« back to all changes in this revision
Viewing changes to src/vm_mips.dasc
- browse files at revision 16
- compare with another revision
- download diff
- download tarball
- view history from revision 16
- Committer: Package Import Robot
- Author(s): Enrico Tassi
- Date: 2012-11-03 14:07:56 UTC
- mfrom: (1.2.1) (15.1.1 experimental)
- Revision ID: package-import@ubuntu.com-20121103140756-z0zcnyrwqlvuc2m5
Tags: 2.0.0+dfsg-1
New upstream release
- files added:
- .pc/.quilt_patches
- .pc/multi-arch-path.patch
- .pc/multi-arch-path.patch/src
- src/host
- src/jit
- files removed:
- .pc/beta9_hotfix1.patch
- .pc/beta9_hotfix1.patch/src
- .pc/multi-arch-paths.patch
- .pc/multi-arch-paths.patch/src
- lib
- files modified:
- .pc/applied-patches
added
removed
src/vm_mips.dasc
1
|// Low-level VM code for MIPS CPUs.
2
|// Bytecode interpreter, fast functions and helper functions.
3
|// Copyright (C) 2005-2012 Mike Pall. See Copyright Notice in luajit.h
4
|
5
|.arch mips
6
|.section code_op, code_sub
7
|
8
|.actionlist build_actionlist
9
|.globals GLOB_
10
|.globalnames globnames
11
|.externnames extnames
12
|
13
|// Note: The ragged indentation of the instructions is intentional.
14
|// The starting columns indicate data dependencies.
15
|
16
|//-----------------------------------------------------------------------
17
|
18
|// Fixed register assignments for the interpreter.
19
|// Don't use: r0 = 0, r26/r27 = reserved, r28 = gp, r29 = sp, r31 = ra
20
|
21
|// The following must be C callee-save (but BASE is often refetched).
22
|.define BASE, r16 // Base of current Lua stack frame.
23
|.define KBASE, r17 // Constants of current Lua function.
24
|.define PC, r18 // Next PC.
25
|.define DISPATCH, r19 // Opcode dispatch table.
26
|.define LREG, r20 // Register holding lua_State (also in SAVE_L).
27
|.define MULTRES, r21 // Size of multi-result: (nresults+1)*8.
28
|// NYI: r22 currently unused.
29
|
30
|.define JGL, r30 // On-trace: global_State + 32768.
31
|
32
|// Constants for type-comparisons, stores and conversions. C callee-save.
33
|.define TISNIL, r30
34
|.define TOBIT, f30 // 2^52 + 2^51.
35
|
36
|// The following temporaries are not saved across C calls, except for RA.
37
|.define RA, r23 // Callee-save.
38
|.define RB, r8
39
|.define RC, r9
40
|.define RD, r10
41
|.define INS, r11
42
|
43
|.define AT, r1 // Assembler temporary.
44
|.define TMP0, r12
45
|.define TMP1, r13
46
|.define TMP2, r14
47
|.define TMP3, r15
48
|
49
|// Calling conventions.
50
|.define CFUNCADDR, r25
51
|.define CARG1, r4
52
|.define CARG2, r5
53
|.define CARG3, r6
54
|.define CARG4, r7
55
|
56
|.define CRET1, r2
57
|.define CRET2, r3
58
|
59
|.define FARG1, f12
60
|.define FARG2, f14
61
|
62
|.define FRET1, f0
63
|.define FRET2, f2
64
|
65
|// Stack layout while in interpreter. Must match with lj_frame.h.
66
|.define CFRAME_SPACE, 112 // Delta for sp.
67
|
68
|.define SAVE_ERRF, 124(sp) // 32 bit C frame info.
69
|.define SAVE_NRES, 120(sp)
70
|.define SAVE_CFRAME, 116(sp)
71
|.define SAVE_L, 112(sp)
72
|//----- 8 byte aligned, ^^^^ 16 byte register save area, owned by interpreter.
73
|.define SAVE_GPR_, 72 // .. 72+10*4: 32 bit GPR saves.
74
|.define SAVE_FPR_, 24 // .. 24+6*8: 64 bit FPR saves.
75
|.define SAVE_PC, 20(sp)
76
|.define ARG5, 16(sp)
77
|.define CSAVE_4, 12(sp)
78
|.define CSAVE_3, 8(sp)
79
|.define CSAVE_2, 4(sp)
80
|.define CSAVE_1, 0(sp)
81
|//----- 8 byte aligned, ^^^^ 16 byte register save area, owned by callee.
82
|
83
|.define ARG5_OFS, 16
84
|.define SAVE_MULTRES, ARG5
85
|
86
|.macro saveregs
87
| addiu sp, sp, -CFRAME_SPACE
88
| sw ra, SAVE_GPR_+9*4(sp)
89
| sw r30, SAVE_GPR_+8*4(sp)
90
| sdc1 f30, SAVE_FPR_+5*8(sp)
91
| sw r23, SAVE_GPR_+7*4(sp)
92
| sw r22, SAVE_GPR_+6*4(sp)
93
| sdc1 f28, SAVE_FPR_+4*8(sp)
94
| sw r21, SAVE_GPR_+5*4(sp)
95
| sw r20, SAVE_GPR_+4*4(sp)
96
| sdc1 f26, SAVE_FPR_+3*8(sp)
97
| sw r19, SAVE_GPR_+3*4(sp)
98
| sw r18, SAVE_GPR_+2*4(sp)
99
| sdc1 f24, SAVE_FPR_+2*8(sp)
100
| sw r17, SAVE_GPR_+1*4(sp)
101
| sw r16, SAVE_GPR_+0*4(sp)
102
| sdc1 f22, SAVE_FPR_+1*8(sp)
103
| sdc1 f20, SAVE_FPR_+0*8(sp)
104
|.endmacro
105
|
106
|.macro restoreregs_ret
107
| lw ra, SAVE_GPR_+9*4(sp)
108
| lw r30, SAVE_GPR_+8*4(sp)
109
| ldc1 f30, SAVE_FPR_+5*8(sp)
110
| lw r23, SAVE_GPR_+7*4(sp)
111
| lw r22, SAVE_GPR_+6*4(sp)
112
| ldc1 f28, SAVE_FPR_+4*8(sp)
113
| lw r21, SAVE_GPR_+5*4(sp)
114
| lw r20, SAVE_GPR_+4*4(sp)
115
| ldc1 f26, SAVE_FPR_+3*8(sp)
116
| lw r19, SAVE_GPR_+3*4(sp)
117
| lw r18, SAVE_GPR_+2*4(sp)
118
| ldc1 f24, SAVE_FPR_+2*8(sp)
119
| lw r17, SAVE_GPR_+1*4(sp)
120
| lw r16, SAVE_GPR_+0*4(sp)
121
| ldc1 f22, SAVE_FPR_+1*8(sp)
122
| ldc1 f20, SAVE_FPR_+0*8(sp)
123
| jr ra
124
| addiu sp, sp, CFRAME_SPACE
125
|.endmacro
126
|
127
|// Type definitions. Some of these are only used for documentation.
128
|.type L, lua_State, LREG
129
|.type GL, global_State
130
|.type TVALUE, TValue
131
|.type GCOBJ, GCobj
132
|.type STR, GCstr
133
|.type TAB, GCtab
134
|.type LFUNC, GCfuncL
135
|.type CFUNC, GCfuncC
136
|.type PROTO, GCproto
137
|.type UPVAL, GCupval
138
|.type NODE, Node
139
|.type NARGS8, int
140
|.type TRACE, GCtrace
141
|
142
|//-----------------------------------------------------------------------
143
|
144
|// Trap for not-yet-implemented parts.
145
|.macro NYI; .long 0xf0f0f0f0; .endmacro
146
|
147
|// Macros to mark delay slots.
148
|.macro ., a; a; .endmacro
149
|.macro ., a,b; a,b; .endmacro
150
|.macro ., a,b,c; a,b,c; .endmacro
151
|
152
|//-----------------------------------------------------------------------
153
|
154
|// Endian-specific defines.
155
|.define FRAME_PC, LJ_ENDIAN_SELECT(-4,-8)
156
|.define FRAME_FUNC, LJ_ENDIAN_SELECT(-8,-4)
157
|.define HI, LJ_ENDIAN_SELECT(4,0)
158
|.define LO, LJ_ENDIAN_SELECT(0,4)
159
|.define OFS_RD, LJ_ENDIAN_SELECT(2,0)
160
|.define OFS_RA, LJ_ENDIAN_SELECT(1,2)
161
|.define OFS_OP, LJ_ENDIAN_SELECT(0,3)
162
|
163
|// Instruction decode.
164
|.macro decode_OP1, dst, ins; andi dst, ins, 0xff; .endmacro
165
|.macro decode_OP4a, dst, ins; andi dst, ins, 0xff; .endmacro
166
|.macro decode_OP4b, dst; sll dst, dst, 2; .endmacro
167
|.macro decode_RC4a, dst, ins; srl dst, ins, 14; .endmacro
168
|.macro decode_RC4b, dst; andi dst, dst, 0x3fc; .endmacro
169
|.macro decode_RD4b, dst; sll dst, dst, 2; .endmacro
170
|.macro decode_RA8a, dst, ins; srl dst, ins, 5; .endmacro
171
|.macro decode_RA8b, dst; andi dst, dst, 0x7f8; .endmacro
172
|.macro decode_RB8a, dst, ins; srl dst, ins, 21; .endmacro
173
|.macro decode_RB8b, dst; andi dst, dst, 0x7f8; .endmacro
174
|.macro decode_RD8a, dst, ins; srl dst, ins, 16; .endmacro
175
|.macro decode_RD8b, dst; sll dst, dst, 3; .endmacro
176
|.macro decode_RDtoRC8, dst, src; andi dst, src, 0x7f8; .endmacro
177
|
178
|// Instruction fetch.
179
|.macro ins_NEXT1
180
| lw INS, 0(PC)
181
| addiu PC, PC, 4
182
|.endmacro
183
|// Instruction decode+dispatch.
184
|.macro ins_NEXT2
185
| decode_OP4a TMP1, INS
186
| decode_OP4b TMP1
187
| addu TMP0, DISPATCH, TMP1
188
| decode_RD8a RD, INS
189
| lw AT, 0(TMP0)
190
| decode_RA8a RA, INS
191
| decode_RD8b RD
192
| jr AT
193
| decode_RA8b RA
194
|.endmacro
195
|.macro ins_NEXT
196
| ins_NEXT1
197
| ins_NEXT2
198
|.endmacro
199
|
200
|// Instruction footer.
201
|.if 1
202
| // Replicated dispatch. Less unpredictable branches, but higher I-Cache use.
203
| .define ins_next, ins_NEXT
204
| .define ins_next_, ins_NEXT
205
| .define ins_next1, ins_NEXT1
206
| .define ins_next2, ins_NEXT2
207
|.else
208
| // Common dispatch. Lower I-Cache use, only one (very) unpredictable branch.
209
| // Affects only certain kinds of benchmarks (and only with -j off).
210
| .macro ins_next
211
| b ->ins_next
212
| .endmacro
213
| .macro ins_next1
214
| .endmacro
215
| .macro ins_next2
216
| b ->ins_next
217
| .endmacro
218
| .macro ins_next_
219
| ->ins_next:
220
| ins_NEXT
221
| .endmacro
222
|.endif
223
|
224
|// Call decode and dispatch.
225
|.macro ins_callt
226
| // BASE = new base, RB = LFUNC/CFUNC, RC = nargs*8, FRAME_PC(BASE) = PC
227
| lw PC, LFUNC:RB->pc
228
| lw INS, 0(PC)
229
| addiu PC, PC, 4
230
| decode_OP4a TMP1, INS
231
| decode_RA8a RA, INS
232
| decode_OP4b TMP1
233
| decode_RA8b RA
234
| addu TMP0, DISPATCH, TMP1
235
| lw TMP0, 0(TMP0)
236
| jr TMP0
237
| addu RA, RA, BASE
238
|.endmacro
239
|
240
|.macro ins_call
241
| // BASE = new base, RB = LFUNC/CFUNC, RC = nargs*8, PC = caller PC
242
| sw PC, FRAME_PC(BASE)
243
| ins_callt
244
|.endmacro
245
|
246
|//-----------------------------------------------------------------------
247
|
248
|.macro branch_RD
249
| srl TMP0, RD, 1
250
| lui AT, (-(BCBIAS_J*4 >> 16) & 65535)
251
| addu TMP0, TMP0, AT
252
| addu PC, PC, TMP0
253
|.endmacro
254
|
255
|// Assumes DISPATCH is relative to GL.
256
#define DISPATCH_GL(field) (GG_DISP2G + (int)offsetof(global_State, field))
257
#define DISPATCH_J(field) (GG_DISP2J + (int)offsetof(jit_State, field))
258
#define GG_DISP2GOT (GG_OFS(got) - GG_OFS(dispatch))
259
#define DISPATCH_GOT(name) (GG_DISP2GOT + 4*LJ_GOT_##name)
260
|
261
#define PC2PROTO(field) ((int)offsetof(GCproto, field)-(int)sizeof(GCproto))
262
|
263
|.macro load_got, func
264
| lw CFUNCADDR, DISPATCH_GOT(func)(DISPATCH)
265
|.endmacro
266
|// Much faster. Sadly, there's no easy way to force the required code layout.
267
|// .macro call_intern, func; bal extern func; .endmacro
268
|.macro call_intern, func; jalr CFUNCADDR; .endmacro
269
|.macro call_extern; jalr CFUNCADDR; .endmacro
270
|.macro jmp_extern; jr CFUNCADDR; .endmacro
271
|
272
|.macro hotcheck, delta, target
273
| srl TMP1, PC, 1
274
| andi TMP1, TMP1, 126
275
| addu TMP1, TMP1, DISPATCH
276
| lhu TMP2, GG_DISP2HOT(TMP1)
277
| addiu TMP2, TMP2, -delta
278
| bltz TMP2, target
279
|. sh TMP2, GG_DISP2HOT(TMP1)
280
|.endmacro
281
|
282
|.macro hotloop
283
| hotcheck HOTCOUNT_LOOP, ->vm_hotloop
284
|.endmacro
285
|
286
|.macro hotcall
287
| hotcheck HOTCOUNT_CALL, ->vm_hotcall
288
|.endmacro
289
|
290
|// Set current VM state. Uses TMP0.
291
|.macro li_vmstate, st; li TMP0, ~LJ_VMST_..st; .endmacro
292
|.macro st_vmstate; sw TMP0, DISPATCH_GL(vmstate)(DISPATCH); .endmacro
293
|
294
|// Move table write barrier back. Overwrites mark and tmp.
295
|.macro barrierback, tab, mark, tmp, target
296
| lw tmp, DISPATCH_GL(gc.grayagain)(DISPATCH)
297
| andi mark, mark, ~LJ_GC_BLACK & 255 // black2gray(tab)
298
| sw tab, DISPATCH_GL(gc.grayagain)(DISPATCH)
299
| sb mark, tab->marked
300
| b target
301
|. sw tmp, tab->gclist
302
|.endmacro
303
|
304
|//-----------------------------------------------------------------------
305
306
/* Generate subroutines used by opcodes and other parts of the VM. */
307
/* The .code_sub section should be last to help static branch prediction. */
308
static void build_subroutines(BuildCtx *ctx)
309
{
310
|.code_sub
311
|
312
|//-----------------------------------------------------------------------
313
|//-- Return handling ----------------------------------------------------
314
|//-----------------------------------------------------------------------
315
|
316
|->vm_returnp:
317
| // See vm_return. Also: TMP2 = previous base.
318
| andi AT, PC, FRAME_P
319
| beqz AT, ->cont_dispatch
320
|. li TMP1, LJ_TTRUE
321
|
322
| // Return from pcall or xpcall fast func.
323
| lw PC, FRAME_PC(TMP2) // Fetch PC of previous frame.
324
| move BASE, TMP2 // Restore caller base.
325
| // Prepending may overwrite the pcall frame, so do it at the end.
326
| sw TMP1, FRAME_PC(RA) // Prepend true to results.
327
| addiu RA, RA, -8
328
|
329
|->vm_returnc:
330
| addiu RD, RD, 8 // RD = (nresults+1)*8.
331
| andi TMP0, PC, FRAME_TYPE
332
| beqz RD, ->vm_unwind_c_eh
333
|. li CRET1, LUA_YIELD
334
| beqz TMP0, ->BC_RET_Z // Handle regular return to Lua.
335
|. move MULTRES, RD
336
|
337
|->vm_return:
338
| // BASE = base, RA = resultptr, RD/MULTRES = (nresults+1)*8, PC = return
339
| // TMP0 = PC & FRAME_TYPE
340
| li TMP2, -8
341
| xori AT, TMP0, FRAME_C
342
| and TMP2, PC, TMP2
343
| bnez AT, ->vm_returnp
344
| subu TMP2, BASE, TMP2 // TMP2 = previous base.
345
|
346
| addiu TMP1, RD, -8
347
| sw TMP2, L->base
348
| li_vmstate C
349
| lw TMP2, SAVE_NRES
350
| addiu BASE, BASE, -8
351
| st_vmstate
352
| beqz TMP1, >2
353
|. sll TMP2, TMP2, 3
354
|1:
355
| addiu TMP1, TMP1, -8
356
| ldc1 f0, 0(RA)
357
| addiu RA, RA, 8
358
| sdc1 f0, 0(BASE)
359
| bnez TMP1, <1
360
|. addiu BASE, BASE, 8
361
|
362
|2:
363
| bne TMP2, RD, >6
364
|3:
365
|. sw BASE, L->top // Store new top.
366
|
367
|->vm_leave_cp:
368
| lw TMP0, SAVE_CFRAME // Restore previous C frame.
369
| move CRET1, r0 // Ok return status for vm_pcall.
370
| sw TMP0, L->cframe
371
|
372
|->vm_leave_unw:
373
| restoreregs_ret
374
|
375
|6:
376
| lw TMP1, L->maxstack
377
| slt AT, TMP2, RD
378
| bnez AT, >7 // Less results wanted?
379
| // More results wanted. Check stack size and fill up results with nil.
380
|. slt AT, BASE, TMP1
381
| beqz AT, >8
382
|. nop
383
| sw TISNIL, HI(BASE)
384
| addiu RD, RD, 8
385
| b <2
386
|. addiu BASE, BASE, 8
387
|
388
|7: // Less results wanted.
389
| subu TMP0, RD, TMP2
390
| subu TMP0, BASE, TMP0 // Either keep top or shrink it.
391
| b <3
392
|. movn BASE, TMP0, TMP2 // LUA_MULTRET+1 case?
393
|
394
|8: // Corner case: need to grow stack for filling up results.
395
| // This can happen if:
396
| // - A C function grows the stack (a lot).
397
| // - The GC shrinks the stack in between.
398
| // - A return back from a lua_call() with (high) nresults adjustment.
399
| load_got lj_state_growstack
400
| move MULTRES, RD
401
| move CARG2, TMP2
402
| call_intern lj_state_growstack // (lua_State *L, int n)
403
|. move CARG1, L
404
| lw TMP2, SAVE_NRES
405
| lw BASE, L->top // Need the (realloced) L->top in BASE.
406
| move RD, MULTRES
407
| b <2
408
|. sll TMP2, TMP2, 3
409
|
410
|->vm_unwind_c: // Unwind C stack, return from vm_pcall.
411
| // (void *cframe, int errcode)
412
| move sp, CARG1
413
| move CRET1, CARG2
414
|->vm_unwind_c_eh: // Landing pad for external unwinder.
415
| lw L, SAVE_L
416
| li TMP0, ~LJ_VMST_C
417
| lw GL:TMP1, L->glref
418
| b ->vm_leave_unw
419
|. sw TMP0, GL:TMP1->vmstate
420
|
421
|->vm_unwind_ff: // Unwind C stack, return from ff pcall.
422
| // (void *cframe)
423
| li AT, -4
424
| and sp, CARG1, AT
425
|->vm_unwind_ff_eh: // Landing pad for external unwinder.
426
| lw L, SAVE_L
427
| lui TMP3, 0x59c0 // TOBIT = 2^52 + 2^51 (float).
428
| li TISNIL, LJ_TNIL
429
| lw BASE, L->base
430
| lw DISPATCH, L->glref // Setup pointer to dispatch table.
431
| mtc1 TMP3, TOBIT
432
| li TMP1, LJ_TFALSE
433
| li_vmstate INTERP
434
| lw PC, FRAME_PC(BASE) // Fetch PC of previous frame.
435
| cvt.d.s TOBIT, TOBIT
436
| addiu RA, BASE, -8 // Results start at BASE-8.
437
| addiu DISPATCH, DISPATCH, GG_G2DISP
438
| sw TMP1, HI(RA) // Prepend false to error message.
439
| st_vmstate
440
| b ->vm_returnc
441
|. li RD, 16 // 2 results: false + error message.
442
|
443
|//-----------------------------------------------------------------------
444
|//-- Grow stack for calls -----------------------------------------------
445
|//-----------------------------------------------------------------------
446
|
447
|->vm_growstack_c: // Grow stack for C function.
448
| b >2
449
|. li CARG2, LUA_MINSTACK
450
|
451
|->vm_growstack_l: // Grow stack for Lua function.
452
| // BASE = new base, RA = BASE+framesize*8, RC = nargs*8, PC = first PC
453
| addu RC, BASE, RC
454
| subu RA, RA, BASE
455
| sw BASE, L->base
456
| addiu PC, PC, 4 // Must point after first instruction.
457
| sw RC, L->top
458
| srl CARG2, RA, 3
459
|2:
460
| // L->base = new base, L->top = top
461
| load_got lj_state_growstack
462
| sw PC, SAVE_PC
463
| call_intern lj_state_growstack // (lua_State *L, int n)
464
|. move CARG1, L
465
| lw BASE, L->base
466
| lw RC, L->top
467
| lw LFUNC:RB, FRAME_FUNC(BASE)
468
| subu RC, RC, BASE
469
| // BASE = new base, RB = LFUNC/CFUNC, RC = nargs*8, FRAME_PC(BASE) = PC
470
| ins_callt // Just retry the call.
471
|
472
|//-----------------------------------------------------------------------
473
|//-- Entry points into the assembler VM ---------------------------------
474
|//-----------------------------------------------------------------------
475
|
476
|->vm_resume: // Setup C frame and resume thread.
477
| // (lua_State *L, TValue *base, int nres1 = 0, ptrdiff_t ef = 0)
478
| saveregs
479
| move L, CARG1
480
| lw DISPATCH, L->glref // Setup pointer to dispatch table.
481
| move BASE, CARG2
482
| lbu TMP1, L->status
483
| sw L, SAVE_L
484
| li PC, FRAME_CP
485
| addiu TMP0, sp, CFRAME_RESUME
486
| addiu DISPATCH, DISPATCH, GG_G2DISP
487
| sw r0, SAVE_NRES
488
| sw r0, SAVE_ERRF
489
| sw TMP0, L->cframe
490
| sw r0, SAVE_CFRAME
491
| beqz TMP1, >3
492
|. sw CARG1, SAVE_PC // Any value outside of bytecode is ok.
493
|
494
| // Resume after yield (like a return).
495
| move RA, BASE
496
| lw BASE, L->base
497
| lw TMP1, L->top
498
| lw PC, FRAME_PC(BASE)
499
| lui TMP3, 0x59c0 // TOBIT = 2^52 + 2^51 (float).
