Repository navigation
Expand file tree
/
Copy pathjit.c
More file actions
565 lines (482 loc) · 17.3 KB
/
Copy pathjit.c
File metadata and controls
565 lines (482 loc) · 17.3 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
#include <stdint.h>
#include <string.h>
#include "pd_api.h"
#include "vecx.h"
#include "jit.h"
/* 6809 dynamic recompiler.
*
* M1: runtime ARM codegen works (emit + clearICache + call). Buffer is in PSRAM.
* M2: Thumb emitter + load/modify/store primitives.
* M3: translate a representative 6809 kernel (loads/stores/AND with N/Z flags) to
* a native block, verify it matches a switch interpreter, then time both for a
* go/no-go speedup. Both run from PSRAM so it isolates the JIT's structural win
* (no per-op fetch/decode/dispatch) from memory speed.
*/
static uint16_t jit_code[512] __attribute__((aligned(8)));
typedef struct {
uint16_t *code;
unsigned len;
unsigned cap;
} jit_buf;
static void emit16(jit_buf *b, unsigned hw)
{
if (b->len < b->cap)
b->code[b->len] = (uint16_t)hw;
b->len++;
}
/* Thumb-1, low registers r0-r7 */
static void emit_movs_imm8(jit_buf *b, unsigned rd, unsigned imm8)
{ emit16(b, 0x2000u | (rd << 8) | (imm8 & 0xffu)); }
static void emit_ldrb_imm(jit_buf *b, unsigned rt, unsigned rn, unsigned imm5)
{ emit16(b, 0x7800u | ((imm5 & 0x1fu) << 6) | (rn << 3) | rt); }
static void emit_strb_imm(jit_buf *b, unsigned rt, unsigned rn, unsigned imm5)
{ emit16(b, 0x7000u | ((imm5 & 0x1fu) << 6) | (rn << 3) | rt); }
static void emit_adds_imm3(jit_buf *b, unsigned rd, unsigned rn, unsigned imm3)
{ emit16(b, 0x1c00u | ((imm3 & 7u) << 6) | (rn << 3) | rd); }
static void emit_subs_imm3(jit_buf *b, unsigned rd, unsigned rn, unsigned imm3)
{ emit16(b, 0x1e00u | ((imm3 & 7u) << 6) | (rn << 3) | rd); }
static void emit_lsr_imm5(jit_buf *b, unsigned rd, unsigned rm, unsigned imm5)
{ emit16(b, 0x0800u | ((imm5 & 0x1fu) << 6) | (rm << 3) | rd); }
static void emit_lsl_imm5(jit_buf *b, unsigned rd, unsigned rm, unsigned imm5)
{ emit16(b, 0x0000u | ((imm5 & 0x1fu) << 6) | (rm << 3) | rd); }
static void emit_ands(jit_buf *b, unsigned rd, unsigned rm)
{ emit16(b, 0x4000u | (rm << 3) | rd); }
static void emit_orrs(jit_buf *b, unsigned rd, unsigned rm)
{ emit16(b, 0x4300u | (rm << 3) | rd); }
static void emit_bics(jit_buf *b, unsigned rd, unsigned rm)
{ emit16(b, 0x4380u | (rm << 3) | rd); }
static void emit_eors(jit_buf *b, unsigned rd, unsigned rm)
{ emit16(b, 0x4040u | (rm << 3) | rd); }
static void emit_adds_reg(jit_buf *b, unsigned rd, unsigned rn, unsigned rm)
{ emit16(b, 0x1800u | (rm << 6) | (rn << 3) | rd); }
static void emit_mov_reg(jit_buf *b, unsigned rd, unsigned rm)
{ emit16(b, 0x0000u | (rm << 3) | rd); } /* lsls rd, rm, #0 */
static void emit_uxtb(jit_buf *b, unsigned rd, unsigned rm)
