summaryrefslogtreecommitdiff
path: root/src/vm_mips.dasc
diff options
context:
space:
mode:
Diffstat (limited to 'src/vm_mips.dasc')
-rw-r--r--src/vm_mips.dasc4224
1 files changed, 4224 insertions, 0 deletions
diff --git a/src/vm_mips.dasc b/src/vm_mips.dasc
new file mode 100644
index 00000000..a7c3b4ae
--- /dev/null
+++ b/src/vm_mips.dasc
@@ -0,0 +1,4224 @@
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. */
308static 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 | andi TMP0, PC, FRAME_TYPE
331 | addiu RD, RD, 8 // RD = (nresults+1)*8.
332 | beqz TMP0, ->BC_RET_Z // Handle regular return to Lua.
333 |. move MULTRES, RD
334 |
335 |->vm_return:
336 | // BASE = base, RA = resultptr, RD/MULTRES = (nresults+1)*8, PC = return
337 | // TMP0 = PC & FRAME_TYPE
338 | li TMP2, -8
339 | xori AT, TMP0, FRAME_C
340 | and TMP2, PC, TMP2
341 | bnez AT, ->vm_returnp
342 | subu TMP2, BASE, TMP2 // TMP2 = previous base.
343 |
344 | addiu TMP1, RD, -8
345 | sw TMP2, L->base
346 | li_vmstate C
347 | lw TMP2, SAVE_NRES
348 | addiu BASE, BASE, -8
349 | st_vmstate
350 | beqz TMP1, >2
351 |. sll TMP2, TMP2, 3
352 |1:
353 | addiu TMP1, TMP1, -8
354 | ldc1 f0, 0(RA)
355 | addiu RA, RA, 8
356 | sdc1 f0, 0(BASE)
357 | bnez TMP1, <1
358 |. addiu BASE, BASE, 8
359 |
360 |2:
361 | bne TMP2, RD, >6
362 |3:
363 |. sw BASE, L->top // Store new top.
364 |
365 |->vm_leave_cp:
366 | lw TMP0, SAVE_CFRAME // Restore previous C frame.
367 | move CRET1, r0 // Ok return status for vm_pcall.
368 | sw TMP0, L->cframe
369 |
370 |->vm_leave_unw:
371 | restoreregs_ret
372 |
373 |6:
374 | lw TMP1, L->maxstack
375 | slt AT, TMP2, RD
376 | bnez AT, >7 // Less results wanted?
377 | // More results wanted. Check stack size and fill up results with nil.
378 |. slt AT, BASE, TMP1
379 | beqz AT, >8
380 |. nop
381 | sw TISNIL, HI(BASE)
382 | addiu RD, RD, 8
383 | b <2
384 |. addiu BASE, BASE, 8
385 |
386 |7: // Less results wanted.
387 | subu TMP0, RD, TMP2
388 | subu TMP0, BASE, TMP0 // Either keep top or shrink it.
389 | b <3
390 |. movn BASE, TMP0, TMP2 // LUA_MULTRET+1 case?
391 |
392 |8: // Corner case: need to grow stack for filling up results.
393 | // This can happen if:
394 | // - A C function grows the stack (a lot).
395 | // - The GC shrinks the stack in between.
396 | // - A return back from a lua_call() with (high) nresults adjustment.
397 | load_got lj_state_growstack
398 | move MULTRES, RD
399 | move CARG2, TMP2
400 | call_intern lj_state_growstack // (lua_State *L, int n)
401 |. move CARG1, L
402 | lw TMP2, SAVE_NRES
403 | lw BASE, L->top // Need the (realloced) L->top in BASE.
404 | move RD, MULTRES
405 | b <2
406 |. sll TMP2, TMP2, 3
407 |
408 |->vm_unwind_c: // Unwind C stack, return from vm_pcall.
409 | // (void *cframe, int errcode)
410 | move sp, CARG1
411 | move CRET1, CARG2
412 |->vm_unwind_c_eh: // Landing pad for external unwinder.
413 | lw L, SAVE_L
414 | li TMP0, ~LJ_VMST_C
415 | lw GL:TMP1, L->glref
416 | b ->vm_leave_unw
417 |. sw TMP0, GL:TMP1->vmstate
418 |
419 |->vm_unwind_ff: // Unwind C stack, return from ff pcall.
420 | // (void *cframe)
421 | li AT, -4
422 | and sp, CARG1, AT
423 |->vm_unwind_ff_eh: // Landing pad for external unwinder.
424 | lw L, SAVE_L
425 | lui TMP3, 0x59c0 // TOBIT = 2^52 + 2^51 (float).
426 | li TISNIL, LJ_TNIL
427 | lw BASE, L->base
428 | lw DISPATCH, L->glref // Setup pointer to dispatch table.
429 | mtc1 TMP3, TOBIT
430 | li TMP1, LJ_TFALSE
431 | li_vmstate INTERP
432 | lw PC, FRAME_PC(BASE) // Fetch PC of previous frame.
433 | cvt.d.s TOBIT, TOBIT
434 | addiu RA, BASE, -8 // Results start at BASE-8.
435 | addiu DISPATCH, DISPATCH, GG_G2DISP
436 | sw TMP1, HI(RA) // Prepend false to error message.
437 | st_vmstate
438 | b ->vm_returnc
439 |. li RD, 16 // 2 results: false + error message.
440 |
441 |//-----------------------------------------------------------------------
442 |//-- Grow stack for calls -----------------------------------------------
443 |//-----------------------------------------------------------------------
444 |
445 |->vm_growstack_c: // Grow stack for C function.
446 | b >2
447 |. li CARG2, LUA_MINSTACK
448 |
449 |->vm_growstack_l: // Grow stack for Lua function.
450 | // BASE = new base, RA = BASE+framesize*8, RC = nargs*8, PC = first PC
451 | addu RC, BASE, RC
452 | subu RA, RA, BASE
453 | sw BASE, L->base
454 | addiu PC, PC, 4 // Must point after first instruction.
455 | sw RC, L->top
456 | srl CARG2, RA, 3
457 |2:
458 | // L->base = new base, L->top = top
459 | load_got lj_state_growstack
460 | sw PC, SAVE_PC
461 | call_intern lj_state_growstack // (lua_State *L, int n)
462 |. move CARG1, L
463 | lw BASE, L->base
464 | lw RC, L->top
465 | lw LFUNC:RB, FRAME_FUNC(BASE)
466 | subu RC, RC, BASE
467 | // BASE = new base, RB = LFUNC/CFUNC, RC = nargs*8, FRAME_PC(BASE) = PC
468 | ins_callt // Just retry the call.
469 |
470 |//-----------------------------------------------------------------------
471 |//-- Entry points into the assembler VM ---------------------------------
472 |//-----------------------------------------------------------------------
473 |
474 |->vm_resume: // Setup C frame and resume thread.
475 | // (lua_State *L, TValue *base, int nres1 = 0, ptrdiff_t ef = 0)
476 | saveregs
477 | move L, CARG1
478 | lw DISPATCH, L->glref // Setup pointer to dispatch table.
479 | move BASE, CARG2
480 | lbu TMP1, L->status
481 | sw L, SAVE_L
482 | li PC, FRAME_CP
483 | addiu TMP0, sp, CFRAME_RESUME
484 | addiu DISPATCH, DISPATCH, GG_G2DISP
485 | sw r0, SAVE_NRES
486 | sw r0, SAVE_ERRF
487 | sw TMP0, L->cframe
488 | sw r0, SAVE_CFRAME
489 | beqz TMP1, >3
490 |. sw CARG1, SAVE_PC // Any value outside of bytecode is ok.
491 |
492 | // Resume after yield (like a return).
493 | move RA, BASE
494 | lw BASE, L->base
495 | lw TMP1, L->top
496 | lw PC, FRAME_PC(BASE)
497 | lui TMP3, 0x59c0 // TOBIT = 2^52 + 2^51 (float).
498 | subu RD, TMP1, BASE
499 | mtc1 TMP3, TOBIT
500 | sb r0, L->status
501 | cvt.d.s TOBIT, TOBIT
502 | li_vmstate INTERP
503 | addiu RD, RD, 8
504 | st_vmstate
505 | move MULTRES, RD
506 | andi TMP0, PC, FRAME_TYPE
507 | beqz TMP0, ->BC_RET_Z
508 |. li TISNIL, LJ_TNIL
509 | b ->vm_return
510 |. nop
511 |
512 |->vm_pcall: // Setup protected C frame and enter VM.
513 | // (lua_State *L, TValue *base, int nres1, ptrdiff_t ef)
514 | saveregs
515 | sw CARG4, SAVE_ERRF
516 | b >1
517 |. li PC, FRAME_CP
518 |
519 |->vm_call: // Setup C frame and enter VM.
520 | // (lua_State *L, TValue *base, int nres1)
521 | saveregs
522 | li PC, FRAME_C
523 |
524 |1: // Entry point for vm_pcall above (PC = ftype).
525 | lw TMP1, L:CARG1->cframe
526 | sw CARG3, SAVE_NRES
527 | move L, CARG1
528 | sw CARG1, SAVE_L
529 | move BASE, CARG2
530 | sw sp, L->cframe // Add our C frame to cframe chain.
531 | lw DISPATCH, L->glref // Setup pointer to dispatch table.
532 | sw CARG1, SAVE_PC // Any value outside of bytecode is ok.
533 | sw TMP1, SAVE_CFRAME
534 | addiu DISPATCH, DISPATCH, GG_G2DISP
535 |
536 |3: // Entry point for vm_cpcall/vm_resume (BASE = base, PC = ftype).
537 | lw TMP2, L->base // TMP2 = old base (used in vmeta_call).
538 | lui TMP3, 0x59c0 // TOBIT = 2^52 + 2^51 (float).
539 | lw TMP1, L->top
540 | mtc1 TMP3, TOBIT
541 | addu PC, PC, BASE
542 | subu NARGS8:RC, TMP1, BASE
543 | subu PC, PC, TMP2 // PC = frame delta + frame type
544 | cvt.d.s TOBIT, TOBIT
545 | li_vmstate INTERP
546 | li TISNIL, LJ_TNIL
547 | st_vmstate
548 |
549 |->vm_call_dispatch:
550 | // TMP2 = old base, BASE = new base, RC = nargs*8, PC = caller PC
551 | lw TMP0, FRAME_PC(BASE)
552 | li AT, LJ_TFUNC
553 | bne TMP0, AT, ->vmeta_call
554 |. lw LFUNC:RB, FRAME_FUNC(BASE)
555 |
556 |->vm_call_dispatch_f:
557 | ins_call
558 | // BASE = new base, RB = func, RC = nargs*8, PC = caller PC
559 |
560 |->vm_cpcall: // Setup protected C frame, call C.
561 | // (lua_State *L, lua_CFunction func, void *ud, lua_CPFunction cp)
562 | saveregs
563 | move L, CARG1
564 | lw TMP0, L:CARG1->stack
565 | sw CARG1, SAVE_L
566 | lw TMP1, L->top
567 | sw CARG1, SAVE_PC // Any value outside of bytecode is ok.
568 | subu TMP0, TMP0, TMP1 // Compute -savestack(L, L->top).
569 | lw TMP1, L->cframe
570 | sw sp, L->cframe // Add our C frame to cframe chain.
571 | sw TMP0, SAVE_NRES // Neg. delta means cframe w/o frame.
572 | sw r0, SAVE_ERRF // No error function.
573 | move CFUNCADDR, CARG4
574 | jalr CARG4 // (lua_State *L, lua_CFunction func, void *ud)
575 |. sw TMP1, SAVE_CFRAME
576 | move BASE, CRET1
577 | lw DISPATCH, L->glref // Setup pointer to dispatch table.
578 | li PC, FRAME_CP
579 | bnez CRET1, <3 // Else continue with the call.
580 |. addiu DISPATCH, DISPATCH, GG_G2DISP
581 | b ->vm_leave_cp // No base? Just remove C frame.
582 |. nop
583 |
584 |//-----------------------------------------------------------------------
585 |//-- Metamethod handling ------------------------------------------------
586 |//-----------------------------------------------------------------------
587 |
588 |// The lj_meta_* functions (except for lj_meta_cat) don't reallocate the
589 |// stack, so BASE doesn't need to be reloaded across these calls.
590 |
591 |//-- Continuation dispatch ----------------------------------------------
592 |
593 |->cont_dispatch:
594 | // BASE = meta base, RA = resultptr, RD = (nresults+1)*8
595 | lw TMP0, -16+LO(BASE) // Continuation.
596 | move RB, BASE
597 | move BASE, TMP2 // Restore caller BASE.
598 | lw LFUNC:TMP1, FRAME_FUNC(TMP2)
599#if LJ_HASFFI
600 | sltiu AT, TMP0, 2
601#endif
602 | lw PC, -16+HI(RB) // Restore PC from [cont|PC].
603 | addu TMP2, RA, RD
604 | lw TMP1, LFUNC:TMP1->pc
605#if LJ_HASFFI
606 | bnez AT, >1
607#endif
608 |. sw TISNIL, -8+HI(TMP2) // Ensure one valid arg.
609 | // BASE = base, RA = resultptr, RB = meta base
610 | jr TMP0 // Jump to continuation.
611 |. lw KBASE, PC2PROTO(k)(TMP1)
612 |
613#if LJ_HASFFI
614 |1:
615 | bnez TMP0, ->cont_ffi_callback // cont = 1: return from FFI callback.
616 | // cont = 0: tailcall from C function.
617 |. addiu TMP1, RB, -16
618 | b ->vm_call_tail
619 |. subu RC, TMP1, BASE
620#endif
621 |
622 |->cont_cat: // RA = resultptr, RB = meta base
623 | lw INS, -4(PC)
624 | addiu CARG2, RB, -16
625 | ldc1 f0, 0(RA)
626 | decode_RB8a MULTRES, INS
627 | decode_RA8a RA, INS
628 | decode_RB8b MULTRES
629 | decode_RA8b RA
630 | addu TMP1, BASE, MULTRES
631 | sw BASE, L->base
632 | subu CARG3, CARG2, TMP1
633 | bne TMP1, CARG2, ->BC_CAT_Z
634 |. sdc1 f0, 0(CARG2)
635 | addu RA, BASE, RA
636 | b ->cont_nop
637 |. sdc1 f0, 0(RA)
638 |
639 |//-- Table indexing metamethods -----------------------------------------
640 |
641 |->vmeta_tgets1:
642 | addiu CARG3, DISPATCH, DISPATCH_GL(tmptv)
643 | li TMP0, LJ_TSTR
644 | sw STR:RC, LO(CARG3)
645 | b >1
646 |. sw TMP0, HI(CARG3)
647 |
648 |->vmeta_tgets:
649 | addiu CARG2, DISPATCH, DISPATCH_GL(tmptv)
650 | li TMP0, LJ_TTAB
651 | sw TAB:RB, LO(CARG2)
652 | addiu CARG3, DISPATCH, DISPATCH_GL(tmptv2)
653 | sw TMP0, HI(CARG2)
654 | li TMP1, LJ_TSTR
655 | sw STR:RC, LO(CARG3)
656 | b >1
657 |. sw TMP1, HI(CARG3)
658 |
659 |->vmeta_tgetb: // TMP0 = index
660 | mtc1 TMP0, f0
661 | cvt.d.w f0, f0
662 | addiu CARG3, DISPATCH, DISPATCH_GL(tmptv)
663 | sdc1 f0, 0(CARG3)
664 |
665 |->vmeta_tgetv:
666 |1:
667 | load_got lj_meta_tget
668 | sw BASE, L->base
669 | sw PC, SAVE_PC
670 | call_intern lj_meta_tget // (lua_State *L, TValue *o, TValue *k)
671 |. move CARG1, L
672 | // Returns TValue * (finished) or NULL (metamethod).
673 | beqz CRET1, >3
674 |. addiu TMP1, BASE, -FRAME_CONT
675 | ldc1 f0, 0(CRET1)
676 | ins_next1
677 | sdc1 f0, 0(RA)
678 | ins_next2
679 |
680 |3: // Call __index metamethod.
681 | // BASE = base, L->top = new base, stack = cont/func/t/k
682 | lw BASE, L->top
683 | sw PC, -16+HI(BASE) // [cont|PC]
684 | subu PC, BASE, TMP1
685 | lw LFUNC:RB, FRAME_FUNC(BASE) // Guaranteed to be a function here.
686 | b ->vm_call_dispatch_f
687 |. li NARGS8:RC, 16 // 2 args for func(t, k).
688 |
689 |//-----------------------------------------------------------------------
690 |
691 |->vmeta_tsets1:
692 | addiu CARG3, DISPATCH, DISPATCH_GL(tmptv)
693 | li TMP0, LJ_TSTR
694 | sw STR:RC, LO(CARG3)
695 | b >1
696 |. sw TMP0, HI(CARG3)
697 |
698 |->vmeta_tsets:
699 | addiu CARG2, DISPATCH, DISPATCH_GL(tmptv)
700 | li TMP0, LJ_TTAB
701 | sw TAB:RB, LO(CARG2)
702 | addiu CARG3, DISPATCH, DISPATCH_GL(tmptv2)
703 | sw TMP0, HI(CARG2)
704 | li TMP1, LJ_TSTR
705 | sw STR:RC, LO(CARG3)
706 | b >1
707 |. sw TMP1, HI(CARG3)
708 |
709 |->vmeta_tsetb: // TMP0 = index
710 | mtc1 TMP0, f0
711 | cvt.d.w f0, f0
712 | addiu CARG3, DISPATCH, DISPATCH_GL(tmptv)
713 | sdc1 f0, 0(CARG3)
714 |
715 |->vmeta_tsetv:
716 |1:
717 | load_got lj_meta_tset
718 | sw BASE, L->base
719 | sw PC, SAVE_PC
720 | call_intern lj_meta_tset // (lua_State *L, TValue *o, TValue *k)
721 |. move CARG1, L
722 | // Returns TValue * (finished) or NULL (metamethod).