500
| subu RD, TMP1, BASE
501
| mtc1 TMP3, TOBIT
502
| sb r0, L->status
503
| cvt.d.s TOBIT, TOBIT
504
| li_vmstate INTERP
505
| addiu RD, RD, 8
506
| st_vmstate
507
| move MULTRES, RD
508
| andi TMP0, PC, FRAME_TYPE
509
| beqz TMP0, ->BC_RET_Z
510
|. li TISNIL, LJ_TNIL
511
| b ->vm_return
512
|. nop
513
|
514
|->vm_pcall: // Setup protected C frame and enter VM.
515
| // (lua_State *L, TValue *base, int nres1, ptrdiff_t ef)
516
| saveregs
517
| sw CARG4, SAVE_ERRF
518
| b >1
519
|. li PC, FRAME_CP
520
|
521
|->vm_call: // Setup C frame and enter VM.
522
| // (lua_State *L, TValue *base, int nres1)
523
| saveregs
524
| li PC, FRAME_C
525
|
526
|1: // Entry point for vm_pcall above (PC = ftype).
527
| lw TMP1, L:CARG1->cframe
528
| sw CARG3, SAVE_NRES
529
| move L, CARG1
530
| sw CARG1, SAVE_L
531
| move BASE, CARG2
532
| sw sp, L->cframe // Add our C frame to cframe chain.
533
| lw DISPATCH, L->glref // Setup pointer to dispatch table.
534
| sw CARG1, SAVE_PC // Any value outside of bytecode is ok.
535
| sw TMP1, SAVE_CFRAME
536
| addiu DISPATCH, DISPATCH, GG_G2DISP
537
|
538
|3: // Entry point for vm_cpcall/vm_resume (BASE = base, PC = ftype).
539
| lw TMP2, L->base // TMP2 = old base (used in vmeta_call).
540
| lui TMP3, 0x59c0 // TOBIT = 2^52 + 2^51 (float).
541
| lw TMP1, L->top
542
| mtc1 TMP3, TOBIT
543
| addu PC, PC, BASE
544
| subu NARGS8:RC, TMP1, BASE
545
| subu PC, PC, TMP2 // PC = frame delta + frame type
546
| cvt.d.s TOBIT, TOBIT
547
| li_vmstate INTERP
548
| li TISNIL, LJ_TNIL
549
| st_vmstate
550
|
551
|->vm_call_dispatch:
552
| // TMP2 = old base, BASE = new base, RC = nargs*8, PC = caller PC
553
| lw TMP0, FRAME_PC(BASE)
554
| li AT, LJ_TFUNC
555
| bne TMP0, AT, ->vmeta_call
556
|. lw LFUNC:RB, FRAME_FUNC(BASE)
557
|
558
|->vm_call_dispatch_f:
559
| ins_call
560
| // BASE = new base, RB = func, RC = nargs*8, PC = caller PC
561
|
562
|->vm_cpcall: // Setup protected C frame, call C.
563
| // (lua_State *L, lua_CFunction func, void *ud, lua_CPFunction cp)
564
| saveregs
565
| move L, CARG1
566
| lw TMP0, L:CARG1->stack
567
| sw CARG1, SAVE_L
568
| lw TMP1, L->top
569
| sw CARG1, SAVE_PC // Any value outside of bytecode is ok.
570
| subu TMP0, TMP0, TMP1 // Compute -savestack(L, L->top).
571
| lw TMP1, L->cframe
572
| sw sp, L->cframe // Add our C frame to cframe chain.
573
| sw TMP0, SAVE_NRES // Neg. delta means cframe w/o frame.
574
| sw r0, SAVE_ERRF // No error function.
575
| move CFUNCADDR, CARG4
576
| jalr CARG4 // (lua_State *L, lua_CFunction func, void *ud)
577
|. sw TMP1, SAVE_CFRAME
578
| move BASE, CRET1
579
| lw DISPATCH, L->glref // Setup pointer to dispatch table.
580
| li PC, FRAME_CP
581
| bnez CRET1, <3 // Else continue with the call.
582
|. addiu DISPATCH, DISPATCH, GG_G2DISP
583
| b ->vm_leave_cp // No base? Just remove C frame.
584
|. nop
585
|
586
|//-----------------------------------------------------------------------
587
|//-- Metamethod handling ------------------------------------------------
588
|//-----------------------------------------------------------------------
589
|
590
|// The lj_meta_* functions (except for lj_meta_cat) don't reallocate the
591
|// stack, so BASE doesn't need to be reloaded across these calls.
592
|
593
|//-- Continuation dispatch ----------------------------------------------
594
|
595
|->cont_dispatch:
596
| // BASE = meta base, RA = resultptr, RD = (nresults+1)*8
597
| lw TMP0, -16+LO(BASE) // Continuation.
598
| move RB, BASE
599
| move BASE, TMP2 // Restore caller BASE.
600
| lw LFUNC:TMP1, FRAME_FUNC(TMP2)
601
|.if FFI
602
| sltiu AT, TMP0, 2
603
|.endif
604
| lw PC, -16+HI(RB) // Restore PC from [cont|PC].
605
| addu TMP2, RA, RD
606
| lw TMP1, LFUNC:TMP1->pc
607
|.if FFI
608
| bnez AT, >1
609
|.endif
610
|. sw TISNIL, -8+HI(TMP2) // Ensure one valid arg.
611
| // BASE = base, RA = resultptr, RB = meta base
612
| jr TMP0 // Jump to continuation.
613
|. lw KBASE, PC2PROTO(k)(TMP1)
614
|
615
|.if FFI
616
|1:
617
| bnez TMP0, ->cont_ffi_callback // cont = 1: return from FFI callback.
618
| // cont = 0: tailcall from C function.
619
|. addiu TMP1, RB, -16
620
| b ->vm_call_tail
621
|. subu RC, TMP1, BASE
622
|.endif
623
|
624
|->cont_cat: // RA = resultptr, RB = meta base
625
| lw INS, -4(PC)
626
| addiu CARG2, RB, -16
627
| ldc1 f0, 0(RA)
628
| decode_RB8a MULTRES, INS
629
| decode_RA8a RA, INS
630
| decode_RB8b MULTRES
631
| decode_RA8b RA
632
| addu TMP1, BASE, MULTRES
633
| sw BASE, L->base
634
| subu CARG3, CARG2, TMP1
635
| bne TMP1, CARG2, ->BC_CAT_Z
636
|. sdc1 f0, 0(CARG2)
637
| addu RA, BASE, RA
638
| b ->cont_nop
639
|. sdc1 f0, 0(RA)
640
|
641
|//-- Table indexing metamethods -----------------------------------------
642
|
643
|->vmeta_tgets1:
644
| addiu CARG3, DISPATCH, DISPATCH_GL(tmptv)
645
| li TMP0, LJ_TSTR
646
| sw STR:RC, LO(CARG3)
647
| b >1
648
|. sw TMP0, HI(CARG3)
649
|
650
|->vmeta_tgets:
651
| addiu CARG2, DISPATCH, DISPATCH_GL(tmptv)
652
| li TMP0, LJ_TTAB
653
| sw TAB:RB, LO(CARG2)
654
| addiu CARG3, DISPATCH, DISPATCH_GL(tmptv2)
655
| sw TMP0, HI(CARG2)
656
| li TMP1, LJ_TSTR
657
| sw STR:RC, LO(CARG3)
658
| b >1
659
|. sw TMP1, HI(CARG3)
660
|
661
|->vmeta_tgetb: // TMP0 = index
662
| mtc1 TMP0, f0
663
| cvt.d.w f0, f0
664
| addiu CARG3, DISPATCH, DISPATCH_GL(tmptv)
665
| sdc1 f0, 0(CARG3)
666
|
667
|->vmeta_tgetv:
668
|1:
669
| load_got lj_meta_tget
670
| sw BASE, L->base
671
| sw PC, SAVE_PC
672
| call_intern lj_meta_tget // (lua_State *L, TValue *o, TValue *k)
673
|. move CARG1, L
674
| // Returns TValue * (finished) or NULL (metamethod).
675
| beqz CRET1, >3
676
|. addiu TMP1, BASE, -FRAME_CONT
677
| ldc1 f0, 0(CRET1)
678
| ins_next1
679
| sdc1 f0, 0(RA)
680
| ins_next2
681
|
682
|3: // Call __index metamethod.
683
| // BASE = base, L->top = new base, stack = cont/func/t/k
684
| lw BASE, L->top
685
| sw PC, -16+HI(BASE) // [cont|PC]
686
| subu PC, BASE, TMP1
687
| lw LFUNC:RB, FRAME_FUNC(BASE) // Guaranteed to be a function here.
688
| b ->vm_call_dispatch_f
689
|. li NARGS8:RC, 16 // 2 args for func(t, k).
690
|
691
|//-----------------------------------------------------------------------
692
|
693
|->vmeta_tsets1:
694
| addiu CARG3, DISPATCH, DISPATCH_GL(tmptv)
695
| li TMP0, LJ_TSTR
696
| sw STR:RC, LO(CARG3)
697
| b >1
698
|. sw TMP0, HI(CARG3)
699
|
700
|->vmeta_tsets:
701
| addiu CARG2, DISPATCH, DISPATCH_GL(tmptv)
702
| li TMP0, LJ_TTAB
703
| sw TAB:RB, LO(CARG2)
704
| addiu CARG3, DISPATCH, DISPATCH_GL(tmptv2)
705
| sw TMP0, HI(CARG2)
706
| li TMP1, LJ_TSTR
707
| sw STR:RC, LO(CARG3)
708
| b >1
709
|. sw TMP1, HI(CARG3)
710
|
711
|->vmeta_tsetb: // TMP0 = index
712
| mtc1 TMP0, f0
713
| cvt.d.w f0, f0
714
| addiu CARG3, DISPATCH, DISPATCH_GL(tmptv)
715
| sdc1 f0, 0(CARG3)
716
|
717
|->vmeta_tsetv:
718
|1:
719
| load_got lj_meta_tset
720
| sw BASE, L->base
721
| sw PC, SAVE_PC
722
| call_intern lj_meta_tset // (lua_State *L, TValue *o, TValue *k)
723
|. move CARG1, L
724
| // Returns TValue * (finished) or NULL (metamethod).
725
| beqz CRET1, >3
726
|. ldc1 f0, 0(RA)
727
| // NOBARRIER: lj_meta_tset ensures the table is not black.
728
| ins_next1
729
| sdc1 f0, 0(CRET1)
730
| ins_next2
731
|
732
|3: // Call __newindex metamethod.
733
| // BASE = base, L->top = new base, stack = cont/func/t/k/(v)
734
| addiu TMP1, BASE, -FRAME_CONT
735
| lw BASE, L->top
736
| sw PC, -16+HI(BASE) // [cont|PC]
737
| subu PC, BASE, TMP1
738
| lw LFUNC:RB, FRAME_FUNC(BASE) // Guaranteed to be a function here.
739
| sdc1 f0, 16(BASE) // Copy value to third argument.
740
| b ->vm_call_dispatch_f
741
|. li NARGS8:RC, 24 // 3 args for func(t, k, v)
742
|
743
|//-- Comparison metamethods ---------------------------------------------
744
|
745
|->vmeta_comp:
746
| // CARG2, CARG3 are already set by BC_ISLT/BC_ISGE/BC_ISLE/BC_ISGT.
747
| load_got lj_meta_comp
748
| addiu PC, PC, -4
749
| sw BASE, L->base
750
| sw PC, SAVE_PC
751
| decode_OP1 CARG4, INS
752
| call_intern lj_meta_comp // (lua_State *L, TValue *o1, *o2, int op)
753
|. move CARG1, L
754
| // Returns 0/1 or TValue * (metamethod).
755
|3:
756
| sltiu AT, CRET1, 2
757
| beqz AT, ->vmeta_binop
758
| negu TMP2, CRET1
759
|4:
760
| lhu RD, OFS_RD(PC)
761
| addiu PC, PC, 4
762
| lui TMP1, (-(BCBIAS_J*4 >> 16) & 65535)
763
| sll RD, RD, 2
764
| addu RD, RD, TMP1
765
| and RD, RD, TMP2
766
| addu PC, PC, RD
767
|->cont_nop:
768
| ins_next
769
|
770
|->cont_ra: // RA = resultptr
771
| lbu TMP1, -4+OFS_RA(PC)
772
| ldc1 f0, 0(RA)
773
| sll TMP1, TMP1, 3
774
| addu TMP1, BASE, TMP1
775
| b ->cont_nop
776
|. sdc1 f0, 0(TMP1)
777
|
778
|->cont_condt: // RA = resultptr
779
| lw TMP0, HI(RA)
780
| sltiu AT, TMP0, LJ_TISTRUECOND
781
| b <4
782
|. negu TMP2, AT // Branch if result is true.
783
|
784
|->cont_condf: // RA = resultptr
785
| lw TMP0, HI(RA)
786
| sltiu AT, TMP0, LJ_TISTRUECOND
787
| b <4
788
|. addiu TMP2, AT, -1 // Branch if result is false.
789
|
790
|->vmeta_equal:
791
| // CARG2, CARG3, CARG4 are already set by BC_ISEQV/BC_ISNEV.
792
| load_got lj_meta_equal
793
| addiu PC, PC, -4
794
| sw BASE, L->base
795
| sw PC, SAVE_PC
796
| call_intern lj_meta_equal // (lua_State *L, GCobj *o1, *o2, int ne)
797
|. move CARG1, L
798
| // Returns 0/1 or TValue * (metamethod).
799
| b <3
800
|. nop
801
|
802
|->vmeta_equal_cd:
803
|.if FFI
804
| load_got lj_meta_equal_cd
805
| move CARG2, INS
806
| addiu PC, PC, -4
807
| sw BASE, L->base
808
| sw PC, SAVE_PC
809
| call_intern lj_meta_equal_cd // (lua_State *L, BCIns op)
810
|. move CARG1, L
811
| // Returns 0/1 or TValue * (metamethod).
812
| b <3
813
|. nop
814
|.endif
815
|
816
|//-- Arithmetic metamethods ---------------------------------------------
817
|
818
|->vmeta_unm:
819
| move CARG4, CARG3
820
|
821
|->vmeta_arith:
822
| load_got lj_meta_arith
823
| decode_OP1 TMP0, INS
824
| sw BASE, L->base
825
| sw PC, SAVE_PC
826
| move CARG2, RA
827
| sw TMP0, ARG5
828
| call_intern lj_meta_arith // (lua_State *L, TValue *ra,*rb,*rc, BCReg op)
829
|. move CARG1, L
830
| // Returns NULL (finished) or TValue * (metamethod).
831
| beqz CRET1, ->cont_nop
832
|. nop
833
|
834
| // Call metamethod for binary op.
835
|->vmeta_binop:
836
| // BASE = old base, CRET1 = new base, stack = cont/func/o1/o2
837
| subu TMP1, CRET1, BASE
838
| sw PC, -16+HI(CRET1) // [cont|PC]
839
| move TMP2, BASE
840
| addiu PC, TMP1, FRAME_CONT
841
| move BASE, CRET1
842
| b ->vm_call_dispatch
843
|. li NARGS8:RC, 16 // 2 args for func(o1, o2).
844
|
845
|->vmeta_len:
846
| // CARG2 already set by BC_LEN.
847
#if LJ_52
848
| move MULTRES, CARG1
849
#endif
850
| load_got lj_meta_len
851
| sw BASE, L->base
852
| sw PC, SAVE_PC
853
| call_intern lj_meta_len // (lua_State *L, TValue *o)
854
|. move CARG1, L
855
| // Returns NULL (retry) or TValue * (metamethod base).
856
#if LJ_52
857
| bnez CRET1, ->vmeta_binop // Binop call for compatibility.
858
|. nop
859
| b ->BC_LEN_Z
860
|. move CARG1, MULTRES
861
#else
862
| b ->vmeta_binop // Binop call for compatibility.
863
|. nop
864
#endif
865
|
866
|//-- Call metamethod ----------------------------------------------------
867
|
868
|->vmeta_call: // Resolve and call __call metamethod.
869
| // TMP2 = old base, BASE = new base, RC = nargs*8
870
| load_got lj_meta_call
871
| sw TMP2, L->base // This is the callers base!
872
| addiu CARG2, BASE, -8
873
| sw PC, SAVE_PC
874
| addu CARG3, BASE, RC
875
| move MULTRES, NARGS8:RC
876
| call_intern lj_meta_call // (lua_State *L, TValue *func, TValue *top)
877
|. move CARG1, L
878
| lw LFUNC:RB, FRAME_FUNC(BASE) // Guaranteed to be a function here.
879
| addiu NARGS8:RC, MULTRES, 8 // Got one more argument now.
880
| ins_call
881
|
882
|->vmeta_callt: // Resolve __call for BC_CALLT.
883
| // BASE = old base, RA = new base, RC = nargs*8
884
| load_got lj_meta_call
885
| sw BASE, L->base
886
| addiu CARG2, RA, -8
887
| sw PC, SAVE_PC
888
| addu CARG3, RA, RC
889
| move MULTRES, NARGS8:RC
890
| call_intern lj_meta_call // (lua_State *L, TValue *func, TValue *top)
891
|. move CARG1, L
892
| lw TMP1, FRAME_PC(BASE)
893
| lw LFUNC:RB, FRAME_FUNC(RA) // Guaranteed to be a function here.
894
| b ->BC_CALLT_Z
895
|. addiu NARGS8:RC, MULTRES, 8 // Got one more argument now.
896
|
897
|//-- Argument coercion for 'for' statement ------------------------------
898
|
899
|->vmeta_for:
900
| load_got lj_meta_for
901
| sw BASE, L->base
902
| move CARG2, RA
903
| sw PC, SAVE_PC
904
| move MULTRES, INS
905
| call_intern lj_meta_for // (lua_State *L, TValue *base)
906
|. move CARG1, L
907
|.if JIT
908
| decode_OP1 TMP0, MULTRES
909
| li AT, BC_JFORI
910
|.endif
911
| decode_RA8a RA, MULTRES
912
| decode_RD8a RD, MULTRES
913
| decode_RA8b RA
914
|.if JIT
915
| beq TMP0, AT, =>BC_JFORI
916
|. decode_RD8b RD
917
| b =>BC_FORI
918
|. nop
919
|.else
920
| b =>BC_FORI
921
|. decode_RD8b RD
922
|.endif
923
|
924
|//-----------------------------------------------------------------------
925
|//-- Fast functions -----------------------------------------------------
926
|//-----------------------------------------------------------------------
927
|
928
|.macro .ffunc, name
929
|->ff_ .. name:
930
|.endmacro
931
|
932
|.macro .ffunc_1, name
933
|->ff_ .. name:
934
| beqz NARGS8:RC, ->fff_fallback
935
|. lw CARG3, HI(BASE)
936
| lw CARG1, LO(BASE)
937
|.endmacro
938
|
939
|.macro .ffunc_2, name
940
|->ff_ .. name:
941
| sltiu AT, NARGS8:RC, 16
942
| lw CARG3, HI(BASE)
943
| bnez AT, ->fff_fallback
944
|. lw CARG4, 8+HI(BASE)
945
| lw CARG1, LO(BASE)
946
| lw CARG2, 8+LO(BASE)
947
|.endmacro
948
|
949
|.macro .ffunc_n, name // Caveat: has delay slot!
950
|->ff_ .. name:
951
| lw CARG3, HI(BASE)
952
| beqz NARGS8:RC, ->fff_fallback
953
|. ldc1 FARG1, 0(BASE)
954
| sltiu AT, CARG3, LJ_TISNUM
955
| beqz AT, ->fff_fallback
956
|.endmacro
957
|
958
|.macro .ffunc_nn, name // Caveat: has delay slot!
959
|->ff_ .. name:
960
| sltiu AT, NARGS8:RC, 16
961
| lw CARG3, HI(BASE)
962
| bnez AT, ->fff_fallback
963
|. lw CARG4, 8+HI(BASE)
964
| ldc1 FARG1, 0(BASE)
965
| ldc1 FARG2, 8(BASE)
966
| sltiu TMP0, CARG3, LJ_TISNUM
967
| sltiu TMP1, CARG4, LJ_TISNUM
968
| and TMP0, TMP0, TMP1
969
| beqz TMP0, ->fff_fallback
970
|.endmacro
971
|
972
|// Inlined GC threshold check. Caveat: uses TMP0 and TMP1 and has delay slot!
973
|.macro ffgccheck
974
| lw TMP0, DISPATCH_GL(gc.total)(DISPATCH)
975
| lw TMP1, DISPATCH_GL(gc.threshold)(DISPATCH)
976
| subu AT, TMP0, TMP1
977
| bgezal AT, ->fff_gcstep
978
|.endmacro
979
|
980
|//-- Base library: checks -----------------------------------------------
981
|
982
|.ffunc_1 assert
983
| sltiu AT, CARG3, LJ_TISTRUECOND
984
| beqz AT, ->fff_fallback
985
|. addiu RA, BASE, -8
986
| lw PC, FRAME_PC(BASE)
987
| addiu RD, NARGS8:RC, 8 // Compute (nresults+1)*8.
988
| addu TMP2, RA, NARGS8:RC
989
| sw CARG3, HI(RA)
990
| addiu TMP1, BASE, 8
991
| beq BASE, TMP2, ->fff_res // Done if exactly 1 argument.
992
|. sw CARG1, LO(RA)
993
|1:
994
| ldc1 f0, 0(TMP1)
995
| sdc1 f0, -8(TMP1)
996
| bne TMP1, TMP2, <1
997
|. addiu TMP1, TMP1, 8
998
| b ->fff_res
999
|. nop
1000
|
1001
|.ffunc type
1002
| lw CARG3, HI(BASE)
1003
| li TMP1, LJ_TISNUM
1004
| beqz NARGS8:RC, ->fff_fallback
1005
|. sltiu TMP0, CARG3, LJ_TISNUM
1006
| movz TMP1, CARG3, TMP0
1007
| not TMP1, TMP1
1008
| sll TMP1, TMP1, 3
1009
| addu TMP1, CFUNC:RB, TMP1
1010
| b ->fff_resn
1011
|. ldc1 FRET1, CFUNC:TMP1->upvalue
1012
|
1013
|//-- Base library: getters and setters ---------------------------------
1014
|
1015
|.ffunc_1 getmetatable
1016
| li AT, LJ_TTAB
1017
| bne CARG3, AT, >6
1018
|. li AT, LJ_TUDATA
1019
|1: // Field metatable must be at same offset for GCtab and GCudata!
1020
| lw TAB:CARG1, TAB:CARG1->metatable
1021
|2:
1022
| lw STR:RC, DISPATCH_GL(gcroot[GCROOT_MMNAME+MM_metatable])(DISPATCH)
1023
| beqz TAB:CARG1, ->fff_restv
1024
|. li CARG3, LJ_TNIL
1025
| lw TMP0, TAB:CARG1->hmask
1026
| li CARG3, LJ_TTAB // Use metatable as default result.
1027
| lw TMP1, STR:RC->hash
1028
| lw NODE:TMP2, TAB:CARG1->node
1029
| and TMP1, TMP1, TMP0 // idx = str->hash & tab->hmask
1030
| sll TMP0, TMP1, 5
1031
| sll TMP1, TMP1, 3
1032
| subu TMP1, TMP0, TMP1
1033
| addu NODE:TMP2, NODE:TMP2, TMP1 // node = tab->node + (idx*32-idx*8)
1034
| li AT, LJ_TSTR
1035
|3: // Rearranged logic, because we expect _not_ to find the key.