{ emit16(b, 0xb2c0u | (rm << 3) | rd); }
static void emit_blx_reg(jit_buf *b, unsigned rm)
{ emit16(b, 0x4780u | (rm << 3)); }
/* Thumb-2 32-bit movw/movt (M7 supports them): load any 16-bit half into rd. */
static void emit_movw(jit_buf *b, unsigned rd, unsigned imm16)
{
emit16(b, 0xf240u | (((imm16 >> 11) & 1u) << 10) | ((imm16 >> 12) & 0xfu));
emit16(b, (((imm16 >> 8) & 7u) << 12) | (rd << 8) | (imm16 & 0xffu));
}
static void emit_movt(jit_buf *b, unsigned rd, unsigned imm16)
{
emit16(b, 0xf2c0u | (((imm16 >> 11) & 1u) << 10) | ((imm16 >> 12) & 0xfu));
emit16(b, (((imm16 >> 8) & 7u) << 12) | (rd << 8) | (imm16 & 0xffu));
}
/* rd = 32-bit absolute addr (Thumb bit included); then blx rd. */
static void emit_call_abs(jit_buf *b, unsigned rd, uintptr_t addr)
{
emit_movw(b, rd, (unsigned)(addr & 0xffffu));
emit_movt(b, rd, (unsigned)((addr >> 16) & 0xffffu));
emit_blx_reg(b, rd);
}
static void emit_bx_lr(jit_buf *b)
{ emit16(b, 0x4770u); }
static void emit_push_r456_lr(jit_buf *b)
{ emit16(b, 0xb570u); } /* push {r4,r5,r6,lr} */
static void emit_pop_r456_pc(jit_buf *b)
{ emit16(b, 0xbd70u); } /* pop {r4,r5,r6,pc} (returns) */
static void *jit_finalize(PlaydateAPI *pd, jit_buf *b)
{
pd->system->clearICache();
return (void *)(((uintptr_t)b->code) | 1u);
}
/* ----------------------------------------------------------------- M3 kernel */
#define KZ 0x04u
#define KN 0x08u
typedef struct {
uint8_t a, b, cc, pad;
uint8_t mem[256];
} kcpu; /* a@0 b@1 cc@2 mem@4 */
static const uint8_t kprog[] = {
0x86, 0x5a, /* LDA #0x5a */
0x97, 0x10, /* STA $10 */
0xd6, 0x10, /* LDB $10 */
0xc4, 0x0f, /* ANDB #0x0f */
0xd7, 0x11, /* STB $11 */
0x96, 0x11, /* LDA $11 */
0x85, 0x08, /* BITA #0x08 */
};
static void ksetnz(kcpu *c, unsigned v)
{
c->cc = (uint8_t)((c->cc & ~(KZ | KN)) | (v ? 0u : KZ) | ((v & 0x80u) ? KN : 0u));
}
static void kinterp(kcpu *c)
{
unsigned pc = 0;
while (pc < sizeof(kprog)) {
unsigned op = kprog[pc++];
switch (op) {
case 0x86: c->a = kprog[pc++]; ksetnz(c, c->a); break;
case 0x97: c->mem[kprog[pc++]] = c->a; ksetnz(c, c->a); break;
case 0xd6: c->b = c->mem[kprog[pc++]]; ksetnz(c, c->b); break;
case 0xc4: c->b &= kprog[pc++]; ksetnz(c, c->b); break;
case 0xd7: c->mem[kprog[pc++]] = c->b; ksetnz(c, c->b); break;
case 0x96: c->a = c->mem[kprog[pc++]]; ksetnz(c, c->a); break;
case 0x85: ksetnz(c, c->a & kprog[pc++]); break;
default: return;
}
}
}
/* set N/Z in r3(cc) from value reg rv (rv must NOT be r4/r5); clobbers r4,r5 */
static void emit_setnz(jit_buf *b, unsigned rv)
{
emit_lsr_imm5(b, 4, rv, 4); /* r4 = rv >> 4 */
emit_movs_imm8(b, 5, 0x08u); /* r5 = 0x08 */
emit_ands(b, 4, 5); /* r4 = (rv>>4)&8 = N bit */
emit_subs_imm3(b, 5, rv, 1); /* r5 = rv - 1 */
emit_lsr_imm5(b, 5, 5, 31); /* r5 = (rv==0) ? 1 : 0 */
emit_lsl_imm5(b, 5, 5, 2); /* r5 = Z bit (0x04 or 0) */