723 | beqz CRET1, >3
724 |. ldc1 f0, 0(RA)
725 | // NOBARRIER: lj_meta_tset ensures the table is not black.
726 | ins_next1
727 | sdc1 f0, 0(CRET1)
728 | ins_next2
729 |
730 |3: // Call __newindex metamethod.
731 | // BASE = base, L->top = new base, stack = cont/func/t/k/(v)
732 | addiu TMP1, BASE, -FRAME_CONT
733 | lw BASE, L->top
734 | sw PC, -16+HI(BASE) // [cont|PC]
735 | subu PC, BASE, TMP1
736 | lw LFUNC:RB, FRAME_FUNC(BASE) // Guaranteed to be a function here.
737 | sdc1 f0, 16(BASE) // Copy value to third argument.
738 | b ->vm_call_dispatch_f
739 |. li NARGS8:RC, 24 // 3 args for func(t, k, v)
740 |
741 |//-- Comparison metamethods ---------------------------------------------
742 |
743 |->vmeta_comp:
744 | // CARG2, CARG3 are already set by BC_ISLT/BC_ISGE/BC_ISLE/BC_ISGT.
745 | load_got lj_meta_comp
746 | addiu PC, PC, -4
747 | sw BASE, L->base
748 | sw PC, SAVE_PC
749 | decode_OP1 CARG4, INS
750 | call_intern lj_meta_comp // (lua_State *L, TValue *o1, *o2, int op)
751 |. move CARG1, L
752 | // Returns 0/1 or TValue * (metamethod).
753 |3:
754 | sltiu AT, CRET1, 2
755 | beqz AT, ->vmeta_binop
756 | negu TMP2, CRET1
757 |4:
758 | lhu RD, OFS_RD(PC)
759 | addiu PC, PC, 4
760 | lui TMP1, (-(BCBIAS_J*4 >> 16) & 65535)
761 | sll RD, RD, 2
762 | addu RD, RD, TMP1
763 | and RD, RD, TMP2
764 | addu PC, PC, RD
765 |->cont_nop:
766 | ins_next
767 |
768 |->cont_ra: // RA = resultptr
769 | lbu TMP1, -4+OFS_RA(PC)
770 | ldc1 f0, 0(RA)
771 | sll TMP1, TMP1, 3
772 | addu TMP1, BASE, TMP1
773 | b ->cont_nop
774 |. sdc1 f0, 0(TMP1)
775 |
776 |->cont_condt: // RA = resultptr
777 | lw TMP0, HI(RA)
778 | sltiu AT, TMP0, LJ_TISTRUECOND
779 | b <4
780 |. negu TMP2, AT // Branch if result is true.
781 |
782 |->cont_condf: // RA = resultptr
783 | lw TMP0, HI(RA)
784 | sltiu AT, TMP0, LJ_TISTRUECOND
785 | b <4
786 |. addiu TMP2, AT, -1 // Branch if result is false.
787 |
788 |->vmeta_equal:
789 | // CARG2, CARG3, CARG4 are already set by BC_ISEQV/BC_ISNEV.
790 | load_got lj_meta_equal
791 | addiu PC, PC, -4
792 | sw BASE, L->base
793 | sw PC, SAVE_PC
794 | call_intern lj_meta_equal // (lua_State *L, GCobj *o1, *o2, int ne)
795 |. move CARG1, L
796 | // Returns 0/1 or TValue * (metamethod).
797 | b <3
798 |. nop
799 |
800 |->vmeta_equal_cd:
801#if LJ_HASFFI
802 | load_got lj_meta_equal_cd
803 | move CARG2, INS
804 | addiu PC, PC, -4
805 | sw BASE, L->base
806 | sw PC, SAVE_PC
807 | call_intern lj_meta_equal_cd // (lua_State *L, BCIns op)
808 |. move CARG1, L
809 | // Returns 0/1 or TValue * (metamethod).
810 | b <3
811 |. nop
812#endif
813 |
814 |//-- Arithmetic metamethods ---------------------------------------------
815 |
816 |->vmeta_unm:
817 | move CARG4, CARG3
818 |
819 |->vmeta_arith:
820 | load_got lj_meta_arith
821 | decode_OP1 TMP0, INS
822 | sw BASE, L->base
823 | sw PC, SAVE_PC
824 | move CARG2, RA
825 | sw TMP0, ARG5
826 | call_intern lj_meta_arith // (lua_State *L, TValue *ra,*rb,*rc, BCReg op)
827 |. move CARG1, L
828 | // Returns NULL (finished) or TValue * (metamethod).
829 | beqz CRET1, ->cont_nop
830 |. nop
831 |
832 | // Call metamethod for binary op.
833 |->vmeta_binop:
834 | // BASE = old base, CRET1 = new base, stack = cont/func/o1/o2
835 | subu TMP1, CRET1, BASE
836 | sw PC, -16+HI(CRET1) // [cont|PC]
837 | move TMP2, BASE
838 | addiu PC, TMP1, FRAME_CONT
839 | move BASE, CRET1
840 | b ->vm_call_dispatch
841 |. li NARGS8:RC, 16 // 2 args for func(o1, o2).
842 |
843 |->vmeta_len:
844 | // CARG2 already set by BC_LEN.
845#ifdef LUAJIT_ENABLE_LUA52COMPAT
846 | move MULTRES, CARG1
847#endif
848 | load_got lj_meta_len
849 | sw BASE, L->base
850 | sw PC, SAVE_PC
851 | call_intern lj_meta_len // (lua_State *L, TValue *o)
852 |. move CARG1, L
853 | // Returns NULL (retry) or TValue * (metamethod base).
854#ifdef LUAJIT_ENABLE_LUA52COMPAT
855 | bnez CRET1, ->vmeta_binop // Binop call for compatibility.
856 |. nop
857 | b ->BC_LEN_Z
858 |. move CARG1, MULTRES
859#else
860 | b ->vmeta_binop // Binop call for compatibility.
861 |. nop
862#endif
863 |
864 |//-- Call metamethod ----------------------------------------------------
865 |
866 |->vmeta_call: // Resolve and call __call metamethod.
867 | // TMP2 = old base, BASE = new base, RC = nargs*8
868 | load_got lj_meta_call
869 | sw TMP2, L->base // This is the callers base!
870 | addiu CARG2, BASE, -8
871 | sw PC, SAVE_PC
872 | addu CARG3, BASE, RC
873 | move MULTRES, NARGS8:RC
874 | call_intern lj_meta_call // (lua_State *L, TValue *func, TValue *top)
875 |. move CARG1, L
876 | lw LFUNC:RB, FRAME_FUNC(BASE) // Guaranteed to be a function here.
877 | addiu NARGS8:RC, MULTRES, 8 // Got one more argument now.
878 | ins_call
879 |
880 |->vmeta_callt: // Resolve __call for BC_CALLT.
881 | // BASE = old base, RA = new base, RC = nargs*8
882 | load_got lj_meta_call
883 | sw BASE, L->base
884 | addiu CARG2, RA, -8
885 | sw PC, SAVE_PC
886 | addu CARG3, RA, RC
887 | move MULTRES, NARGS8:RC
888 | call_intern lj_meta_call // (lua_State *L, TValue *func, TValue *top)
889 |. move CARG1, L
890 | lw TMP1, FRAME_PC(BASE)
891 | lw LFUNC:RB, FRAME_FUNC(RA) // Guaranteed to be a function here.
892 | b ->BC_CALLT_Z
893 |. addiu NARGS8:RC, MULTRES, 8 // Got one more argument now.
894 |
895 |//-- Argument coercion for 'for' statement ------------------------------
896 |
897 |->vmeta_for:
898 | load_got lj_meta_for
899 | sw BASE, L->base
900 | move CARG2, RA
901 | sw PC, SAVE_PC
902 | move MULTRES, INS
903 | call_intern lj_meta_for // (lua_State *L, TValue *base)
904 |. move CARG1, L
905#if LJ_HASJIT
906 | decode_OP1 TMP0, MULTRES
907 | li AT, BC_JFORI
908#endif
909 | decode_RA8a RA, MULTRES
910 | decode_RD8a RD, MULTRES
911 | decode_RA8b RA
912#if LJ_HASJIT
913 | beq TMP0, AT, =>BC_JFORI
914 |. decode_RD8b RD
915 | b =>BC_FORI
916 |. nop
917#else
918 | b =>BC_FORI
919 |. decode_RD8b RD
920#endif
921 |
922 |//-----------------------------------------------------------------------
923 |//-- Fast functions -----------------------------------------------------
924 |//-----------------------------------------------------------------------
925 |
926 |.macro .ffunc, name
927 |->ff_ .. name:
928 |.endmacro
929 |
930 |.macro .ffunc_1, name
931 |->ff_ .. name:
932 | beqz NARGS8:RC, ->fff_fallback
933 |. lw CARG3, HI(BASE)
934 | lw CARG1, LO(BASE)
935 |.endmacro
936 |
937 |.macro .ffunc_2, name
938 |->ff_ .. name:
939 | sltiu AT, NARGS8:RC, 16
940 | lw CARG3, HI(BASE)
941 | bnez AT, ->fff_fallback
942 |. lw CARG4, 8+HI(BASE)
943 | lw CARG1, LO(BASE)
944 | lw CARG2, 8+LO(BASE)
945 |.endmacro
946 |
947 |.macro .ffunc_n, name // Caveat: has delay slot!
948 |->ff_ .. name:
949 | lw CARG3, HI(BASE)
950 | beqz NARGS8:RC, ->fff_fallback
951 |. ldc1 FARG1, 0(BASE)
952 | sltiu AT, CARG3, LJ_TISNUM
953 | beqz AT, ->fff_fallback
954 |.endmacro
955 |
956 |.macro .ffunc_nn, name // Caveat: has delay slot!
957 |->ff_ .. name:
958 | sltiu AT, NARGS8:RC, 16
959 | lw CARG3, HI(BASE)
960 | bnez AT, ->fff_fallback
961 |. lw CARG4, 8+HI(BASE)
962 | ldc1 FARG1, 0(BASE)
963 | ldc1 FARG2, 8(BASE)
964 | sltiu TMP0, CARG3, LJ_TISNUM
965 | sltiu TMP1, CARG4, LJ_TISNUM
966 | and TMP0, TMP0, TMP1
967 | beqz TMP0, ->fff_fallback
968 |.endmacro
969 |
970 |// Inlined GC threshold check. Caveat: uses TMP0 and TMP1 and has delay slot!
971 |.macro ffgccheck
972 | lw TMP0, DISPATCH_GL(gc.total)(DISPATCH)
973 | lw TMP1, DISPATCH_GL(gc.threshold)(DISPATCH)
974 | subu AT, TMP0, TMP1
975 | bgezal AT, ->fff_gcstep
976 |.endmacro
977 |
978 |//-- Base library: checks -----------------------------------------------
979 |
980 |.ffunc_1 assert
981 | sltiu AT, CARG3, LJ_TISTRUECOND
982 | beqz AT, ->fff_fallback
983 |. addiu RA, BASE, -8
984 | lw PC, FRAME_PC(BASE)
985 | addiu RD, NARGS8:RC, 8 // Compute (nresults+1)*8.
986 | addu TMP2, RA, NARGS8:RC
987 | sw CARG3, HI(RA)
988 | addiu TMP1, BASE, 8
989 | beq BASE, TMP2, ->fff_res // Done if exactly 1 argument.
990 |. sw CARG1, LO(RA)
991 |1:
992 | ldc1 f0, 0(TMP1)
993 | sdc1 f0, -8(TMP1)
994 | bne TMP1, TMP2, <1
995 |. addiu TMP1, TMP1, 8
996 | b ->fff_res
997 |. nop
998 |
999 |.ffunc type
1000 | lw CARG3, HI(BASE)
1001 | li TMP1, LJ_TISNUM
1002 | beqz NARGS8:RC, ->fff_fallback
1003 |. sltiu TMP0, CARG3, LJ_TISNUM
1004 | movz TMP1, CARG3, TMP0
1005 | not TMP1, TMP1
1006 | sll TMP1, TMP1, 3
1007 | addu TMP1, CFUNC:RB, TMP1
1008 | b ->fff_resn
1009 |. ldc1 FRET1, CFUNC:TMP1->upvalue
1010 |
1011 |//-- Base library: getters and setters ---------------------------------
1012 |
1013 |.ffunc_1 getmetatable
1014 | li AT, LJ_TTAB
1015 | bne CARG3, AT, >6
1016 |. li AT, LJ_TUDATA
1017 |1: // Field metatable must be at same offset for GCtab and GCudata!
1018 | lw TAB:CARG1, TAB:CARG1->metatable
1019 |2:
1020 | lw STR:RC, DISPATCH_GL(gcroot[GCROOT_MMNAME+MM_metatable])(DISPATCH)
1021 | beqz TAB:CARG1, ->fff_restv
1022 |. li CARG3, LJ_TNIL
1023 | lw TMP0, TAB:CARG1->hmask
1024 | li CARG3, LJ_TTAB // Use metatable as default result.
1025 | lw TMP1, STR:RC->hash
1026 | lw NODE:TMP2, TAB:CARG1->node
1027 | and TMP1, TMP1, TMP0 // idx = str->hash & tab->hmask
1028 | sll TMP0, TMP1, 5
1029 | sll TMP1, TMP1, 3
1030 | subu TMP1, TMP0, TMP1
1031 | addu NODE:TMP2, NODE:TMP2, TMP1 // node = tab->node + (idx*32-idx*8)
1032 | li AT, LJ_TSTR
1033 |3: // Rearranged logic, because we expect _not_ to find the key.
1034 | lw CARG4, offsetof(Node, key)+HI(NODE:TMP2)
1035 | lw TMP0, offsetof(Node, key)+LO(NODE:TMP2)
1036 | lw NODE:TMP3, NODE:TMP2->next
1037 | bne CARG4, AT, >4
1038 |. lw CARG2, offsetof(Node, val)+HI(NODE:TMP2)
1039 | beq TMP0, STR:RC, >5
1040 |. lw TMP1, offsetof(Node, val)+LO(NODE:TMP2)
1041 |4:
1042 | beqz NODE:TMP3, ->fff_restv // Not found, keep default result.
1043 |. move NODE:TMP2, NODE:TMP3
1044 | b <3
1045 |. nop
1046 |5:
1047 | beq CARG2, TISNIL, ->fff_restv // Ditto for nil value.
1048 |. nop
1049 | move CARG3, CARG2 // Return value of mt.__metatable.
1050 | b ->fff_restv
1051 |. move CARG1, TMP1
1052 |
1053 |6:
1054 | beq CARG3, AT, <1
1055 |. sltiu TMP0, CARG3, LJ_TISNUM
1056 | li TMP1, LJ_TISNUM
1057 | movz TMP1, CARG3, TMP0
1058 | not TMP1, TMP1
1059 | sll TMP1, TMP1, 2
1060 | addu TMP1, DISPATCH, TMP1
1061 | b <2
1062 |. lw TAB:CARG1, DISPATCH_GL(gcroot[GCROOT_BASEMT])(TMP1)
1063 |
1064 |.ffunc_2 setmetatable
1065 | // Fast path: no mt for table yet and not clearing the mt.
1066 | li AT, LJ_TTAB
1067 | bne CARG3, AT, ->fff_fallback
1068 |. addiu CARG4, CARG4, -LJ_TTAB
1069 | lw TAB:TMP1, TAB:CARG1->metatable
1070 | lbu TMP3, TAB:CARG1->marked
1071 | or AT, CARG4, TAB:TMP1
1072 | bnez AT, ->fff_fallback
1073 |. andi AT, TMP3, LJ_GC_BLACK // isblack(table)
1074 | beqz AT, ->fff_restv
1075 |. sw TAB:CARG2, TAB:CARG1->metatable
1076 | barrierback TAB:CARG1, TMP3, TMP0, ->fff_restv
1077 |
1078 |.ffunc rawget
1079 | lw CARG4, HI(BASE)
1080 | sltiu AT, NARGS8:RC, 16
1081 | lw TAB:CARG2, LO(BASE)
1082 | load_got lj_tab_get
1083 | addiu CARG4, CARG4, -LJ_TTAB
1084 | or AT, AT, CARG4
1085 | bnez AT, ->fff_fallback
1086 | addiu CARG3, BASE, 8
1087 | call_intern lj_tab_get // (lua_State *L, GCtab *t, cTValue *key)
1088 |. move CARG1, L
1089 | // Returns cTValue *.
1090 | b ->fff_resn
1091 |. ldc1 FRET1, 0(CRET1)
1092 |
1093 |//-- Base library: conversions ------------------------------------------
1094 |
1095 |.ffunc tonumber
1096 | // Only handles the number case inline (without a base argument).
1097 | lw CARG1, HI(BASE)
1098 | xori AT, NARGS8:RC, 8
1099 | sltiu CARG1, CARG1, LJ_TISNUM
1100 | movn CARG1, r0, AT
1101 | beqz CARG1, ->fff_fallback // Exactly one number argument.
1102 |. ldc1 FRET1, 0(BASE)
1103 | b ->fff_resn
1104 |. nop
1105 |
1106 |.ffunc_1 tostring
1107 | // Only handles the string or number case inline.
1108 | li AT, LJ_TSTR
1109 | // A __tostring method in the string base metatable is ignored.
1110 | beq CARG3, AT, ->fff_restv // String key?
1111 | // Handle numbers inline, unless a number base metatable is present.
