count + 7) >> 3) << 3;
76 int i;
77 if (mode & 0x01000000) {
78 int32_t word = cpu->memory.load32(cpu, source, &cpu->cycles);
79 for (i = 0; i < count; i += 4) {
80 cpu->memory.store32(cpu, dest + ((i + 0) << 2), word, 0);
81 cpu->memory.store32(cpu, dest + ((i + 1) << 2), word, 0);
82 cpu->memory.store32(cpu, dest + ((i + 2) << 2), word, 0);
83 cpu->memory.store32(cpu, dest + ((i + 3) << 2), word, 0);
84 cpu->cycles += storeCycles;
85 cpu->irqh.processEvents(cpu);
86 }
87 } else {
88 int loadCycles = cpu->memory.waitMultiple(cpu, source, 4);
89 for (i = 0; i < count; i += 4) {
90 int32_t word0 = cpu->memory.load32(cpu, source + ((i + 0) << 2), 0);
91 int32_t word1 = cpu->memory.load32(cpu, source + ((i + 1) << 2), 0);
92 int32_t word2 = cpu->memory.load32(cpu, source + ((i + 2) << 2), 0);
93 int32_t word3 = cpu->memory.load32(cpu, source + ((i + 3) << 2), 0);
94 cpu->cycles += loadCycles;
95 cpu->irqh.processEvents(cpu);
96 cpu->memory.store32(cpu, dest + ((i + 0) << 2), word0, 0);
97 cpu->memory.store32(cpu, dest + ((i + 1) << 2), word1, 0);
98 cpu->memory.store32(cpu, dest + ((i + 2) << 2), word2, 0);
99 cpu->memory.store32(cpu, dest + ((i + 3) << 2), word3, 0);
100 cpu->cycles += storeCycles;
101 cpu->irqh.processEvents(cpu);
102 }
103 }
104}
105
106static void _BgAffineSet(struct GBA* gba) {
107 struct ARMCore* cpu = gba->cpu;
108 int i = cpu->gprs[2];
109 float ox, oy;
110 float cx, cy;
111 float sx, sy;
112 float theta;
113 int offset = cpu->gprs[0];
114 int destination = cpu->gprs[1];
115 int diff = cpu->gprs[3];
116 (void)(diff); // Are we supposed to use this?
117 float a, b, c, d;
118 float rx, ry;
119 while (i--) {
120 // [ sx 0 0 ] [ cos(theta) -sin(theta) 0 ] [ 1 0 cx - ox ] [ A B rx ]
121 // [ 0 sy 0 ] * [ sin(theta) cos(theta) 0 ] * [ 0 1 cy - oy ] = [ C D ry ]
122 // [ 0 0 1 ] [ 0 0 1 ] [ 0 0 1 ] [ 0 0 1 ]
123 ox = cpu->memory.load32(cpu, offset, 0) / 256.f;
124 oy = cpu->memory.load32(cpu, offset + 4, 0) / 256.f;
125 cx = cpu->memory.load16(cpu, offset + 8, 0);
126 cy = cpu->memory.load16(cpu, offset + 10, 0);
127 sx = cpu->memory.load16(cpu, offset + 12, 0) / 256.f;
128 sy = cpu->memory.load16(cpu, offset + 14, 0) / 256.f;
129 theta = (cpu->memory.loadU16(cpu, offset + 16, 0) >> 8) / 128.f * M_PI;
130 offset += 20;
131 // Rotation
132 a = d = cosf(theta);
133 b = c = sinf(theta);
134 // Scale
135 a *= sx;
136 b *= -sx;
137 c *= sy;
138 d *= sy;
139 // Translate
140 rx = ox - (a * cx + b * cy);
141 ry = oy - (c * cx + d * cy);
142 cpu->memory.store16(cpu, destination, a * 256, 0);
143 cpu->memory.store16(cpu, destination + 2, b * 256, 0);
144 cpu->memory.store16(cpu, destination + 4, c * 256, 0);
145 cpu->memory.store16(cpu, destination + 6, d * 256, 0);
146 cpu->memory.store32(cpu, destination + 8, rx * 256, 0);
147 cpu->memory.store32(cpu, destination + 12, ry * 256, 0);
148 destination += 16;
149 }
150}
151
152static void _ObjAffineSet(struct GBA* gba) {
153 struct ARMCore* cpu = gba->cpu;
154 int i = cpu->gprs[2];
155 float sx, sy;
156 float theta;
157 int offset = cpu->gprs[0];
158 int destination = cpu->gprs[1];
159 int diff = cpu->gprs[3];