emit_orrs(b, 4, 5); /* r4 = N | Z */
emit_movs_imm8(b, 5, 0x0cu); /* r5 = 0x0c */
emit_bics(b, 3, 5); /* r3 &= ~0x0c */
emit_orrs(b, 3, 4); /* r3 |= N|Z */
}
/* r0=state, r1=a, r2=b, r3=cc, r4/r5=setnz temps, r6=BITA temp */
static void jit_translate_kernel(jit_buf *b)
{
unsigned pc = 0;
emit_push_r456_lr(b);
emit_ldrb_imm(b, 1, 0, 0);
emit_ldrb_imm(b, 2, 0, 1);
emit_ldrb_imm(b, 3, 0, 2);
while (pc < sizeof(kprog)) {
unsigned op = kprog[pc++];
switch (op) {
case 0x86: emit_movs_imm8(b, 1, kprog[pc++]); emit_setnz(b, 1); break;
case 0x97: emit_strb_imm(b, 1, 0, 4 + kprog[pc++]); emit_setnz(b, 1); break;
case 0xd6: emit_ldrb_imm(b, 2, 0, 4 + kprog[pc++]); emit_setnz(b, 2); break;
case 0xc4: emit_movs_imm8(b, 4, kprog[pc++]); emit_ands(b, 2, 4); emit_setnz(b, 2); break;
case 0xd7: emit_strb_imm(b, 2, 0, 4 + kprog[pc++]); emit_setnz(b, 2); break;
case 0x96: emit_ldrb_imm(b, 1, 0, 4 + kprog[pc++]); emit_setnz(b, 1); break;
case 0x85: emit_movs_imm8(b, 6, kprog[pc++]); emit_ands(b, 6, 1); emit_setnz(b, 6); break;
default: pc = sizeof(kprog); break;
}
}
emit_strb_imm(b, 1, 0, 0);
emit_strb_imm(b, 2, 0, 1);
emit_strb_imm(b, 3, 0, 2);
emit_pop_r456_pc(b);
}
static void jit_m3_benchmark(PlaydateAPI *pd)
{
jit_buf b;
void (*blk)(kcpu *);
kcpu ci, cj;
unsigned i, ok;
uint32_t t0, t1, t2;
const unsigned K = 200000u;
b.code = jit_code;
b.len = 0;
b.cap = sizeof(jit_code) / sizeof(jit_code[0]);
jit_translate_kernel(&b);
blk = (void (*)(kcpu *))jit_finalize(pd, &b);
/* correctness: same start state -> same end state */
memset(&ci, 0, sizeof(ci));
memset(&cj, 0, sizeof(cj));
kinterp(&ci);
blk(&cj);
ok = (ci.a == cj.a && ci.b == cj.b && ci.cc == cj.cc &&
memcmp(ci.mem, cj.mem, sizeof(ci.mem)) == 0);
pd->system->logToConsole(
"cranktrex jit: M3 verify interp(a=%02x b=%02x cc=%02x) jit(a=%02x b=%02x cc=%02x) %s (%u hw)",
ci.a, ci.b, ci.cc, cj.a, cj.b, cj.cc, ok ? "OK" : "MISMATCH", b.len);
if (!ok)
return;
t0 = pd->system->getCurrentTimeMilliseconds();
for (i = 0; i < K; i++)
kinterp(&ci);
t1 = pd->system->getCurrentTimeMilliseconds();
for (i = 0; i < K; i++)
blk(&cj);
t2 = pd->system->getCurrentTimeMilliseconds();
{
uint32_t interp_ms = t1 - t0;
uint32_t jit_ms = t2 - t1;
uint32_t ratio_x100 = jit_ms ? (interp_ms * 100u) / jit_ms : 0u;
pd->system->logToConsole(
"cranktrex jit: M3 bench K=%u interp=%ums jit=%ums speedup=%u.%02ux (both PSRAM)",
K, (unsigned)interp_ms, (unsigned)jit_ms,
(unsigned)(ratio_x100 / 100u), (unsigned)(ratio_x100 % 100u));
}
}
/* ----------------------------------------- M5: DTCM placement (the crux) */
/* Same kernel as M3, but emitted into a stack buffer (SRAM/DTCM at 0x20000000,
* the fast region the relocated hot core runs from) instead of the static buffer
* (PSRAM at 0x60000000). If DTCM is much faster, that's the path; it's also what a
