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