1112 |. lw TMP1, DISPATCH_GL(gcroot[GCROOT_BASEMT_NUM])(DISPATCH)
1113 | sltiu TMP0, CARG3, LJ_TISNUM
1114 | sltiu TMP1, TMP1, 1
1115 | and TMP0, TMP0, TMP1
1116 | beqz TMP0, ->fff_fallback
1117 |. sw BASE, L->base // Add frame since C call can throw.
1118 | ffgccheck
1119 |. sw PC, SAVE_PC // Redundant (but a defined value).
1120 | load_got lj_str_fromnum
1121 | move CARG1, L
1122 | call_intern lj_str_fromnum // (lua_State *L, lua_Number *np)
1123 |. move CARG2, BASE
1124 | // Returns GCstr *.
1125 | li CARG3, LJ_TSTR
1126 | b ->fff_restv
1127 |. move CARG1, CRET1
1128 |
1129 |//-- Base library: iterators -------------------------------------------
1130 |
1131 |.ffunc next
1132 | lw CARG1, HI(BASE)
1133 | lw TAB:CARG2, LO(BASE)
1134 | beqz NARGS8:RC, ->fff_fallback
1135 |. addu TMP2, BASE, NARGS8:RC
1136 | li AT, LJ_TTAB
1137 | sw TISNIL, HI(TMP2) // Set missing 2nd arg to nil.
1138 | bne CARG1, AT, ->fff_fallback
1139 |. lw PC, FRAME_PC(BASE)
1140 | load_got lj_tab_next
1141 | sw BASE, L->base // Add frame since C call can throw.
1142 | sw BASE, L->top // Dummy frame length is ok.
1143 | addiu CARG3, BASE, 8
1144 | sw PC, SAVE_PC
1145 | call_intern lj_tab_next // (lua_State *L, GCtab *t, TValue *key)
1146 |. move CARG1, L
1147 | // Returns 0 at end of traversal.
1148 | beqz CRET1, ->fff_restv // End of traversal: return nil.
1149 |. li CARG3, LJ_TNIL
1150 | ldc1 f0, 8(BASE) // Copy key and value to results.
1151 | addiu RA, BASE, -8
1152 | ldc1 f2, 16(BASE)
1153 | li RD, (2+1)*8
1154 | sdc1 f0, 0(RA)
1155 | b ->fff_res
1156 |. sdc1 f2, 8(RA)
1157 |
1158 |.ffunc_1 pairs
1159 | li AT, LJ_TTAB
1160 | bne CARG3, AT, ->fff_fallback
1161 |. lw PC, FRAME_PC(BASE)
1162#ifdef LUAJIT_ENABLE_LUA52COMPAT
1163 | lw TAB:TMP2, TAB:CARG1->metatable
1164 | ldc1 f0, CFUNC:RB->upvalue[0]
1165 | bnez TAB:TMP2, ->fff_fallback
1166#else
1167 | ldc1 f0, CFUNC:RB->upvalue[0]
1168#endif
1169 |. addiu RA, BASE, -8
1170 | sw TISNIL, 8+HI(BASE)
1171 | li RD, (3+1)*8
1172 | b ->fff_res
1173 |. sdc1 f0, 0(RA)
1174 |
1175 |.ffunc ipairs_aux
1176 | sltiu AT, NARGS8:RC, 16
1177 | lw CARG3, HI(BASE)
1178 | lw TAB:CARG1, LO(BASE)
1179 | lw CARG4, 8+HI(BASE)
1180 | bnez AT, ->fff_fallback
1181 |. ldc1 FARG2, 8(BASE)
1182 | addiu CARG3, CARG3, -LJ_TTAB
1183 | sltiu AT, CARG4, LJ_TISNUM
1184 | li TMP0, 1
1185 | movn AT, r0, CARG3
1186 | mtc1 TMP0, FARG1
1187 | beqz AT, ->fff_fallback
1188 |. lw PC, FRAME_PC(BASE)
1189 | cvt.w.d FRET1, FARG2
1190 | cvt.d.w FARG1, FARG1
1191 | lw TMP0, TAB:CARG1->asize
1192 | lw TMP1, TAB:CARG1->array
1193 | mfc1 TMP2, FRET1
1194 | addiu RA, BASE, -8
1195 | add.d FARG2, FARG2, FARG1
1196 | addiu TMP2, TMP2, 1
1197 | sltu AT, TMP2, TMP0
1198 | sll TMP3, TMP2, 3
1199 | addu TMP3, TMP1, TMP3
1200 | beqz AT, >2 // Not in array part?
1201 |. sdc1 FARG2, 0(RA)
1202 | lw TMP2, HI(TMP3)
1203 | ldc1 f0, 0(TMP3)
1204 |1:
1205 | beq TMP2, TISNIL, ->fff_res // End of iteration, return 0 results.
1206 |. li RD, (0+1)*8
1207 | li RD, (2+1)*8
1208 | b ->fff_res
1209 |. sdc1 f0, 8(RA)
1210 |2: // Check for empty hash part first. Otherwise call C function.
1211 | lw TMP0, TAB:CARG1->hmask
1212 | load_got lj_tab_getinth
1213 | beqz TMP0, ->fff_res
1214 |. li RD, (0+1)*8
1215 | call_intern lj_tab_getinth // (GCtab *t, int32_t key)
1216 |. move CARG2, TMP2
1217 | // Returns cTValue * or NULL.
1218 | beqz CRET1, ->fff_res
1219 |. li RD, (0+1)*8
1220 | lw TMP2, HI(CRET1)
1221 | b <1
1222 |. ldc1 f0, 0(CRET1)
1223 |
1224 |.ffunc_1 ipairs
1225 | li AT, LJ_TTAB
1226 | bne CARG3, AT, ->fff_fallback
1227 |. lw PC, FRAME_PC(BASE)
1228#ifdef LUAJIT_ENABLE_LUA52COMPAT
1229 | lw TAB:TMP2, TAB:CARG1->metatable
1230 | ldc1 f0, CFUNC:RB->upvalue[0]
1231 | bnez TAB:TMP2, ->fff_fallback
1232#else
1233 | ldc1 f0, CFUNC:RB->upvalue[0]
1234#endif
1235 |. addiu RA, BASE, -8
1236 | sw r0, 8+HI(BASE)
1237 | sw r0, 8+LO(BASE)
1238 | li RD, (3+1)*8
1239 | b ->fff_res
1240 |. sdc1 f0, 0(RA)
1241 |
1242 |//-- Base library: catch errors ----------------------------------------
1243 |
1244 |.ffunc pcall
1245 | lbu TMP3, DISPATCH_GL(hookmask)(DISPATCH)
1246 | beqz NARGS8:RC, ->fff_fallback
1247 | move TMP2, BASE
1248 | addiu BASE, BASE, 8
1249 | // Remember active hook before pcall.
1250 | srl TMP3, TMP3, HOOK_ACTIVE_SHIFT
1251 | andi TMP3, TMP3, 1
1252 | addiu PC, TMP3, 8+FRAME_PCALL
1253 | b ->vm_call_dispatch
1254 |. addiu NARGS8:RC, NARGS8:RC, -8
1255 |
1256 |.ffunc xpcall
1257 | sltiu AT, NARGS8:RC, 16
1258 | lw CARG4, 8+HI(BASE)
1259 | bnez AT, ->fff_fallback
1260 |. ldc1 FARG2, 8(BASE)
1261 | ldc1 FARG1, 0(BASE)
1262 | lbu TMP1, DISPATCH_GL(hookmask)(DISPATCH)
1263 | li AT, LJ_TFUNC
1264 | move TMP2, BASE
1265 | bne CARG4, AT, ->fff_fallback // Traceback must be a function.
1266 | addiu BASE, BASE, 16
1267 | // Remember active hook before pcall.
1268 | srl TMP3, TMP3, HOOK_ACTIVE_SHIFT
1269 | sdc1 FARG2, 0(TMP2) // Swap function and traceback.
1270 | andi TMP3, TMP3, 1
1271 | sdc1 FARG1, 8(TMP2)
1272 | addiu PC, TMP3, 16+FRAME_PCALL
1273 | b ->vm_call_dispatch
1274 |. addiu NARGS8:RC, NARGS8:RC, -16
1275 |
1276 |//-- Coroutine library --------------------------------------------------
1277 |
1278 |.macro coroutine_resume_wrap, resume
1279 |.if resume
1280 |.ffunc_1 coroutine_resume
1281 | li AT, LJ_TTHREAD
1282 | bne CARG3, AT, ->fff_fallback
1283 |.else
1284 |.ffunc coroutine_wrap_aux
1285 | lw L:CARG1, CFUNC:RB->upvalue[0].gcr
1286 |.endif
1287 | lbu TMP0, L:CARG1->status
1288 | lw TMP1, L:CARG1->cframe
1289 | lw CARG2, L:CARG1->top
1290 | lw TMP2, L:CARG1->base
1291 | addiu TMP3, TMP0, -LUA_YIELD
1292 | bgtz TMP3, ->fff_fallback // st > LUA_YIELD?
1293 |. xor TMP2, TMP2, CARG2
1294 | bnez TMP1, ->fff_fallback // cframe != 0?
1295 |. or AT, TMP2, TMP0
1296 | lw TMP0, L:CARG1->maxstack
1297 | beqz AT, ->fff_fallback // base == top && st == 0?
1298 |. lw PC, FRAME_PC(BASE)
1299 | addu TMP2, CARG2, NARGS8:RC
1300 | sltu AT, TMP0, TMP2
1301 | bnez AT, ->fff_fallback // Stack overflow?
1302 |. sw PC, SAVE_PC
1303 | sw BASE, L->base
1304 |1:
1305 |.if resume
1306 | addiu BASE, BASE, 8 // Keep resumed thread in stack for GC.
1307 | addiu NARGS8:RC, NARGS8:RC, -8
1308 | addiu TMP2, TMP2, -8
1309 |.endif
1310 | sw TMP2, L:CARG1->top
1311 | addu TMP1, BASE, NARGS8:RC
1312 | move CARG3, CARG2
1313 | sw BASE, L->top
1314 |2: // Move args to coroutine.
1315 | ldc1 f0, 0(BASE)
1316 | sltu AT, BASE, TMP1
1317 | beqz AT, >3
1318 |. addiu BASE, BASE, 8
1319 | sdc1 f0, 0(CARG3)
1320 | b <2
1321 |. addiu CARG3, CARG3, 8
1322 |3:
1323 | bal ->vm_resume // (lua_State *L, TValue *base, 0, 0)
1324 |. move L:RA, L:CARG1
1325 | // Returns thread status.
1326 |4:
1327 | lw TMP2, L:RA->base
1328 | sltiu AT, CRET1, LUA_YIELD+1
1329 | lw TMP3, L:RA->top
1330 | li_vmstate INTERP
1331 | lw BASE, L->base
1332 | st_vmstate
1333 | beqz AT, >8
1334 |. subu RD, TMP3, TMP2
1335 | lw TMP0, L->maxstack
1336 | beqz RD, >6 // No results?
1337 |. addu TMP1, BASE, RD
1338 | sltu AT, TMP0, TMP1
1339 | bnez AT, >9 // Need to grow stack?
1340 |. addu TMP3, TMP2, RD
1341 | sw TMP2, L:RA->top // Clear coroutine stack.
1342 | move TMP1, BASE
1343 |5: // Move results from coroutine.
1344 | ldc1 f0, 0(TMP2)
1345 | addiu TMP2, TMP2, 8
1346 | sltu AT, TMP2, TMP3
1347 | sdc1 f0, 0(TMP1)
1348 | bnez AT, <5
1349 |. addiu TMP1, TMP1, 8
1350 |6:
1351 | andi TMP0, PC, FRAME_TYPE
1352 |.if resume
1353 | li TMP1, LJ_TTRUE
1354 | addiu RA, BASE, -8
1355 | sw TMP1, -8+HI(BASE) // Prepend true to results.
1356 | addiu RD, RD, 16
1357 |.else
1358 | move RA, BASE
1359 | addiu RD, RD, 8
1360 |.endif
1361 |7:
1362 | sw PC, SAVE_PC
1363 | beqz TMP0, ->BC_RET_Z
1364 |. move MULTRES, RD
1365 | b ->vm_return
1366 |. nop
1367 |
1368 |8: // Coroutine returned with error (at co->top-1).
1369 |.if resume
1370 | addiu TMP3, TMP3, -8
1371 | li TMP1, LJ_TFALSE
1372 | ldc1 f0, 0(TMP3)
1373 | sw TMP3, L:RA->top // Remove error from coroutine stack.
1374 | li RD, (2+1)*8
1375 | sw TMP1, -8+HI(BASE) // Prepend false to results.
1376 | addiu RA, BASE, -8
1377 | sdc1 f0, 0(BASE) // Copy error message.
1378 | b <7
1379 |. andi TMP0, PC, FRAME_TYPE
1380 |.else
1381 | load_got lj_ffh_coroutine_wrap_err
1382 | move CARG2, L:RA
1383 | call_intern lj_ffh_coroutine_wrap_err // (lua_State *L, lua_State *co)
1384 |. move CARG1, L
1385 |.endif
1386 |
1387 |9: // Handle stack expansion on return from yield.
1388 | load_got lj_state_growstack
1389 | srl CARG2, RD, 3
1390 | call_intern lj_state_growstack // (lua_State *L, int n)
1391 |. move CARG1, L
1392 | b <4
1393 |. li CRET1, 0
1394 |.endmacro
1395 |
1396 | coroutine_resume_wrap 1 // coroutine.resume
1397 | coroutine_resume_wrap 0 // coroutine.wrap
1398 |
1399 |.ffunc coroutine_yield
1400 | lw TMP0, L->cframe
1401 | addu TMP1, BASE, NARGS8:RC
1402 | sw BASE, L->base
1403 | andi TMP0, TMP0, CFRAME_RESUME
1404 | sw TMP1, L->top
1405 | beqz TMP0, ->fff_fallback
1406 |. li CRET1, LUA_YIELD
1407 | sw r0, L->cframe
1408 | b ->vm_leave_unw
1409 |. sb CRET1, L->status
1410 |
1411 |//-- Math library -------------------------------------------------------
1412 |
1413 |.ffunc_n math_abs
1414 |. abs.d FRET1, FARG1
1415 |->fff_resn:
1416 | lw PC, FRAME_PC(BASE)
1417 | addiu RA, BASE, -8
1418 | b ->fff_res1
1419 |. sdc1 FRET1, -8(BASE)
1420 |
1421 |->fff_restv:
1422 | // CARG3/CARG1 = TValue result.
1423 | lw PC, FRAME_PC(BASE)
1424 | sw CARG3, -8+HI(BASE)
1425 | addiu RA, BASE, -8
1426 | sw CARG1, -8+LO(BASE)
1427 |->fff_res1:
1428 | // RA = results, PC = return.
1429 | li RD, (1+1)*8
1430 |->fff_res:
1431 | // RA = results, RD = (nresults+1)*8, PC = return.
1432 | andi TMP0, PC, FRAME_TYPE
1433 | bnez TMP0, ->vm_return
1434 |. move MULTRES, RD
1435 | lw INS, -4(PC)
1436 | decode_RB8a RB, INS
1437 | decode_RB8b RB
1438 |5:
1439 | sltu AT, RD, RB
1440 | bnez AT, >6 // More results expected?
1441 |. decode_RA8a TMP0, INS
1442 | decode_RA8b TMP0
1443 | ins_next1
1444 | // Adjust BASE. KBASE is assumed to be set for the calling frame.
1445 | subu BASE, RA, TMP0
1446 | ins_next2
1447 |
1448 |6: // Fill up results with nil.