* JIT block needs to beat the DTCM interpreter. */
static void jit_m5_dtcm(PlaydateAPI *pd)
{
uint16_t dtcm_code[256] __attribute__((aligned(8)));
jit_buf b;
void (*blk)(kcpu *);
kcpu ci, cj;
unsigned i, ok;
uint32_t t0, t1, t2;
const unsigned K = 200000u;
b.code = dtcm_code;
b.len = 0;
b.cap = sizeof(dtcm_code) / sizeof(dtcm_code[0]);
jit_translate_kernel(&b);
pd->system->clearICache();
blk = (void (*)(kcpu *))(((uintptr_t)dtcm_code) | 1u);
memset(&ci, 0, sizeof(ci));
memset(&cj, 0, sizeof(cj));
kinterp(&ci);
blk(&cj);
ok = (ci.a == cj.a && ci.b == cj.b && ci.cc == cj.cc &&
memcmp(ci.mem, cj.mem, sizeof(ci.mem)) == 0);
pd->system->logToConsole("cranktrex jit: M5 DTCM block at %p verify %s",
(void *)dtcm_code, ok ? "OK" : "BAD");
if (!ok)
return;
t0 = pd->system->getCurrentTimeMilliseconds();
for (i = 0; i < K; i++)
kinterp(&ci);
t1 = pd->system->getCurrentTimeMilliseconds();
for (i = 0; i < K; i++)
blk(&cj);
t2 = pd->system->getCurrentTimeMilliseconds();
{
uint32_t im = t1 - t0, jm = t2 - t1;
uint32_t rx = jm ? (im * 100u) / jm : 0u;
pd->system->logToConsole(
"cranktrex jit: M5 DTCM bench K=%u interp(PSRAM)=%ums jit(DTCM)=%ums speedup=%u.%02ux (cf jit-PSRAM was 95ms)",
K, (unsigned)im, (unsigned)jm, (unsigned)(rx / 100u), (unsigned)(rx % 100u));
}
}
/* ------------------------------------------------- M4: full-flag ADD codegen */
/* CC: H=0x20 N=0x08 Z=0x04 V=0x02 C=0x01 */
static unsigned ref_addb(unsigned bval, unsigned imm, unsigned *cc_out)
{
unsigned sum = bval + imm;
unsigned r = sum & 0xffu;
unsigned cc = 0;
if (sum & 0x100u) cc |= 0x01u; /* C */
if (((bval & 0xfu) + (imm & 0xfu)) & 0x10u) cc |= 0x20u; /* H */
if ((bval ^ r) & (imm ^ r) & 0x80u) cc |= 0x02u; /* V */
if (r & 0x80u) cc |= 0x08u; /* N */
if (r == 0u) cc |= 0x04u; /* Z */
*cc_out = cc;
return r;
}
/* void addb(uint8_t *s) // s[0]=b s[1]=cc ; b += imm with full 6809 flags.
* r0=s, r2=b, r3=cc, r4/r5/r6 = temps (r5 holds the result). */
static void jit_translate_addb_imm(jit_buf *b, unsigned imm)
{
emit_push_r456_lr(b);
emit_ldrb_imm(b, 2, 0, 0); /* r2 = b */
emit_ldrb_imm(b, 3, 0, 1); /* r3 = cc */
emit_movs_imm8(b, 4, 0x2fu); /* clear H,N,Z,V,C */
emit_bics(b, 3, 4);
emit_movs_imm8(b, 4, imm & 0xffu);
emit_adds_reg(b, 5, 2, 4); /* r5 = b + imm (9-bit)*/
emit_lsr_imm5(b, 4, 5, 8); /* C = sum>>8 */
emit_orrs(b, 3, 4);
emit_movs_imm8(b, 4, 0xffu);
emit_ands(b, 5, 4); /* r5 = result */
emit_mov_reg(b, 4, 2); /* H: r4 = b&0xF */
emit_movs_imm8(b, 6, 0x0fu);
emit_ands(b, 4, 6);
emit_movs_imm8(b, 6, imm & 0x0fu);
emit_adds_reg(b, 4, 4, 6); /* + imm&0xF */
emit_lsr_imm5(b, 4, 4, 4); /* nibble carry */
emit_lsl_imm5(b, 4, 4, 5); /* -> 0x20 */
emit_orrs(b, 3, 4);
emit_mov_reg(b, 4, 2); /* V: r4 = b ^ result */
emit_eors(b, 4, 5);