1036
| lw CARG4, offsetof(Node, key)+HI(NODE:TMP2)
1037
| lw TMP0, offsetof(Node, key)+LO(NODE:TMP2)
1038
| lw NODE:TMP3, NODE:TMP2->next
1039
| bne CARG4, AT, >4
1040
|. lw CARG2, offsetof(Node, val)+HI(NODE:TMP2)
1041
| beq TMP0, STR:RC, >5
1042
|. lw TMP1, offsetof(Node, val)+LO(NODE:TMP2)
1043
|4:
1044
| beqz NODE:TMP3, ->fff_restv // Not found, keep default result.
1045
|. move NODE:TMP2, NODE:TMP3
1046
| b <3
1047
|. nop
1048
|5:
1049
| beq CARG2, TISNIL, ->fff_restv // Ditto for nil value.
1050
|. nop
1051
| move CARG3, CARG2 // Return value of mt.__metatable.
1052
| b ->fff_restv
1053
|. move CARG1, TMP1
1054
|
1055
|6:
1056
| beq CARG3, AT, <1
1057
|. sltiu TMP0, CARG3, LJ_TISNUM
1058
| li TMP1, LJ_TISNUM
1059
| movz TMP1, CARG3, TMP0
1060
| not TMP1, TMP1
1061
| sll TMP1, TMP1, 2
1062
| addu TMP1, DISPATCH, TMP1
1063
| b <2
1064
|. lw TAB:CARG1, DISPATCH_GL(gcroot[GCROOT_BASEMT])(TMP1)
1065
|
1066
|.ffunc_2 setmetatable
1067
| // Fast path: no mt for table yet and not clearing the mt.
1068
| li AT, LJ_TTAB
1069
| bne CARG3, AT, ->fff_fallback
1070
|. addiu CARG4, CARG4, -LJ_TTAB
1071
| lw TAB:TMP1, TAB:CARG1->metatable
1072
| lbu TMP3, TAB:CARG1->marked
1073
| or AT, CARG4, TAB:TMP1
1074
| bnez AT, ->fff_fallback
1075
|. andi AT, TMP3, LJ_GC_BLACK // isblack(table)
1076
| beqz AT, ->fff_restv
1077
|. sw TAB:CARG2, TAB:CARG1->metatable
1078
| barrierback TAB:CARG1, TMP3, TMP0, ->fff_restv
1079
|
1080
|.ffunc rawget
1081
| lw CARG4, HI(BASE)
1082
| sltiu AT, NARGS8:RC, 16
1083
| lw TAB:CARG2, LO(BASE)
1084
| load_got lj_tab_get
1085
| addiu CARG4, CARG4, -LJ_TTAB
1086
| or AT, AT, CARG4
1087
| bnez AT, ->fff_fallback
1088
| addiu CARG3, BASE, 8
1089
| call_intern lj_tab_get // (lua_State *L, GCtab *t, cTValue *key)
1090
|. move CARG1, L
1091
| // Returns cTValue *.
1092
| b ->fff_resn
1093
|. ldc1 FRET1, 0(CRET1)
1094
|
1095
|//-- Base library: conversions ------------------------------------------
1096
|
1097
|.ffunc tonumber
1098
| // Only handles the number case inline (without a base argument).
1099
| lw CARG1, HI(BASE)
1100
| xori AT, NARGS8:RC, 8
1101
| sltiu CARG1, CARG1, LJ_TISNUM
1102
| movn CARG1, r0, AT
1103
| beqz CARG1, ->fff_fallback // Exactly one number argument.
1104
|. ldc1 FRET1, 0(BASE)
1105
| b ->fff_resn
1106
|. nop
1107
|
1108
|.ffunc_1 tostring
1109
| // Only handles the string or number case inline.
1110
| li AT, LJ_TSTR
1111
| // A __tostring method in the string base metatable is ignored.
1112
| beq CARG3, AT, ->fff_restv // String key?
1113
| // Handle numbers inline, unless a number base metatable is present.
1114
|. lw TMP1, DISPATCH_GL(gcroot[GCROOT_BASEMT_NUM])(DISPATCH)
1115
| sltiu TMP0, CARG3, LJ_TISNUM
1116
| sltiu TMP1, TMP1, 1
1117
| and TMP0, TMP0, TMP1
1118
| beqz TMP0, ->fff_fallback
1119
|. sw BASE, L->base // Add frame since C call can throw.
1120
| ffgccheck
1121
|. sw PC, SAVE_PC // Redundant (but a defined value).
1122
| load_got lj_str_fromnum
1123
| move CARG1, L
1124
| call_intern lj_str_fromnum // (lua_State *L, lua_Number *np)
1125
|. move CARG2, BASE
1126
| // Returns GCstr *.
1127
| li CARG3, LJ_TSTR
1128
| b ->fff_restv
1129
|. move CARG1, CRET1
1130
|
1131
|//-- Base library: iterators -------------------------------------------
1132
|
1133
|.ffunc next
1134
| lw CARG1, HI(BASE)
1135
| lw TAB:CARG2, LO(BASE)
1136
| beqz NARGS8:RC, ->fff_fallback
1137
|. addu TMP2, BASE, NARGS8:RC
1138
| li AT, LJ_TTAB
1139
| sw TISNIL, HI(TMP2) // Set missing 2nd arg to nil.
1140
| bne CARG1, AT, ->fff_fallback
1141
|. lw PC, FRAME_PC(BASE)
1142
| load_got lj_tab_next
1143
| sw BASE, L->base // Add frame since C call can throw.
1144
| sw BASE, L->top // Dummy frame length is ok.
1145
| addiu CARG3, BASE, 8
1146
| sw PC, SAVE_PC
1147
| call_intern lj_tab_next // (lua_State *L, GCtab *t, TValue *key)
1148
|. move CARG1, L
1149
| // Returns 0 at end of traversal.
1150
| beqz CRET1, ->fff_restv // End of traversal: return nil.
1151
|. li CARG3, LJ_TNIL
1152
| ldc1 f0, 8(BASE) // Copy key and value to results.
1153
| addiu RA, BASE, -8
1154
| ldc1 f2, 16(BASE)
1155
| li RD, (2+1)*8
1156
| sdc1 f0, 0(RA)
1157
| b ->fff_res
1158
|. sdc1 f2, 8(RA)
1159
|
1160
|.ffunc_1 pairs
1161
| li AT, LJ_TTAB
1162
| bne CARG3, AT, ->fff_fallback
1163
|. lw PC, FRAME_PC(BASE)
1164
#if LJ_52
1165
| lw TAB:TMP2, TAB:CARG1->metatable
1166
| ldc1 f0, CFUNC:RB->upvalue[0]
1167
| bnez TAB:TMP2, ->fff_fallback
1168
#else
1169
| ldc1 f0, CFUNC:RB->upvalue[0]
1170
#endif
1171
|. addiu RA, BASE, -8
1172
| sw TISNIL, 8+HI(BASE)
1173
| li RD, (3+1)*8
1174
| b ->fff_res
1175
|. sdc1 f0, 0(RA)
1176
|
1177
|.ffunc ipairs_aux
1178
| sltiu AT, NARGS8:RC, 16
1179
| lw CARG3, HI(BASE)
1180
| lw TAB:CARG1, LO(BASE)
1181
| lw CARG4, 8+HI(BASE)
1182
| bnez AT, ->fff_fallback
1183
|. ldc1 FARG2, 8(BASE)
1184
| addiu CARG3, CARG3, -LJ_TTAB
1185
| sltiu AT, CARG4, LJ_TISNUM
1186
| li TMP0, 1
1187
| movn AT, r0, CARG3
1188
| mtc1 TMP0, FARG1
1189
| beqz AT, ->fff_fallback
1190
|. lw PC, FRAME_PC(BASE)
1191
| cvt.w.d FRET1, FARG2
1192
| cvt.d.w FARG1, FARG1
1193
| lw TMP0, TAB:CARG1->asize
1194
| lw TMP1, TAB:CARG1->array
1195
| mfc1 TMP2, FRET1
1196
| addiu RA, BASE, -8
1197
| add.d FARG2, FARG2, FARG1
1198
| addiu TMP2, TMP2, 1
1199
| sltu AT, TMP2, TMP0
1200
| sll TMP3, TMP2, 3
1201
| addu TMP3, TMP1, TMP3
1202
| beqz AT, >2 // Not in array part?
1203
|. sdc1 FARG2, 0(RA)
1204
| lw TMP2, HI(TMP3)
1205
| ldc1 f0, 0(TMP3)
1206
|1:
1207
| beq TMP2, TISNIL, ->fff_res // End of iteration, return 0 results.
1208
|. li RD, (0+1)*8
1209
| li RD, (2+1)*8
1210
| b ->fff_res
1211
|. sdc1 f0, 8(RA)
1212
|2: // Check for empty hash part first. Otherwise call C function.
1213
| lw TMP0, TAB:CARG1->hmask
1214
| load_got lj_tab_getinth
1215
| beqz TMP0, ->fff_res
1216
|. li RD, (0+1)*8
1217
| call_intern lj_tab_getinth // (GCtab *t, int32_t key)
1218
|. move CARG2, TMP2
1219
| // Returns cTValue * or NULL.
1220
| beqz CRET1, ->fff_res
1221
|. li RD, (0+1)*8
1222
| lw TMP2, HI(CRET1)
1223
| b <1
1224
|. ldc1 f0, 0(CRET1)
1225
|
1226
|.ffunc_1 ipairs
1227
| li AT, LJ_TTAB
1228
| bne CARG3, AT, ->fff_fallback
1229
|. lw PC, FRAME_PC(BASE)
1230
#if LJ_52
1231
| lw TAB:TMP2, TAB:CARG1->metatable
1232
| ldc1 f0, CFUNC:RB->upvalue[0]
1233
| bnez TAB:TMP2, ->fff_fallback
1234
#else
1235
| ldc1 f0, CFUNC:RB->upvalue[0]
1236
#endif
1237
|. addiu RA, BASE, -8
1238
| sw r0, 8+HI(BASE)
1239
| sw r0, 8+LO(BASE)
1240
| li RD, (3+1)*8
1241
| b ->fff_res
1242
|. sdc1 f0, 0(RA)
1243
|
1244
|//-- Base library: catch errors ----------------------------------------
1245
|
1246
|.ffunc pcall
1247
| lbu TMP3, DISPATCH_GL(hookmask)(DISPATCH)
1248
| beqz NARGS8:RC, ->fff_fallback
1249
| move TMP2, BASE
1250
| addiu BASE, BASE, 8
1251
| // Remember active hook before pcall.
1252
| srl TMP3, TMP3, HOOK_ACTIVE_SHIFT
1253
| andi TMP3, TMP3, 1
1254
| addiu PC, TMP3, 8+FRAME_PCALL
1255
| b ->vm_call_dispatch
1256
|. addiu NARGS8:RC, NARGS8:RC, -8
1257
|
1258
|.ffunc xpcall
1259
| sltiu AT, NARGS8:RC, 16
1260
| lw CARG4, 8+HI(BASE)
1261
| bnez AT, ->fff_fallback
1262
|. ldc1 FARG2, 8(BASE)
1263
| ldc1 FARG1, 0(BASE)
1264
| lbu TMP1, DISPATCH_GL(hookmask)(DISPATCH)
1265
| li AT, LJ_TFUNC
1266
| move TMP2, BASE
1267
| bne CARG4, AT, ->fff_fallback // Traceback must be a function.
1268
| addiu BASE, BASE, 16
1269
| // Remember active hook before pcall.
1270
| srl TMP3, TMP3, HOOK_ACTIVE_SHIFT
1271
| sdc1 FARG2, 0(TMP2) // Swap function and traceback.
1272
| andi TMP3, TMP3, 1
1273
| sdc1 FARG1, 8(TMP2)
1274
| addiu PC, TMP3, 16+FRAME_PCALL
1275
| b ->vm_call_dispatch
1276
|. addiu NARGS8:RC, NARGS8:RC, -16
1277
|
1278
|//-- Coroutine library --------------------------------------------------
1279
|
1280
|.macro coroutine_resume_wrap, resume
1281
|.if resume
1282
|.ffunc_1 coroutine_resume
1283
| li AT, LJ_TTHREAD
1284
| bne CARG3, AT, ->fff_fallback
1285
|.else
1286
|.ffunc coroutine_wrap_aux
1287
| lw L:CARG1, CFUNC:RB->upvalue[0].gcr
1288
|.endif
1289
| lbu TMP0, L:CARG1->status
1290
| lw TMP1, L:CARG1->cframe
1291
| lw CARG2, L:CARG1->top
1292
| lw TMP2, L:CARG1->base
1293
| addiu TMP3, TMP0, -LUA_YIELD
1294
| bgtz TMP3, ->fff_fallback // st > LUA_YIELD?
1295
|. xor TMP2, TMP2, CARG2
1296
| bnez TMP1, ->fff_fallback // cframe != 0?
1297
|. or AT, TMP2, TMP0
1298
| lw TMP0, L:CARG1->maxstack
1299
| beqz AT, ->fff_fallback // base == top && st == 0?
1300
|. lw PC, FRAME_PC(BASE)
1301
| addu TMP2, CARG2, NARGS8:RC
1302
| sltu AT, TMP0, TMP2
1303
| bnez AT, ->fff_fallback // Stack overflow?
1304
|. sw PC, SAVE_PC
1305
| sw BASE, L->base
1306
|1:
1307
|.if resume
1308
| addiu BASE, BASE, 8 // Keep resumed thread in stack for GC.
1309
| addiu NARGS8:RC, NARGS8:RC, -8
1310
| addiu TMP2, TMP2, -8
1311
|.endif
1312
| sw TMP2, L:CARG1->top
1313
| addu TMP1, BASE, NARGS8:RC
1314
| move CARG3, CARG2
1315
| sw BASE, L->top
1316
|2: // Move args to coroutine.
1317
| ldc1 f0, 0(BASE)
1318
| sltu AT, BASE, TMP1
1319
| beqz AT, >3
1320
|. addiu BASE, BASE, 8
1321
| sdc1 f0, 0(CARG3)
1322
| b <2
1323
|. addiu CARG3, CARG3, 8
1324
|3:
1325
| bal ->vm_resume // (lua_State *L, TValue *base, 0, 0)
1326
|. move L:RA, L:CARG1
1327
| // Returns thread status.
1328
|4:
1329
| lw TMP2, L:RA->base
1330
| sltiu AT, CRET1, LUA_YIELD+1
1331
| lw TMP3, L:RA->top
1332
| li_vmstate INTERP
1333
| lw BASE, L->base
1334
| st_vmstate
1335
| beqz AT, >8
1336
|. subu RD, TMP3, TMP2
1337
| lw TMP0, L->maxstack
1338
| beqz RD, >6 // No results?
1339
|. addu TMP1, BASE, RD
1340
| sltu AT, TMP0, TMP1
1341
| bnez AT, >9 // Need to grow stack?
1342
|. addu TMP3, TMP2, RD
1343
| sw TMP2, L:RA->top // Clear coroutine stack.
1344
| move TMP1, BASE
1345
|5: // Move results from coroutine.
1346
| ldc1 f0, 0(TMP2)
1347
| addiu TMP2, TMP2, 8
1348
| sltu AT, TMP2, TMP3
1349
| sdc1 f0, 0(TMP1)
1350
| bnez AT, <5
1351
|. addiu TMP1, TMP1, 8
1352
|6:
1353
| andi TMP0, PC, FRAME_TYPE
1354
|.if resume
1355
| li TMP1, LJ_TTRUE
1356
| addiu RA, BASE, -8
1357
| sw TMP1, -8+HI(BASE) // Prepend true to results.
1358
| addiu RD, RD, 16
1359
|.else
1360
| move RA, BASE
1361
| addiu RD, RD, 8
1362
|.endif
1363
|7:
1364
| sw PC, SAVE_PC
1365
| beqz TMP0, ->BC_RET_Z
1366
|. move MULTRES, RD
1367
| b ->vm_return
1368
|. nop
1369
|
1370
|8: // Coroutine returned with error (at co->top-1).
1371
|.if resume
1372
| addiu TMP3, TMP3, -8
1373
| li TMP1, LJ_TFALSE
1374
| ldc1 f0, 0(TMP3)
1375
| sw TMP3, L:RA->top // Remove error from coroutine stack.
1376
| li RD, (2+1)*8
1377
| sw TMP1, -8+HI(BASE) // Prepend false to results.
1378
| addiu RA, BASE, -8
1379
| sdc1 f0, 0(BASE) // Copy error message.
1380
| b <7
1381
|. andi TMP0, PC, FRAME_TYPE
1382
|.else
1383
| load_got lj_ffh_coroutine_wrap_err
1384
| move CARG2, L:RA
1385
| call_intern lj_ffh_coroutine_wrap_err // (lua_State *L, lua_State *co)
1386
|. move CARG1, L
1387
|.endif
1388
|
1389
|9: // Handle stack expansion on return from yield.
1390
| load_got lj_state_growstack
1391
| srl CARG2, RD, 3
1392
| call_intern lj_state_growstack // (lua_State *L, int n)
1393
|. move CARG1, L
1394
| b <4
1395
|. li CRET1, 0
1396
|.endmacro
1397
|
1398
| coroutine_resume_wrap 1 // coroutine.resume
1399
| coroutine_resume_wrap 0 // coroutine.wrap
1400
|
1401
|.ffunc coroutine_yield
1402
| lw TMP0, L->cframe
1403
| addu TMP1, BASE, NARGS8:RC
1404
| sw BASE, L->base
1405
| andi TMP0, TMP0, CFRAME_RESUME
1406
| sw TMP1, L->top
1407
| beqz TMP0, ->fff_fallback
1408
|. li CRET1, LUA_YIELD
1409
| sw r0, L->cframe
1410
| b ->vm_leave_unw
1411
|. sb CRET1, L->status
1412
|
1413
|//-- Math library -------------------------------------------------------
1414
|
1415
|.ffunc_n math_abs
1416
|. abs.d FRET1, FARG1
1417
|->fff_resn:
1418
| lw PC, FRAME_PC(BASE)
1419
| addiu RA, BASE, -8
1420
| b ->fff_res1
1421
|. sdc1 FRET1, -8(BASE)
1422
|
1423
|->fff_restv:
1424
| // CARG3/CARG1 = TValue result.
1425
| lw PC, FRAME_PC(BASE)
1426
| sw CARG3, -8+HI(BASE)
1427
| addiu RA, BASE, -8
1428
| sw CARG1, -8+LO(BASE)
1429
|->fff_res1:
1430
| // RA = results, PC = return.
1431
| li RD, (1+1)*8
1432
|->fff_res:
1433
| // RA = results, RD = (nresults+1)*8, PC = return.
1434
| andi TMP0, PC, FRAME_TYPE
1435
| bnez TMP0, ->vm_return
1436
|. move MULTRES, RD
1437
| lw INS, -4(PC)
1438
| decode_RB8a RB, INS
1439
| decode_RB8b RB
1440
|5:
1441
| sltu AT, RD, RB
1442
| bnez AT, >6 // More results expected?
1443
|. decode_RA8a TMP0, INS
1444
| decode_RA8b TMP0
1445
| ins_next1
1446
| // Adjust BASE. KBASE is assumed to be set for the calling frame.
1447
| subu BASE, RA, TMP0
1448
| ins_next2
1449
|
1450
|6: // Fill up results with nil.
1451
| addu TMP1, RA, RD
1452
| addiu RD, RD, 8
1453
| b <5
1454
|. sw TISNIL, -8+HI(TMP1)
1455
|
1456
|.macro math_extern, func
1457
|->ff_math_ .. func:
1458
| lw CARG3, HI(BASE)
1459
| beqz NARGS8:RC, ->fff_fallback
1460
|. load_got func
1461
| sltiu AT, CARG3, LJ_TISNUM
1462
| beqz AT, ->fff_fallback
1463
|. nop
1464
| call_extern
1465
|. ldc1 FARG1, 0(BASE)
1466
| b ->fff_resn
1467
|. nop
1468
|.endmacro
1469
|
1470
|.macro math_extern2, func
1471
| .ffunc_nn math_ .. func
1472
|. load_got func
1473
| call_extern
1474
|. nop
1475
| b ->fff_resn
1476
|. nop
1477
|.endmacro
1478
|
1479
|.macro math_round, func
1480
| .ffunc_n math_ .. func
1481
|. nop
1482
| bal ->vm_ .. func
1483
|. nop
1484
| b ->fff_resn
1485
|. nop
1486
|.endmacro
1487
|
1488
| math_round floor
1489
| math_round ceil
1490
|
1491
|.ffunc math_log
1492
| lw CARG3, HI(BASE)
1493
| li AT, 8
1494
| bne NARGS8:RC, AT, ->fff_fallback // Exactly 1 argument.
1495
|. load_got log
1496
| sltiu AT, CARG3, LJ_TISNUM
1497
| beqz AT, ->fff_fallback
1498
|. nop
1499
| call_extern
1500
|. ldc1 FARG1, 0(BASE)
1501
| b ->fff_resn
1502
|. nop
1503
|
1504
| math_extern log10
1505
| math_extern exp
1506
| math_extern sin
1507
| math_extern cos
1508
| math_extern tan
1509
| math_extern asin
1510
| math_extern acos
1511
| math_extern atan
1512
| math_extern sinh
1513
| math_extern cosh
1514
| math_extern tanh
1515
| math_extern2 pow
1516
| math_extern2 atan2
1517
| math_extern2 fmod
1518
|
1519
|.ffunc_n math_sqrt
1520
|. sqrt.d FRET1, FARG1
1521
| b ->fff_resn
1522
|. nop
1523
|
1524
|->ff_math_deg:
1525
|.ffunc_n math_rad
1526
|. ldc1 FARG2, CFUNC:RB->upvalue[0]
1527
| b ->fff_resn
1528
|. mul.d FRET1, FARG1, FARG2
1529
|
1530
|.ffunc_nn math_ldexp
1531
| cvt.w.d FARG2, FARG2
1532
| load_got ldexp
1533
| mfc1 CARG3, FARG2
1534
| call_extern
1535
|. nop
1536
| b ->fff_resn
1537
|. nop
1538
|
1539
|.ffunc_n math_frexp
1540
| load_got frexp
1541
| lw PC, FRAME_PC(BASE)
1542
| call_extern
1543
|. addiu CARG3, DISPATCH, DISPATCH_GL(tmptv)
1544
| lw TMP1, DISPATCH_GL(tmptv)(DISPATCH)
1545
| addiu RA, BASE, -8
1546
| mtc1 TMP1, FARG2
1547
| sdc1 FRET1, 0(RA)
1548
| cvt.d.w FARG2, FARG2
1549
| sdc1 FARG2, 8(RA)
1550
| b ->fff_res
1551
|. li RD, (2+1)*8
1552
|
1553
|.ffunc_n math_modf
1554
| load_got modf
1555
| lw PC, FRAME_PC(BASE)
1556
| call_extern
1557
|. addiu CARG3, BASE, -8
1558
| addiu RA, BASE, -8
1559
| sdc1 FRET1, 0(BASE)
1560
| b ->fff_res
1561
|. li RD, (2+1)*8
1562
|
1563
|.macro math_minmax, name, ismax
1564
|->ff_ .. name:
1565
| lw CARG3, HI(BASE)
1566
| beqz NARGS8:RC, ->fff_fallback
1567
|. ldc1 FRET1, 0(BASE)
1568
| sltiu AT, CARG3, LJ_TISNUM
1569
| beqz AT, ->fff_fallback
1570
|. addu TMP2, BASE, NARGS8:RC
1571
| addiu TMP1, BASE, 8
1572
| beq TMP1, TMP2, ->fff_resn
1573
|1:
1574
|. lw CARG3, HI(TMP1)
1575
| ldc1 FARG1, 0(TMP1)
1576
| addiu TMP1, TMP1, 8
1577
| sltiu AT, CARG3, LJ_TISNUM
1578
| beqz AT, ->fff_fallback
1579
|.if ismax
1580
|. c.olt.d FARG1, FRET1
1581
|.else
1582
|. c.olt.d FRET1, FARG1
1583
|.endif
1584
| bne TMP1, TMP2, <1
1585
|. movf.d FRET1, FARG1
1586
| b ->fff_resn
1587
|. nop
1588
|.endmacro
1589
|
1590
| math_minmax math_min, 0
1591
| math_minmax math_max, 1
1592
|
1593
|//-- String library -----------------------------------------------------
1594
|
1595
|.ffunc_1 string_len
1596
| li AT, LJ_TSTR
1597
| bne CARG3, AT, ->fff_fallback
1598
|. nop
1599
| b ->fff_resi
1600
|. lw CRET1, STR:CARG1->len
1601
|
1602
|.ffunc string_byte // Only handle the 1-arg case here.