1449 | addu TMP1, RA, RD
1450 | addiu RD, RD, 8
1451 | b <5
1452 |. sw TISNIL, -8+HI(TMP1)
1453 |
1454 |.macro math_extern, func
1455 |->ff_math_ .. func:
1456 | lw CARG3, HI(BASE)
1457 | beqz NARGS8:RC, ->fff_fallback
1458 |. load_got func
1459 | sltiu AT, CARG3, LJ_TISNUM
1460 | beqz AT, ->fff_fallback
1461 |. nop
1462 | call_extern
1463 |. ldc1 FARG1, 0(BASE)
1464 | b ->fff_resn
1465 |. nop
1466 |.endmacro
1467 |
1468 |.macro math_extern2, func
1469 | .ffunc_nn math_ .. func
1470 |. load_got func
1471 | call_extern
1472 |. nop
1473 | b ->fff_resn
1474 |. nop
1475 |.endmacro
1476 |
1477 |.macro math_round, func
1478 | .ffunc_n math_ .. func
1479 |. nop
1480 | bal ->vm_ .. func
1481 |. nop
1482 | b ->fff_resn
1483 |. nop
1484 |.endmacro
1485 |
1486 | math_round floor
1487 | math_round ceil
1488 |
1489 | math_extern log
1490 | math_extern log10
1491 | math_extern exp
1492 | math_extern sin
1493 | math_extern cos
1494 | math_extern tan
1495 | math_extern asin
1496 | math_extern acos
1497 | math_extern atan
1498 | math_extern sinh
1499 | math_extern cosh
1500 | math_extern tanh
1501 | math_extern2 pow
1502 | math_extern2 atan2
1503 | math_extern2 fmod
1504 |
1505 |.ffunc_n math_sqrt
1506 |. sqrt.d FRET1, FARG1
1507 | b ->fff_resn
1508 |. nop
1509 |
1510 |->ff_math_deg:
1511 |.ffunc_n math_rad
1512 |. ldc1 FARG2, CFUNC:RB->upvalue[0]
1513 | b ->fff_resn
1514 |. mul.d FRET1, FARG1, FARG2
1515 |
1516 |.ffunc_nn math_ldexp
1517 | cvt.w.d FARG2, FARG2
1518 | load_got ldexp
1519 | mfc1 CARG3, FARG2
1520 | call_extern
1521 |. nop
1522 | b ->fff_resn
1523 |. nop
1524 |
1525 |.ffunc_n math_frexp
1526 | load_got frexp
1527 | lw PC, FRAME_PC(BASE)
1528 | call_extern
1529 |. addiu CARG3, DISPATCH, DISPATCH_GL(tmptv)
1530 | lw TMP1, DISPATCH_GL(tmptv)(DISPATCH)
1531 | addiu RA, BASE, -8
1532 | mtc1 TMP1, FARG2
1533 | sdc1 FRET1, 0(RA)
1534 | cvt.d.w FARG2, FARG2
1535 | sdc1 FARG2, 8(RA)
1536 | b ->fff_res
1537 |. li RD, (2+1)*8
1538 |
1539 |.ffunc_n math_modf
1540 | load_got modf
1541 | lw PC, FRAME_PC(BASE)
1542 | call_extern
1543 |. addiu CARG3, BASE, -8
1544 | addiu RA, BASE, -8
1545 | sdc1 FRET1, 0(BASE)
1546 | b ->fff_res
1547 |. li RD, (2+1)*8
1548 |
1549 |.macro math_minmax, name, ismax
1550 |->ff_ .. name:
1551 | lw CARG3, HI(BASE)
1552 | beqz NARGS8:RC, ->fff_fallback
1553 |. ldc1 FRET1, 0(BASE)
1554 | sltiu AT, CARG3, LJ_TISNUM
1555 | beqz AT, ->fff_fallback
1556 |. addu TMP2, BASE, NARGS8:RC
1557 | addiu TMP1, BASE, 8
1558 | beq TMP1, TMP2, ->fff_resn
1559 |1:
1560 |. lw CARG3, HI(TMP1)
1561 | ldc1 FARG1, 0(TMP1)
1562 | addiu TMP1, TMP1, 8
1563 | sltiu AT, CARG3, LJ_TISNUM
1564 | beqz AT, ->fff_fallback
1565 |.if ismax
1566 |. c.olt.d FARG1, FRET1
1567 |.else
1568 |. c.olt.d FRET1, FARG1
1569 |.endif
1570 | bne TMP1, TMP2, <1
1571 |. movf.d FRET1, FARG1
1572 | b ->fff_resn
1573 |. nop
1574 |.endmacro
1575 |
1576 | math_minmax math_min, 0
1577 | math_minmax math_max, 1
1578 |
1579 |//-- String library -----------------------------------------------------
1580 |
1581 |.ffunc_1 string_len
1582 | li AT, LJ_TSTR
1583 | bne CARG3, AT, ->fff_fallback
1584 |. nop
1585 | b ->fff_resi
1586 |. lw CRET1, STR:CARG1->len
1587 |
1588 |.ffunc string_byte // Only handle the 1-arg case here.
1589 | lw CARG3, HI(BASE)
1590 | lw STR:CARG1, LO(BASE)
1591 | xori AT, NARGS8:RC, 8
1592 | addiu CARG3, CARG3, -LJ_TSTR
1593 | or AT, AT, CARG3
1594 | bnez AT, ->fff_fallback // Need exactly 1 string argument.
1595 |. nop
1596 | lw TMP0, STR:CARG1->len
1597 | lbu TMP1, STR:CARG1[1] // Access is always ok (NUL at end).
1598 | addiu RA, BASE, -8
1599 | sltu RD, r0, TMP0
1600 | mtc1 TMP1, f0
1601 | addiu RD, RD, 1
1602 | cvt.d.w f0, f0
1603 | lw PC, FRAME_PC(BASE)
1604 | sll RD, RD, 3 // RD = ((str->len != 0)+1)*8
1605 | b ->fff_res
1606 |. sdc1 f0, 0(RA)
1607 |
1608 |.ffunc string_char // Only handle the 1-arg case here.
1609 | ffgccheck
1610 | lw CARG3, HI(BASE)
1611 | ldc1 FARG1, 0(BASE)
1612 | li AT, 8
1613 | bne NARGS8:RC, AT, ->fff_fallback // Exactly 1 argument.
1614 |. sltiu AT, CARG3, LJ_TISNUM
1615 | beqz AT, ->fff_fallback
1616 |. li CARG3, 1
1617 | cvt.w.d FARG1, FARG1
1618 | addiu CARG2, sp, ARG5_OFS
1619 | sltiu AT, TMP0, 256
1620 | mfc1 TMP0, FARG1
1621 | beqz AT, ->fff_fallback
1622 |. sw TMP0, ARG5
1623 |->fff_newstr:
1624 | load_got lj_str_new
1625 | sw BASE, L->base
1626 | sw PC, SAVE_PC
1627 | call_intern lj_str_new // (lua_State *L, char *str, size_t l)
1628 |. move CARG1, L
1629 | // Returns GCstr *.
1630 | lw BASE, L->base
1631 | move CARG1, CRET1
1632 | b ->fff_restv
1633 |. li CARG3, LJ_TSTR
1634 |
1635 |.ffunc string_sub
1636 | ffgccheck
1637 | addiu AT, NARGS8:RC, -16
1638 | lw CARG3, 16+HI(BASE)
1639 | ldc1 f0, 16(BASE)
1640 | lw TMP0, HI(BASE)
1641 | lw STR:CARG1, LO(BASE)
1642 | bltz AT, ->fff_fallback
1643 | lw CARG2, 8+HI(BASE)
1644 | ldc1 f2, 8(BASE)
1645 | beqz AT, >1
1646 |. li CARG4, -1
1647 | cvt.w.d f0, f0
1648 | sltiu AT, CARG3, LJ_TISNUM
1649 | beqz AT, ->fff_fallback
1650 |. mfc1 CARG4, f0
1651 |1:
1652 | sltiu AT, CARG2, LJ_TISNUM
1653 | beqz AT, ->fff_fallback
1654 |. li AT, LJ_TSTR
1655 | cvt.w.d f2, f2
1656 | bne TMP0, AT, ->fff_fallback
1657 |. lw CARG2, STR:CARG1->len
1658 | mfc1 CARG3, f2
1659 | // STR:CARG1 = str, CARG2 = str->len, CARG3 = start, CARG4 = end
1660 | slt AT, CARG4, r0
1661 | addiu TMP0, CARG2, 1
1662 | addu TMP1, CARG4, TMP0
1663 | slt TMP3, CARG3, r0
1664 | movn CARG4, TMP1, AT // if (end < 0) end += len+1
1665 | addu TMP1, CARG3, TMP0
1666 | movn CARG3, TMP1, TMP3 // if (start < 0) start += len+1
1667 | li TMP2, 1
1668 | slt AT, CARG4, r0
1669 | slt TMP3, r0, CARG3
1670 | movn CARG4, r0, AT // if (end < 0) end = 0
1671 | movz CARG3, TMP2, TMP3 // if (start < 1) start = 1
1672 | slt AT, CARG2, CARG4
1673 | movn CARG4, CARG2, AT // if (end > len) end = len
1674 | addu CARG2, STR:CARG1, CARG3
1675 | subu CARG3, CARG4, CARG3 // len = end - start
1676 | addiu CARG2, CARG2, sizeof(GCstr)-1
1677 | bgez CARG3, ->fff_newstr
1678 |. addiu CARG3, CARG3, 1 // len++
1679 |->fff_emptystr: // Return empty string.
1680 | addiu STR:CARG1, DISPATCH, DISPATCH_GL(strempty)
1681 | b ->fff_restv
1682 |. li CARG3, LJ_TSTR
1683 |
1684 |.ffunc string_rep // Only handle the 1-char case inline.
1685 | ffgccheck
1686 | lw TMP0, HI(BASE)
1687 | sltiu AT, NARGS8:RC, 16
1688 | lw CARG4, 8+HI(BASE)
1689 | lw STR:CARG1, LO(BASE)
1690 | addiu TMP0, TMP0, -LJ_TSTR
1691 | ldc1 f0, 8(BASE)
1692 | or AT, AT, TMP0
1693 | bnez AT, ->fff_fallback
1694 |. sltiu AT, CARG4, LJ_TISNUM
1695 | cvt.w.d f0, f0
1696 | beqz AT, ->fff_fallback
1697 |. lw TMP0, STR:CARG1->len
1698 | mfc1 CARG3, f0
1699 | lw TMP1, DISPATCH_GL(tmpbuf.sz)(DISPATCH)
1700 | li AT, 1
1701 | blez CARG3, ->fff_emptystr // Count <= 0?
1702 |. sltu AT, AT, TMP0
1703 | beqz TMP0, ->fff_emptystr // Zero length string?
1704 |. sltu TMP0, TMP1, CARG3
1705 | or AT, AT, TMP0
1706 | lw CARG2, DISPATCH_GL(tmpbuf.buf)(DISPATCH)
1707 | bnez AT, ->fff_fallback // Fallback for > 1-char strings.
1708 |. lbu TMP0, STR:CARG1[1]
1709 | addu TMP2, CARG2, CARG3
1710 |1: // Fill buffer with char. Yes, this is suboptimal code (do you care?).
1711 | addiu TMP2, TMP2, -1
1712 | sltu AT, CARG2, TMP2
1713 | bnez AT, <1
1714 |. sb TMP0, 0(TMP2)
1715 | b ->fff_newstr
1716 |. nop
1717 |
1718 |.ffunc string_reverse
1719 | ffgccheck
1720 | lw CARG3, HI(BASE)
1721 | lw STR:CARG1, LO(BASE)
1722 | beqz NARGS8:RC, ->fff_fallback
1723 |. li AT, LJ_TSTR
1724 | bne CARG3, AT, ->fff_fallback
1725 |. lw TMP1, DISPATCH_GL(tmpbuf.sz)(DISPATCH)
1726 | lw CARG3, STR:CARG1->len
1727 | addiu CARG1, STR:CARG1, #STR
1728 | lw CARG2, DISPATCH_GL(tmpbuf.buf)(DISPATCH)
1729 | sltu AT, TMP1, CARG3
1730 | bnez AT, ->fff_fallback
1731 |. addu TMP3, CARG1, CARG3
1732 | addu CARG4, CARG2, CARG3
1733 |1: // Reverse string copy.
1734 | lbu TMP1, 0(CARG1)
1735 | sltu AT, CARG1, TMP3
1736 | beqz AT, ->fff_newstr
1737 |. addiu CARG1, CARG1, 1
1738 | addiu CARG4, CARG4, -1
1739 | b <1
1740 | sb TMP1, 0(CARG4)
1741 |
1742 |.macro ffstring_case, name, lo
1743 | .ffunc name
1744 | ffgccheck
1745 | lw CARG3, HI(BASE)
1746 | lw STR:CARG1, LO(BASE)
1747 | beqz NARGS8:RC, ->fff_fallback
1748 |. li AT, LJ_TSTR
1749 | bne CARG3, AT, ->fff_fallback
1750 |. lw TMP1, DISPATCH_GL(tmpbuf.sz)(DISPATCH)
1751 | lw CARG3, STR:CARG1->len
1752 | addiu CARG1, STR:CARG1, #STR
1753 | lw CARG2, DISPATCH_GL(tmpbuf.buf)(DISPATCH)
1754 | sltu AT, TMP1, CARG3
1755 | bnez AT, ->fff_fallback
1756 |. addu TMP3, CARG1, CARG3
1757 | move CARG4, CARG2
1758 |1: // ASCII case conversion.
1759 | lbu TMP1, 0(CARG1)
1760 | sltu AT, CARG1, TMP3
1761 | beqz AT, ->fff_newstr
1762 |. addiu TMP0, TMP1, -lo
1763 | xori TMP2, TMP1, 0x20
1764 | sltiu AT, TMP0, 26
1765 | movn TMP1, TMP2, AT
1766 | addiu CARG1, CARG1, 1
1767 | sb TMP1, 0(CARG4)
1768 | b <1
1769 |. addiu CARG4, CARG4, 1
1770 |.endmacro
1771 |
1772 |ffstring_case string_lower, 65
1773 |ffstring_case string_upper, 97
1774 |
1775 |//-- Table library ------------------------------------------------------
1776 |
1777 |.ffunc_1 table_getn
1778 | li AT, LJ_TTAB
1779 | bne CARG3, AT, ->fff_fallback
1780 |. load_got lj_tab_len
1781 | call_intern lj_tab_len // (GCtab *t)
1782 |. nop
1783 | // Returns uint32_t (but less than 2^31).
1784 | b ->fff_resi
1785 |. nop
1786 |
1787 |//-- Bit library --------------------------------------------------------
1788 |
1789 |.macro .ffunc_bit, name
1790 | .ffunc_n bit_..name
1791 |. add.d FARG1, FARG1, TOBIT
1792 | mfc1 CRET1, FARG1
1793 |.endmacro
1794 |
1795 |.macro .ffunc_bit_op, name, ins
1796 | .ffunc_bit name
1797 | addiu TMP1, BASE, 8
1798 | addu TMP2, BASE, NARGS8:RC
1799 |1:
1800 | lw CARG4, HI(TMP1)
1801 | beq TMP1, TMP2, ->fff_resi
1802 |. ldc1 FARG1, 0(TMP1)
1803 | sltiu AT, CARG4, LJ_TISNUM
1804 | beqz AT, ->fff_fallback
1805 | add.d FARG1, FARG1, TOBIT
1806 | mfc1 CARG2, FARG1
1807 | ins CRET1, CRET1, CARG2
1808 | b <1
1809 |. addiu TMP1, TMP1, 8
1810 |.endmacro
1811 |
1812 |.ffunc_bit_op band, and
1813 |.ffunc_bit_op bor, or
1814 |.ffunc_bit_op bxor, xor
1815 |
1816 |.ffunc_bit bswap
1817 | srl TMP0, CRET1, 24
1818 | srl TMP2, CRET1, 8
1819 | sll TMP1, CRET1, 24
1820 | andi TMP2, TMP2, 0xff00
1821 | or TMP0, TMP0, TMP1
1822 | andi CRET1, CRET1, 0xff00
1823 | or TMP0, TMP0, TMP2
1824 | sll CRET1, CRET1, 8
1825 | b ->fff_resi
1826 |. or CRET1, TMP0, CRET1
1827 |
1828 |.ffunc_bit bnot
1829 | b ->fff_resi
1830 |. not CRET1, CRET1
1831 |
1832 |.macro .ffunc_bit_sh, name, ins, shmod
1833 | .ffunc_nn bit_..name
1834 |. add.d FARG1, FARG1, TOBIT
1835 | add.d FARG2, FARG2, TOBIT
1836 | mfc1 CARG1, FARG1
1837 | mfc1 CARG2, FARG2
1838 |.if shmod == 1
1839 | li AT, 32
1840 | subu TMP0, AT, CARG2
1841 | sllv CARG2, CARG1, CARG2
1842 | srlv CARG1, CARG1, TMP0
1843 |.elif shmod == 2
1844 | li AT, 32
1845 | subu TMP0, AT, CARG2
1846 | srlv CARG2, CARG1, CARG2
1847 | sllv CARG1, CARG1, TMP0
1848 |.endif
1849 | b ->fff_resi
1850 |. ins CRET1, CARG1, CARG2
1851 |.endmacro
1852 |
1853 |.ffunc_bit_sh lshift, sllv, 0
1854 |.ffunc_bit_sh rshift, srlv, 0
1855 |.ffunc_bit_sh arshift, srav, 0
1856 |// Can't use rotrv, since it's only in MIPS32R2.
1857 |.ffunc_bit_sh rol, or, 1
1858 |.ffunc_bit_sh ror, or, 2
1859 |
1860 |.ffunc_bit tobit
1861 |->fff_resi:
1862 | mtc1 CRET1, FRET1
1863 | b ->fff_resn
1864 |. cvt.d.w FRET1, FRET1
1865 |
1866 |//-----------------------------------------------------------------------
1867 |
1868 |->fff_fallback: // Call fast function fallback handler.
1869 | // BASE = new base, RB = CFUNC, RC = nargs*8
1870 | lw TMP3, CFUNC:RB->f
1871 | addu TMP1, BASE, NARGS8:RC
1872 | lw PC, FRAME_PC(BASE) // Fallback may overwrite PC.
1873 | addiu TMP0, TMP1, 8*LUA_MINSTACK
1874 | lw TMP2, L->maxstack
1875 | sw PC, SAVE_PC // Redundant (but a defined value).
1876 | sltu AT, TMP2, TMP0
1877 | sw BASE, L->base
1878 | sw TMP1, L->top
1879 | bnez AT, >5 // Need to grow stack.
1880 |. move CFUNCADDR, TMP3
1881 | jalr TMP3 // (lua_State *L)
1882 |. move CARG1, L
1883 | // Either throws an error, or recovers and returns -1, 0 or nresults+1.
1884 | lw BASE, L->base
1885 | sll RD, CRET1, 3
1886 | bgtz CRET1, ->fff_res // Returned nresults+1?
1887 |. addiu RA, BASE, -8
1888 |1: // Returned 0 or -1: retry fast path.
1889 | lw TMP0, L->top
1890 | lw LFUNC:RB, FRAME_FUNC(BASE)
1891 | bnez CRET1, ->vm_call_tail // Returned -1?
1892 |. subu NARGS8:RC, TMP0, BASE
1893 | ins_callt // Returned 0: retry fast path.
1894 |
1895 |// Reconstruct previous base for vmeta_call during tailcall.
1896 |->vm_call_tail:
1897 | andi TMP0, PC, FRAME_TYPE
1898 | li AT, -4
1899 | bnez TMP0, >3
1900 |. and TMP1, PC, AT
1901 | lbu TMP1, OFS_RA(PC)
1902 | sll TMP1, TMP1, 3
1903 | addiu TMP1, TMP1, 8
1904 |3:
1905 | b ->vm_call_dispatch // Resolve again for tailcall.