160 float a, b, c, d;
161 while (i--) {
162 // [ sx 0 ] [ cos(theta) -sin(theta) ] [ A B ]
163 // [ 0 sy ] * [ sin(theta) cos(theta) ] = [ C D ]
164 sx = cpu->memory.load16(cpu, offset, 0) / 256.f;
165 sy = cpu->memory.load16(cpu, offset + 2, 0) / 256.f;
166 theta = (cpu->memory.loadU16(cpu, offset + 4, 0) >> 8) / 128.f * M_PI;
167 offset += 6;
168 // Rotation
169 a = d = cosf(theta);
170 b = c = sinf(theta);
171 // Scale
172 a *= sx;
173 b *= -sx;
174 c *= sy;
175 d *= sy;
176 cpu->memory.store16(cpu, destination, a * 256, 0);
177 cpu->memory.store16(cpu, destination + diff, b * 256, 0);
178 cpu->memory.store16(cpu, destination + diff * 2, c * 256, 0);
179 cpu->memory.store16(cpu, destination + diff * 3, d * 256, 0);
180 destination += diff * 4;
181 }
182}
183
184static void _MidiKey2Freq(struct GBA* gba) {
185 struct ARMCore* cpu = gba->cpu;
186 uint32_t key = cpu->memory.load32(cpu, cpu->gprs[0] + 4, 0);
187 cpu->gprs[0] = key / powf(2, (180.f - cpu->gprs[1] - cpu->gprs[2] / 256.f) / 12.f);
188}
189
190void GBASwi16(struct ARMCore* cpu, int immediate) {
191 struct GBA* gba = (struct GBA*) cpu->master;
192 if (gba->memory.fullBios) {
193 ARMRaiseSWI(cpu);
194 return;
195 }
196 switch (immediate) {
197 case 0x1:
198 _RegisterRamReset(gba);
199 break;
200 case 0x2:
201 GBAHalt(gba);
202 break;
203 case 0x05:
204 // VBlankIntrWait
205 cpu->gprs[0] = 1;
206 cpu->gprs[1] = 1;
207 // Fall through:
208 case 0x04:
209 // IntrWait
210 gba->memory.io[REG_IME >> 1] = 1;
211 if (!cpu->gprs[0] && gba->memory.io[REG_IF >> 1] & cpu->gprs[1]) {
212 break;
213 }
214 gba->memory.io[REG_IF >> 1] = 0;
215 ARMRaiseSWI(cpu);
216 break;
217 case 0x6:
218 {
219 div_t result = div(cpu->gprs[0], cpu->gprs[1]);
220 cpu->gprs[0] = result.quot;
221 cpu->gprs[1] = result.rem;
222 cpu->gprs[3] = abs(result.quot);
223 }
224 break;
225 case 0x7:
226 {
227 div_t result = div(cpu->gprs[1], cpu->gprs[0]);
228 cpu->gprs[0] = result.quot;
229 cpu->gprs[1] = result.rem;
230 cpu->gprs[3] = abs(result.quot);
231 }
232 break;
233 case 0x8:
234 cpu->gprs[0] = sqrt(cpu->gprs[0]);
235 break;
236 case 0xA:
237 cpu->gprs[0] = atan2f(cpu->gprs[1] / 16384.f, cpu->gprs[0] / 16384.f) / (2 * M_PI) * 0x10000;
238 break;
239 case 0xB:
240 _CpuSet(gba);
241 break;
242 case 0xC:
243 _FastCpuSet(gba);
244 break;
245 case 0xD:
246 cpu->gprs[0] = GBAChecksum(gba->memory.bios, SIZE_BIOS);
247 case 0xE:
248 _BgAffineSet(gba);
249 break;
250 case 0xF:
251 _ObjAffineSet(gba);
252 break;
253 case 0x11:
254 case 0x12:
255 if (cpu->gprs[0] < BASE_WORKING_RAM) {
256 GBALog(gba, GBA_LOG_GAME_ERROR, "Bad LZ77 source");
257 break;
258 }
259 switch (cpu->gprs[1] >> BASE_OFFSET) {
260 case REGION_WORKING_RAM:
261 _unLz77(gba, cpu->gprs[0], &((uint8_t*) gba->memory.wram)[(cpu->gprs[1] & (SIZE_WORKING_RAM - 1))]);
262 break;
263 case REGION_WORKING_IRAM:
264 _unLz77(gba, cpu->gprs[0], &((uint8_t*) gba->memory.iwram)[(cpu->gprs[1] & (SIZE_WORKING_IRAM - 1))]);
265 break;
266 case REGION_VRAM:
267 _unLz77(gba, cpu->gprs[0], &((uint8_t*) gba->video.renderer->vram)[(cpu->gprs[1] & 0x0001FFFF)]);
268 break;
269 default:
270 GBALog(gba, GBA_LOG_GAME_ERROR, "Bad LZ77 destination");
271 break;
272 }
273 break;
274 case 0x13:
275 if (cpu->gprs[0] < BASE_WORKING_RAM) {
276 GBALog(gba, GBA_LOG_GAME_ERROR, "Bad Huffman source");
277 break;
278 }
279 switch (cpu->gprs[1] >> BASE_OFFSET) {
280 case REGION_WORKING_RAM:
281 _unHuffman(gba, cpu->gprs[0], &((uint32_t*) gba->memory.wram)[(cpu->gprs[1] & (SIZE_WORKING_RAM - 3)) >> 2]);
282 break;
283 case REGION_WORKING_IRAM:
284 _unHuffman(gba, cpu->gprs[0], &((uint32_t*) gba->memory.iwram)[(cpu->gprs[1] & (SIZE_WORKING_IRAM - 3)) >> 2]);
285 break;
286 case REGION_VRAM:
287 _unHuffman(gba, cpu->gprs[0], &((uint32_t*) gba->video.renderer->vram)[(cpu->gprs[1] & 0x0001FFFC) >> 2]);
288 break;
289 default:
290 GBALog(gba, GBA_LOG_GAME_ERROR, "Bad Huffman destination");
291 break;
292 }
293 break;
294 case 0x14:
295 case 0x15:
296 if (cpu->gprs[0] < BASE_WORKING_RAM) {
297 GBALog(gba, GBA_LOG_GAME_ERROR, "Bad RL source");
298 break;
299 }
300 switch (cpu->gprs[1] >> BASE_OFFSET) {
301 case REGION_WORKING_RAM:
302 _unRl(gba, cpu->gprs[0], &((uint8_t*) gba->memory.wram)[(cpu->gprs[1] & (SIZE_WORKING_RAM - 1))]);
303 break;
304 case REGION_WORKING_IRAM:
305 _unRl(gba, cpu->gprs[0], &((uint8_t*) gba->memory.iwram)[(cpu->gprs[1] & (SIZE_WORKING_IRAM - 1))]);
306 break;
307 case REGION_VRAM:
308 _unRl(gba, cpu->gprs[0], &((uint8_t*) gba->video.renderer->vram)[(cpu->gprs[1] & 0x0001FFFF)]);
309 break;
310 default:
311 GBALog(gba, GBA_LOG_GAME_ERROR, "Bad RL destination");
312 break;
313 }
314 break;
315 case 0x1F:
316 _MidiKey2Freq(gba);
317 break;
318 default:
319 GBALog(gba, GBA_LOG_STUB, "Stub software interrupt: %02x", immediate);
320 }
321}
322
323void GBASwi32(struct ARMCore* cpu, int immediate) {
324 GBASwi16(cpu, immediate >> 16);
325}
326
327uint32_t GBAChecksum(uint32_t* memory, size_t size) {
328 size_t i;
329 uint32_t sum = 0;
330 for (i = 0; i < size; i += 4) {
331 sum += memory[i >> 2];
332 }
333 return sum;
334}
335
336static void _unLz77(struct GBA* gba, uint32_t source, uint8_t* dest) {
337 struct ARMCore* cpu = gba->cpu;
338 int remaining = (cpu->memory.load32(cpu, source, 0) & 0xFFFFFF00) >> 8;
339 // We assume the signature byte (0x10) is correct
340 int blockheader;
341 uint32_t sPointer = source + 4;
342 uint8_t* dPointer = dest;
343 int blocksRemaining = 0;
344 int block;
345 uint8_t* disp;
346 int bytes;
347 while (remaining > 0) {
348 if (blocksRemaining) {
349 if (blockheader & 0x80) {
350 // Compressed
351 block = cpu->memory.loadU8(cpu, sPointer, 0) | (cpu->memory.loadU8(cpu, sPointer + 1, 0) << 8);
352 sPointer += 2;
353 disp = dPointer - (((block & 0x000F) << 8) | ((block & 0xFF00) >> 8)) - 1;
354 bytes = ((block & 0x00F0) >> 4) + 3;
355 while (bytes-- && remaining) {
356 --remaining;
357 *dPointer = *disp;
358 ++disp;
359 ++dPointer;
360 }
361 } else {
362 // Uncompressed
363 *dPointer = cpu->memory.loadU8(cpu, sPointer++, 0);
364 ++dPointer;
365 --remaining;
366 }
367 blockheader <<= 1;
368 --blocksRemaining;
369 } else {
370 blockheader = cpu->memory.loadU8(cpu, sPointer++, 0);
371 blocksRemaining = 8;
372 }
373 }