1603
| lw CARG3, HI(BASE)
1604
| lw STR:CARG1, LO(BASE)
1605
| xori AT, NARGS8:RC, 8
1606
| addiu CARG3, CARG3, -LJ_TSTR
1607
| or AT, AT, CARG3
1608
| bnez AT, ->fff_fallback // Need exactly 1 string argument.
1609
|. nop
1610
| lw TMP0, STR:CARG1->len
1611
| lbu TMP1, STR:CARG1[1] // Access is always ok (NUL at end).
1612
| addiu RA, BASE, -8
1613
| sltu RD, r0, TMP0
1614
| mtc1 TMP1, f0
1615
| addiu RD, RD, 1
1616
| cvt.d.w f0, f0
1617
| lw PC, FRAME_PC(BASE)
1618
| sll RD, RD, 3 // RD = ((str->len != 0)+1)*8
1619
| b ->fff_res
1620
|. sdc1 f0, 0(RA)
1621
|
1622
|.ffunc string_char // Only handle the 1-arg case here.
1623
| ffgccheck
1624
| lw CARG3, HI(BASE)
1625
| ldc1 FARG1, 0(BASE)
1626
| li AT, 8
1627
| bne NARGS8:RC, AT, ->fff_fallback // Exactly 1 argument.
1628
|. sltiu AT, CARG3, LJ_TISNUM
1629
| beqz AT, ->fff_fallback
1630
|. li CARG3, 1
1631
| cvt.w.d FARG1, FARG1
1632
| addiu CARG2, sp, ARG5_OFS
1633
| sltiu AT, TMP0, 256
1634
| mfc1 TMP0, FARG1
1635
| beqz AT, ->fff_fallback
1636
|. sw TMP0, ARG5
1637
|->fff_newstr:
1638
| load_got lj_str_new
1639
| sw BASE, L->base
1640
| sw PC, SAVE_PC
1641
| call_intern lj_str_new // (lua_State *L, char *str, size_t l)
1642
|. move CARG1, L
1643
| // Returns GCstr *.
1644
| lw BASE, L->base
1645
| move CARG1, CRET1
1646
| b ->fff_restv
1647
|. li CARG3, LJ_TSTR
1648
|
1649
|.ffunc string_sub
1650
| ffgccheck
1651
| addiu AT, NARGS8:RC, -16
1652
| lw CARG3, 16+HI(BASE)
1653
| ldc1 f0, 16(BASE)
1654
| lw TMP0, HI(BASE)
1655
| lw STR:CARG1, LO(BASE)
1656
| bltz AT, ->fff_fallback
1657
| lw CARG2, 8+HI(BASE)
1658
| ldc1 f2, 8(BASE)
1659
| beqz AT, >1
1660
|. li CARG4, -1
1661
| cvt.w.d f0, f0
1662
| sltiu AT, CARG3, LJ_TISNUM
1663
| beqz AT, ->fff_fallback
1664
|. mfc1 CARG4, f0
1665
|1:
1666
| sltiu AT, CARG2, LJ_TISNUM
1667
| beqz AT, ->fff_fallback
1668
|. li AT, LJ_TSTR
1669
| cvt.w.d f2, f2
1670
| bne TMP0, AT, ->fff_fallback
1671
|. lw CARG2, STR:CARG1->len
1672
| mfc1 CARG3, f2
1673
| // STR:CARG1 = str, CARG2 = str->len, CARG3 = start, CARG4 = end
1674
| slt AT, CARG4, r0
1675
| addiu TMP0, CARG2, 1
1676
| addu TMP1, CARG4, TMP0
1677
| slt TMP3, CARG3, r0
1678
| movn CARG4, TMP1, AT // if (end < 0) end += len+1
1679
| addu TMP1, CARG3, TMP0
1680
| movn CARG3, TMP1, TMP3 // if (start < 0) start += len+1
1681
| li TMP2, 1
1682
| slt AT, CARG4, r0
1683
| slt TMP3, r0, CARG3
1684
| movn CARG4, r0, AT // if (end < 0) end = 0
1685
| movz CARG3, TMP2, TMP3 // if (start < 1) start = 1
1686
| slt AT, CARG2, CARG4
1687
| movn CARG4, CARG2, AT // if (end > len) end = len
1688
| addu CARG2, STR:CARG1, CARG3
1689
| subu CARG3, CARG4, CARG3 // len = end - start
1690
| addiu CARG2, CARG2, sizeof(GCstr)-1
1691
| bgez CARG3, ->fff_newstr
1692
|. addiu CARG3, CARG3, 1 // len++
1693
|->fff_emptystr: // Return empty string.
1694
| addiu STR:CARG1, DISPATCH, DISPATCH_GL(strempty)
1695
| b ->fff_restv
1696
|. li CARG3, LJ_TSTR
1697
|
1698
|.ffunc string_rep // Only handle the 1-char case inline.
1699
| ffgccheck
1700
| lw TMP0, HI(BASE)
1701
| addiu AT, NARGS8:RC, -16 // Exactly 2 arguments.
1702
| lw CARG4, 8+HI(BASE)
1703
| lw STR:CARG1, LO(BASE)
1704
| addiu TMP0, TMP0, -LJ_TSTR
1705
| ldc1 f0, 8(BASE)
1706
| or AT, AT, TMP0
1707
| bnez AT, ->fff_fallback
1708
|. sltiu AT, CARG4, LJ_TISNUM
1709
| cvt.w.d f0, f0
1710
| beqz AT, ->fff_fallback
1711
|. lw TMP0, STR:CARG1->len
1712
| mfc1 CARG3, f0
1713
| lw TMP1, DISPATCH_GL(tmpbuf.sz)(DISPATCH)
1714
| li AT, 1
1715
| blez CARG3, ->fff_emptystr // Count <= 0?
1716
|. sltu AT, AT, TMP0
1717
| beqz TMP0, ->fff_emptystr // Zero length string?
1718
|. sltu TMP0, TMP1, CARG3
1719
| or AT, AT, TMP0
1720
| lw CARG2, DISPATCH_GL(tmpbuf.buf)(DISPATCH)
1721
| bnez AT, ->fff_fallback // Fallback for > 1-char strings.
1722
|. lbu TMP0, STR:CARG1[1]
1723
| addu TMP2, CARG2, CARG3
1724
|1: // Fill buffer with char. Yes, this is suboptimal code (do you care?).
1725
| addiu TMP2, TMP2, -1
1726
| sltu AT, CARG2, TMP2
1727
| bnez AT, <1
1728
|. sb TMP0, 0(TMP2)
1729
| b ->fff_newstr
1730
|. nop
1731
|
1732
|.ffunc string_reverse
1733
| ffgccheck
1734
| lw CARG3, HI(BASE)
1735
| lw STR:CARG1, LO(BASE)
1736
| beqz NARGS8:RC, ->fff_fallback
1737
|. li AT, LJ_TSTR
1738
| bne CARG3, AT, ->fff_fallback
1739
|. lw TMP1, DISPATCH_GL(tmpbuf.sz)(DISPATCH)
1740
| lw CARG3, STR:CARG1->len
1741
| addiu CARG1, STR:CARG1, #STR
1742
| lw CARG2, DISPATCH_GL(tmpbuf.buf)(DISPATCH)
1743
| sltu AT, TMP1, CARG3
1744
| bnez AT, ->fff_fallback
1745
|. addu TMP3, CARG1, CARG3
1746
| addu CARG4, CARG2, CARG3
1747
|1: // Reverse string copy.
1748
| lbu TMP1, 0(CARG1)
1749
| sltu AT, CARG1, TMP3
1750
| beqz AT, ->fff_newstr
1751
|. addiu CARG1, CARG1, 1
1752
| addiu CARG4, CARG4, -1
1753
| b <1
1754
| sb TMP1, 0(CARG4)
1755
|
1756
|.macro ffstring_case, name, lo
1757
| .ffunc name
1758
| ffgccheck
1759
| lw CARG3, HI(BASE)
1760
| lw STR:CARG1, LO(BASE)
1761
| beqz NARGS8:RC, ->fff_fallback
1762
|. li AT, LJ_TSTR
1763
| bne CARG3, AT, ->fff_fallback
1764
|. lw TMP1, DISPATCH_GL(tmpbuf.sz)(DISPATCH)
1765
| lw CARG3, STR:CARG1->len
1766
| addiu CARG1, STR:CARG1, #STR
1767
| lw CARG2, DISPATCH_GL(tmpbuf.buf)(DISPATCH)
1768
| sltu AT, TMP1, CARG3
1769
| bnez AT, ->fff_fallback
1770
|. addu TMP3, CARG1, CARG3
1771
| move CARG4, CARG2
1772
|1: // ASCII case conversion.
1773
| lbu TMP1, 0(CARG1)
1774
| sltu AT, CARG1, TMP3
1775
| beqz AT, ->fff_newstr
1776
|. addiu TMP0, TMP1, -lo
1777
| xori TMP2, TMP1, 0x20
1778
| sltiu AT, TMP0, 26
1779
| movn TMP1, TMP2, AT
1780
| addiu CARG1, CARG1, 1
1781
| sb TMP1, 0(CARG4)
1782
| b <1
1783
|. addiu CARG4, CARG4, 1
1784
|.endmacro
1785
|
1786
|ffstring_case string_lower, 65
1787
|ffstring_case string_upper, 97
1788
|
1789
|//-- Table library ------------------------------------------------------
1790
|
1791
|.ffunc_1 table_getn
1792
| li AT, LJ_TTAB
1793
| bne CARG3, AT, ->fff_fallback
1794
|. load_got lj_tab_len
1795
| call_intern lj_tab_len // (GCtab *t)
1796
|. nop
1797
| // Returns uint32_t (but less than 2^31).
1798
| b ->fff_resi
1799
|. nop
1800
|
1801
|//-- Bit library --------------------------------------------------------
1802
|
1803
|.macro .ffunc_bit, name
1804
| .ffunc_n bit_..name
1805
|. add.d FARG1, FARG1, TOBIT
1806
| mfc1 CRET1, FARG1
1807
|.endmacro
1808
|
1809
|.macro .ffunc_bit_op, name, ins
1810
| .ffunc_bit name
1811
| addiu TMP1, BASE, 8
1812
| addu TMP2, BASE, NARGS8:RC
1813
|1:
1814
| lw CARG4, HI(TMP1)
1815
| beq TMP1, TMP2, ->fff_resi
1816
|. ldc1 FARG1, 0(TMP1)
1817
| sltiu AT, CARG4, LJ_TISNUM
1818
| beqz AT, ->fff_fallback
1819
| add.d FARG1, FARG1, TOBIT
1820
| mfc1 CARG2, FARG1
1821
| ins CRET1, CRET1, CARG2
1822
| b <1
1823
|. addiu TMP1, TMP1, 8
1824
|.endmacro
1825
|
1826
|.ffunc_bit_op band, and
1827
|.ffunc_bit_op bor, or
1828
|.ffunc_bit_op bxor, xor
1829
|
1830
|.ffunc_bit bswap
1831
| srl TMP0, CRET1, 24
1832
| srl TMP2, CRET1, 8
1833
| sll TMP1, CRET1, 24
1834
| andi TMP2, TMP2, 0xff00
1835
| or TMP0, TMP0, TMP1
1836
| andi CRET1, CRET1, 0xff00
1837
| or TMP0, TMP0, TMP2
1838
| sll CRET1, CRET1, 8
1839
| b ->fff_resi
1840
|. or CRET1, TMP0, CRET1
1841
|
1842
|.ffunc_bit bnot
1843
| b ->fff_resi
1844
|. not CRET1, CRET1
1845
|
1846
|.macro .ffunc_bit_sh, name, ins, shmod
1847
| .ffunc_nn bit_..name
1848
|. add.d FARG1, FARG1, TOBIT
1849
| add.d FARG2, FARG2, TOBIT
1850
| mfc1 CARG1, FARG1
1851
| mfc1 CARG2, FARG2
1852
|.if shmod == 1
1853
| li AT, 32
1854
| subu TMP0, AT, CARG2
1855
| sllv CARG2, CARG1, CARG2
1856
| srlv CARG1, CARG1, TMP0
1857
|.elif shmod == 2
1858
| li AT, 32
1859
| subu TMP0, AT, CARG2
1860
| srlv CARG2, CARG1, CARG2
1861
| sllv CARG1, CARG1, TMP0
1862
|.endif
1863
| b ->fff_resi
1864
|. ins CRET1, CARG1, CARG2
1865
|.endmacro
1866
|
1867
|.ffunc_bit_sh lshift, sllv, 0
1868
|.ffunc_bit_sh rshift, srlv, 0
1869
|.ffunc_bit_sh arshift, srav, 0
1870
|// Can't use rotrv, since it's only in MIPS32R2.
1871
|.ffunc_bit_sh rol, or, 1
1872
|.ffunc_bit_sh ror, or, 2
1873
|
1874
|.ffunc_bit tobit
1875
|->fff_resi:
1876
| mtc1 CRET1, FRET1
1877
| b ->fff_resn
1878
|. cvt.d.w FRET1, FRET1
1879
|
1880
|//-----------------------------------------------------------------------
1881
|
1882
|->fff_fallback: // Call fast function fallback handler.
1883
| // BASE = new base, RB = CFUNC, RC = nargs*8
1884
| lw TMP3, CFUNC:RB->f
1885
| addu TMP1, BASE, NARGS8:RC
1886
| lw PC, FRAME_PC(BASE) // Fallback may overwrite PC.
1887
| addiu TMP0, TMP1, 8*LUA_MINSTACK
1888
| lw TMP2, L->maxstack
1889
| sw PC, SAVE_PC // Redundant (but a defined value).
1890
| sltu AT, TMP2, TMP0
1891
| sw BASE, L->base
1892
| sw TMP1, L->top
1893
| bnez AT, >5 // Need to grow stack.
1894
|. move CFUNCADDR, TMP3
1895
| jalr TMP3 // (lua_State *L)
1896
|. move CARG1, L
1897
| // Either throws an error, or recovers and returns -1, 0 or nresults+1.
1898
| lw BASE, L->base
1899
| sll RD, CRET1, 3
1900
| bgtz CRET1, ->fff_res // Returned nresults+1?
1901
|. addiu RA, BASE, -8
1902
|1: // Returned 0 or -1: retry fast path.
1903
| lw TMP0, L->top
1904
| lw LFUNC:RB, FRAME_FUNC(BASE)
1905
| bnez CRET1, ->vm_call_tail // Returned -1?
1906
|. subu NARGS8:RC, TMP0, BASE
1907
| ins_callt // Returned 0: retry fast path.
1908
|
1909
|// Reconstruct previous base for vmeta_call during tailcall.
1910
|->vm_call_tail:
1911
| andi TMP0, PC, FRAME_TYPE
1912
| li AT, -4
1913
| bnez TMP0, >3
1914
|. and TMP1, PC, AT
1915
| lbu TMP1, OFS_RA(PC)
1916
| sll TMP1, TMP1, 3
1917
| addiu TMP1, TMP1, 8
1918
|3:
1919
| b ->vm_call_dispatch // Resolve again for tailcall.
1920
|. subu TMP2, BASE, TMP1
1921
|
1922
|5: // Grow stack for fallback handler.
1923
| load_got lj_state_growstack
1924
| li CARG2, LUA_MINSTACK
1925
| call_intern lj_state_growstack // (lua_State *L, int n)
1926
|. move CARG1, L
1927
| lw BASE, L->base
1928
| b <1
1929
|. li CRET1, 0 // Force retry.
1930
|
1931
|->fff_gcstep: // Call GC step function.
1932
| // BASE = new base, RC = nargs*8
1933
| move MULTRES, ra
1934
| load_got lj_gc_step
1935
| sw BASE, L->base
1936
| addu TMP0, BASE, NARGS8:RC
1937
| sw PC, SAVE_PC // Redundant (but a defined value).
1938
| sw TMP0, L->top
1939
| call_intern lj_gc_step // (lua_State *L)
1940
|. move CARG1, L
1941
| lw BASE, L->base
1942
| move ra, MULTRES
1943
| lw TMP0, L->top
1944
| lw CFUNC:RB, FRAME_FUNC(BASE)
1945
| jr ra
1946
|. subu NARGS8:RC, TMP0, BASE
1947
|
1948
|//-----------------------------------------------------------------------
1949
|//-- Special dispatch targets -------------------------------------------
1950
|//-----------------------------------------------------------------------
1951
|
1952
|->vm_record: // Dispatch target for recording phase.
1953
|.if JIT
1954
| lbu TMP3, DISPATCH_GL(hookmask)(DISPATCH)
1955
| andi AT, TMP3, HOOK_VMEVENT // No recording while in vmevent.
1956
| bnez AT, >5
1957
| // Decrement the hookcount for consistency, but always do the call.
1958
|. lw TMP2, DISPATCH_GL(hookcount)(DISPATCH)
1959
| andi AT, TMP3, HOOK_ACTIVE
1960
| bnez AT, >1
1961
|. addiu TMP2, TMP2, -1
1962
| andi AT, TMP3, LUA_MASKLINE|LUA_MASKCOUNT
1963
| beqz AT, >1
1964
|. nop
1965
| b >1
1966
|. sw TMP2, DISPATCH_GL(hookcount)(DISPATCH)
1967
|.endif
1968
|
1969
|->vm_rethook: // Dispatch target for return hooks.
1970
| lbu TMP3, DISPATCH_GL(hookmask)(DISPATCH)
1971
| andi AT, TMP3, HOOK_ACTIVE // Hook already active?
1972
| beqz AT, >1
1973
|5: // Re-dispatch to static ins.
1974
|. lw AT, GG_DISP2STATIC(TMP0) // Assumes TMP0 holds DISPATCH+OP*4.
1975
| jr AT
1976
|. nop
1977
|
1978
|->vm_inshook: // Dispatch target for instr/line hooks.
1979
| lbu TMP3, DISPATCH_GL(hookmask)(DISPATCH)
1980
| lw TMP2, DISPATCH_GL(hookcount)(DISPATCH)
1981
| andi AT, TMP3, HOOK_ACTIVE // Hook already active?
1982
| bnez AT, <5
1983
|. andi AT, TMP3, LUA_MASKLINE|LUA_MASKCOUNT
1984
| beqz AT, <5
1985
|. addiu TMP2, TMP2, -1
1986
| beqz TMP2, >1
1987
|. sw TMP2, DISPATCH_GL(hookcount)(DISPATCH)
1988
| andi AT, TMP3, LUA_MASKLINE
1989
| beqz AT, <5
1990
|1:
1991
|. load_got lj_dispatch_ins
1992
| sw MULTRES, SAVE_MULTRES
1993
| move CARG2, PC
1994
| sw BASE, L->base
1995
| // SAVE_PC must hold the _previous_ PC. The callee updates it with PC.
1996
| call_intern lj_dispatch_ins // (lua_State *L, const BCIns *pc)
1997
|. move CARG1, L
1998
|3:
1999
| lw BASE, L->base
2000
|4: // Re-dispatch to static ins.
2001
| lw INS, -4(PC)
2002
| decode_OP4a TMP1, INS
2003
| decode_OP4b TMP1
2004
| addu TMP0, DISPATCH, TMP1
2005
| decode_RD8a RD, INS
2006
| lw AT, GG_DISP2STATIC(TMP0)
2007
| decode_RA8a RA, INS
2008
| decode_RD8b RD
2009
| jr AT
2010
| decode_RA8b RA
2011
|
2012
|->cont_hook: // Continue from hook yield.
2013
| addiu PC, PC, 4
2014
| b <4
2015
|. lw MULTRES, -24+LO(RB) // Restore MULTRES for *M ins.
2016
|
2017
|->vm_hotloop: // Hot loop counter underflow.
2018
|.if JIT
2019
| lw LFUNC:TMP1, FRAME_FUNC(BASE)
2020
| addiu CARG1, DISPATCH, GG_DISP2J
2021
| sw PC, SAVE_PC
2022
| lw TMP1, LFUNC:TMP1->pc
2023
| move CARG2, PC
2024
| sw L, DISPATCH_J(L)(DISPATCH)
2025
| lbu TMP1, PC2PROTO(framesize)(TMP1)
2026
| load_got lj_trace_hot
2027
| sw BASE, L->base
2028
| sll TMP1, TMP1, 3
2029
| addu TMP1, BASE, TMP1
2030
| call_intern lj_trace_hot // (jit_State *J, const BCIns *pc)
2031
|. sw TMP1, L->top
2032
| b <3
2033
|. nop
2034
|.endif
2035
|
2036
|->vm_callhook: // Dispatch target for call hooks.
2037
|.if JIT
2038
| b >1
2039
|.endif
2040
|. move CARG2, PC
2041
|
2042
|->vm_hotcall: // Hot call counter underflow.
2043
|.if JIT
2044
| ori CARG2, PC, 1
2045
|1:
2046
|.endif
2047
| load_got lj_dispatch_call
2048
| addu TMP0, BASE, RC
2049
| sw PC, SAVE_PC
2050
| sw BASE, L->base
2051
| subu RA, RA, BASE
2052
| sw TMP0, L->top
2053
| call_intern lj_dispatch_call // (lua_State *L, const BCIns *pc)
2054
|. move CARG1, L
2055
| // Returns ASMFunction.
2056
| lw BASE, L->base
2057
| lw TMP0, L->top
2058
| sw r0, SAVE_PC // Invalidate for subsequent line hook.
2059
| subu NARGS8:RC, TMP0, BASE
2060
| addu RA, BASE, RA
2061
| lw LFUNC:RB, FRAME_FUNC(BASE)
2062
| jr CRET1
2063
|. lw INS, -4(PC)
2064
|
2065
|//-----------------------------------------------------------------------
2066
|//-- Trace exit handler -------------------------------------------------
2067
|//-----------------------------------------------------------------------
2068
|
2069
|.macro savex_, a, b
2070
| sdc1 f..a, 16+a*8(sp)
2071
| sw r..a, 16+32*8+a*4(sp)
2072
| sw r..b, 16+32*8+b*4(sp)
2073
|.endmacro
2074
|
2075
|->vm_exit_handler:
2076
|.if JIT
2077
| addiu sp, sp, -(16+32*8+32*4)
2078
| savex_ 0, 1
2079
| savex_ 2, 3
2080
| savex_ 4, 5
2081
| savex_ 6, 7
2082
| savex_ 8, 9
2083
| savex_ 10, 11
2084
| savex_ 12, 13
2085
| savex_ 14, 15
2086
| savex_ 16, 17
2087
| savex_ 18, 19
2088
| savex_ 20, 21
2089
| savex_ 22, 23
2090
| savex_ 24, 25
2091
| savex_ 26, 27
2092
| sdc1 f28, 16+28*8(sp)
2093
| sw r28, 16+32*8+28*4(sp)
2094
| sdc1 f30, 16+30*8(sp)
2095
| sw r30, 16+32*8+30*4(sp)
2096
| sw r0, 16+32*8+31*4(sp) // Clear RID_TMP.