emit_movs_imm8(b, 6, imm & 0xffu);
emit_eors(b, 6, 5); /* r6 = imm ^ result */
emit_ands(b, 4, 6);
emit_movs_imm8(b, 6, 0x80u);
emit_ands(b, 4, 6);
emit_lsr_imm5(b, 4, 4, 6); /* 0x80>>6 -> 0x02 */
emit_orrs(b, 3, 4);
emit_lsr_imm5(b, 4, 5, 4); /* N = (result>>4)&8 */
emit_movs_imm8(b, 6, 0x08u);
emit_ands(b, 4, 6);
emit_orrs(b, 3, 4);
emit_subs_imm3(b, 4, 5, 1); /* Z = (result==0)?4:0 */
emit_lsr_imm5(b, 4, 4, 31);
emit_lsl_imm5(b, 4, 4, 2);
emit_orrs(b, 3, 4);
emit_mov_reg(b, 2, 5); /* b = result */
emit_strb_imm(b, 2, 0, 0);
emit_strb_imm(b, 3, 0, 1);
emit_pop_r456_pc(b);
}
static void jit_m4_addflags(PlaydateAPI *pd)
{
static const unsigned imms[] = { 0x01u, 0x0fu, 0x7fu, 0x80u, 0xffu };
unsigned ti, bv, fails = 0, tested = 0;
for (ti = 0; ti < sizeof(imms) / sizeof(imms[0]); ti++) {
jit_buf jb;
void (*addb)(uint8_t *);
jb.code = jit_code;
jb.len = 0;
jb.cap = sizeof(jit_code) / sizeof(jit_code[0]);
jit_translate_addb_imm(&jb, imms[ti]);
addb = (void (*)(uint8_t *))jit_finalize(pd, &jb);
for (bv = 0; bv < 256u; bv++) {
uint8_t s[2];
unsigned rcc, rres;
s[0] = (uint8_t)bv;
s[1] = 0;
addb(s);
rres = ref_addb(bv, imms[ti], &rcc);
tested++;
if (s[0] != (uint8_t)rres || s[1] != (uint8_t)rcc)
fails++;
}
}
pd->system->logToConsole("cranktrex jit: M4 ADDB full-flags %u/%u match %s",
tested - fails, tested, fails == 0 ? "ALL OK" : "FAILS!");
}
/* --------------------------------------- M6: real memory ops (honest number) */
/* Load/store kernel hitting the REAL vecx_read8/vecx_write8 (full address decode),
* so the per-op cost includes what the JIT can't eliminate. Registers a,cc live in
* callee-saved r4,r5 so the memory calls don't clobber them. op: 0=LDA 1=STA. */
typedef struct { uint8_t a, cc; } kcpu6;
typedef struct { uint8_t op; uint16_t addr; } kop6;
static const kop6 m6prog[] = {
{ 0, 0xc850 }, { 1, 0xc851 }, { 0, 0xc852 },
{ 1, 0xc853 }, { 0, 0xc851 }, { 1, 0xc850 },
};
#define M6N (sizeof(m6prog) / sizeof(m6prog[0]))
static void m6_init_mem(void)
{
vecx_write8(0xc850u, 0x5au);
vecx_write8(0xc851u, 0x00u);
vecx_write8(0xc852u, 0x80u);
vecx_write8(0xc853u, 0x01u);
}
static void kinterp6(kcpu6 *c)
{
unsigned i;
for (i = 0; i < M6N; i++) {
if (m6prog[i].op == 0)
c->a = vecx_read8(m6prog[i].addr);
else
vecx_write8(m6prog[i].addr, c->a);
c->cc = (uint8_t)((c->cc & ~0x0cu) | (c->a ? 0u : 0x04u) | ((c->a & 0x80u) ? 0x08u : 0u));
}
}
/* set N/Z in r5(cc) from r4(a); temps r0,r1 (caller-saved, free between calls) */
static void emit_setnz6(jit_buf *b)
{
emit_lsr_imm5(b, 0, 4, 4);
emit_movs_imm8(b, 1, 0x08u);
emit_ands(b, 0, 1);
emit_subs_imm3(b, 1, 4, 1);
emit_lsr_imm5(b, 1, 1, 31);
emit_lsl_imm5(b, 1, 1, 2);
emit_orrs(b, 0, 1);
emit_movs_imm8(b, 1, 0x0cu);
emit_bics(b, 5, 1);
emit_orrs(b, 5, 0);
}