1906 |. subu TMP2, BASE, TMP1
1907 |
1908 |5: // Grow stack for fallback handler.
1909 | load_got lj_state_growstack
1910 | li CARG2, LUA_MINSTACK
1911 | call_intern lj_state_growstack // (lua_State *L, int n)
1912 |. move CARG1, L
1913 | lw BASE, L->base
1914 | b <1
1915 |. li CRET1, 0 // Force retry.
1916 |
1917 |->fff_gcstep: // Call GC step function.
1918 | // BASE = new base, RC = nargs*8
1919 | move MULTRES, ra
1920 | load_got lj_gc_step
1921 | sw BASE, L->base
1922 | addu TMP0, BASE, NARGS8:RC
1923 | sw PC, SAVE_PC // Redundant (but a defined value).
1924 | sw TMP0, L->top
1925 | call_intern lj_gc_step // (lua_State *L)
1926 |. move CARG1, L
1927 | lw BASE, L->base
1928 | move ra, MULTRES
1929 | lw TMP0, L->top
1930 | lw CFUNC:RB, FRAME_FUNC(BASE)
1931 | jr ra
1932 |. subu NARGS8:RC, TMP0, BASE
1933 |
1934 |//-----------------------------------------------------------------------
1935 |//-- Special dispatch targets -------------------------------------------
1936 |//-----------------------------------------------------------------------
1937 |
1938 |->vm_record: // Dispatch target for recording phase.
1939#if LJ_HASJIT
1940 | lbu TMP3, DISPATCH_GL(hookmask)(DISPATCH)
1941 | andi AT, TMP3, HOOK_VMEVENT // No recording while in vmevent.
1942 | bnez AT, >5
1943 | // Decrement the hookcount for consistency, but always do the call.
1944 |. lw TMP2, DISPATCH_GL(hookcount)(DISPATCH)
1945 | andi AT, TMP3, HOOK_ACTIVE
1946 | bnez AT, >1
1947 |. addiu TMP2, TMP2, -1
1948 | andi AT, TMP3, LUA_MASKLINE|LUA_MASKCOUNT
1949 | beqz AT, >1
1950 |. nop
1951 | b >1
1952 |. sw TMP2, DISPATCH_GL(hookcount)(DISPATCH)
1953#endif
1954 |
1955 |->vm_rethook: // Dispatch target for return hooks.
1956 | lbu TMP3, DISPATCH_GL(hookmask)(DISPATCH)
1957 | andi AT, TMP3, HOOK_ACTIVE // Hook already active?
1958 | beqz AT, >1
1959 |5: // Re-dispatch to static ins.
1960 |. lw AT, GG_DISP2STATIC(TMP0) // Assumes TMP0 holds DISPATCH+OP*4.
1961 | jr AT
1962 |. nop
1963 |
1964 |->vm_inshook: // Dispatch target for instr/line hooks.
1965 | lbu TMP3, DISPATCH_GL(hookmask)(DISPATCH)
1966 | lw TMP2, DISPATCH_GL(hookcount)(DISPATCH)
1967 | andi AT, TMP3, HOOK_ACTIVE // Hook already active?
1968 | bnez AT, <5
1969 |. andi AT, TMP3, LUA_MASKLINE|LUA_MASKCOUNT
1970 | beqz AT, <5
1971 |. addiu TMP2, TMP2, -1
1972 | beqz TMP2, >1
1973 |. sw TMP2, DISPATCH_GL(hookcount)(DISPATCH)
1974 | andi AT, TMP3, LUA_MASKLINE
1975 | beqz AT, <5
1976 |1:
1977 |. load_got lj_dispatch_ins
1978 | sw MULTRES, SAVE_MULTRES
1979 | move CARG2, PC
1980 | sw BASE, L->base
1981 | // SAVE_PC must hold the _previous_ PC. The callee updates it with PC.
1982 | call_intern lj_dispatch_ins // (lua_State *L, const BCIns *pc)
1983 |. move CARG1, L
1984 |3:
1985 | lw BASE, L->base
1986 |4: // Re-dispatch to static ins.
1987 | lw INS, -4(PC)
1988 | decode_OP4a TMP1, INS
1989 | decode_OP4b TMP1
1990 | addu TMP0, DISPATCH, TMP1
1991 | decode_RD8a RD, INS
1992 | lw AT, GG_DISP2STATIC(TMP0)
1993 | decode_RA8a RA, INS
1994 | decode_RD8b RD
1995 | jr AT
1996 | decode_RA8b RA
1997 |
1998 |->cont_hook: // Continue from hook yield.
1999 | addiu PC, PC, 4
2000 | b <4
2001 |. lw MULTRES, -24+LO(RB) // Restore MULTRES for *M ins.
2002 |
2003 |->vm_hotloop: // Hot loop counter underflow.
2004#if LJ_HASJIT
2005 | lw LFUNC:TMP1, FRAME_FUNC(BASE)
2006 | addiu CARG1, DISPATCH, GG_DISP2J
2007 | sw PC, SAVE_PC
2008 | lw TMP1, LFUNC:TMP1->pc
2009 | move CARG2, PC
2010 | sw L, DISPATCH_J(L)(DISPATCH)
2011 | lbu TMP1, PC2PROTO(framesize)(TMP1)
2012 | load_got lj_trace_hot
2013 | sw BASE, L->base
2014 | sll TMP1, TMP1, 3
2015 | addu TMP1, BASE, TMP1
2016 | call_intern lj_trace_hot // (jit_State *J, const BCIns *pc)
2017 |. sw TMP1, L->top
2018 | b <3
2019 |. nop
2020#endif
2021 |
2022 |->vm_callhook: // Dispatch target for call hooks.
2023#if LJ_HASJIT
2024 | b >1
2025#endif
2026 |. move CARG2, PC
2027 |
2028 |->vm_hotcall: // Hot call counter underflow.
2029#if LJ_HASJIT
2030 | ori CARG2, PC, 1
2031 |1:
2032#endif
2033 | load_got lj_dispatch_call
2034 | addu TMP0, BASE, RC
2035 | sw PC, SAVE_PC
2036 | sw BASE, L->base
2037 | subu RA, RA, BASE
2038 | sw TMP0, L->top
2039 | call_intern lj_dispatch_call // (lua_State *L, const BCIns *pc)
2040 |. move CARG1, L
2041 | // Returns ASMFunction.
2042 | lw BASE, L->base
2043 | lw TMP0, L->top
2044 | sw r0, SAVE_PC // Invalidate for subsequent line hook.
2045 | subu NARGS8:RC, TMP0, BASE
2046 | addu RA, BASE, RA
2047 | lw LFUNC:RB, FRAME_FUNC(BASE)
2048 | jr CRET1
2049 |. lw INS, -4(PC)
2050 |
2051 |//-----------------------------------------------------------------------
2052 |//-- Trace exit handler -------------------------------------------------
2053 |//-----------------------------------------------------------------------
2054 |
2055 |.macro savex_, a, b
2056 | sdc1 f..a, 16+a*8(sp)
2057 | sw r..a, 16+32*8+a*4(sp)
2058 | sw r..b, 16+32*8+b*4(sp)
2059 |.endmacro
2060 |
2061 |->vm_exit_handler:
2062#if LJ_HASJIT
2063 | addiu sp, sp, -(16+32*8+32*4)
2064 | savex_ 0, 1
2065 | savex_ 2, 3
2066 | savex_ 4, 5
2067 | savex_ 6, 7
2068 | savex_ 8, 9
2069 | savex_ 10, 11
2070 | savex_ 12, 13
2071 | savex_ 14, 15
2072 | savex_ 16, 17
2073 | savex_ 18, 19
2074 | savex_ 20, 21
2075 | savex_ 22, 23
2076 | savex_ 24, 25
2077 | savex_ 26, 27
2078 | sdc1 f28, 16+28*8(sp)
2079 | sw r28, 16+32*8+28*4(sp)
2080 | sdc1 f30, 16+30*8(sp)
2081 | sw r30, 16+32*8+30*4(sp)
2082 | sw r0, 16+32*8+31*4(sp) // Clear RID_TMP.
2083 | li_vmstate EXIT
2084 | addiu TMP2, sp, 16+32*8+32*4 // Recompute original value of sp.
2085 | addiu DISPATCH, JGL, -GG_DISP2G-32768
2086 | lw TMP1, 0(TMP2) // Load exit number.
2087 | st_vmstate
2088 | sw TMP2, 16+32*8+29*4(sp) // Store sp in RID_SP.
2089 | lw L, DISPATCH_GL(jit_L)(DISPATCH)
2090 | lw BASE, DISPATCH_GL(jit_base)(DISPATCH)
2091 | load_got lj_trace_exit
2092 | sw L, DISPATCH_J(L)(DISPATCH)
2093 | sw ra, DISPATCH_J(parent)(DISPATCH) // Store trace number.
2094 | sw TMP1, DISPATCH_J(exitno)(DISPATCH) // Store exit number.
2095 | addiu CARG1, DISPATCH, GG_DISP2J
2096 | sw BASE, L->base
2097 | call_intern lj_trace_exit // (jit_State *J, ExitState *ex)
2098 |. addiu CARG2, sp, 16
2099 | // Returns MULTRES (unscaled) or negated error code.
2100 | lw TMP1, L->cframe
2101 | li AT, -4
2102 | lw BASE, L->base
2103 | and sp, TMP1, AT
2104 | lw PC, SAVE_PC // Get SAVE_PC.
2105 | b >1
2106 |. sw L, SAVE_L // Set SAVE_L (on-trace resume/yield).
2107#endif
2108 |->vm_exit_interp:
2109#if LJ_HASJIT
2110 | // CRET1 = MULTRES or negated error code, BASE, PC and JGL set.
2111 | lw L, SAVE_L
2112 | addiu DISPATCH, JGL, -GG_DISP2G-32768
2113 |1:
2114 | bltz CRET1, >3 // Check for error from exit.
2115 |. lw LFUNC:TMP1, FRAME_FUNC(BASE)
2116 | lui TMP3, 0x59c0 // TOBIT = 2^52 + 2^51 (float).
2117 | sll MULTRES, CRET1, 3
2118 | li TISNIL, LJ_TNIL
2119 | sw MULTRES, SAVE_MULTRES
2120 | mtc1 TMP3, TOBIT
2121 | lw TMP1, LFUNC:TMP1->pc
2122 | sw r0, DISPATCH_GL(jit_L)(DISPATCH)
2123 | lw KBASE, PC2PROTO(k)(TMP1)
2124 | cvt.d.s TOBIT, TOBIT
2125 | // Modified copy of ins_next which handles function header dispatch, too.
2126 | lw INS, 0(PC)
2127 | addiu PC, PC, 4
2128 | // Assumes TISNIL == ~LJ_VMST_INTERP == -1
2129 | sw TISNIL, DISPATCH_GL(vmstate)(DISPATCH)
2130 | decode_OP4a TMP1, INS
2131 | decode_OP4b TMP1
2132 | sltiu TMP2, TMP1, BC_FUNCF*4 // Function header?
2133 | addu TMP0, DISPATCH, TMP1
2134 | decode_RD8a RD, INS
2135 | lw AT, 0(TMP0)
2136 | decode_RA8a RA, INS
2137 | beqz TMP2, >2
2138 |. decode_RA8b RA
2139 | jr AT
2140 |. decode_RD8b RD
2141 |2:
2142 | addiu RC, MULTRES, -8
2143 | jr AT
2144 |. add RA, RA, BASE
2145 |
2146 |3: // Rethrow error from the right C frame.
2147 | load_got lj_err_throw
2148 | negu CARG2, CRET1
2149 | call_intern lj_err_throw // (lua_State *L, int errcode)
2150 |. move CARG1, L
2151#endif
2152 |
2153 |//-----------------------------------------------------------------------
2154 |//-- Math helper functions ----------------------------------------------
2155 |//-----------------------------------------------------------------------
2156 |
2157 |// Modifies AT, TMP0, FRET1, FRET2, f4. Keeps all others incl. FARG1, FARG2.
2158 |.macro vm_round, func
2159 | lui TMP0, 0x4330 // Hiword of 2^52 (double).
2160 | mtc1 r0, f4
2161 | mtc1 TMP0, f5
2162 | abs.d FRET2, FARG1 // |x|
2163 | mfc1 AT, f13
2164 | c.olt.d 0, FRET2, f4
2165 | add.d FRET1, FRET2, f4 // (|x| + 2^52) - 2^52
2166 | bc1f 0, >1 // Truncate only if |x| < 2^52.
2167 |. sub.d FRET1, FRET1, f4
2168 | slt AT, AT, r0
2169 |.if "func" == "ceil"
2170 | lui TMP0, 0xbff0 // Hiword of -1 (double). Preserves -0.
2171 |.else
2172 | lui TMP0, 0x3ff0 // Hiword of +1 (double).
2173 |.endif
2174 |.if "func" == "trunc"
2175 | mtc1 TMP0, f5
2176 | c.olt.d 0, FRET2, FRET1 // |x| < result?
2177 | sub.d FRET2, FRET1, f4
2178 | movt.d FRET1, FRET2, 0 // If yes, subtract +1.
2179 | neg.d FRET2, FRET1
2180 | jr ra
2181 |. movn.d FRET1, FRET2, AT // Merge sign bit back in.
2182 |.else
2183 | neg.d FRET2, FRET1
2184 | mtc1 TMP0, f5
2185 | movn.d FRET1, FRET2, AT // Merge sign bit back in.
2186 |.if "func" == "ceil"
2187 | c.olt.d 0, FRET1, FARG1 // x > result?
2188 |.else
2189 | c.olt.d 0, FARG1, FRET1 // x < result?
2190 |.endif
2191 | sub.d FRET2, FRET1, f4 // If yes, subtract +-1.
2192 | jr ra
2193 |. movt.d FRET1, FRET2, 0
2194 |.endif
2195 |1:
2196 | jr ra
2197 |. mov.d FRET1, FARG1
2198 |.endmacro
2199 |
2200 |->vm_floor:
2201 | vm_round floor
2202 |->vm_ceil:
2203 | vm_round ceil
2204 |->vm_trunc:
2205#if LJ_HASJIT
2206 | vm_round trunc
2207#endif
2208 |
2209 |//-----------------------------------------------------------------------
2210 |//-- Miscellaneous functions --------------------------------------------
2211 |//-----------------------------------------------------------------------
2212 |
2213 |//-----------------------------------------------------------------------
2214 |//-- FFI helper functions -----------------------------------------------
2215 |//-----------------------------------------------------------------------
2216 |
2217 |// Handler for callback functions. Callback slot number in r1, g in r2.
2218 |->vm_ffi_callback:
2219#if LJ_HASFFI
2220 |.type CTSTATE, CTState, PC
2221 | saveregs
2222 | lw CTSTATE, GL:r2->ctype_state
2223 | addiu DISPATCH, r2, GG_G2DISP
2224 | load_got lj_ccallback_enter
2225 | sw r1, CTSTATE->cb.slot
2226 | sw CARG1, CTSTATE->cb.gpr[0]
2227 | sw CARG2, CTSTATE->cb.gpr[1]
2228 | sdc1 FARG1, CTSTATE->cb.fpr[0]
2229 | sw CARG3, CTSTATE->cb.gpr[2]
2230 | sw CARG4, CTSTATE->cb.gpr[3]
2231 | sdc1 FARG2, CTSTATE->cb.fpr[1]
2232 | addiu TMP0, sp, CFRAME_SPACE+16
2233 | sw TMP0, CTSTATE->cb.stack
2234 | sw r0, SAVE_PC // Any value outside of bytecode is ok.
2235 | move CARG2, sp
2236 | call_intern lj_ccallback_enter // (CTState *cts, void *cf)
2237 |. move CARG1, CTSTATE
2238 | // Returns lua_State *.
2239 | lw BASE, L:CRET1->base
2240 | lw RC, L:CRET1->top
2241 | move L, CRET1
2242 | lui TMP3, 0x59c0 // TOBIT = 2^52 + 2^51 (float).
2243 | lw LFUNC:RB, FRAME_FUNC(BASE)
2244 | mtc1 TMP3, TOBIT
2245 | li_vmstate INTERP
2246 | li TISNIL, LJ_TNIL
2247 | subu RC, RC, BASE
2248 | st_vmstate
2249 | cvt.d.s TOBIT, TOBIT
2250 | ins_callt
2251#endif
2252 |
2253 |->cont_ffi_callback: // Return from FFI callback.
2254#if LJ_HASFFI
2255 | load_got lj_ccallback_leave
2256 | lw CTSTATE, DISPATCH_GL(ctype_state)(DISPATCH)
2257 | sw BASE, L->base
2258 | sw RB, L->top
2259 | sw L, CTSTATE->L
2260 | move CARG2, RA
2261 | call_intern lj_ccallback_leave // (CTState *cts, TValue *o)
2262 |. move CARG1, CTSTATE
2263 | lw CRET1, CTSTATE->cb.gpr[0]
2264 | ldc1 FRET1, CTSTATE->cb.fpr[0]
2265 | lw CRET2, CTSTATE->cb.gpr[1]
2266 | b ->vm_leave_unw
2267 |. ldc1 FRET2, CTSTATE->cb.fpr[1]
2268#endif
2269 |
2270 |->vm_ffi_call: // Call C function via FFI.
2271 | // Caveat: needs special frame unwinding, see below.