2097
| li_vmstate EXIT
2098
| addiu TMP2, sp, 16+32*8+32*4 // Recompute original value of sp.
2099
| addiu DISPATCH, JGL, -GG_DISP2G-32768
2100
| lw TMP1, 0(TMP2) // Load exit number.
2101
| st_vmstate
2102
| sw TMP2, 16+32*8+29*4(sp) // Store sp in RID_SP.
2103
| lw L, DISPATCH_GL(jit_L)(DISPATCH)
2104
| lw BASE, DISPATCH_GL(jit_base)(DISPATCH)
2105
| load_got lj_trace_exit
2106
| sw L, DISPATCH_J(L)(DISPATCH)
2107
| sw ra, DISPATCH_J(parent)(DISPATCH) // Store trace number.
2108
| sw TMP1, DISPATCH_J(exitno)(DISPATCH) // Store exit number.
2109
| addiu CARG1, DISPATCH, GG_DISP2J
2110
| sw BASE, L->base
2111
| call_intern lj_trace_exit // (jit_State *J, ExitState *ex)
2112
|. addiu CARG2, sp, 16
2113
| // Returns MULTRES (unscaled) or negated error code.
2114
| lw TMP1, L->cframe
2115
| li AT, -4
2116
| lw BASE, L->base
2117
| and sp, TMP1, AT
2118
| lw PC, SAVE_PC // Get SAVE_PC.
2119
| b >1
2120
|. sw L, SAVE_L // Set SAVE_L (on-trace resume/yield).
2121
|.endif
2122
|->vm_exit_interp:
2123
|.if JIT
2124
| // CRET1 = MULTRES or negated error code, BASE, PC and JGL set.
2125
| lw L, SAVE_L
2126
| addiu DISPATCH, JGL, -GG_DISP2G-32768
2127
|1:
2128
| bltz CRET1, >3 // Check for error from exit.
2129
|. lw LFUNC:TMP1, FRAME_FUNC(BASE)
2130
| lui TMP3, 0x59c0 // TOBIT = 2^52 + 2^51 (float).
2131
| sll MULTRES, CRET1, 3
2132
| li TISNIL, LJ_TNIL
2133
| sw MULTRES, SAVE_MULTRES
2134
| mtc1 TMP3, TOBIT
2135
| lw TMP1, LFUNC:TMP1->pc
2136
| sw r0, DISPATCH_GL(jit_L)(DISPATCH)
2137
| lw KBASE, PC2PROTO(k)(TMP1)
2138
| cvt.d.s TOBIT, TOBIT
2139
| // Modified copy of ins_next which handles function header dispatch, too.
2140
| lw INS, 0(PC)
2141
| addiu PC, PC, 4
2142
| // Assumes TISNIL == ~LJ_VMST_INTERP == -1
2143
| sw TISNIL, DISPATCH_GL(vmstate)(DISPATCH)
2144
| decode_OP4a TMP1, INS
2145
| decode_OP4b TMP1
2146
| sltiu TMP2, TMP1, BC_FUNCF*4 // Function header?
2147
| addu TMP0, DISPATCH, TMP1
2148
| decode_RD8a RD, INS
2149
| lw AT, 0(TMP0)
2150
| decode_RA8a RA, INS
2151
| beqz TMP2, >2
2152
|. decode_RA8b RA
2153
| jr AT
2154
|. decode_RD8b RD
2155
|2:
2156
| addiu RC, MULTRES, -8
2157
| jr AT
2158
|. add RA, RA, BASE
2159
|
2160
|3: // Rethrow error from the right C frame.
2161
| load_got lj_err_throw
2162
| negu CARG2, CRET1
2163
| call_intern lj_err_throw // (lua_State *L, int errcode)
2164
|. move CARG1, L
2165
|.endif
2166
|
2167
|//-----------------------------------------------------------------------
2168
|//-- Math helper functions ----------------------------------------------
2169
|//-----------------------------------------------------------------------
2170
|
2171
|// Modifies AT, TMP0, FRET1, FRET2, f4. Keeps all others incl. FARG1.
2172
|.macro vm_round, func
2173
| lui TMP0, 0x4330 // Hiword of 2^52 (double).
2174
| mtc1 r0, f4
2175
| mtc1 TMP0, f5
2176
| abs.d FRET2, FARG1 // |x|
2177
| mfc1 AT, f13
2178
| c.olt.d 0, FRET2, f4
2179
| add.d FRET1, FRET2, f4 // (|x| + 2^52) - 2^52
2180
| bc1f 0, >1 // Truncate only if |x| < 2^52.
2181
|. sub.d FRET1, FRET1, f4
2182
| slt AT, AT, r0
2183
|.if "func" == "ceil"
2184
| lui TMP0, 0xbff0 // Hiword of -1 (double). Preserves -0.
2185
|.else
2186
| lui TMP0, 0x3ff0 // Hiword of +1 (double).
2187
|.endif
2188
|.if "func" == "trunc"
2189
| mtc1 TMP0, f5
2190
| c.olt.d 0, FRET2, FRET1 // |x| < result?
2191
| sub.d FRET2, FRET1, f4
2192
| movt.d FRET1, FRET2, 0 // If yes, subtract +1.
2193
| neg.d FRET2, FRET1
2194
| jr ra
2195
|. movn.d FRET1, FRET2, AT // Merge sign bit back in.
2196
|.else
2197
| neg.d FRET2, FRET1
2198
| mtc1 TMP0, f5
2199
| movn.d FRET1, FRET2, AT // Merge sign bit back in.
2200
|.if "func" == "ceil"
2201
| c.olt.d 0, FRET1, FARG1 // x > result?
2202
|.else
2203
| c.olt.d 0, FARG1, FRET1 // x < result?
2204
|.endif
2205
| sub.d FRET2, FRET1, f4 // If yes, subtract +-1.
2206
| jr ra
2207
|. movt.d FRET1, FRET2, 0
2208
|.endif
2209
|1:
2210
| jr ra
2211
|. mov.d FRET1, FARG1
2212
|.endmacro
2213
|
2214
|->vm_floor:
2215
| vm_round floor
2216
|->vm_ceil:
2217
| vm_round ceil
2218
|->vm_trunc:
2219
|.if JIT
2220
| vm_round trunc
2221
|.endif
2222
|
2223
|//-----------------------------------------------------------------------
2224
|//-- Miscellaneous functions --------------------------------------------
2225
|//-----------------------------------------------------------------------
2226
|
2227
|//-----------------------------------------------------------------------
2228
|//-- FFI helper functions -----------------------------------------------
2229
|//-----------------------------------------------------------------------
2230
|
2231
|// Handler for callback functions. Callback slot number in r1, g in r2.
2232
|->vm_ffi_callback:
2233
|.if FFI
2234
|.type CTSTATE, CTState, PC
2235
| saveregs
2236
| lw CTSTATE, GL:r2->ctype_state
2237
| addiu DISPATCH, r2, GG_G2DISP
2238
| load_got lj_ccallback_enter
2239
| sw r1, CTSTATE->cb.slot
2240
| sw CARG1, CTSTATE->cb.gpr[0]
2241
| sw CARG2, CTSTATE->cb.gpr[1]
2242
| sdc1 FARG1, CTSTATE->cb.fpr[0]
2243
| sw CARG3, CTSTATE->cb.gpr[2]
2244
| sw CARG4, CTSTATE->cb.gpr[3]
2245
| sdc1 FARG2, CTSTATE->cb.fpr[1]
2246
| addiu TMP0, sp, CFRAME_SPACE+16
2247
| sw TMP0, CTSTATE->cb.stack
2248
| sw r0, SAVE_PC // Any value outside of bytecode is ok.
2249
| move CARG2, sp
2250
| call_intern lj_ccallback_enter // (CTState *cts, void *cf)
2251
|. move CARG1, CTSTATE
2252
| // Returns lua_State *.
2253
| lw BASE, L:CRET1->base
2254
| lw RC, L:CRET1->top
2255
| move L, CRET1
2256
| lui TMP3, 0x59c0 // TOBIT = 2^52 + 2^51 (float).
2257
| lw LFUNC:RB, FRAME_FUNC(BASE)
2258
| mtc1 TMP3, TOBIT
2259
| li_vmstate INTERP
2260
| li TISNIL, LJ_TNIL
2261
| subu RC, RC, BASE
2262
| st_vmstate
2263
| cvt.d.s TOBIT, TOBIT
2264
| ins_callt
2265
|.endif
2266
|
2267
|->cont_ffi_callback: // Return from FFI callback.
2268
|.if FFI
2269
| load_got lj_ccallback_leave
2270
| lw CTSTATE, DISPATCH_GL(ctype_state)(DISPATCH)
2271
| sw BASE, L->base
2272
| sw RB, L->top
2273
| sw L, CTSTATE->L
2274
| move CARG2, RA
2275
| call_intern lj_ccallback_leave // (CTState *cts, TValue *o)
2276
|. move CARG1, CTSTATE
2277
| lw CRET1, CTSTATE->cb.gpr[0]
2278
| ldc1 FRET1, CTSTATE->cb.fpr[0]
2279
| lw CRET2, CTSTATE->cb.gpr[1]
2280
| b ->vm_leave_unw
2281
|. ldc1 FRET2, CTSTATE->cb.fpr[1]
2282
|.endif
2283
|
2284
|->vm_ffi_call: // Call C function via FFI.
2285
| // Caveat: needs special frame unwinding, see below.
2286
|.if FFI
2287
| .type CCSTATE, CCallState, CARG1
2288
| lw TMP1, CCSTATE->spadj
2289
| lbu CARG2, CCSTATE->nsp
2290
| move TMP2, sp
2291
| subu sp, sp, TMP1
2292
| sw ra, -4(TMP2)
2293
| sll CARG2, CARG2, 2
2294
| sw r16, -8(TMP2)
2295
| sw CCSTATE, -12(TMP2)
2296
| move r16, TMP2
2297
| addiu TMP1, CCSTATE, offsetof(CCallState, stack)
2298
| addiu TMP2, sp, 16
2299
| beqz CARG2, >2
2300
|. addu TMP3, TMP1, CARG2
2301
|1:
2302
| lw TMP0, 0(TMP1)
2303
| addiu TMP1, TMP1, 4
2304
| sltu AT, TMP1, TMP3
2305
| sw TMP0, 0(TMP2)
2306
| bnez AT, <1
2307
|. addiu TMP2, TMP2, 4
2308
|2:
2309
| lw CFUNCADDR, CCSTATE->func
2310
| lw CARG2, CCSTATE->gpr[1]
2311
| lw CARG3, CCSTATE->gpr[2]
2312
| lw CARG4, CCSTATE->gpr[3]
2313
| ldc1 FARG1, CCSTATE->fpr[0]
2314
| ldc1 FARG2, CCSTATE->fpr[1]
2315
| jalr CFUNCADDR
2316
|. lw CARG1, CCSTATE->gpr[0] // Do this last, since CCSTATE is CARG1.
2317
| lw CCSTATE:TMP1, -12(r16)
2318
| lw TMP2, -8(r16)
2319
| lw ra, -4(r16)
2320
| sw CRET1, CCSTATE:TMP1->gpr[0]
2321
| sw CRET2, CCSTATE:TMP1->gpr[1]
2322
| sdc1 FRET1, CCSTATE:TMP1->fpr[0]
2323
| sdc1 FRET2, CCSTATE:TMP1->fpr[1]
2324
| move sp, r16
2325
| jr ra
2326
|. move r16, TMP2
2327
|.endif
2328
|// Note: vm_ffi_call must be the last function in this object file!
2329
|
2330
|//-----------------------------------------------------------------------
2331
}
2332
2333
/* Generate the code for a single instruction. */
2334
static void build_ins(BuildCtx *ctx, BCOp op, int defop)
2335
{
2336
int vk = 0;
2337
|=>defop:
2338
2339
switch (op) {
2340
2341
/* -- Comparison ops ---------------------------------------------------- */
2342
2343
/* Remember: all ops branch for a true comparison, fall through otherwise. */
2344
2345
case BC_ISLT: case BC_ISGE: case BC_ISLE: case BC_ISGT:
2346
| // RA = src1*8, RD = src2*8, JMP with RD = target
2347
| addu CARG2, BASE, RA
2348
| addu CARG3, BASE, RD
2349
| lw TMP0, HI(CARG2)
2350
| lw TMP1, HI(CARG3)
2351
| ldc1 f0, 0(CARG2)
2352
| ldc1 f2, 0(CARG3)
2353
| sltiu TMP0, TMP0, LJ_TISNUM
2354
| sltiu TMP1, TMP1, LJ_TISNUM
2355
| lhu TMP2, OFS_RD(PC)
2356
| and TMP0, TMP0, TMP1
2357
| addiu PC, PC, 4
2358
| beqz TMP0, ->vmeta_comp
2359
|. lui TMP1, (-(BCBIAS_J*4 >> 16) & 65535)
2360
| decode_RD4b TMP2
2361
| addu TMP2, TMP2, TMP1
2362
if (op == BC_ISLT || op == BC_ISGE) {
2363
| c.olt.d f0, f2
2364
} else {
2365
| c.ole.d f0, f2
2366
}
2367
if (op == BC_ISLT || op == BC_ISLE) {
2368
| movf TMP2, r0
2369
} else {
2370
| movt TMP2, r0
2371
}
2372
| addu PC, PC, TMP2
2373
|1:
2374
| ins_next
2375
break;
2376
2377
case BC_ISEQV: case BC_ISNEV:
2378
vk = op == BC_ISEQV;
2379
| // RA = src1*8, RD = src2*8, JMP with RD = target
2380
| addu RA, BASE, RA
2381
| addiu PC, PC, 4
2382
| lw TMP0, HI(RA)
2383
| ldc1 f0, 0(RA)
2384
| addu RD, BASE, RD
2385
| lhu TMP2, -4+OFS_RD(PC)
2386
| lw TMP1, HI(RD)
2387
| ldc1 f2, 0(RD)
2388
| lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535)
2389
| sltiu AT, TMP0, LJ_TISNUM
2390
| sltiu CARG1, TMP1, LJ_TISNUM
2391
| decode_RD4b TMP2
2392
| and AT, AT, CARG1
2393
| beqz AT, >5
2394
|. addu TMP2, TMP2, TMP3
2395
| c.eq.d f0, f2
2396
if (vk) {
2397
| movf TMP2, r0
2398
} else {
2399
| movt TMP2, r0
2400
}
2401
|1:
2402
| addu PC, PC, TMP2
2403
| ins_next
2404
|5: // Either or both types are not numbers.
2405
| lw CARG2, LO(RA)
2406
| lw CARG3, LO(RD)
2407
|.if FFI
2408
| li TMP3, LJ_TCDATA
2409
| beq TMP0, TMP3, ->vmeta_equal_cd
2410
|.endif
2411
|. sltiu AT, TMP0, LJ_TISPRI // Not a primitive?
2412
|.if FFI
2413
| beq TMP1, TMP3, ->vmeta_equal_cd
2414
|.endif
2415
|. xor TMP3, CARG2, CARG3 // Same tv?
2416
| xor TMP1, TMP1, TMP0 // Same type?
2417
| sltiu CARG1, TMP0, LJ_TISTABUD+1 // Table or userdata?
2418
| movz TMP3, r0, AT // Ignore tv if primitive.
2419
| movn CARG1, r0, TMP1 // Tab/ud and same type?
2420
| or AT, TMP1, TMP3 // Same type && (pri||same tv).
2421
| movz CARG1, r0, AT
2422
| beqz CARG1, <1 // Done if not tab/ud or not same type or same tv.
2423
if (vk) {
2424
|. movn TMP2, r0, AT
2425
} else {
2426
|. movz TMP2, r0, AT
2427
}
2428
| // Different tables or userdatas. Need to check __eq metamethod.
2429
| // Field metatable must be at same offset for GCtab and GCudata!
2430
| lw TAB:TMP1, TAB:CARG2->metatable
2431
| beqz TAB:TMP1, <1 // No metatable?
2432
|. nop
2433
| lbu TMP1, TAB:TMP1->nomm
2434
| andi TMP1, TMP1, 1<<MM_eq
2435
| bnez TMP1, <1 // Or 'no __eq' flag set?
2436
|. nop
2437
| b ->vmeta_equal // Handle __eq metamethod.
2438
|. li CARG4, 1-vk // ne = 0 or 1.
2439
break;
2440
2441
case BC_ISEQS: case BC_ISNES:
2442
vk = op == BC_ISEQS;
2443
| // RA = src*8, RD = str_const*8 (~), JMP with RD = target
2444
| addu RA, BASE, RA
2445
| addiu PC, PC, 4
2446
| lw TMP0, HI(RA)
2447
| srl RD, RD, 1
2448
| lw STR:TMP3, LO(RA)
2449
| subu RD, KBASE, RD
2450
| lhu TMP2, -4+OFS_RD(PC)
2451
|.if FFI
2452
| li AT, LJ_TCDATA
2453
| beq TMP0, AT, ->vmeta_equal_cd
2454
|.endif
2455
|. lw STR:TMP1, -4(RD) // KBASE-4-str_const*4
2456
| addiu TMP0, TMP0, -LJ_TSTR
2457
| decode_RD4b TMP2
2458
| xor TMP1, STR:TMP1, STR:TMP3
2459
| or TMP0, TMP0, TMP1
2460
| lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535)
2461
| addu TMP2, TMP2, TMP3
2462
if (vk) {
2463
| movn TMP2, r0, TMP0
2464
} else {
2465
| movz TMP2, r0, TMP0
2466
}
2467
| addu PC, PC, TMP2
2468
| ins_next
2469
break;
2470
2471
case BC_ISEQN: case BC_ISNEN:
2472
vk = op == BC_ISEQN;
2473
| // RA = src*8, RD = num_const*8, JMP with RD = target
2474
| addu RA, BASE, RA
2475
| addiu PC, PC, 4
2476
| lw TMP0, HI(RA)
2477
| ldc1 f0, 0(RA)
2478
| addu RD, KBASE, RD
2479
| lhu TMP2, -4+OFS_RD(PC)
2480
| ldc1 f2, 0(RD)
2481
| lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535)
2482
| sltiu AT, TMP0, LJ_TISNUM
2483
| decode_RD4b TMP2
2484
|.if FFI
2485
| beqz AT, >5
2486
|.else
2487
| beqz AT, >1
2488
|.endif
2489
|. addu TMP2, TMP2, TMP3
2490
| c.eq.d f0, f2
2491
if (vk) {
2492
| movf TMP2, r0
2493
| addu PC, PC, TMP2
2494
|1:
2495
} else {
2496
| movt TMP2, r0
2497
|1:
2498
| addu PC, PC, TMP2
2499
}
2500
| ins_next
2501
|.if FFI
2502
|5:
2503
| li AT, LJ_TCDATA
2504
| beq TMP0, AT, ->vmeta_equal_cd
2505
|. nop
2506
| b <1
2507
|. nop
2508
|.endif
2509
break;
2510
2511
case BC_ISEQP: case BC_ISNEP:
2512
vk = op == BC_ISEQP;
2513
| // RA = src*8, RD = primitive_type*8 (~), JMP with RD = target
2514
| addu RA, BASE, RA
2515
| srl TMP1, RD, 3
2516
| lw TMP0, HI(RA)
2517
| lhu TMP2, OFS_RD(PC)
2518
| not TMP1, TMP1
2519
| addiu PC, PC, 4
2520
|.if FFI
2521
| li AT, LJ_TCDATA
2522
| beq TMP0, AT, ->vmeta_equal_cd
2523
|.endif
2524
|. xor TMP0, TMP0, TMP1
2525
| decode_RD4b TMP2
2526
| lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535)
2527
| addu TMP2, TMP2, TMP3
2528
if (vk) {
2529
| movn TMP2, r0, TMP0
2530
} else {
2531
| movz TMP2, r0, TMP0
2532
}
2533
| addu PC, PC, TMP2
2534
| ins_next
2535
break;
2536
2537
/* -- Unary test and copy ops ------------------------------------------- */
2538
2539
case BC_ISTC: case BC_ISFC: case BC_IST: case BC_ISF:
2540
| // RA = dst*8 or unused, RD = src*8, JMP with RD = target
2541
| addu RD, BASE, RD
2542
| lhu TMP2, OFS_RD(PC)
2543
| lw TMP0, HI(RD)
2544
| addiu PC, PC, 4
2545
if (op == BC_IST || op == BC_ISF) {
2546
| sltiu TMP0, TMP0, LJ_TISTRUECOND
2547
| decode_RD4b TMP2
2548
| lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535)
2549
| addu TMP2, TMP2, TMP3
2550
if (op == BC_IST) {
2551
| movz TMP2, r0, TMP0
2552
} else {
2553
| movn TMP2, r0, TMP0
2554
}
2555
| addu PC, PC, TMP2
2556
} else {
2557
| sltiu TMP0, TMP0, LJ_TISTRUECOND
2558
| ldc1 f0, 0(RD)
2559
if (op == BC_ISTC) {
2560
| beqz TMP0, >1
2561
} else {
2562
| bnez TMP0, >1
2563
}
2564
|. addu RA, BASE, RA
2565
| decode_RD4b TMP2
2566
| lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535)
2567
| addu TMP2, TMP2, TMP3
2568
| sdc1 f0, 0(RA)
2569
| addu PC, PC, TMP2
2570
|1:
2571
}
2572
| ins_next
2573
break;
2574
2575
/* -- Unary ops --------------------------------------------------------- */
2576
2577
case BC_MOV:
2578
| // RA = dst*8, RD = src*8
2579
| addu RD, BASE, RD
2580
| addu RA, BASE, RA
2581
| ldc1 f0, 0(RD)
2582
| ins_next1
2583
| sdc1 f0, 0(RA)
2584
| ins_next2
2585
break;
2586
case BC_NOT:
2587
| // RA = dst*8, RD = src*8
2588
| addu RD, BASE, RD
2589
| addu RA, BASE, RA
2590
| lw TMP0, HI(RD)
2591
| li TMP1, LJ_TFALSE
2592
| sltiu TMP0, TMP0, LJ_TISTRUECOND
2593
| addiu TMP1, TMP0, LJ_TTRUE
2594
| ins_next1
2595
| sw TMP1, HI(RA)
2596
| ins_next2
2597
break;
2598
case BC_UNM:
2599
| // RA = dst*8, RD = src*8
2600
| addu CARG3, BASE, RD
2601
| addu RA, BASE, RA
2602
| lw TMP0, HI(CARG3)
2603
| ldc1 f0, 0(CARG3)
2604
| sltiu AT, TMP0, LJ_TISNUM
2605
| beqz AT, ->vmeta_unm
2606
|. neg.d f0, f0
2607
| ins_next1
2608
| sdc1 f0, 0(RA)
2609
| ins_next2
2610
break;
2611
case BC_LEN:
2612
| // RA = dst*8, RD = src*8
2613
| addu CARG2, BASE, RD
2614
| addu RA, BASE, RA
2615
| lw TMP0, HI(CARG2)
2616
| lw CARG1, LO(CARG2)
2617
| li AT, LJ_TSTR
2618
| bne TMP0, AT, >2
2619
|. li AT, LJ_TTAB
2620
| lw CRET1, STR:CARG1->len
2621
|1:
2622
| mtc1 CRET1, f0
2623
| cvt.d.w f0, f0
2624
| ins_next1
2625
| sdc1 f0, 0(RA)
2626
| ins_next2
2627
|2:
2628
| bne TMP0, AT, ->vmeta_len
2629
|. nop
2630
#if LJ_52
2631
| lw TAB:TMP2, TAB:CARG1->metatable
2632
| bnez TAB:TMP2, >9
2633
|. nop
2634
|3:
2635
#endif
2636
|->BC_LEN_Z:
2637
| load_got lj_tab_len
2638
| call_intern lj_tab_len // (GCtab *t)
2639
|. nop
2640
| // Returns uint32_t (but less than 2^31).