/* r4=a, r5=cc, r6=state ; calls vecx_read8/write8 (preserve r4-r6) */
static void jit_translate_m6(jit_buf *b)
{
unsigned i;
emit_push_r456_lr(b);
emit_mov_reg(b, 6, 0); /* r6 = state ptr */
emit_ldrb_imm(b, 4, 6, 0); /* r4 = a */
emit_ldrb_imm(b, 5, 6, 1); /* r5 = cc */
for (i = 0; i < M6N; i++) {
if (m6prog[i].op == 0) { /* LDA addr */
emit_movw(b, 0, m6prog[i].addr); /* r0 = addr */
emit_call_abs(b, 1, (uintptr_t)&vecx_read8);
emit_uxtb(b, 4, 0); /* a = (byte)result */
} else { /* STA addr */
emit_movw(b, 0, m6prog[i].addr); /* r0 = addr */
emit_mov_reg(b, 1, 4); /* r1 = a (data) */
emit_call_abs(b, 2, (uintptr_t)&vecx_write8);
}
emit_setnz6(b);
}
emit_strb_imm(b, 4, 6, 0);
emit_strb_imm(b, 5, 6, 1);
emit_pop_r456_pc(b);
}
static void jit_m6_realmem(PlaydateAPI *pd)
{
jit_buf b;
void (*blk)(kcpu6 *);
kcpu6 ci, cj;
unsigned i, ok;
uint32_t t0, t1, t2;
const unsigned K = 100000u;
b.code = jit_code;
b.len = 0;
b.cap = sizeof(jit_code) / sizeof(jit_code[0]);
jit_translate_m6(&b);
blk = (void (*)(kcpu6 *))jit_finalize(pd, &b);
memset(&ci, 0, sizeof(ci));
memset(&cj, 0, sizeof(cj));
m6_init_mem();
kinterp6(&ci);
m6_init_mem();
blk(&cj);
ok = (ci.a == cj.a && ci.cc == cj.cc);
pd->system->logToConsole(
"cranktrex jit: M6 realmem verify interp(a=%02x cc=%02x) jit(a=%02x cc=%02x) %s (%u hw)",
ci.a, ci.cc, cj.a, cj.cc, ok ? "OK" : "MISMATCH", b.len);
if (!ok)
return;
t0 = pd->system->getCurrentTimeMilliseconds();
for (i = 0; i < K; i++)
kinterp6(&ci);
t1 = pd->system->getCurrentTimeMilliseconds();
for (i = 0; i < K; i++)
blk(&cj);
t2 = pd->system->getCurrentTimeMilliseconds();
{
uint32_t im = t1 - t0, jm = t2 - t1;
uint32_t rx = jm ? (im * 100u) / jm : 0u;
pd->system->logToConsole(
"cranktrex jit: M6 realmem K=%u interp=%ums jit=%ums speedup=%u.%02ux (real vecx_read8/write8)",
K, (unsigned)im, (unsigned)jm, (unsigned)(rx / 100u), (unsigned)(rx % 100u));
}
}
void jit_selftest(PlaydateAPI *pd)
{
jit_buf b;
/* M1 */
b.code = jit_code;
b.len = 0;
b.cap = sizeof(jit_code) / sizeof(jit_code[0]);
emit_movs_imm8(&b, 0, 0x42u);
emit_bx_lr(&b);
{
unsigned (*fn)(void) = (unsigned (*)(void))jit_finalize(pd, &b);
unsigned r = fn();
pd->system->logToConsole("cranktrex jit: M1 return-0x42 -> 0x%x %s", r, r == 0x42u ? "OK" : "WRONG");
}
/* M2 */
b.code = jit_code;
b.len = 0;
b.cap = sizeof(jit_code) / sizeof(jit_code[0]);
emit_ldrb_imm(&b, 1, 0, 0);
emit_adds_imm3(&b, 1, 1, 1);
emit_strb_imm(&b, 1, 0, 0);
emit_bx_lr(&b);
{
void (*inc)(uint8_t *) = (void (*)(uint8_t *))jit_finalize(pd, &b);
volatile uint8_t v = 0x10u;
inc((uint8_t *)&v);
inc((uint8_t *)&v);
pd->system->logToConsole("cranktrex jit: M2 inc x2 (0x10) -> 0x%x %s", (unsigned)v, v == 0x12u ? "OK" : "WRONG");
}
/* M3 */
jit_m3_benchmark(pd);
/* M4 */
jit_m4_addflags(pd);
/* M5 */
jit_m5_dtcm(pd);
/* M6 */
jit_m6_realmem(pd);
}