2272#if LJ_HASFFI
2273 | .type CCSTATE, CCallState, CARG1
2274 | lw TMP1, CCSTATE->spadj
2275 | lbu CARG2, CCSTATE->nsp
2276 | move TMP2, sp
2277 | subu sp, sp, TMP1
2278 | sw ra, -4(TMP2)
2279 | sll CARG2, CARG2, 2
2280 | sw r16, -8(TMP2)
2281 | sw CCSTATE, -12(TMP2)
2282 | move r16, TMP2
2283 | addiu TMP1, CCSTATE, offsetof(CCallState, stack)
2284 | addiu TMP2, sp, 16
2285 | beqz CARG2, >2
2286 |. addu TMP3, TMP1, CARG2
2287 |1:
2288 | lw TMP0, 0(TMP1)
2289 | addiu TMP1, TMP1, 4
2290 | sltu AT, TMP1, TMP3
2291 | sw TMP0, 0(TMP2)
2292 | bnez AT, <1
2293 |. addiu TMP2, TMP2, 4
2294 |2:
2295 | lw CFUNCADDR, CCSTATE->func
2296 | lw CARG2, CCSTATE->gpr[1]
2297 | lw CARG3, CCSTATE->gpr[2]
2298 | lw CARG4, CCSTATE->gpr[3]
2299 | ldc1 FARG1, CCSTATE->fpr[0]
2300 | ldc1 FARG2, CCSTATE->fpr[1]
2301 | jalr CFUNCADDR
2302 |. lw CARG1, CCSTATE->gpr[0] // Do this last, since CCSTATE is CARG1.
2303 | lw CCSTATE:TMP1, -12(r16)
2304 | lw TMP2, -8(r16)
2305 | lw ra, -4(r16)
2306 | sw CRET1, CCSTATE:TMP1->gpr[0]
2307 | sw CRET2, CCSTATE:TMP1->gpr[1]
2308 | sdc1 FRET1, CCSTATE:TMP1->fpr[0]
2309 | sdc1 FRET2, CCSTATE:TMP1->fpr[1]
2310 | move sp, r16
2311 | jr ra
2312 |. move r16, TMP2
2313#endif
2314 |// Note: vm_ffi_call must be the last function in this object file!
2315 |
2316 |//-----------------------------------------------------------------------
2317}
2318
2319/* Generate the code for a single instruction. */
2320static void build_ins(BuildCtx *ctx, BCOp op, int defop)
2321{
2322 int vk = 0;
2323 |=>defop:
2324
2325 switch (op) {
2326
2327 /* -- Comparison ops ---------------------------------------------------- */
2328
2329 /* Remember: all ops branch for a true comparison, fall through otherwise. */
2330
2331 case BC_ISLT: case BC_ISGE: case BC_ISLE: case BC_ISGT:
2332 | // RA = src1*8, RD = src2*8, JMP with RD = target
2333 | addu CARG2, BASE, RA
2334 | addu CARG3, BASE, RD
2335 | lw TMP0, HI(CARG2)
2336 | lw TMP1, HI(CARG3)
2337 | ldc1 f0, 0(CARG2)
2338 | ldc1 f2, 0(CARG3)
2339 | sltiu TMP0, TMP0, LJ_TISNUM
2340 | sltiu TMP1, TMP1, LJ_TISNUM
2341 | lhu TMP2, OFS_RD(PC)
2342 | and TMP0, TMP0, TMP1
2343 | addiu PC, PC, 4
2344 | beqz TMP0, ->vmeta_comp
2345 |. lui TMP1, (-(BCBIAS_J*4 >> 16) & 65535)
2346 | decode_RD4b TMP2
2347 | addu TMP2, TMP2, TMP1
2348 if (op == BC_ISLT || op == BC_ISGE) {
2349 | c.olt.d f0, f2
2350 } else {
2351 | c.ole.d f0, f2
2352 }
2353 if (op == BC_ISLT || op == BC_ISLE) {
2354 | movf TMP2, r0
2355 } else {
2356 | movt TMP2, r0
2357 }
2358 | addu PC, PC, TMP2
2359 |1:
2360 | ins_next
2361 break;
2362
2363 case BC_ISEQV: case BC_ISNEV:
2364 vk = op == BC_ISEQV;
2365 | // RA = src1*8, RD = src2*8, JMP with RD = target
2366 | addu RA, BASE, RA
2367 | addiu PC, PC, 4
2368 | lw TMP0, HI(RA)
2369 | ldc1 f0, 0(RA)
2370 | addu RD, BASE, RD
2371 | lhu TMP2, -4+OFS_RD(PC)
2372 | lw TMP1, HI(RD)
2373 | ldc1 f2, 0(RD)
2374 | lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535)
2375 | sltiu AT, TMP0, LJ_TISNUM
2376 | sltiu CARG1, TMP1, LJ_TISNUM
2377 | decode_RD4b TMP2
2378 | and AT, AT, CARG1
2379 | beqz AT, >5
2380 |. addu TMP2, TMP2, TMP3
2381 | c.eq.d f0, f2
2382 if (vk) {
2383 | movf TMP2, r0
2384 } else {
2385 | movt TMP2, r0
2386 }
2387 |1:
2388 | addu PC, PC, TMP2
2389 | ins_next
2390 |5: // Either or both types are not numbers.
2391 | lw CARG2, LO(RA)
2392 | lw CARG3, LO(RD)
2393 if (LJ_HASFFI) {
2394 | li TMP3, LJ_TCDATA
2395 | beq TMP0, TMP3, ->vmeta_equal_cd
2396 }
2397 |. sltiu AT, TMP0, LJ_TISPRI // Not a primitive?
2398 if (LJ_HASFFI) {
2399 | beq TMP1, TMP3, ->vmeta_equal_cd
2400 }
2401 |. xor TMP3, CARG2, CARG3 // Same tv?
2402 | xor TMP1, TMP1, TMP0 // Same type?
2403 | sltiu CARG1, TMP0, LJ_TISTABUD+1 // Table or userdata?
2404 | movz TMP3, r0, AT // Ignore tv if primitive.
2405 | movn CARG1, r0, TMP1 // Tab/ud and same type?
2406 | or AT, TMP1, TMP3 // Same type && (pri||same tv).
2407 | movz CARG1, r0, AT
2408 | beqz CARG1, <1 // Done if not tab/ud or not same type or same tv.
2409 if (vk) {
2410 |. movn TMP2, r0, AT
2411 } else {
2412 |. movz TMP2, r0, AT
2413 }
2414 | // Different tables or userdatas. Need to check __eq metamethod.
2415 | // Field metatable must be at same offset for GCtab and GCudata!
2416 | lw TAB:TMP1, TAB:CARG2->metatable
2417 | beqz TAB:TMP1, <1 // No metatable?
2418 |. nop
2419 | lbu TMP1, TAB:TMP1->nomm
2420 | andi TMP1, TMP1, 1<<MM_eq
2421 | bnez TMP1, <1 // Or 'no __eq' flag set?
2422 |. nop
2423 | b ->vmeta_equal // Handle __eq metamethod.
2424 |. li CARG4, 1-vk // ne = 0 or 1.
2425 break;
2426
2427 case BC_ISEQS: case BC_ISNES:
2428 vk = op == BC_ISEQS;
2429 | // RA = src*8, RD = str_const*8 (~), JMP with RD = target
2430 | addu RA, BASE, RA
2431 | addiu PC, PC, 4
2432 | lw TMP0, HI(RA)
2433 | srl RD, RD, 1
2434 | lw STR:TMP3, LO(RA)
2435 | subu RD, KBASE, RD
2436 | lhu TMP2, -4+OFS_RD(PC)
2437 if (LJ_HASFFI) {
2438 | li AT, LJ_TCDATA
2439 | beq TMP0, AT, ->vmeta_equal_cd
2440 }
2441 |. lw STR:TMP1, -4(RD) // KBASE-4-str_const*4
2442 | addiu TMP0, TMP0, -LJ_TSTR
2443 | decode_RD4b TMP2
2444 | xor TMP1, STR:TMP1, STR:TMP3
2445 | or TMP0, TMP0, TMP1
2446 | lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535)
2447 | addu TMP2, TMP2, TMP3
2448 if (vk) {
2449 | movn TMP2, r0, TMP0
2450 } else {
2451 | movz TMP2, r0, TMP0
2452 }
2453 | addu PC, PC, TMP2
2454 | ins_next
2455 break;
2456
2457 case BC_ISEQN: case BC_ISNEN:
2458 vk = op == BC_ISEQN;
2459 | // RA = src*8, RD = num_const*8, JMP with RD = target
2460 | addu RA, BASE, RA
2461 | addiu PC, PC, 4
2462 | lw TMP0, HI(RA)
2463 | ldc1 f0, 0(RA)
2464 | addu RD, KBASE, RD
2465 | lhu TMP2, -4+OFS_RD(PC)
2466 | ldc1 f2, 0(RD)
2467 | lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535)
2468 | sltiu AT, TMP0, LJ_TISNUM
2469 | decode_RD4b TMP2
2470 if (LJ_HASFFI) {
2471 | beqz AT, >5
2472 } else {
2473 | beqz AT, >1
2474 }
2475 |. addu TMP2, TMP2, TMP3
2476 | c.eq.d f0, f2
2477 if (vk) {
2478 | movf TMP2, r0
2479 | addu PC, PC, TMP2
2480 |1:
2481 } else {
2482 | movt TMP2, r0
2483 |1:
2484 | addu PC, PC, TMP2
2485 }
2486 | ins_next
2487 if (LJ_HASFFI) {
2488 |5:
2489 | li AT, LJ_TCDATA
2490 | beq TMP0, AT, ->vmeta_equal_cd
2491 |. nop
2492 | b <1
2493 |. nop
2494 }
2495 break;
2496
2497 case BC_ISEQP: case BC_ISNEP:
2498 vk = op == BC_ISEQP;
2499 | // RA = src*8, RD = primitive_type*8 (~), JMP with RD = target
2500 | addu RA, BASE, RA
2501 | srl TMP1, RD, 3
2502 | lw TMP0, HI(RA)
2503 | lhu TMP2, OFS_RD(PC)
2504 | not TMP1, TMP1
2505 | addiu PC, PC, 4
2506 if (LJ_HASFFI) {
2507 | li AT, LJ_TCDATA
2508 | beq TMP0, AT, ->vmeta_equal_cd
2509 }
2510 |. xor TMP0, TMP0, TMP1
2511 | decode_RD4b TMP2
2512 | lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535)
2513 | addu TMP2, TMP2, TMP3
2514 if (vk) {
2515 | movn TMP2, r0, TMP0
2516 } else {
2517 | movz TMP2, r0, TMP0
2518 }
2519 | addu PC, PC, TMP2
2520 | ins_next
2521 break;
2522
2523 /* -- Unary test and copy ops ------------------------------------------- */
2524
2525 case BC_ISTC: case BC_ISFC: case BC_IST: case BC_ISF:
2526 | // RA = dst*8 or unused, RD = src*8, JMP with RD = target
2527 | addu RD, BASE, RD
2528 | lhu TMP2, OFS_RD(PC)
2529 | lw TMP0, HI(RD)
2530 | addiu PC, PC, 4
2531 if (op == BC_IST || op == BC_ISF) {
2532 | sltiu TMP0, TMP0, LJ_TISTRUECOND
2533 | decode_RD4b TMP2
2534 | lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535)
2535 | addu TMP2, TMP2, TMP3
2536 if (op == BC_IST) {
2537 | movz TMP2, r0, TMP0
2538 } else {
2539 | movn TMP2, r0, TMP0
2540 }
2541 | addu PC, PC, TMP2
2542 } else {
2543 | sltiu TMP0, TMP0, LJ_TISTRUECOND
2544 | ldc1 f0, 0(RD)
2545 if (op == BC_ISTC) {
2546 | beqz TMP0, >1
2547 } else {
2548 | bnez TMP0, >1
2549 }
2550 |. addu RA, BASE, RA
2551 | decode_RD4b TMP2
2552 | lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535)
2553 | addu TMP2, TMP2, TMP3
2554 | sdc1 f0, 0(RA)
2555 | addu PC, PC, TMP2
2556 |1:
2557 }
2558 | ins_next
2559 break;
2560
2561 /* -- Unary ops --------------------------------------------------------- */
2562
2563 case BC_MOV:
2564 | // RA = dst*8, RD = src*8
2565 | addu RD, BASE, RD
2566 | addu RA, BASE, RA
2567 | ldc1 f0, 0(RD)
2568 | ins_next1
2569 | sdc1 f0, 0(RA)
2570 | ins_next2
2571 break;
2572 case BC_NOT:
2573 | // RA = dst*8, RD = src*8
2574 | addu RD, BASE, RD
2575 | addu RA, BASE, RA
2576 | lw TMP0, HI(RD)
2577 | li TMP1, LJ_TFALSE
2578 | sltiu TMP0, TMP0, LJ_TISTRUECOND
2579 | addiu TMP1, TMP0, LJ_TTRUE
2580 | ins_next1
2581 | sw TMP1, HI(RA)
2582 | ins_next2
2583 break;
2584 case BC_UNM:
2585 | // RA = dst*8, RD = src*8
2586 | addu CARG3, BASE, RD
2587 | addu RA, BASE, RA
2588 | lw TMP0, HI(CARG3)
2589 | ldc1 f0, 0(CARG3)
2590 | sltiu AT, TMP0, LJ_TISNUM
2591 | beqz AT, ->vmeta_unm
2592 |. neg.d f0, f0
2593 | ins_next1
2594 | sdc1 f0, 0(RA)
2595 | ins_next2
2596 break;
2597 case BC_LEN:
2598 | // RA = dst*8, RD = src*8
2599 | addu CARG2, BASE, RD
2600 | addu RA, BASE, RA
2601 | lw TMP0, HI(CARG2)
2602 | lw CARG1, LO(CARG2)
2603 | li AT, LJ_TSTR
2604 | bne TMP0, AT, >2
2605 |. li AT, LJ_TTAB
2606 | lw CRET1, STR:CARG1->len
2607 |1:
2608 | mtc1 CRET1, f0
2609 | cvt.d.w f0, f0
2610 | ins_next1
2611 | sdc1 f0, 0(RA)
2612 | ins_next2
2613 |2:
2614 | bne TMP0, AT, ->vmeta_len
2615 |. nop
2616#ifdef LUAJIT_ENABLE_LUA52COMPAT
2617 | lw TAB:TMP2, TAB:CARG1->metatable
2618 | bnez TAB:TMP2, >9
2619 |. nop
2620 |3:
2621#endif
2622 |->BC_LEN_Z:
2623 | load_got lj_tab_len
2624 | call_intern lj_tab_len // (GCtab *t)
2625 |. nop
2626 | // Returns uint32_t (but less than 2^31).
2627 | b <1
2628 |. nop
2629#ifdef LUAJIT_ENABLE_LUA52COMPAT
2630 |9:
2631 | lbu TMP0, TAB:TMP2->nomm
2632 | andi TMP0, TMP0, 1<<MM_len
2633 | bnez TMP0, <3 // 'no __len' flag set: done.
2634 |. nop
2635 | b ->vmeta_len
2636 |. nop
2637#endif
2638 break;
2639
2640 /* -- Binary ops -------------------------------------------------------- */
2641
2642 |.macro ins_arithpre
2643 ||vk = ((int)op - BC_ADDVN) / (BC_ADDNV-BC_ADDVN);
2644 | decode_RB8a RB, INS
2645 | decode_RB8b RB
2646 | decode_RDtoRC8 RC, RD
2647 | // RA = dst*8, RB = src1*8, RC = src2*8 | num_const*8
2648 ||switch (vk) {
2649 ||case 0:
2650 | addu CARG3, BASE, RB
2651 | addu CARG4, KBASE, RC
2652 | lw TMP1, HI(CARG3)
2653 | ldc1 f20, 0(CARG3)
2654 | ldc1 f22, 0(CARG4)
2655 | sltiu AT, TMP1, LJ_TISNUM
2656 || break;
2657 ||case 1:
2658 | addu CARG4, BASE, RB
2659 | addu CARG3, KBASE, RC
2660 | lw TMP1, HI(CARG4)
2661 | ldc1 f22, 0(CARG4)
2662 | ldc1 f20, 0(CARG3)
2663 | sltiu AT, TMP1, LJ_TISNUM
2664 || break;
2665 ||default:
2666 | addu CARG3, BASE, RB
2667 | addu CARG4, BASE, RC
2668 | lw TMP1, HI(CARG3)
2669 | lw TMP2, HI(CARG4)
2670 | ldc1 f20, 0(CARG3)
2671 | ldc1 f22, 0(CARG4)
2672 | sltiu AT, TMP1, LJ_TISNUM
2673 | sltiu TMP0, TMP2, LJ_TISNUM
2674 | and AT, AT, TMP0
2675 || break;
2676 ||}
2677 | beqz AT, ->vmeta_arith
2678 |. addu RA, BASE, RA
2679 |.endmacro
2680 |
2681 |.macro fpmod, a, b, c
2682 |->BC_MODVN_Z:
2683 | bal ->vm_floor // floor(b/c)
2684 |. div.d FARG1, b, c
2685 | mul.d a, FRET1, c
2686 | sub.d a, b, a // b - floor(b/c)*c
2687 |.endmacro
2688 |
2689 |.macro ins_arith, ins
2690 | ins_arithpre
2691 |.if "ins" == "fpmod_"
2692 | b ->BC_MODVN_Z // Avoid 3 copies. It's slow anyway.