2641
| b <1
2642
|. nop
2643
#if LJ_52
2644
|9:
2645
| lbu TMP0, TAB:TMP2->nomm
2646
| andi TMP0, TMP0, 1<<MM_len
2647
| bnez TMP0, <3 // 'no __len' flag set: done.
2648
|. nop
2649
| b ->vmeta_len
2650
|. nop
2651
#endif
2652
break;
2653
2654
/* -- Binary ops -------------------------------------------------------- */
2655
2656
|.macro ins_arithpre
2657
||vk = ((int)op - BC_ADDVN) / (BC_ADDNV-BC_ADDVN);
2658
| decode_RB8a RB, INS
2659
| decode_RB8b RB
2660
| decode_RDtoRC8 RC, RD
2661
| // RA = dst*8, RB = src1*8, RC = src2*8 | num_const*8
2662
||switch (vk) {
2663
||case 0:
2664
| addu CARG3, BASE, RB
2665
| addu CARG4, KBASE, RC
2666
| lw TMP1, HI(CARG3)
2667
| ldc1 f20, 0(CARG3)
2668
| ldc1 f22, 0(CARG4)
2669
| sltiu AT, TMP1, LJ_TISNUM
2670
|| break;
2671
||case 1:
2672
| addu CARG4, BASE, RB
2673
| addu CARG3, KBASE, RC
2674
| lw TMP1, HI(CARG4)
2675
| ldc1 f22, 0(CARG4)
2676
| ldc1 f20, 0(CARG3)
2677
| sltiu AT, TMP1, LJ_TISNUM
2678
|| break;
2679
||default:
2680
| addu CARG3, BASE, RB
2681
| addu CARG4, BASE, RC
2682
| lw TMP1, HI(CARG3)
2683
| lw TMP2, HI(CARG4)
2684
| ldc1 f20, 0(CARG3)
2685
| ldc1 f22, 0(CARG4)
2686
| sltiu AT, TMP1, LJ_TISNUM
2687
| sltiu TMP0, TMP2, LJ_TISNUM
2688
| and AT, AT, TMP0
2689
|| break;
2690
||}
2691
| beqz AT, ->vmeta_arith
2692
|. addu RA, BASE, RA
2693
|.endmacro
2694
|
2695
|.macro fpmod, a, b, c
2696
|->BC_MODVN_Z:
2697
| bal ->vm_floor // floor(b/c)
2698
|. div.d FARG1, b, c
2699
| mul.d a, FRET1, c
2700
| sub.d a, b, a // b - floor(b/c)*c
2701
|.endmacro
2702
|
2703
|.macro ins_arith, ins
2704
| ins_arithpre
2705
|.if "ins" == "fpmod_"
2706
| b ->BC_MODVN_Z // Avoid 3 copies. It's slow anyway.
2707
|. nop
2708
|.else
2709
| ins f0, f20, f22
2710
| ins_next1
2711
| sdc1 f0, 0(RA)
2712
| ins_next2
2713
|.endif
2714
|.endmacro
2715
2716
case BC_ADDVN: case BC_ADDNV: case BC_ADDVV:
2717
| ins_arith add.d
2718
break;
2719
case BC_SUBVN: case BC_SUBNV: case BC_SUBVV:
2720
| ins_arith sub.d
2721
break;
2722
case BC_MULVN: case BC_MULNV: case BC_MULVV:
2723
| ins_arith mul.d
2724
break;
2725
case BC_DIVVN: case BC_DIVNV: case BC_DIVVV:
2726
| ins_arith div.d
2727
break;
2728
case BC_MODVN:
2729
| ins_arith fpmod
2730
break;
2731
case BC_MODNV: case BC_MODVV:
2732
| ins_arith fpmod_
2733
break;
2734
case BC_POW:
2735
| decode_RB8a RB, INS
2736
| decode_RB8b RB
2737
| decode_RDtoRC8 RC, RD
2738
| addu CARG3, BASE, RB
2739
| addu CARG4, BASE, RC
2740
| lw TMP1, HI(CARG3)
2741
| lw TMP2, HI(CARG4)
2742
| ldc1 FARG1, 0(CARG3)
2743
| ldc1 FARG2, 0(CARG4)
2744
| sltiu AT, TMP1, LJ_TISNUM
2745
| sltiu TMP0, TMP2, LJ_TISNUM
2746
| and AT, AT, TMP0
2747
| load_got pow
2748
| beqz AT, ->vmeta_arith
2749
|. addu RA, BASE, RA
2750
| call_extern
2751
|. nop
2752
| ins_next1
2753
| sdc1 FRET1, 0(RA)
2754
| ins_next2
2755
break;
2756
2757
case BC_CAT:
2758
| // RA = dst*8, RB = src_start*8, RC = src_end*8
2759
| decode_RB8a RB, INS
2760
| decode_RB8b RB
2761
| decode_RDtoRC8 RC, RD
2762
| subu CARG3, RC, RB
2763
| sw BASE, L->base
2764
| addu CARG2, BASE, RC
2765
| move MULTRES, RB
2766
|->BC_CAT_Z:
2767
| load_got lj_meta_cat
2768
| srl CARG3, CARG3, 3
2769
| sw PC, SAVE_PC
2770
| call_intern lj_meta_cat // (lua_State *L, TValue *top, int left)
2771
|. move CARG1, L
2772
| // Returns NULL (finished) or TValue * (metamethod).
2773
| bnez CRET1, ->vmeta_binop
2774
|. lw BASE, L->base
2775
| addu RB, BASE, MULTRES
2776
| ldc1 f0, 0(RB)
2777
| addu RA, BASE, RA
2778
| ins_next1
2779
| sdc1 f0, 0(RA) // Copy result from RB to RA.
2780
| ins_next2
2781
break;
2782
2783
/* -- Constant ops ------------------------------------------------------ */
2784
2785
case BC_KSTR:
2786
| // RA = dst*8, RD = str_const*8 (~)
2787
| srl TMP1, RD, 1
2788
| subu TMP1, KBASE, TMP1
2789
| ins_next1
2790
| lw TMP0, -4(TMP1) // KBASE-4-str_const*4
2791
| addu RA, BASE, RA
2792
| li TMP2, LJ_TSTR
2793
| sw TMP0, LO(RA)
2794
| sw TMP2, HI(RA)
2795
| ins_next2
2796
break;
2797
case BC_KCDATA:
2798
|.if FFI
2799
| // RA = dst*8, RD = cdata_const*8 (~)
2800
| srl TMP1, RD, 1
2801
| subu TMP1, KBASE, TMP1
2802
| ins_next1
2803
| lw TMP0, -4(TMP1) // KBASE-4-cdata_const*4
2804
| addu RA, BASE, RA
2805
| li TMP2, LJ_TCDATA
2806
| sw TMP0, LO(RA)
2807
| sw TMP2, HI(RA)
2808
| ins_next2
2809
|.endif
2810
break;
2811
case BC_KSHORT:
2812
| // RA = dst*8, RD = int16_literal*8
2813
| sra RD, INS, 16
2814
| mtc1 RD, f0
2815
| addu RA, BASE, RA
2816
| cvt.d.w f0, f0
2817
| ins_next1
2818
| sdc1 f0, 0(RA)
2819
| ins_next2
2820
break;
2821
case BC_KNUM:
2822
| // RA = dst*8, RD = num_const*8
2823
| addu RD, KBASE, RD
2824
| addu RA, BASE, RA
2825
| ldc1 f0, 0(RD)
2826
| ins_next1
2827
| sdc1 f0, 0(RA)
2828
| ins_next2
2829
break;
2830
case BC_KPRI:
2831
| // RA = dst*8, RD = primitive_type*8 (~)
2832
| srl TMP1, RD, 3
2833
| addu RA, BASE, RA
2834
| not TMP0, TMP1
2835
| ins_next1
2836
| sw TMP0, HI(RA)
2837
| ins_next2
2838
break;
2839
case BC_KNIL:
2840
| // RA = base*8, RD = end*8
2841
| addu RA, BASE, RA
2842
| sw TISNIL, HI(RA)
2843
| addiu RA, RA, 8
2844
| addu RD, BASE, RD
2845
|1:
2846
| sw TISNIL, HI(RA)
2847
| slt AT, RA, RD
2848
| bnez AT, <1
2849
|. addiu RA, RA, 8
2850
| ins_next_
2851
break;
2852
2853
/* -- Upvalue and function ops ------------------------------------------ */
2854
2855
case BC_UGET:
2856
| // RA = dst*8, RD = uvnum*8
2857
| lw LFUNC:RB, FRAME_FUNC(BASE)
2858
| srl RD, RD, 1
2859
| addu RD, RD, LFUNC:RB
2860
| lw UPVAL:RB, LFUNC:RD->uvptr
2861
| ins_next1
2862
| lw TMP1, UPVAL:RB->v
2863
| ldc1 f0, 0(TMP1)
2864
| addu RA, BASE, RA
2865
| sdc1 f0, 0(RA)
2866
| ins_next2
2867
break;
2868
case BC_USETV:
2869
| // RA = uvnum*8, RD = src*8
2870
| lw LFUNC:RB, FRAME_FUNC(BASE)
2871
| srl RA, RA, 1
2872
| addu RD, BASE, RD
2873
| addu RA, RA, LFUNC:RB
2874
| ldc1 f0, 0(RD)
2875
| lw UPVAL:RB, LFUNC:RA->uvptr
2876
| lbu TMP3, UPVAL:RB->marked
2877
| lw CARG2, UPVAL:RB->v
2878
| andi TMP3, TMP3, LJ_GC_BLACK // isblack(uv)
2879
| lbu TMP0, UPVAL:RB->closed
2880
| lw TMP2, HI(RD)
2881
| sdc1 f0, 0(CARG2)
2882
| li AT, LJ_GC_BLACK|1
2883
| or TMP3, TMP3, TMP0
2884
| beq TMP3, AT, >2 // Upvalue is closed and black?
2885
|. addiu TMP2, TMP2, -(LJ_TISNUM+1)
2886
|1:
2887
| ins_next
2888
|
2889
|2: // Check if new value is collectable.
2890
| sltiu AT, TMP2, LJ_TISGCV - (LJ_TISNUM+1)
2891
| beqz AT, <1 // tvisgcv(v)
2892
|. lw TMP1, LO(RD)
2893
| lbu TMP3, GCOBJ:TMP1->gch.marked
2894
| andi TMP3, TMP3, LJ_GC_WHITES // iswhite(v)
2895
| beqz TMP3, <1
2896
|. load_got lj_gc_barrieruv
2897
| // Crossed a write barrier. Move the barrier forward.
2898
| call_intern lj_gc_barrieruv // (global_State *g, TValue *tv)
2899
|. addiu CARG1, DISPATCH, GG_DISP2G
2900
| b <1
2901
|. nop
2902
break;
2903
case BC_USETS:
2904
| // RA = uvnum*8, RD = str_const*8 (~)
2905
| lw LFUNC:RB, FRAME_FUNC(BASE)
2906
| srl RA, RA, 1
2907
| srl TMP1, RD, 1
2908
| addu RA, RA, LFUNC:RB
2909
| subu TMP1, KBASE, TMP1
2910
| lw UPVAL:RB, LFUNC:RA->uvptr
2911
| lw STR:TMP1, -4(TMP1) // KBASE-4-str_const*4
2912
| lbu TMP2, UPVAL:RB->marked
2913
| lw CARG2, UPVAL:RB->v
2914
| lbu TMP3, STR:TMP1->marked
2915
| andi AT, TMP2, LJ_GC_BLACK // isblack(uv)
2916
| lbu TMP2, UPVAL:RB->closed
2917
| li TMP0, LJ_TSTR
2918
| sw STR:TMP1, LO(CARG2)
2919
| bnez AT, >2
2920
|. sw TMP0, HI(CARG2)
2921
|1:
2922
| ins_next
2923
|
2924
|2: // Check if string is white and ensure upvalue is closed.
2925
| beqz TMP2, <1
2926
|. andi AT, TMP3, LJ_GC_WHITES // iswhite(str)
2927
| beqz AT, <1
2928
|. load_got lj_gc_barrieruv
2929
| // Crossed a write barrier. Move the barrier forward.
2930
| call_intern lj_gc_barrieruv // (global_State *g, TValue *tv)
2931
|. addiu CARG1, DISPATCH, GG_DISP2G
2932
| b <1
2933
|. nop
2934
break;
2935
case BC_USETN:
2936
| // RA = uvnum*8, RD = num_const*8
2937
| lw LFUNC:RB, FRAME_FUNC(BASE)
2938
| srl RA, RA, 1
2939
| addu RD, KBASE, RD
2940
| addu RA, RA, LFUNC:RB
2941
| ldc1 f0, 0(RD)
2942
| lw UPVAL:RB, LFUNC:RA->uvptr
2943
| ins_next1
2944
| lw TMP1, UPVAL:RB->v
2945
| sdc1 f0, 0(TMP1)
2946
| ins_next2
2947
break;
2948
case BC_USETP:
2949
| // RA = uvnum*8, RD = primitive_type*8 (~)
2950
| lw LFUNC:RB, FRAME_FUNC(BASE)
2951
| srl RA, RA, 1
2952
| srl TMP0, RD, 3
2953
| addu RA, RA, LFUNC:RB
2954
| not TMP0, TMP0
2955
| lw UPVAL:RB, LFUNC:RA->uvptr
2956
| ins_next1
2957
| lw TMP1, UPVAL:RB->v
2958
| sw TMP0, HI(TMP1)
2959
| ins_next2
2960
break;
2961
2962
case BC_UCLO:
2963
| // RA = level*8, RD = target
2964
| lw TMP2, L->openupval
2965
| branch_RD // Do this first since RD is not saved.
2966
| load_got lj_func_closeuv
2967
| sw BASE, L->base
2968
| beqz TMP2, >1
2969
|. move CARG1, L
2970
| call_intern lj_func_closeuv // (lua_State *L, TValue *level)
2971
|. addu CARG2, BASE, RA
2972
| lw BASE, L->base
2973
|1:
2974
| ins_next
2975
break;
2976
2977
case BC_FNEW:
2978
| // RA = dst*8, RD = proto_const*8 (~) (holding function prototype)
2979
| srl TMP1, RD, 1
2980
| load_got lj_func_newL_gc
2981
| subu TMP1, KBASE, TMP1
2982
| lw CARG3, FRAME_FUNC(BASE)
2983
| lw CARG2, -4(TMP1) // KBASE-4-tab_const*4
2984
| sw BASE, L->base
2985
| sw PC, SAVE_PC
2986
| // (lua_State *L, GCproto *pt, GCfuncL *parent)
2987
| call_intern lj_func_newL_gc
2988
|. move CARG1, L
2989
| // Returns GCfuncL *.
2990
| lw BASE, L->base
2991
| li TMP0, LJ_TFUNC
2992
| ins_next1
2993
| addu RA, BASE, RA
2994
| sw TMP0, HI(RA)
2995
| sw LFUNC:CRET1, LO(RA)
2996
| ins_next2
2997
break;
2998
2999
/* -- Table ops --------------------------------------------------------- */
3000
3001
case BC_TNEW:
3002
case BC_TDUP:
3003
| // RA = dst*8, RD = (hbits|asize)*8 | tab_const*8 (~)
3004
| lw TMP0, DISPATCH_GL(gc.total)(DISPATCH)
3005
| lw TMP1, DISPATCH_GL(gc.threshold)(DISPATCH)
3006
| sw BASE, L->base
3007
| sw PC, SAVE_PC
3008
| sltu AT, TMP0, TMP1
3009
| beqz AT, >5
3010
|1:
3011
if (op == BC_TNEW) {
3012
| load_got lj_tab_new
3013
| srl CARG2, RD, 3
3014
| andi CARG2, CARG2, 0x7ff
3015
| li TMP0, 0x801
3016
| addiu AT, CARG2, -0x7ff
3017
| srl CARG3, RD, 14
3018
| movz CARG2, TMP0, AT
3019
| // (lua_State *L, int32_t asize, uint32_t hbits)
3020
| call_intern lj_tab_new
3021
|. move CARG1, L
3022
| // Returns Table *.
3023
} else {
3024
| load_got lj_tab_dup
3025
| srl TMP1, RD, 1
3026
| subu TMP1, KBASE, TMP1
3027
| move CARG1, L
3028
| call_intern lj_tab_dup // (lua_State *L, Table *kt)
3029
|. lw CARG2, -4(TMP1) // KBASE-4-str_const*4
3030
| // Returns Table *.
3031
}
3032
| lw BASE, L->base
3033
| ins_next1
3034
| addu RA, BASE, RA
3035
| li TMP0, LJ_TTAB
3036
| sw TAB:CRET1, LO(RA)
3037
| sw TMP0, HI(RA)
3038
| ins_next2
3039
|5:
3040
| load_got lj_gc_step_fixtop
3041
| move MULTRES, RD
3042
| call_intern lj_gc_step_fixtop // (lua_State *L)
3043
|. move CARG1, L
3044
| b <1
3045
|. move RD, MULTRES
3046
break;
3047
3048
case BC_GGET:
3049
| // RA = dst*8, RD = str_const*8 (~)
3050
case BC_GSET:
3051
| // RA = src*8, RD = str_const*8 (~)
3052
| lw LFUNC:TMP2, FRAME_FUNC(BASE)
3053
| srl TMP1, RD, 1
3054
| subu TMP1, KBASE, TMP1
3055
| lw TAB:RB, LFUNC:TMP2->env
3056
| lw STR:RC, -4(TMP1) // KBASE-4-str_const*4
3057
if (op == BC_GGET) {
3058
| b ->BC_TGETS_Z
3059
} else {
3060
| b ->BC_TSETS_Z
3061
}
3062
|. addu RA, BASE, RA
3063
break;
3064
3065
case BC_TGETV:
3066
| // RA = dst*8, RB = table*8, RC = key*8
3067
| decode_RB8a RB, INS
3068
| decode_RB8b RB
3069
| decode_RDtoRC8 RC, RD
3070
| addu CARG2, BASE, RB
3071
| addu CARG3, BASE, RC
3072
| lw TMP1, HI(CARG2)
3073
| lw TMP2, HI(CARG3)
3074
| lw TAB:RB, LO(CARG2)
3075
| li AT, LJ_TTAB
3076
| ldc1 f0, 0(CARG3)
3077
| bne TMP1, AT, ->vmeta_tgetv
3078
|. addu RA, BASE, RA
3079
| sltiu AT, TMP2, LJ_TISNUM
3080
| beqz AT, >5
3081
|. li AT, LJ_TSTR
3082
|
3083
| // Convert number key to integer, check for integerness and range.
3084
| cvt.w.d f2, f0
3085
| lw TMP0, TAB:RB->asize
3086
| mfc1 TMP2, f2
3087
| cvt.d.w f4, f2
3088
| lw TMP1, TAB:RB->array
3089
| c.eq.d f0, f4
3090
| sltu AT, TMP2, TMP0
3091
| movf AT, r0
3092
| sll TMP2, TMP2, 3
3093
| beqz AT, ->vmeta_tgetv // Integer key and in array part?
3094
|. addu TMP2, TMP1, TMP2
3095
| lw TMP0, HI(TMP2)
3096
| beq TMP0, TISNIL, >2
3097
|. ldc1 f0, 0(TMP2)
3098
|1:
3099
| ins_next1
3100
| sdc1 f0, 0(RA)
3101
| ins_next2
3102
|
3103
|2: // Check for __index if table value is nil.
3104
| lw TAB:TMP2, TAB:RB->metatable
3105
| beqz TAB:TMP2, <1 // No metatable: done.
3106
|. nop
3107
| lbu TMP0, TAB:TMP2->nomm
3108
| andi TMP0, TMP0, 1<<MM_index
3109
| bnez TMP0, <1 // 'no __index' flag set: done.
3110
|. nop
3111
| b ->vmeta_tgetv
3112
|. nop
3113
|
3114
|5:
3115
| bne TMP2, AT, ->vmeta_tgetv
3116
|. lw STR:RC, LO(CARG3)
3117
| b ->BC_TGETS_Z // String key?
3118
|. nop
3119
break;
3120
case BC_TGETS:
3121
| // RA = dst*8, RB = table*8, RC = str_const*4 (~)
3122
| decode_RB8a RB, INS
3123
| decode_RB8b RB
3124
| addu CARG2, BASE, RB
3125
| decode_RC4a RC, INS
3126
| lw TMP0, HI(CARG2)
3127
| decode_RC4b RC
3128
| li AT, LJ_TTAB
3129
| lw TAB:RB, LO(CARG2)
3130
| subu CARG3, KBASE, RC
3131
| lw STR:RC, -4(CARG3) // KBASE-4-str_const*4
3132
| bne TMP0, AT, ->vmeta_tgets1
3133
|. addu RA, BASE, RA
3134
|->BC_TGETS_Z:
3135
| // TAB:RB = GCtab *, STR:RC = GCstr *, RA = dst*8
3136
| lw TMP0, TAB:RB->hmask
3137
| lw TMP1, STR:RC->hash
3138
| lw NODE:TMP2, TAB:RB->node
3139
| and TMP1, TMP1, TMP0 // idx = str->hash & tab->hmask
3140
| sll TMP0, TMP1, 5
3141
| sll TMP1, TMP1, 3
3142
| subu TMP1, TMP0, TMP1
3143
| addu NODE:TMP2, NODE:TMP2, TMP1 // node = tab->node + (idx*32-idx*8)
3144
|1:
3145
| lw CARG1, offsetof(Node, key)+HI(NODE:TMP2)
3146
| lw TMP0, offsetof(Node, key)+LO(NODE:TMP2)
3147
| lw NODE:TMP1, NODE:TMP2->next
3148
| lw CARG2, offsetof(Node, val)+HI(NODE:TMP2)
3149
| addiu CARG1, CARG1, -LJ_TSTR
3150
| xor TMP0, TMP0, STR:RC
3151
| or AT, CARG1, TMP0
3152
| bnez AT, >4
3153
|. lw TAB:TMP3, TAB:RB->metatable
3154
| beq CARG2, TISNIL, >5 // Key found, but nil value?
3155
|. lw CARG1, offsetof(Node, val)+LO(NODE:TMP2)
3156
|3:
3157
| ins_next1
3158
| sw CARG2, HI(RA)
3159
| sw CARG1, LO(RA)
3160
| ins_next2
3161
|
3162
|4: // Follow hash chain.
3163
| bnez NODE:TMP1, <1
3164
|. move NODE:TMP2, NODE:TMP1
3165
| // End of hash chain: key not found, nil result.
3166
|
3167
|5: // Check for __index if table value is nil.
3168
| beqz TAB:TMP3, <3 // No metatable: done.
3169
|. li CARG2, LJ_TNIL
3170
| lbu TMP0, TAB:TMP3->nomm
3171
| andi TMP0, TMP0, 1<<MM_index
3172
| bnez TMP0, <3 // 'no __index' flag set: done.