374}
375
376static void _unHuffman(struct GBA* gba, uint32_t source, uint32_t* dest) {
377 struct ARMCore* cpu = gba->cpu;
378 source = source & 0xFFFFFFFC;
379 uint32_t header = cpu->memory.load32(cpu, source, 0);
380 int remaining = header >> 8;
381 int bits = header & 0xF;
382 if (32 % bits) {
383 GBALog(gba, GBA_LOG_STUB, "Unimplemented unaligned Huffman");
384 return;
385 }
386 int padding = (4 - remaining) & 0x3;
387 remaining &= 0xFFFFFFFC;
388 // We assume the signature byte (0x20) is correct
389 //var tree = [];
390 int treesize = (cpu->memory.loadU8(cpu, source + 4, 0) << 1) + 1;
391 int block = 0;
392 uint32_t treeBase = source + 5;
393 uint32_t sPointer = source + 5 + treesize;
394 uint32_t* dPointer = dest;
395 uint32_t nPointer = treeBase;
396 union HuffmanNode {
397 struct {
398 unsigned offset : 6;
399 unsigned rTerm : 1;
400 unsigned lTerm : 1;
401 };
402 uint8_t packed;
403 } node;
404 int bitsRemaining;
405 int readBits;
406 int bitsSeen = 0;
407 node.packed = cpu->memory.load8(cpu, nPointer, 0);
408 while (remaining > 0) {
409 uint32_t bitstream = cpu->memory.load32(cpu, sPointer, 0);
410 sPointer += 4;
411 for (bitsRemaining = 32; bitsRemaining > 0; --bitsRemaining, bitstream <<= 1) {
412 uint32_t next = (nPointer & ~1) + node.offset * 2 + 2;
413 if (bitstream & 0x80000000) {
414 // Go right
415 if (node.rTerm) {
416 readBits = cpu->memory.load8(cpu, next + 1, 0);
417 } else {
418 nPointer = next + 1;
419 node.packed = cpu->memory.load8(cpu, nPointer, 0);
420 continue;
421 }
422 } else {
423 // Go left
424 if (node.lTerm) {
425 readBits = cpu->memory.load8(cpu, next, 0);
426 } else {
427 nPointer = next;
428 node.packed = cpu->memory.load8(cpu, nPointer, 0);
429 continue;
430 }
431 }
432
433 block |= (readBits & ((1 << bits) - 1)) << bitsSeen;
434 bitsSeen += bits;
435 nPointer = treeBase;
436 node.packed = cpu->memory.load8(cpu, nPointer, 0);
437 if (bitsSeen == 32) {
438 bitsSeen = 0;
439 *dPointer = block;
440 ++dPointer;
441 remaining -= 4;
442 block = 0;
443 }
444 }
445
446 }
447 if (padding) {
448 *dPointer = block;
449 }
450}
451
452static void _unRl(struct GBA* gba, uint32_t source, uint8_t* dest) {
453 struct ARMCore* cpu = gba->cpu;
454 source = source & 0xFFFFFFFC;
455 int remaining = (cpu->memory.load32(cpu, source, 0) & 0xFFFFFF00) >> 8;
456 int padding = (4 - remaining) & 0x3;
457 // We assume the signature byte (0x30) is correct
458 int blockheader;
459 int block;
460 uint32_t sPointer = source + 4;
461 uint8_t* dPointer = dest;
462 while (remaining > 0) {
463 blockheader = cpu->memory.loadU8(cpu, sPointer++, 0);
464 if (blockheader & 0x80) {
465 // Compressed
466 blockheader &= 0x7F;
467 blockheader += 3;
468 block = cpu->memory.loadU8(cpu, sPointer++, 0);
469 while (blockheader-- && remaining) {
470 --remaining;
471 *dPointer = block;
472 ++dPointer;
473 }
474 } else {
475 // Uncompressed
476 blockheader++;
477 while (blockheader-- && remaining) {
478 --remaining;
479 *dPointer = cpu->memory.loadU8(cpu, sPointer++, 0);
480 ++dPointer;
481 }
482 }
483 }
484 while (padding--) {
485 *dPointer = 0;
486 ++dPointer;
487 }
488}