2693 |. nop
2694 |.else
2695 | ins f0, f20, f22
2696 | ins_next1
2697 | sdc1 f0, 0(RA)
2698 | ins_next2
2699 |.endif
2700 |.endmacro
2701
2702 case BC_ADDVN: case BC_ADDNV: case BC_ADDVV:
2703 | ins_arith add.d
2704 break;
2705 case BC_SUBVN: case BC_SUBNV: case BC_SUBVV:
2706 | ins_arith sub.d
2707 break;
2708 case BC_MULVN: case BC_MULNV: case BC_MULVV:
2709 | ins_arith mul.d
2710 break;
2711 case BC_DIVVN: case BC_DIVNV: case BC_DIVVV:
2712 | ins_arith div.d
2713 break;
2714 case BC_MODVN:
2715 | ins_arith fpmod
2716 break;
2717 case BC_MODNV: case BC_MODVV:
2718 | ins_arith fpmod_
2719 break;
2720 case BC_POW:
2721 | decode_RB8a RB, INS
2722 | decode_RB8b RB
2723 | decode_RDtoRC8 RC, RD
2724 | addu CARG3, BASE, RB
2725 | addu CARG4, BASE, RC
2726 | lw TMP1, HI(CARG3)
2727 | lw TMP2, HI(CARG4)
2728 | ldc1 FARG1, 0(CARG3)
2729 | ldc1 FARG2, 0(CARG4)
2730 | sltiu AT, TMP1, LJ_TISNUM
2731 | sltiu TMP0, TMP2, LJ_TISNUM
2732 | and AT, AT, TMP0
2733 | load_got pow
2734 | beqz AT, ->vmeta_arith
2735 |. addu RA, BASE, RA
2736 | call_extern
2737 |. nop
2738 | ins_next1
2739 | sdc1 FRET1, 0(RA)
2740 | ins_next2
2741 break;
2742
2743 case BC_CAT:
2744 | // RA = dst*8, RB = src_start*8, RC = src_end*8
2745 | decode_RB8a RB, INS
2746 | decode_RB8b RB
2747 | decode_RDtoRC8 RC, RD
2748 | subu CARG3, RC, RB
2749 | sw BASE, L->base
2750 | addu CARG2, BASE, RC
2751 | move MULTRES, RB
2752 |->BC_CAT_Z:
2753 | load_got lj_meta_cat
2754 | srl CARG3, CARG3, 3
2755 | sw PC, SAVE_PC
2756 | call_intern lj_meta_cat // (lua_State *L, TValue *top, int left)
2757 |. move CARG1, L
2758 | // Returns NULL (finished) or TValue * (metamethod).
2759 | bnez CRET1, ->vmeta_binop
2760 |. lw BASE, L->base
2761 | addu RB, BASE, MULTRES
2762 | ldc1 f0, 0(RB)
2763 | addu RA, BASE, RA
2764 | ins_next1
2765 | sdc1 f0, 0(RA) // Copy result from RB to RA.
2766 | ins_next2
2767 break;
2768
2769 /* -- Constant ops ------------------------------------------------------ */
2770
2771 case BC_KSTR:
2772 | // RA = dst*8, RD = str_const*8 (~)
2773 | srl TMP1, RD, 1
2774 | subu TMP1, KBASE, TMP1
2775 | ins_next1
2776 | lw TMP0, -4(TMP1) // KBASE-4-str_const*4
2777 | addu RA, BASE, RA
2778 | li TMP2, LJ_TSTR
2779 | sw TMP0, LO(RA)
2780 | sw TMP2, HI(RA)
2781 | ins_next2
2782 break;
2783 case BC_KCDATA:
2784#if LJ_HASFFI
2785 | // RA = dst*8, RD = cdata_const*8 (~)
2786 | srl TMP1, RD, 1
2787 | subu TMP1, KBASE, TMP1
2788 | ins_next1
2789 | lw TMP0, -4(TMP1) // KBASE-4-cdata_const*4
2790 | addu RA, BASE, RA
2791 | li TMP2, LJ_TCDATA
2792 | sw TMP0, LO(RA)
2793 | sw TMP2, HI(RA)
2794 | ins_next2
2795#endif
2796 break;
2797 case BC_KSHORT:
2798 | // RA = dst*8, RD = int16_literal*8
2799 | sra RD, INS, 16
2800 | mtc1 RD, f0
2801 | addu RA, BASE, RA
2802 | cvt.d.w f0, f0
2803 | ins_next1
2804 | sdc1 f0, 0(RA)
2805 | ins_next2
2806 break;
2807 case BC_KNUM:
2808 | // RA = dst*8, RD = num_const*8
2809 | addu RD, KBASE, RD
2810 | addu RA, BASE, RA
2811 | ldc1 f0, 0(RD)
2812 | ins_next1
2813 | sdc1 f0, 0(RA)
2814 | ins_next2
2815 break;
2816 case BC_KPRI:
2817 | // RA = dst*8, RD = primitive_type*8 (~)
2818 | srl TMP1, RD, 3
2819 | addu RA, BASE, RA
2820 | not TMP0, TMP1
2821 | ins_next1
2822 | sw TMP0, HI(RA)
2823 | ins_next2
2824 break;
2825 case BC_KNIL:
2826 | // RA = base*8, RD = end*8
2827 | addu RA, BASE, RA
2828 | sw TISNIL, HI(RA)
2829 | addiu RA, RA, 8
2830 | addu RD, BASE, RD
2831 |1:
2832 | sw TISNIL, HI(RA)
2833 | slt AT, RA, RD
2834 | bnez AT, <1
2835 |. addiu RA, RA, 8
2836 | ins_next_
2837 break;
2838
2839 /* -- Upvalue and function ops ------------------------------------------ */
2840
2841 case BC_UGET:
2842 | // RA = dst*8, RD = uvnum*8
2843 | lw LFUNC:RB, FRAME_FUNC(BASE)
2844 | srl RD, RD, 1
2845 | addu RD, RD, LFUNC:RB
2846 | lw UPVAL:RB, LFUNC:RD->uvptr
2847 | ins_next1
2848 | lw TMP1, UPVAL:RB->v
2849 | ldc1 f0, 0(TMP1)
2850 | addu RA, BASE, RA
2851 | sdc1 f0, 0(RA)
2852 | ins_next2
2853 break;
2854 case BC_USETV:
2855 | // RA = uvnum*8, RD = src*8
2856 | lw LFUNC:RB, FRAME_FUNC(BASE)
2857 | srl RA, RA, 1
2858 | addu RD, BASE, RD
2859 | addu RA, RA, LFUNC:RB
2860 | ldc1 f0, 0(RD)
2861 | lw UPVAL:RB, LFUNC:RA->uvptr
2862 | lbu TMP3, UPVAL:RB->marked
2863 | lw CARG2, UPVAL:RB->v
2864 | andi TMP3, TMP3, LJ_GC_BLACK // isblack(uv)
2865 | lbu TMP0, UPVAL:RB->closed
2866 | lw TMP2, HI(RD)
2867 | sdc1 f0, 0(CARG2)
2868 | li AT, LJ_GC_BLACK|1
2869 | or TMP3, TMP3, TMP0
2870 | beq TMP3, AT, >2 // Upvalue is closed and black?
2871 |. addiu TMP2, TMP2, -(LJ_TISNUM+1)
2872 |1:
2873 | ins_next
2874 |
2875 |2: // Check if new value is collectable.
2876 | sltiu AT, TMP2, LJ_TISGCV - (LJ_TISNUM+1)
2877 | beqz AT, <1 // tvisgcv(v)
2878 |. lw TMP1, LO(RD)
2879 | lbu TMP3, GCOBJ:TMP1->gch.marked
2880 | andi TMP3, TMP3, LJ_GC_WHITES // iswhite(v)
2881 | beqz TMP3, <1
2882 |. load_got lj_gc_barrieruv
2883 | // Crossed a write barrier. Move the barrier forward.
2884 | call_intern lj_gc_barrieruv // (global_State *g, TValue *tv)
2885 |. addiu CARG1, DISPATCH, GG_DISP2G
2886 | b <1
2887 |. nop
2888 break;
2889 case BC_USETS:
2890 | // RA = uvnum*8, RD = str_const*8 (~)
2891 | lw LFUNC:RB, FRAME_FUNC(BASE)
2892 | srl RA, RA, 1
2893 | srl TMP1, RD, 1
2894 | addu RA, RA, LFUNC:RB
2895 | subu TMP1, KBASE, TMP1
2896 | lw UPVAL:RB, LFUNC:RA->uvptr
2897 | lw STR:TMP1, -4(TMP1) // KBASE-4-str_const*4
2898 | lbu TMP2, UPVAL:RB->marked
2899 | lw CARG2, UPVAL:RB->v
2900 | lbu TMP3, STR:TMP1->marked
2901 | andi AT, TMP2, LJ_GC_BLACK // isblack(uv)
2902 | lbu TMP2, UPVAL:RB->closed
2903 | li TMP0, LJ_TSTR
2904 | sw STR:TMP1, LO(CARG2)
2905 | bnez AT, >2
2906 |. sw TMP0, HI(CARG2)
2907 |1:
2908 | ins_next
2909 |
2910 |2: // Check if string is white and ensure upvalue is closed.
2911 | beqz TMP2, <1
2912 |. andi AT, TMP3, LJ_GC_WHITES // iswhite(str)
2913 | beqz AT, <1
2914 |. load_got lj_gc_barrieruv
2915 | // Crossed a write barrier. Move the barrier forward.
2916 | call_intern lj_gc_barrieruv // (global_State *g, TValue *tv)
2917 |. addiu CARG1, DISPATCH, GG_DISP2G
2918 | b <1
2919 |. nop
2920 break;
2921 case BC_USETN:
2922 | // RA = uvnum*8, RD = num_const*8
2923 | lw LFUNC:RB, FRAME_FUNC(BASE)
2924 | srl RA, RA, 1
2925 | addu RD, KBASE, RD
2926 | addu RA, RA, LFUNC:RB
2927 | ldc1 f0, 0(RD)
2928 | lw UPVAL:RB, LFUNC:RA->uvptr
2929 | ins_next1
2930 | lw TMP1, UPVAL:RB->v
2931 | sdc1 f0, 0(TMP1)
2932 | ins_next2
2933 break;
2934 case BC_USETP:
2935 | // RA = uvnum*8, RD = primitive_type*8 (~)
2936 | lw LFUNC:RB, FRAME_FUNC(BASE)
2937 | srl RA, RA, 1
2938 | srl TMP0, RD, 3
2939 | addu RA, RA, LFUNC:RB
2940 | not TMP0, TMP0
2941 | lw UPVAL:RB, LFUNC:RA->uvptr
2942 | ins_next1
2943 | lw TMP1, UPVAL:RB->v
2944 | sw TMP0, HI(TMP1)
2945 | ins_next2
2946 break;
2947
2948 case BC_UCLO:
2949 | // RA = level*8, RD = target
2950 | lw TMP2, L->openupval
2951 | branch_RD // Do this first since RD is not saved.
2952 | load_got lj_func_closeuv
2953 | sw BASE, L->base
2954 | beqz TMP2, >1
2955 |. move CARG1, L
2956 | call_intern lj_func_closeuv // (lua_State *L, TValue *level)
2957 |. addu CARG2, BASE, RA
2958 | lw BASE, L->base
2959 |1:
2960 | ins_next
2961 break;
2962
2963 case BC_FNEW:
2964 | // RA = dst*8, RD = proto_const*8 (~) (holding function prototype)
2965 | srl TMP1, RD, 1
2966 | load_got lj_func_newL_gc
2967 | subu TMP1, KBASE, TMP1
2968 | lw CARG3, FRAME_FUNC(BASE)
2969 | lw CARG2, -4(TMP1) // KBASE-4-tab_const*4
2970 | sw BASE, L->base
2971 | sw PC, SAVE_PC
2972 | // (lua_State *L, GCproto *pt, GCfuncL *parent)
2973 | call_intern lj_func_newL_gc
2974 |. move CARG1, L
2975 | // Returns GCfuncL *.
2976 | lw BASE, L->base
2977 | li TMP0, LJ_TFUNC
2978 | ins_next1
2979 | addu RA, BASE, RA
2980 | sw TMP0, HI(RA)
2981 | sw LFUNC:CRET1, LO(RA)
2982 | ins_next2
2983 break;
2984
2985 /* -- Table ops --------------------------------------------------------- */
2986
2987 case BC_TNEW:
2988 case BC_TDUP:
2989 | // RA = dst*8, RD = (hbits|asize)*8 | tab_const*8 (~)
2990 | lw TMP0, DISPATCH_GL(gc.total)(DISPATCH)
2991 | lw TMP1, DISPATCH_GL(gc.threshold)(DISPATCH)
2992 | sw BASE, L->base
2993 | sw PC, SAVE_PC
2994 | sltu AT, TMP0, TMP1
2995 | beqz AT, >5
2996 |1:
2997 if (op == BC_TNEW) {
2998 | load_got lj_tab_new
2999 | srl CARG2, RD, 3
3000 | andi CARG2, CARG2, 0x7ff
3001 | li TMP0, 0x801
3002 | addiu AT, CARG2, -0x7ff
3003 | srl CARG3, RD, 14
3004 | movz CARG2, TMP0, AT
3005 | // (lua_State *L, int32_t asize, uint32_t hbits)
3006 | call_intern lj_tab_new
3007 |. move CARG1, L
3008 | // Returns Table *.
3009 } else {
3010 | load_got lj_tab_dup
3011 | srl TMP1, RD, 1
3012 | subu TMP1, KBASE, TMP1
3013 | move CARG1, L
3014 | call_intern lj_tab_dup // (lua_State *L, Table *kt)
3015 |. lw CARG2, -4(TMP1) // KBASE-4-str_const*4
3016 | // Returns Table *.
3017 }
3018 | lw BASE, L->base
3019 | ins_next1
3020 | addu RA, BASE, RA
3021 | li TMP0, LJ_TTAB
3022 | sw TAB:CRET1, LO(RA)
3023 | sw TMP0, HI(RA)
3024 | ins_next2
3025 |5:
3026 | load_got lj_gc_step_fixtop
3027 | move MULTRES, RD
3028 | call_intern lj_gc_step_fixtop // (lua_State *L)
3029 |. move CARG1, L
3030 | b <1
3031 |. move RD, MULTRES
3032 break;
3033
3034 case BC_GGET:
3035 | // RA = dst*8, RD = str_const*8 (~)
3036 case BC_GSET:
3037 | // RA = src*8, RD = str_const*8 (~)
3038 | lw LFUNC:TMP2, FRAME_FUNC(BASE)
3039 | srl TMP1, RD, 1
3040 | subu TMP1, KBASE, TMP1
3041 | lw TAB:RB, LFUNC:TMP2->env
3042 | lw STR:RC, -4(TMP1) // KBASE-4-str_const*4
3043 if (op == BC_GGET) {
3044 | b ->BC_TGETS_Z
3045 } else {
3046 | b ->BC_TSETS_Z
3047 }
3048 |. addu RA, BASE, RA
3049 break;
3050
3051 case BC_TGETV:
3052 | // RA = dst*8, RB = table*8, RC = key*8
3053 | decode_RB8a RB, INS
3054 | decode_RB8b RB
3055 | decode_RDtoRC8 RC, RD
3056 | addu CARG2, BASE, RB
3057 | addu CARG3, BASE, RC
3058 | lw TMP1, HI(CARG2)
3059 | lw TMP2, HI(CARG3)
3060 | lw TAB:RB, LO(CARG2)
3061 | li AT, LJ_TTAB
3062 | ldc1 f0, 0(CARG3)
3063 | bne TMP1, AT, ->vmeta_tgetv
3064 |. addu RA, BASE, RA
3065 | sltiu AT, TMP2, LJ_TISNUM
3066 | beqz AT, >5
3067 |. li AT, LJ_TSTR
3068 |
3069 | // Convert number key to integer, check for integerness and range.
3070 | cvt.w.d f2, f0
3071 | lw TMP0, TAB:RB->asize
3072 | mfc1 TMP2, f2
3073 | cvt.d.w f4, f2
3074 | lw TMP1, TAB:RB->array
3075 | c.eq.d f0, f4
3076 | sltu AT, TMP2, TMP0
3077 | movf AT, r0
3078 | sll TMP2, TMP2, 3
3079 | beqz AT, ->vmeta_tgetv // Integer key and in array part?
3080 |. addu TMP2, TMP1, TMP2
3081 | lw TMP0, HI(TMP2)
3082 | beq TMP0, TISNIL, >2
3083 |. ldc1 f0, 0(TMP2)
3084 |1:
3085 | ins_next1
3086 | sdc1 f0, 0(RA)
3087 | ins_next2
3088 |
3089 |2: // Check for __index if table value is nil.
3090 | lw TAB:TMP2, TAB:RB->metatable
3091 | beqz TAB:TMP2, <1 // No metatable: done.
3092 |. nop
3093 | lbu TMP0, TAB:TMP2->nomm
3094 | andi TMP0, TMP0, 1<<MM_index
3095 | bnez TMP0, <1 // 'no __index' flag set: done.
3096 |. nop
3097 | b ->vmeta_tgetv
3098 |. nop
3099 |
3100 |5:
3101 | bne TMP2, AT, ->vmeta_tgetv
3102 |. lw STR:RC, LO(CARG3)
3103 | b ->BC_TGETS_Z // String key?