3173
|. nop
3174
| b ->vmeta_tgets
3175
|. nop
3176
break;
3177
case BC_TGETB:
3178
| // RA = dst*8, RB = table*8, RC = index*8
3179
| decode_RB8a RB, INS
3180
| decode_RB8b RB
3181
| addu CARG2, BASE, RB
3182
| decode_RDtoRC8 RC, RD
3183
| lw CARG1, HI(CARG2)
3184
| li AT, LJ_TTAB
3185
| lw TAB:RB, LO(CARG2)
3186
| addu RA, BASE, RA
3187
| bne CARG1, AT, ->vmeta_tgetb
3188
|. srl TMP0, RC, 3
3189
| lw TMP1, TAB:RB->asize
3190
| lw TMP2, TAB:RB->array
3191
| sltu AT, TMP0, TMP1
3192
| beqz AT, ->vmeta_tgetb
3193
|. addu RC, TMP2, RC
3194
| lw TMP1, HI(RC)
3195
| beq TMP1, TISNIL, >5
3196
|. ldc1 f0, 0(RC)
3197
|1:
3198
| ins_next1
3199
| sdc1 f0, 0(RA)
3200
| ins_next2
3201
|
3202
|5: // Check for __index if table value is nil.
3203
| lw TAB:TMP2, TAB:RB->metatable
3204
| beqz TAB:TMP2, <1 // No metatable: done.
3205
|. nop
3206
| lbu TMP1, TAB:TMP2->nomm
3207
| andi TMP1, TMP1, 1<<MM_index
3208
| bnez TMP1, <1 // 'no __index' flag set: done.
3209
|. nop
3210
| b ->vmeta_tgetb // Caveat: preserve TMP0!
3211
|. nop
3212
break;
3213
3214
case BC_TSETV:
3215
| // RA = src*8, RB = table*8, RC = key*8
3216
| decode_RB8a RB, INS
3217
| decode_RB8b RB
3218
| decode_RDtoRC8 RC, RD
3219
| addu CARG2, BASE, RB
3220
| addu CARG3, BASE, RC
3221
| lw TMP1, HI(CARG2)
3222
| lw TMP2, HI(CARG3)
3223
| lw TAB:RB, LO(CARG2)
3224
| li AT, LJ_TTAB
3225
| ldc1 f0, 0(CARG3)
3226
| bne TMP1, AT, ->vmeta_tsetv
3227
|. addu RA, BASE, RA
3228
| sltiu AT, TMP2, LJ_TISNUM
3229
| beqz AT, >5
3230
|. li AT, LJ_TSTR
3231
|
3232
| // Convert number key to integer, check for integerness and range.
3233
| cvt.w.d f2, f0
3234
| lw TMP0, TAB:RB->asize
3235
| mfc1 TMP2, f2
3236
| cvt.d.w f4, f2
3237
| lw TMP1, TAB:RB->array
3238
| c.eq.d f0, f4
3239
| sltu AT, TMP2, TMP0
3240
| movf AT, r0
3241
| sll TMP2, TMP2, 3
3242
| beqz AT, ->vmeta_tsetv // Integer key and in array part?
3243
|. addu TMP1, TMP1, TMP2
3244
| lbu TMP3, TAB:RB->marked
3245
| lw TMP0, HI(TMP1)
3246
| beq TMP0, TISNIL, >3
3247
|. ldc1 f0, 0(RA)
3248
|1:
3249
| andi AT, TMP3, LJ_GC_BLACK // isblack(table)
3250
| bnez AT, >7
3251
|. sdc1 f0, 0(TMP1)
3252
|2:
3253
| ins_next
3254
|
3255
|3: // Check for __newindex if previous value is nil.
3256
| lw TAB:TMP2, TAB:RB->metatable
3257
| beqz TAB:TMP2, <1 // No metatable: done.
3258
|. nop
3259
| lbu TMP2, TAB:TMP2->nomm
3260
| andi TMP2, TMP2, 1<<MM_newindex
3261
| bnez TMP2, <1 // 'no __newindex' flag set: done.
3262
|. nop
3263
| b ->vmeta_tsetv
3264
|. nop
3265
|
3266
|5:
3267
| bne TMP2, AT, ->vmeta_tsetv
3268
|. lw STR:RC, LO(CARG3)
3269
| b ->BC_TSETS_Z // String key?
3270
|. nop
3271
|
3272
|7: // Possible table write barrier for the value. Skip valiswhite check.
3273
| barrierback TAB:RB, TMP3, TMP0, <2
3274
break;
3275
case BC_TSETS:
3276
| // RA = src*8, RB = table*8, RC = str_const*8 (~)
3277
| decode_RB8a RB, INS
3278
| decode_RB8b RB
3279
| addu CARG2, BASE, RB
3280
| decode_RC4a RC, INS
3281
| lw TMP0, HI(CARG2)
3282
| decode_RC4b RC
3283
| li AT, LJ_TTAB
3284
| subu CARG3, KBASE, RC
3285
| lw TAB:RB, LO(CARG2)
3286
| lw STR:RC, -4(CARG3) // KBASE-4-str_const*4
3287
| bne TMP0, AT, ->vmeta_tsets1
3288
|. addu RA, BASE, RA
3289
|->BC_TSETS_Z:
3290
| // TAB:RB = GCtab *, STR:RC = GCstr *, RA = BASE+src*8
3291
| lw TMP0, TAB:RB->hmask
3292
| lw TMP1, STR:RC->hash
3293
| lw NODE:TMP2, TAB:RB->node
3294
| sb r0, TAB:RB->nomm // Clear metamethod cache.
3295
| and TMP1, TMP1, TMP0 // idx = str->hash & tab->hmask
3296
| sll TMP0, TMP1, 5
3297
| sll TMP1, TMP1, 3
3298
| subu TMP1, TMP0, TMP1
3299
| addu NODE:TMP2, NODE:TMP2, TMP1 // node = tab->node + (idx*32-idx*8)
3300
| ldc1 f20, 0(RA)
3301
|1:
3302
| lw CARG1, offsetof(Node, key)+HI(NODE:TMP2)
3303
| lw TMP0, offsetof(Node, key)+LO(NODE:TMP2)
3304
| li AT, LJ_TSTR
3305
| lw NODE:TMP1, NODE:TMP2->next
3306
| bne CARG1, AT, >5
3307
|. lw CARG2, offsetof(Node, val)+HI(NODE:TMP2)
3308
| bne TMP0, STR:RC, >5
3309
|. lbu TMP3, TAB:RB->marked
3310
| beq CARG2, TISNIL, >4 // Key found, but nil value?
3311
|. lw TAB:TMP0, TAB:RB->metatable
3312
|2:
3313
| andi AT, TMP3, LJ_GC_BLACK // isblack(table)
3314
| bnez AT, >7
3315
|. sdc1 f20, NODE:TMP2->val
3316
|3:
3317
| ins_next
3318
|
3319
|4: // Check for __newindex if previous value is nil.
3320
| beqz TAB:TMP0, <2 // No metatable: done.
3321
|. nop
3322
| lbu TMP0, TAB:TMP0->nomm
3323
| andi TMP0, TMP0, 1<<MM_newindex
3324
| bnez TMP0, <2 // 'no __newindex' flag set: done.
3325
|. nop
3326
| b ->vmeta_tsets
3327
|. nop
3328
|
3329
|5: // Follow hash chain.
3330
| bnez NODE:TMP1, <1
3331
|. move NODE:TMP2, NODE:TMP1
3332
| // End of hash chain: key not found, add a new one
3333
|
3334
| // But check for __newindex first.
3335
| lw TAB:TMP2, TAB:RB->metatable
3336
| beqz TAB:TMP2, >6 // No metatable: continue.
3337
|. addiu CARG3, DISPATCH, DISPATCH_GL(tmptv)
3338
| lbu TMP0, TAB:TMP2->nomm
3339
| andi TMP0, TMP0, 1<<MM_newindex
3340
| beqz TMP0, ->vmeta_tsets // 'no __newindex' flag NOT set: check.
3341
|. li AT, LJ_TSTR
3342
|6:
3343
| load_got lj_tab_newkey
3344
| sw STR:RC, LO(CARG3)
3345
| sw AT, HI(CARG3)
3346
| sw BASE, L->base
3347
| move CARG2, TAB:RB
3348
| sw PC, SAVE_PC
3349
| call_intern lj_tab_newkey // (lua_State *L, GCtab *t, TValue *k
3350
|. move CARG1, L
3351
| // Returns TValue *.
3352
| lw BASE, L->base
3353
| b <3 // No 2nd write barrier needed.
3354
|. sdc1 f20, 0(CRET1)
3355
|
3356
|7: // Possible table write barrier for the value. Skip valiswhite check.
3357
| barrierback TAB:RB, TMP3, TMP0, <3
3358
break;
3359
case BC_TSETB:
3360
| // RA = src*8, RB = table*8, RC = index*8
3361
| decode_RB8a RB, INS
3362
| decode_RB8b RB
3363
| addu CARG2, BASE, RB
3364
| decode_RDtoRC8 RC, RD
3365
| lw CARG1, HI(CARG2)
3366
| li AT, LJ_TTAB
3367
| lw TAB:RB, LO(CARG2)
3368
| addu RA, BASE, RA
3369
| bne CARG1, AT, ->vmeta_tsetb
3370
|. srl TMP0, RC, 3
3371
| lw TMP1, TAB:RB->asize
3372
| lw TMP2, TAB:RB->array
3373
| sltu AT, TMP0, TMP1
3374
| beqz AT, ->vmeta_tsetb
3375
|. addu RC, TMP2, RC
3376
| lw TMP1, HI(RC)
3377
| lbu TMP3, TAB:RB->marked
3378
| beq TMP1, TISNIL, >5
3379
|. ldc1 f0, 0(RA)
3380
|1:
3381
| andi AT, TMP3, LJ_GC_BLACK // isblack(table)
3382
| bnez AT, >7
3383
|. sdc1 f0, 0(RC)
3384
|2:
3385
| ins_next
3386
|
3387
|5: // Check for __newindex if previous value is nil.
3388
| lw TAB:TMP2, TAB:RB->metatable
3389
| beqz TAB:TMP2, <1 // No metatable: done.
3390
|. nop
3391
| lbu TMP1, TAB:TMP2->nomm
3392
| andi TMP1, TMP1, 1<<MM_newindex
3393
| bnez TMP1, <1 // 'no __newindex' flag set: done.
3394
|. nop
3395
| b ->vmeta_tsetb // Caveat: preserve TMP0!
3396
|. nop
3397
|
3398
|7: // Possible table write barrier for the value. Skip valiswhite check.
3399
| barrierback TAB:RB, TMP3, TMP0, <2
3400
break;
3401
3402
case BC_TSETM:
3403
| // RA = base*8 (table at base-1), RD = num_const*8 (start index)
3404
| addu RA, BASE, RA
3405
|1:
3406
| addu TMP3, KBASE, RD
3407
| lw TAB:CARG2, -8+LO(RA) // Guaranteed to be a table.
3408
| addiu TMP0, MULTRES, -8
3409
| lw TMP3, LO(TMP3) // Integer constant is in lo-word.
3410
| beqz TMP0, >4 // Nothing to copy?
3411
|. srl CARG3, TMP0, 3
3412
| addu CARG3, CARG3, TMP3
3413
| lw TMP2, TAB:CARG2->asize
3414
| sll TMP1, TMP3, 3
3415
| lbu TMP3, TAB:CARG2->marked
3416
| lw CARG1, TAB:CARG2->array
3417
| sltu AT, TMP2, CARG3
3418
| bnez AT, >5
3419
|. addu TMP2, RA, TMP0
3420
| addu TMP1, TMP1, CARG1
3421
| andi TMP0, TMP3, LJ_GC_BLACK // isblack(table)
3422
|3: // Copy result slots to table.
3423
| ldc1 f0, 0(RA)
3424
| addiu RA, RA, 8
3425
| sltu AT, RA, TMP2
3426
| sdc1 f0, 0(TMP1)
3427
| bnez AT, <3
3428
|. addiu TMP1, TMP1, 8
3429
| bnez TMP0, >7
3430
|. nop
3431
|4:
3432
| ins_next
3433
|
3434
|5: // Need to resize array part.
3435
| load_got lj_tab_reasize
3436
| sw BASE, L->base
3437
| sw PC, SAVE_PC
3438
| move BASE, RD
3439
| call_intern lj_tab_reasize // (lua_State *L, GCtab *t, int nasize)
3440
|. move CARG1, L
3441
| // Must not reallocate the stack.
3442
| move RD, BASE
3443
| b <1
3444
|. lw BASE, L->base // Reload BASE for lack of a saved register.
3445
|
3446
|7: // Possible table write barrier for any value. Skip valiswhite check.
3447
| barrierback TAB:CARG2, TMP3, TMP0, <4
3448
break;
3449
3450
/* -- Calls and vararg handling ----------------------------------------- */
3451
3452
case BC_CALLM:
3453
| // RA = base*8, (RB = (nresults+1)*8,) RC = extra_nargs*8
3454
| decode_RDtoRC8 NARGS8:RC, RD
3455
| b ->BC_CALL_Z
3456
|. addu NARGS8:RC, NARGS8:RC, MULTRES
3457
break;
3458
case BC_CALL:
3459
| // RA = base*8, (RB = (nresults+1)*8,) RC = (nargs+1)*8
3460
| decode_RDtoRC8 NARGS8:RC, RD
3461
|->BC_CALL_Z:
3462
| move TMP2, BASE
3463
| addu BASE, BASE, RA
3464
| li AT, LJ_TFUNC
3465
| lw TMP0, HI(BASE)
3466
| lw LFUNC:RB, LO(BASE)
3467
| addiu BASE, BASE, 8
3468
| bne TMP0, AT, ->vmeta_call
3469
|. addiu NARGS8:RC, NARGS8:RC, -8
3470
| ins_call
3471
break;
3472
3473
case BC_CALLMT:
3474
| // RA = base*8, (RB = 0,) RC = extra_nargs*8
3475
| addu NARGS8:RD, NARGS8:RD, MULTRES // BC_CALLT gets RC from RD.
3476
| // Fall through. Assumes BC_CALLT follows.
3477
break;
3478
case BC_CALLT:
3479
| // RA = base*8, (RB = 0,) RC = (nargs+1)*8
3480
| addu RA, BASE, RA
3481
| li AT, LJ_TFUNC
3482
| lw TMP0, HI(RA)
3483
| lw LFUNC:RB, LO(RA)
3484
| move NARGS8:RC, RD
3485
| lw TMP1, FRAME_PC(BASE)
3486
| addiu RA, RA, 8
3487
| bne TMP0, AT, ->vmeta_callt
3488
|. addiu NARGS8:RC, NARGS8:RC, -8
3489
|->BC_CALLT_Z:
3490
| andi TMP0, TMP1, FRAME_TYPE // Caveat: preserve TMP0 until the 'or'.
3491
| lbu TMP3, LFUNC:RB->ffid
3492
| bnez TMP0, >7
3493
|. xori TMP2, TMP1, FRAME_VARG
3494
|1:
3495
| sw LFUNC:RB, FRAME_FUNC(BASE) // Copy function down, but keep PC.
3496
| sltiu AT, TMP3, 2 // (> FF_C) Calling a fast function?
3497
| move TMP2, BASE
3498
| beqz NARGS8:RC, >3
3499
|. move TMP3, NARGS8:RC
3500
|2:
3501
| ldc1 f0, 0(RA)
3502
| addiu RA, RA, 8
3503
| addiu TMP3, TMP3, -8
3504
| sdc1 f0, 0(TMP2)
3505
| bnez TMP3, <2
3506
|. addiu TMP2, TMP2, 8
3507
|3:
3508
| or TMP0, TMP0, AT
3509
| beqz TMP0, >5
3510
|. nop
3511
|4:
3512
| ins_callt
3513
|
3514
|5: // Tailcall to a fast function with a Lua frame below.
3515
| lw INS, -4(TMP1)
3516
| decode_RA8a RA, INS
3517
| decode_RA8b RA
3518
| subu TMP1, BASE, RA
3519
| lw LFUNC:TMP1, -8+FRAME_FUNC(TMP1)
3520
| lw TMP1, LFUNC:TMP1->pc
3521
| b <4
3522
|. lw KBASE, PC2PROTO(k)(TMP1) // Need to prepare KBASE.
3523
|
3524
|7: // Tailcall from a vararg function.
3525
| andi AT, TMP2, FRAME_TYPEP
3526
| bnez AT, <1 // Vararg frame below?
3527
|. subu TMP2, BASE, TMP2 // Relocate BASE down.
3528
| move BASE, TMP2
3529
| lw TMP1, FRAME_PC(TMP2)
3530
| b <1
3531
|. andi TMP0, TMP1, FRAME_TYPE
3532
break;
3533
3534
case BC_ITERC:
3535
| // RA = base*8, (RB = (nresults+1)*8, RC = (nargs+1)*8 ((2+1)*8))
3536
| move TMP2, BASE
3537
| addu BASE, BASE, RA
3538
| li AT, LJ_TFUNC
3539
| lw TMP1, -24+HI(BASE)
3540
| lw LFUNC:RB, -24+LO(BASE)
3541
| ldc1 f2, -8(BASE)
3542
| ldc1 f0, -16(BASE)
3543
| sw TMP1, HI(BASE) // Copy callable.
3544
| sw LFUNC:RB, LO(BASE)
3545
| sdc1 f2, 16(BASE) // Copy control var.
3546
| sdc1 f0, 8(BASE) // Copy state.
3547
| addiu BASE, BASE, 8
3548
| bne TMP1, AT, ->vmeta_call
3549
|. li NARGS8:RC, 16 // Iterators get 2 arguments.
3550
| ins_call
3551
break;
3552
3553
case BC_ITERN:
3554
| // RA = base*8, (RB = (nresults+1)*8, RC = (nargs+1)*8 (2+1)*8)
3555
|.if JIT
3556
| // NYI: add hotloop, record BC_ITERN.
3557
|.endif
3558
| addu RA, BASE, RA
3559
| lw TAB:RB, -16+LO(RA)
3560
| lw RC, -8+LO(RA) // Get index from control var.
3561
| lw TMP0, TAB:RB->asize
3562
| lw TMP1, TAB:RB->array
3563
| addiu PC, PC, 4
3564
|1: // Traverse array part.
3565
| sltu AT, RC, TMP0
3566
| beqz AT, >5 // Index points after array part?
3567
|. sll TMP3, RC, 3
3568
| addu TMP3, TMP1, TMP3
3569
| lw TMP2, HI(TMP3)
3570
| ldc1 f0, 0(TMP3)
3571
| mtc1 RC, f2
3572
| lhu RD, -4+OFS_RD(PC)
3573
| beq TMP2, TISNIL, <1 // Skip holes in array part.
3574
|. addiu RC, RC, 1
3575
| cvt.d.w f2, f2
3576
| lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535)
3577
| sdc1 f0, 8(RA)
3578
| decode_RD4b RD
3579
| addu RD, RD, TMP3
3580
| sw RC, -8+LO(RA) // Update control var.
3581
| addu PC, PC, RD
3582
| sdc1 f2, 0(RA)
3583
|3:
3584
| ins_next
3585
|
3586
|5: // Traverse hash part.
3587
| lw TMP1, TAB:RB->hmask
3588
| subu RC, RC, TMP0
3589
| lw TMP2, TAB:RB->node
3590
|6:
3591
| sltu AT, TMP1, RC // End of iteration? Branch to ITERL+1.
3592
| bnez AT, <3
3593
|. sll TMP3, RC, 5
3594
| sll RB, RC, 3
3595
| subu TMP3, TMP3, RB
3596
| addu NODE:TMP3, TMP3, TMP2
3597
| lw RB, HI(NODE:TMP3)
3598
| ldc1 f0, 0(NODE:TMP3)
3599
| lhu RD, -4+OFS_RD(PC)
3600
| beq RB, TISNIL, <6 // Skip holes in hash part.
3601
|. addiu RC, RC, 1
3602
| ldc1 f2, NODE:TMP3->key
3603
| lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535)
3604
| sdc1 f0, 8(RA)
3605
| addu RC, RC, TMP0
3606
| decode_RD4b RD
3607
| addu RD, RD, TMP3
3608
| sdc1 f2, 0(RA)
3609
| addu PC, PC, RD
3610
| b <3
3611
|. sw RC, -8+LO(RA) // Update control var.
3612
break;
3613
3614
case BC_ISNEXT:
3615
| // RA = base*8, RD = target (points to ITERN)
3616
| addu RA, BASE, RA
3617
| lw TMP0, -24+HI(RA)
3618
| lw CFUNC:TMP1, -24+LO(RA)
3619
| lw TMP2, -16+HI(RA)
3620
| lw TMP3, -8+HI(RA)
3621
| li AT, LJ_TFUNC
3622
| bne TMP0, AT, >5
3623
|. addiu TMP2, TMP2, -LJ_TTAB
3624
| lbu TMP1, CFUNC:TMP1->ffid
3625
| addiu TMP3, TMP3, -LJ_TNIL
3626
| srl TMP0, RD, 1
3627
| or TMP2, TMP2, TMP3
3628
| addiu TMP1, TMP1, -FF_next_N
3629
| addu TMP0, PC, TMP0
3630
| or TMP1, TMP1, TMP2
3631
| bnez TMP1, >5
3632
|. lui TMP2, (-(BCBIAS_J*4 >> 16) & 65535)
3633
| addu PC, TMP0, TMP2
3634
| lui TMP1, 0xfffe
3635
| ori TMP1, TMP1, 0x7fff
3636
| sw r0, -8+LO(RA) // Initialize control var.
3637
| sw TMP1, -8+HI(RA)
3638
|1:
3639
| ins_next
3640
|5: // Despecialize bytecode if any of the checks fail.
3641
| li TMP3, BC_JMP
3642
| li TMP1, BC_ITERC
3643
| sb TMP3, -4+OFS_OP(PC)
3644
| addu PC, TMP0, TMP2
3645
| b <1
3646
|. sb TMP1, OFS_OP(PC)
3647
break;
3648
3649
case BC_VARG:
3650
| // RA = base*8, RB = (nresults+1)*8, RC = numparams*8
3651
| lw TMP0, FRAME_PC(BASE)
3652
| decode_RDtoRC8 RC, RD
3653
| decode_RB8a RB, INS
3654
| addu RC, BASE, RC
3655
| decode_RB8b RB
3656
| addu RA, BASE, RA
3657
| addiu RC, RC, FRAME_VARG
3658
| addu TMP2, RA, RB
3659
| addiu TMP3, BASE, -8 // TMP3 = vtop
3660
| subu RC, RC, TMP0 // RC = vbase
3661
| // Note: RC may now be even _above_ BASE if nargs was < numparams.
3662
| beqz RB, >5 // Copy all varargs?
3663
|. subu TMP1, TMP3, RC
3664
| addiu TMP2, TMP2, -16
3665
|1: // Copy vararg slots to destination slots.
3666
| lw CARG1, HI(RC)
3667
| sltu AT, RC, TMP3
3668
| lw CARG2, LO(RC)
3669
| addiu RC, RC, 8
3670
| movz CARG1, TISNIL, AT
3671
| sw CARG1, HI(RA)
3672
| sw CARG2, LO(RA)
3673
| sltu AT, RA, TMP2
3674
| bnez AT, <1
3675
|. addiu RA, RA, 8
3676
|3:
3677
| ins_next
3678
|
3679
|5: // Copy all varargs.
3680
| lw TMP0, L->maxstack
3681
| blez TMP1, <3 // No vararg slots?
3682
|. li MULTRES, 8 // MULTRES = (0+1)*8
3683
| addu TMP2, RA, TMP1
3684
| sltu AT, TMP0, TMP2
3685
| bnez AT, >7
3686
|. addiu MULTRES, TMP1, 8
3687
|6:
3688
| ldc1 f0, 0(RC)
3689
| addiu RC, RC, 8
3690
| sdc1 f0, 0(RA)
3691
| sltu AT, RC, TMP3
3692
| bnez AT, <6 // More vararg slots?
3693
|. addiu RA, RA, 8
3694
| b <3
3695
|. nop
3696
|
3697
|7: // Grow stack for varargs.
3698
| load_got lj_state_growstack
3699
| sw RA, L->top
3700
| subu RA, RA, BASE
3701
| sw BASE, L->base
3702
| subu BASE, RC, BASE // Need delta, because BASE may change.
3703
| sw PC, SAVE_PC
3704
| srl CARG2, TMP1, 3
3705
| call_intern lj_state_growstack // (lua_State *L, int n)
3706
|. move CARG1, L
3707
| move RC, BASE
3708
| lw BASE, L->base
3709
| addu RA, BASE, RA
3710
| addu RC, BASE, RC
3711
| b <6
3712
|. addiu TMP3, BASE, -8
3713
break;
3714
3715
/* -- Returns ----------------------------------------------------------- */
3716
3717
case BC_RETM:
3718
| // RA = results*8, RD = extra_nresults*8
3719
| addu RD, RD, MULTRES // MULTRES >= 8, so RD >= 8.
3720
| // Fall through. Assumes BC_RET follows.
3721
break;
3722
3723
case BC_RET:
3724
| // RA = results*8, RD = (nresults+1)*8
3725
| lw PC, FRAME_PC(BASE)
3726
| addu RA, BASE, RA
3727
| move MULTRES, RD
3728
|1:
3729
| andi TMP0, PC, FRAME_TYPE
3730
| bnez TMP0, ->BC_RETV_Z
3731
|. xori TMP1, PC, FRAME_VARG
3732
|
3733
|->BC_RET_Z:
3734
| // BASE = base, RA = resultptr, RD = (nresults+1)*8, PC = return
3735
| lw INS, -4(PC)
3736
| addiu TMP2, BASE, -8
3737
| addiu RC, RD, -8
3738
| decode_RA8a TMP0, INS
3739
| decode_RB8a RB, INS
3740
| decode_RA8b TMP0
3741
| decode_RB8b RB
3742
| addu TMP3, TMP2, RB
3743
| beqz RC, >3
3744
|. subu BASE, TMP2, TMP0
3745
|2:
3746
| ldc1 f0, 0(RA)
3747
| addiu RA, RA, 8
3748
| addiu RC, RC, -8
3749
| sdc1 f0, 0(TMP2)
3750
| bnez RC, <2
3751
|. addiu TMP2, TMP2, 8
3752
|3:
3753
| addiu TMP3, TMP3, -8
3754
|5:
3755
| sltu AT, TMP2, TMP3
3756
| bnez AT, >6
3757
|. lw LFUNC:TMP1, FRAME_FUNC(BASE)
3758
| ins_next1
3759
| lw TMP1, LFUNC:TMP1->pc
3760
| lw KBASE, PC2PROTO(k)(TMP1)
3761
| ins_next2
3762
|
3763
|6: // Fill up results with nil.
3764
| sw TISNIL, HI(TMP2)
3765
| b <5
3766
|. addiu TMP2, TMP2, 8
3767
|
3768
|->BC_RETV_Z: // Non-standard return case.
3769
| andi TMP2, TMP1, FRAME_TYPEP
3770
| bnez TMP2, ->vm_return
3771
|. nop
3772
| // Return from vararg function: relocate BASE down.
3773
| subu BASE, BASE, TMP1
3774
| b <1
3775
|. lw PC, FRAME_PC(BASE)
3776
break;
3777
3778
case BC_RET0: case BC_RET1:
3779
| // RA = results*8, RD = (nresults+1)*8
3780
| lw PC, FRAME_PC(BASE)
3781
| addu RA, BASE, RA
3782
| move MULTRES, RD
3783
| andi TMP0, PC, FRAME_TYPE
3784
| bnez TMP0, ->BC_RETV_Z
3785
|. xori TMP1, PC, FRAME_VARG
3786
|
3787
| lw INS, -4(PC)
3788
| addiu TMP2, BASE, -8
3789
if (op == BC_RET1) {
3790
| ldc1 f0, 0(RA)
3791
}
3792
| decode_RB8a RB, INS
3793
| decode_RA8a RA, INS
3794
| decode_RB8b RB
3795
| decode_RA8b RA
3796
if (op == BC_RET1) {
3797
| sdc1 f0, 0(TMP2)
3798
}
3799
| subu BASE, TMP2, RA
3800
|5:
3801
| sltu AT, RD, RB
3802
| bnez AT, >6
3803
|. lw LFUNC:TMP1, FRAME_FUNC(BASE)
3804
| ins_next1
3805
| lw TMP1, LFUNC:TMP1->pc
3806
| lw KBASE, PC2PROTO(k)(TMP1)
3807
| ins_next2
3808
|
3809
|6: // Fill up results with nil.
3810
| addiu TMP2, TMP2, 8
3811
| addiu RD, RD, 8
3812
| b <5
3813
if (op == BC_RET1) {
3814
|. sw TISNIL, HI(TMP2)
3815
} else {
3816
|. sw TISNIL, -8+HI(TMP2)
3817
}
3818
break;
3819
3820
/* -- Loops and branches ------------------------------------------------ */
3821
3822
case BC_FORL:
3823
|.if JIT
3824
| hotloop
3825
|.endif
3826
| // Fall through. Assumes BC_IFORL follows.
3827
break;
3828
3829
case BC_JFORI:
3830
case BC_JFORL:
3831
#if !LJ_HASJIT
3832
break;
3833
#endif
3834
case BC_FORI:
3835
case BC_IFORL:
3836
| // RA = base*8, RD = target (after end of loop or start of loop)
3837
vk = (op == BC_IFORL || op == BC_JFORL);
3838
| addu RA, BASE, RA
3839
if (vk) {
3840
| ldc1 f0, FORL_IDX*8(RA)
3841
| ldc1 f4, FORL_STEP*8(RA)
3842
| ldc1 f2, FORL_STOP*8(RA)
3843
| lw TMP3, FORL_STEP*8+HI(RA)
3844
| add.d f0, f0, f4
3845
| sdc1 f0, FORL_IDX*8(RA)
3846
} else {
3847
| lw TMP1, FORL_IDX*8+HI(RA)
3848
| lw TMP3, FORL_STEP*8+HI(RA)
3849
| lw TMP2, FORL_STOP*8+HI(RA)
3850
| sltiu TMP1, TMP1, LJ_TISNUM
3851
| sltiu TMP0, TMP3, LJ_TISNUM
3852
| sltiu TMP2, TMP2, LJ_TISNUM
3853
| and TMP1, TMP1, TMP0
3854
| and TMP1, TMP1, TMP2
3855
| ldc1 f0, FORL_IDX*8(RA)
3856
| beqz TMP1, ->vmeta_for
3857
|. ldc1 f2, FORL_STOP*8(RA)
3858
}
3859
if (op != BC_JFORL) {
3860
| srl RD, RD, 1
3861
| lui TMP0, (-(BCBIAS_J*4 >> 16) & 65535)
3862
}
3863
| c.le.d 0, f0, f2
3864
| c.le.d 1, f2, f0
3865
| sdc1 f0, FORL_EXT*8(RA)
3866
if (op == BC_JFORI) {
3867
| li TMP1, 1
3868
| li TMP2, 1
3869
| addu TMP0, RD, TMP0
3870
| slt TMP3, TMP3, r0
3871
| movf TMP1, r0, 0
3872
| addu PC, PC, TMP0
3873
| movf TMP2, r0, 1
3874
| lhu RD, -4+OFS_RD(PC)
3875
| movn TMP1, TMP2, TMP3
3876
| bnez TMP1, =>BC_JLOOP
3877
|. decode_RD8b RD
3878
} else if (op == BC_JFORL) {
3879
| li TMP1, 1
3880
| li TMP2, 1
3881
| slt TMP3, TMP3, r0
3882
| movf TMP1, r0, 0
3883
| movf TMP2, r0, 1
3884
| movn TMP1, TMP2, TMP3
3885
| bnez TMP1, =>BC_JLOOP
3886
|. nop
3887
} else {
3888
| addu TMP1, RD, TMP0
3889
| slt TMP3, TMP3, r0
3890
| move TMP2, TMP1
3891
if (op == BC_FORI) {
3892
| movt TMP1, r0, 0
3893
| movt TMP2, r0, 1
3894
} else {
3895
| movf TMP1, r0, 0
3896
| movf TMP2, r0, 1
3897
}
3898
| movn TMP1, TMP2, TMP3
3899
| addu PC, PC, TMP1
3900
}
3901
| ins_next
3902
break;
3903
3904
case BC_ITERL:
3905
|.if JIT
3906
| hotloop
3907
|.endif
3908
| // Fall through. Assumes BC_IITERL follows.
3909
break;
3910
3911
case BC_JITERL:
3912
#if !LJ_HASJIT
3913
break;
3914
#endif
3915
case BC_IITERL:
3916
| // RA = base*8, RD = target
3917
| addu RA, BASE, RA
3918
| lw TMP1, HI(RA)
3919
| beq TMP1, TISNIL, >1 // Stop if iterator returned nil.
3920
|. lw TMP2, LO(RA)
3921
if (op == BC_JITERL) {
3922
| sw TMP1, -8+HI(RA)
3923
| b =>BC_JLOOP
3924
|. sw TMP2, -8+LO(RA)
3925
} else {
3926
| branch_RD // Otherwise save control var + branch.
3927
| sw TMP1, -8+HI(RA)
3928
| sw TMP2, -8+LO(RA)
3929
}
3930
|1:
3931
| ins_next
3932
break;
3933
3934
case BC_LOOP:
3935
| // RA = base*8, RD = target (loop extent)
3936
| // Note: RA/RD is only used by trace recorder to determine scope/extent
3937
| // This opcode does NOT jump, it's only purpose is to detect a hot loop.
3938
|.if JIT
3939
| hotloop
3940
|.endif
3941
| // Fall through. Assumes BC_ILOOP follows.
3942
break;
3943
3944
case BC_ILOOP:
3945
| // RA = base*8, RD = target (loop extent)
3946
| ins_next
3947
break;
3948
3949
case BC_JLOOP:
3950
|.if JIT
3951
| // RA = base*8 (ignored), RD = traceno*8
3952
| lw TMP1, DISPATCH_J(trace)(DISPATCH)
3953
| srl RD, RD, 1
3954
| li AT, 0
3955
| addu TMP1, TMP1, RD
3956
| // Traces on MIPS don't store the trace number, so use 0.
3957
| sw AT, DISPATCH_GL(vmstate)(DISPATCH)
3958
| lw TRACE:TMP2, 0(TMP1)
3959
| sw BASE, DISPATCH_GL(jit_base)(DISPATCH)
3960
| sw L, DISPATCH_GL(jit_L)(DISPATCH)
3961
| lw TMP2, TRACE:TMP2->mcode
3962
| jr TMP2
3963
|. addiu JGL, DISPATCH, GG_DISP2G+32768
3964
|.endif
3965
break;
3966
3967
case BC_JMP:
3968
| // RA = base*8 (only used by trace recorder), RD = target
3969
| branch_RD
3970
| ins_next
3971
break;
3972
3973
/* -- Function headers -------------------------------------------------- */
3974
3975
case BC_FUNCF:
3976
|.if JIT
3977
| hotcall
3978
|.endif
3979
case BC_FUNCV: /* NYI: compiled vararg functions. */
3980
| // Fall through. Assumes BC_IFUNCF/BC_IFUNCV follow.
3981
break;
3982
3983
case BC_JFUNCF:
3984
#if !LJ_HASJIT
3985
break;
3986
#endif
3987
case BC_IFUNCF:
3988
| // BASE = new base, RA = BASE+framesize*8, RB = LFUNC, RC = nargs*8
3989
| lw TMP2, L->maxstack
3990
| lbu TMP1, -4+PC2PROTO(numparams)(PC)
3991
| lw KBASE, -4+PC2PROTO(k)(PC)
3992
| sltu AT, TMP2, RA
3993
| bnez AT, ->vm_growstack_l
3994
|. sll TMP1, TMP1, 3
3995
if (op != BC_JFUNCF) {
3996
| ins_next1
3997
}
3998
|2:
3999
| sltu AT, NARGS8:RC, TMP1 // Check for missing parameters.
4000
| bnez AT, >3
4001
|. addu AT, BASE, NARGS8:RC
4002
if (op == BC_JFUNCF) {
4003
| decode_RD8a RD, INS
4004
| b =>BC_JLOOP
4005
|. decode_RD8b RD
4006
} else {
4007
| ins_next2
4008
}
4009
|
4010
|3: // Clear missing parameters.
4011
| sw TISNIL, HI(AT)
4012
| b <2
4013
|. addiu NARGS8:RC, NARGS8:RC, 8
4014
break;
4015
4016
case BC_JFUNCV:
4017
#if !LJ_HASJIT
4018
break;
4019
#endif
4020
| NYI // NYI: compiled vararg functions
4021
break; /* NYI: compiled vararg functions. */
4022
4023
case BC_IFUNCV:
4024
| // BASE = new base, RA = BASE+framesize*8, RB = LFUNC, RC = nargs*8
4025
| addu TMP1, BASE, RC
4026
| lw TMP2, L->maxstack
4027
| addu TMP0, RA, RC
4028
| sw LFUNC:RB, LO(TMP1) // Store copy of LFUNC.
4029
| addiu TMP3, RC, 8+FRAME_VARG
4030
| sltu AT, TMP0, TMP2
4031
| lw KBASE, -4+PC2PROTO(k)(PC)
4032
| beqz AT, ->vm_growstack_l
4033
|. sw TMP3, HI(TMP1) // Store delta + FRAME_VARG.
4034
| lbu TMP2, -4+PC2PROTO(numparams)(PC)
4035
| move RA, BASE
4036
| move RC, TMP1
4037
| ins_next1
4038
| beqz TMP2, >3
4039
|. addiu BASE, TMP1, 8
4040
|1:
4041
| lw TMP0, HI(RA)
4042
| lw TMP3, LO(RA)
4043
| sltu AT, RA, RC // Less args than parameters?
4044
| move CARG1, TMP0
4045
| movz TMP0, TISNIL, AT // Clear missing parameters.
4046
| movn CARG1, TISNIL, AT // Clear old fixarg slot (help the GC).
4047
| sw TMP3, 8+LO(TMP1)
4048
| addiu TMP2, TMP2, -1
4049
| sw TMP0, 8+HI(TMP1)
4050
| addiu TMP1, TMP1, 8
4051
| sw CARG1, HI(RA)
4052
| bnez TMP2, <1
4053
|. addiu RA, RA, 8
4054
|3:
4055
| ins_next2
4056
break;
4057
4058
case BC_FUNCC:
4059
case BC_FUNCCW:
4060
| // BASE = new base, RA = BASE+framesize*8, RB = CFUNC, RC = nargs*8
4061
if (op == BC_FUNCC) {
4062
| lw CFUNCADDR, CFUNC:RB->f
4063
} else {
4064
| lw CFUNCADDR, DISPATCH_GL(wrapf)(DISPATCH)
4065
}
4066
| addu TMP1, RA, NARGS8:RC
4067
| lw TMP2, L->maxstack
4068
| addu RC, BASE, NARGS8:RC
4069
| sw BASE, L->base
4070
| sltu AT, TMP2, TMP1
4071
| sw RC, L->top
4072
| li_vmstate C
4073
if (op == BC_FUNCCW) {
4074
| lw CARG2, CFUNC:RB->f
4075
}
4076
| bnez AT, ->vm_growstack_c // Need to grow stack.
4077
|. move CARG1, L
4078
| jalr CFUNCADDR // (lua_State *L [, lua_CFunction f])
4079
|. st_vmstate
4080
| // Returns nresults.
4081
| lw BASE, L->base
4082
| sll RD, CRET1, 3
4083
| lw TMP1, L->top
4084
| li_vmstate INTERP
4085
| lw PC, FRAME_PC(BASE) // Fetch PC of caller.
4086
| subu RA, TMP1, RD // RA = L->top - nresults*8
4087
| b ->vm_returnc
4088
|. st_vmstate
4089
break;
4090
4091
/* ---------------------------------------------------------------------- */
4092
4093
default:
4094
fprintf(stderr, "Error: undefined opcode BC_%s\n", bc_names[op]);
4095
exit(2);
4096
break;
4097
}
4098
}
4099
4100
static int build_backend(BuildCtx *ctx)
4101
{
4102
int op;
4103
4104
dasm_growpc(Dst, BC__MAX);
4105
4106
build_subroutines(ctx);
4107
4108
|.code_op
4109
for (op = 0; op < BC__MAX; op++)
4110
build_ins(ctx, (BCOp)op, op);
4111
4112
return BC__MAX;
4113
}
4114
4115
/* Emit pseudo frame-info for all assembler functions. */
4116
static void emit_asm_debug(BuildCtx *ctx)
4117
{
4118
int fcofs = (int)((uint8_t *)ctx->glob[GLOB_vm_ffi_call] - ctx->code);
4119
int i;
4120
switch (ctx->mode) {
4121
case BUILD_elfasm:
4122
fprintf(ctx->fp, "\t.section .debug_frame,\"\",@progbits\n");
4123
fprintf(ctx->fp,
4124
".Lframe0:\n"
4125
"\t.4byte .LECIE0-.LSCIE0\n"
4126
".LSCIE0:\n"
4127
"\t.4byte 0xffffffff\n"
4128
"\t.byte 0x1\n"
4129
"\t.string \"\"\n"
4130
"\t.uleb128 0x1\n"
4131
"\t.sleb128 -4\n"
4132
"\t.byte 31\n"
4133
"\t.byte 0xc\n\t.uleb128 29\n\t.uleb128 0\n"
4134
"\t.align 2\n"
4135
".LECIE0:\n\n");
4136
fprintf(ctx->fp,
4137
".LSFDE0:\n"
4138
"\t.4byte .LEFDE0-.LASFDE0\n"
4139
".LASFDE0:\n"
4140
"\t.4byte .Lframe0\n"
4141
"\t.4byte .Lbegin\n"
4142
"\t.4byte %d\n"
4143
"\t.byte 0xe\n\t.uleb128 %d\n"
4144
"\t.byte 0x9f\n\t.sleb128 1\n"
4145
"\t.byte 0x9e\n\t.sleb128 2\n",
4146
fcofs, CFRAME_SIZE);
4147
for (i = 23; i >= 16; i--)
4148
fprintf(ctx->fp, "\t.byte %d\n\t.uleb128 %d\n", 0x80+i, 26-i);
4149
for (i = 30; i >= 20; i -= 2)
4150
fprintf(ctx->fp, "\t.byte %d\n\t.uleb128 %d\n", 0x80+32+i, 42-i);
4151
fprintf(ctx->fp,
4152
"\t.align 2\n"
4153
".LEFDE0:\n\n");
4154
#if LJ_HASFFI
4155
fprintf(ctx->fp,
4156
".LSFDE1:\n"
4157
"\t.4byte .LEFDE1-.LASFDE1\n"
4158
".LASFDE1:\n"
4159
"\t.4byte .Lframe0\n"
4160
"\t.4byte lj_vm_ffi_call\n"
4161
"\t.4byte %d\n"
4162
"\t.byte 0x9f\n\t.uleb128 1\n"
4163
"\t.byte 0x90\n\t.uleb128 2\n"
4164
"\t.byte 0xd\n\t.uleb128 0x10\n"
4165
"\t.align 2\n"
4166
".LEFDE1:\n\n", (int)ctx->codesz - fcofs);
4167
#endif
4168
fprintf(ctx->fp, "\t.section .eh_frame,\"aw\",@progbits\n");
4169
fprintf(ctx->fp,
4170
"\t.globl lj_err_unwind_dwarf\n"
4171
".Lframe1:\n"
4172
"\t.4byte .LECIE1-.LSCIE1\n"
4173
".LSCIE1:\n"
4174
"\t.4byte 0\n"
4175
"\t.byte 0x1\n"
4176
"\t.string \"zPR\"\n"
4177
"\t.uleb128 0x1\n"
4178
"\t.sleb128 -4\n"
4179
"\t.byte 31\n"
4180
"\t.uleb128 6\n" /* augmentation length */
4181
"\t.byte 0\n"
4182
"\t.4byte lj_err_unwind_dwarf\n"
4183
"\t.byte 0\n"
4184
"\t.byte 0xc\n\t.uleb128 29\n\t.uleb128 0\n"
4185
"\t.align 2\n"
4186
".LECIE1:\n\n");
4187
fprintf(ctx->fp,
4188
".LSFDE2:\n"
4189
"\t.4byte .LEFDE2-.LASFDE2\n"
4190
".LASFDE2:\n"
4191
"\t.4byte .LASFDE2-.Lframe1\n"
4192
"\t.4byte .Lbegin\n"
4193
"\t.4byte %d\n"
4194
"\t.uleb128 0\n" /* augmentation length */
4195
"\t.byte 0xe\n\t.uleb128 %d\n"
4196
"\t.byte 0x9f\n\t.sleb128 1\n"
4197
"\t.byte 0x9e\n\t.sleb128 2\n",
4198
fcofs, CFRAME_SIZE);
4199
for (i = 23; i >= 16; i--)
4200
fprintf(ctx->fp, "\t.byte %d\n\t.uleb128 %d\n", 0x80+i, 26-i);
4201
for (i = 30; i >= 20; i -= 2)
4202
fprintf(ctx->fp, "\t.byte %d\n\t.uleb128 %d\n", 0x80+32+i, 42-i);
4203
fprintf(ctx->fp,
4204
"\t.align 2\n"
4205
".LEFDE2:\n\n");
4206
#if LJ_HASFFI
4207
fprintf(ctx->fp,
4208
".Lframe2:\n"
4209
"\t.4byte .LECIE2-.LSCIE2\n"
4210
".LSCIE2:\n"
4211
"\t.4byte 0\n"
4212
"\t.byte 0x1\n"
4213
"\t.string \"zR\"\n"
4214
"\t.uleb128 0x1\n"
4215
"\t.sleb128 -4\n"
4216
"\t.byte 31\n"
4217
"\t.uleb128 1\n" /* augmentation length */
4218
"\t.byte 0\n"
4219
"\t.byte 0xc\n\t.uleb128 29\n\t.uleb128 0\n"
4220
"\t.align 2\n"
4221
".LECIE2:\n\n");
4222
fprintf(ctx->fp,
4223
".LSFDE3:\n"
4224
"\t.4byte .LEFDE3-.LASFDE3\n"
4225
".LASFDE3:\n"
4226
"\t.4byte .LASFDE3-.Lframe2\n"
4227
"\t.4byte lj_vm_ffi_call\n"
4228
"\t.4byte %d\n"
4229
"\t.uleb128 0\n" /* augmentation length */
4230
"\t.byte 0x9f\n\t.uleb128 1\n"
4231
"\t.byte 0x90\n\t.uleb128 2\n"
4232
"\t.byte 0xd\n\t.uleb128 0x10\n"
4233
"\t.align 2\n"
4234
".LEFDE3:\n\n", (int)ctx->codesz - fcofs);
4235
#endif
4236
break;
4237
default:
4238
break;
4239
}
4240
}
4241
Loggerhead is a web-based interface for Breezy
Version: 2.0.1