3104 |. nop
3105 break;
3106 case BC_TGETS:
3107 | // RA = dst*8, RB = table*8, RC = str_const*4 (~)
3108 | decode_RB8a RB, INS
3109 | decode_RB8b RB
3110 | addu CARG2, BASE, RB
3111 | decode_RC4a RC, INS
3112 | lw TMP0, HI(CARG2)
3113 | decode_RC4b RC
3114 | li AT, LJ_TTAB
3115 | lw TAB:RB, LO(CARG2)
3116 | subu CARG3, KBASE, RC
3117 | lw STR:RC, -4(CARG3) // KBASE-4-str_const*4
3118 | bne TMP0, AT, ->vmeta_tgets1
3119 |. addu RA, BASE, RA
3120 |->BC_TGETS_Z:
3121 | // TAB:RB = GCtab *, STR:RC = GCstr *, RA = dst*8
3122 | lw TMP0, TAB:RB->hmask
3123 | lw TMP1, STR:RC->hash
3124 | lw NODE:TMP2, TAB:RB->node
3125 | and TMP1, TMP1, TMP0 // idx = str->hash & tab->hmask
3126 | sll TMP0, TMP1, 5
3127 | sll TMP1, TMP1, 3
3128 | subu TMP1, TMP0, TMP1
3129 | addu NODE:TMP2, NODE:TMP2, TMP1 // node = tab->node + (idx*32-idx*8)
3130 |1:
3131 | lw CARG1, offsetof(Node, key)+HI(NODE:TMP2)
3132 | lw TMP0, offsetof(Node, key)+LO(NODE:TMP2)
3133 | lw NODE:TMP1, NODE:TMP2->next
3134 | lw CARG2, offsetof(Node, val)+HI(NODE:TMP2)
3135 | addiu CARG1, CARG1, -LJ_TSTR
3136 | xor TMP0, TMP0, STR:RC
3137 | or AT, CARG1, TMP0
3138 | bnez AT, >4
3139 |. lw TAB:TMP3, TAB:RB->metatable
3140 | beq CARG2, TISNIL, >5 // Key found, but nil value?
3141 |. lw CARG1, offsetof(Node, val)+LO(NODE:TMP2)
3142 |3:
3143 | ins_next1
3144 | sw CARG2, HI(RA)
3145 | sw CARG1, LO(RA)
3146 | ins_next2
3147 |
3148 |4: // Follow hash chain.
3149 | bnez NODE:TMP1, <1
3150 |. move NODE:TMP2, NODE:TMP1
3151 | // End of hash chain: key not found, nil result.
3152 |
3153 |5: // Check for __index if table value is nil.
3154 | beqz TAB:TMP3, <3 // No metatable: done.
3155 |. li CARG2, LJ_TNIL
3156 | lbu TMP0, TAB:TMP3->nomm
3157 | andi TMP0, TMP0, 1<<MM_index
3158 | bnez TMP0, <3 // 'no __index' flag set: done.
3159 |. nop
3160 | b ->vmeta_tgets
3161 |. nop
3162 break;
3163 case BC_TGETB:
3164 | // RA = dst*8, RB = table*8, RC = index*8
3165 | decode_RB8a RB, INS
3166 | decode_RB8b RB
3167 | addu CARG2, BASE, RB
3168 | decode_RDtoRC8 RC, RD
3169 | lw CARG1, HI(CARG2)
3170 | li AT, LJ_TTAB
3171 | lw TAB:RB, LO(CARG2)
3172 | addu RA, BASE, RA
3173 | bne CARG1, AT, ->vmeta_tgetb
3174 |. srl TMP0, RC, 3
3175 | lw TMP1, TAB:RB->asize
3176 | lw TMP2, TAB:RB->array
3177 | sltu AT, TMP0, TMP1
3178 | beqz AT, ->vmeta_tgetb
3179 |. addu RC, TMP2, RC
3180 | lw TMP1, HI(RC)
3181 | beq TMP1, TISNIL, >5
3182 |. ldc1 f0, 0(RC)
3183 |1:
3184 | ins_next1
3185 | sdc1 f0, 0(RA)
3186 | ins_next2
3187 |
3188 |5: // Check for __index if table value is nil.
3189 | lw TAB:TMP2, TAB:RB->metatable
3190 | beqz TAB:TMP2, <1 // No metatable: done.
3191 |. nop
3192 | lbu TMP1, TAB:TMP2->nomm
3193 | andi TMP1, TMP1, 1<<MM_index
3194 | bnez TMP1, <1 // 'no __index' flag set: done.
3195 |. nop
3196 | b ->vmeta_tgetb // Caveat: preserve TMP0!
3197 |. nop
3198 break;
3199
3200 case BC_TSETV:
3201 | // RA = src*8, RB = table*8, RC = key*8
3202 | decode_RB8a RB, INS
3203 | decode_RB8b RB
3204 | decode_RDtoRC8 RC, RD
3205 | addu CARG2, BASE, RB
3206 | addu CARG3, BASE, RC
3207 | lw TMP1, HI(CARG2)
3208 | lw TMP2, HI(CARG3)
3209 | lw TAB:RB, LO(CARG2)
3210 | li AT, LJ_TTAB
3211 | ldc1 f0, 0(CARG3)
3212 | bne TMP1, AT, ->vmeta_tsetv
3213 |. addu RA, BASE, RA
3214 | sltiu AT, TMP2, LJ_TISNUM
3215 | beqz AT, >5
3216 |. li AT, LJ_TSTR
3217 |
3218 | // Convert number key to integer, check for integerness and range.
3219 | cvt.w.d f2, f0
3220 | lw TMP0, TAB:RB->asize
3221 | mfc1 TMP2, f2
3222 | cvt.d.w f4, f2
3223 | lw TMP1, TAB:RB->array
3224 | c.eq.d f0, f4
3225 | sltu AT, TMP2, TMP0
3226 | movf AT, r0
3227 | sll TMP2, TMP2, 3
3228 | beqz AT, ->vmeta_tsetv // Integer key and in array part?
3229 |. addu TMP1, TMP1, TMP2
3230 | lbu TMP3, TAB:RB->marked
3231 | lw TMP0, HI(TMP1)
3232 | beq TMP0, TISNIL, >3
3233 |. ldc1 f0, 0(RA)
3234 |1:
3235 | andi AT, TMP3, LJ_GC_BLACK // isblack(table)
3236 | bnez AT, >7
3237 |. sdc1 f0, 0(TMP1)
3238 |2:
3239 | ins_next
3240 |
3241 |3: // Check for __newindex if previous value is nil.
3242 | lw TAB:TMP2, TAB:RB->metatable
3243 | beqz TAB:TMP2, <1 // No metatable: done.
3244 |. nop
3245 | lbu TMP2, TAB:TMP2->nomm
3246 | andi TMP2, TMP2, 1<<MM_newindex
3247 | bnez TMP2, <1 // 'no __newindex' flag set: done.
3248 |. nop
3249 | b ->vmeta_tsetv
3250 |. nop
3251 |
3252 |5:
3253 | bne TMP2, AT, ->vmeta_tsetv
3254 |. lw STR:RC, LO(CARG3)
3255 | b ->BC_TSETS_Z // String key?
3256 |. nop
3257 |
3258 |7: // Possible table write barrier for the value. Skip valiswhite check.
3259 | barrierback TAB:RB, TMP3, TMP0, <2
3260 break;
3261 case BC_TSETS:
3262 | // RA = src*8, RB = table*8, RC = str_const*8 (~)
3263 | decode_RB8a RB, INS
3264 | decode_RB8b RB
3265 | addu CARG2, BASE, RB
3266 | decode_RC4a RC, INS
3267 | lw TMP0, HI(CARG2)
3268 | decode_RC4b RC
3269 | li AT, LJ_TTAB
3270 | subu CARG3, KBASE, RC
3271 | lw TAB:RB, LO(CARG2)
3272 | lw STR:RC, -4(CARG3) // KBASE-4-str_const*4
3273 | bne TMP0, AT, ->vmeta_tsets1
3274 |. addu RA, BASE, RA
3275 |->BC_TSETS_Z:
3276 | // TAB:RB = GCtab *, STR:RC = GCstr *, RA = BASE+src*8
3277 | lw TMP0, TAB:RB->hmask
3278 | lw TMP1, STR:RC->hash
3279 | lw NODE:TMP2, TAB:RB->node
3280 | sb r0, TAB:RB->nomm // Clear metamethod cache.
3281 | and TMP1, TMP1, TMP0 // idx = str->hash & tab->hmask
3282 | sll TMP0, TMP1, 5
3283 | sll TMP1, TMP1, 3
3284 | subu TMP1, TMP0, TMP1
3285 | addu NODE:TMP2, NODE:TMP2, TMP1 // node = tab->node + (idx*32-idx*8)
3286 | ldc1 f20, 0(RA)
3287 |1:
3288 | lw CARG1, offsetof(Node, key)+HI(NODE:TMP2)
3289 | lw TMP0, offsetof(Node, key)+LO(NODE:TMP2)
3290 | li AT, LJ_TSTR
3291 | lw NODE:TMP1, NODE:TMP2->next
3292 | bne CARG1, AT, >5
3293 |. lw CARG2, offsetof(Node, val)+HI(NODE:TMP2)
3294 | bne TMP0, STR:RC, >5
3295 |. lbu TMP3, TAB:RB->marked
3296 | beq CARG2, TISNIL, >4 // Key found, but nil value?
3297 |. lw TAB:TMP0, TAB:RB->metatable
3298 |2:
3299 | andi AT, TMP3, LJ_GC_BLACK // isblack(table)
3300 | bnez AT, >7
3301 |. sdc1 f20, NODE:TMP2->val
3302 |3:
3303 | ins_next
3304 |
3305 |4: // Check for __newindex if previous value is nil.
3306 | beqz TAB:TMP0, <2 // No metatable: done.
3307 |. nop
3308 | lbu TMP0, TAB:TMP0->nomm
3309 | andi TMP0, TMP0, 1<<MM_newindex
3310 | bnez TMP0, <2 // 'no __newindex' flag set: done.
3311 |. nop
3312 | b ->vmeta_tsets
3313 |. nop
3314 |
3315 |5: // Follow hash chain.
3316 | bnez NODE:TMP1, <1
3317 |. move NODE:TMP2, NODE:TMP1
3318 | // End of hash chain: key not found, add a new one
3319 |
3320 | // But check for __newindex first.
3321 | lw TAB:TMP2, TAB:RB->metatable
3322 | beqz TAB:TMP2, >6 // No metatable: continue.
3323 |. addiu CARG3, DISPATCH, DISPATCH_GL(tmptv)
3324 | lbu TMP0, TAB:TMP2->nomm
3325 | andi TMP0, TMP0, 1<<MM_newindex
3326 | beqz TMP0, ->vmeta_tsets // 'no __newindex' flag NOT set: check.
3327 |. li AT, LJ_TSTR
3328 |6:
3329 | load_got lj_tab_newkey
3330 | sw STR:RC, LO(CARG3)
3331 | sw AT, HI(CARG3)
3332 | sw BASE, L->base
3333 | move CARG2, TAB:RB
3334 | sw PC, SAVE_PC
3335 | call_intern lj_tab_newkey // (lua_State *L, GCtab *t, TValue *k
3336 |. move CARG1, L
3337 | // Returns TValue *.
3338 | lw BASE, L->base
3339 | b <3 // No 2nd write barrier needed.
3340 |. sdc1 f20, 0(CRET1)
3341 |
3342 |7: // Possible table write barrier for the value. Skip valiswhite check.
3343 | barrierback TAB:RB, TMP3, TMP0, <3
3344 break;
3345 case BC_TSETB:
3346 | // RA = src*8, RB = table*8, RC = index*8
3347 | decode_RB8a RB, INS
3348 | decode_RB8b RB
3349 | addu CARG2, BASE, RB
3350 | decode_RDtoRC8 RC, RD
3351 | lw CARG1, HI(CARG2)
3352 | li AT, LJ_TTAB
3353 | lw TAB:RB, LO(CARG2)
3354 | addu RA, BASE, RA
3355 | bne CARG1, AT, ->vmeta_tsetb
3356 |. srl TMP0, RC, 3
3357 | lw TMP1, TAB:RB->asize
3358 | lw TMP2, TAB:RB->array
3359 | sltu AT, TMP0, TMP1
3360 | beqz AT, ->vmeta_tsetb
3361 |. addu RC, TMP2, RC
3362 | lw TMP1, HI(RC)
3363 | lbu TMP3, TAB:RB->marked
3364 | beq TMP1, TISNIL, >5
3365 |. ldc1 f0, 0(RA)
3366 |1:
3367 | andi AT, TMP3, LJ_GC_BLACK // isblack(table)
3368 | bnez AT, >7
3369 |. sdc1 f0, 0(RC)
3370 |2:
3371 | ins_next
3372 |
3373 |5: // Check for __newindex if previous value is nil.
3374 | lw TAB:TMP2, TAB:RB->metatable
3375 | beqz TAB:TMP2, <1 // No metatable: done.
3376 |. nop
3377 | lbu TMP1, TAB:TMP2->nomm
3378 | andi TMP1, TMP1, 1<<MM_newindex
3379 | bnez TMP1, <1 // 'no __newindex' flag set: done.
3380 |. nop
3381 | b ->vmeta_tsetb // Caveat: preserve TMP0!
3382 |. nop
3383 |
3384 |7: // Possible table write barrier for the value. Skip valiswhite check.
3385 | barrierback TAB:RB, TMP3, TMP0, <2
3386 break;
3387
3388 case BC_TSETM:
3389 | // RA = base*8 (table at base-1), RD = num_const*8 (start index)
3390 | addu RA, BASE, RA
3391 |1:
3392 | addu TMP3, KBASE, RD
3393 | lw TAB:CARG2, -8+LO(RA) // Guaranteed to be a table.
3394 | addiu TMP0, MULTRES, -8
3395 | lw TMP3, LO(TMP3) // Integer constant is in lo-word.
3396 | beqz TMP0, >4 // Nothing to copy?
3397 |. srl CARG3, TMP0, 3
3398 | addu CARG3, CARG3, TMP3
3399 | lw TMP2, TAB:CARG2->asize
3400 | sll TMP1, TMP3, 3
3401 | lbu TMP3, TAB:CARG2->marked
3402 | lw CARG1, TAB:CARG2->array
3403 | sltu AT, TMP2, CARG3
3404 | bnez AT, >5
3405 |. addu TMP2, RA, TMP0
3406 | addu TMP1, TMP1, CARG1
3407 | andi TMP0, TMP3, LJ_GC_BLACK // isblack(table)
3408 |3: // Copy result slots to table.
3409 | ldc1 f0, 0(RA)
3410 | addiu RA, RA, 8
3411 | sltu AT, RA, TMP2
3412 | sdc1 f0, 0(TMP1)
3413 | bnez AT, <3
3414 |. addiu TMP1, TMP1, 8
3415 | bnez TMP0, >7
3416 |. nop
3417 |4:
3418 | ins_next
3419 |
3420 |5: // Need to resize array part.
3421 | load_got lj_tab_reasize
3422 | sw BASE, L->base
3423 | sw PC, SAVE_PC
3424 | move BASE, RD
3425 | call_intern lj_tab_reasize // (lua_State *L, GCtab *t, int nasize)
3426 |. move CARG1, L
3427 | // Must not reallocate the stack.
3428 | move RD, BASE
3429 | b <1
3430 |. lw BASE, L->base // Reload BASE for lack of a saved register.
3431 |
3432 |7: // Possible table write barrier for any value. Skip valiswhite check.
3433 | barrierback TAB:CARG2, TMP3, TMP0, <4
3434 break;
3435
3436 /* -- Calls and vararg handling ----------------------------------------- */
3437
3438 case BC_CALLM:
3439 | // RA = base*8, (RB = (nresults+1)*8,) RC = extra_nargs*8
3440 | decode_RDtoRC8 NARGS8:RC, RD
3441 | b ->BC_CALL_Z
3442 |. addu NARGS8:RC, NARGS8:RC, MULTRES
3443 break;
3444 case BC_CALL:
3445 | // RA = base*8, (RB = (nresults+1)*8,) RC = (nargs+1)*8
3446 | decode_RDtoRC8 NARGS8:RC, RD
3447 |->BC_CALL_Z:
3448 | move TMP2, BASE
3449 | addu BASE, BASE, RA
3450 | li AT, LJ_TFUNC
3451 | lw TMP0, HI(BASE)
3452 | lw LFUNC:RB, LO(BASE)
3453 | addiu BASE, BASE, 8
3454 | bne TMP0, AT, ->vmeta_call
3455 |. addiu NARGS8:RC, NARGS8:RC, -8
3456 | ins_call
3457 break;
3458
3459 case BC_CALLMT:
3460 | // RA = base*8, (RB = 0,) RC = extra_nargs*8
3461 | addu NARGS8:RD, NARGS8:RD, MULTRES // BC_CALLT gets RC from RD.
3462 | // Fall through. Assumes BC_CALLT follows.
3463 break;
3464 case BC_CALLT:
3465 | // RA = base*8, (RB = 0,) RC = (nargs+1)*8
3466 | addu RA, BASE, RA
3467 | li AT, LJ_TFUNC
3468 | lw TMP0, HI(RA)
3469 | lw LFUNC:RB, LO(RA)
3470 | move NARGS8:RC, RD
3471 | lw TMP1, FRAME_PC(BASE)
3472 | addiu RA, RA, 8
3473 | bne TMP0, AT, ->vmeta_callt
3474 |. addiu NARGS8:RC, NARGS8:RC, -8
3475 |->BC_CALLT_Z:
3476 | andi TMP0, TMP1, FRAME_TYPE // Caveat: preserve TMP0 until the 'or'.
3477 | lbu TMP3, LFUNC:RB->ffid
3478 | bnez TMP0, >7
3479 |. xori TMP2, TMP1, FRAME_VARG
3480 |1:
3481 | sw LFUNC:RB, FRAME_FUNC(BASE) // Copy function down, but keep PC.
3482 | sltiu AT, TMP3, 2 // (> FF_C) Calling a fast function?
3483 | move TMP2, BASE
3484 | beqz NARGS8:RC, >3
3485 |. move TMP3, NARGS8:RC
3486 |2:
3487 | ldc1 f0, 0(RA)
3488 | addiu RA, RA, 8
3489 | addiu TMP3, TMP3, -8
3490 | sdc1 f0, 0(TMP2)
3491 | bnez TMP3, <2
3492 |. addiu TMP2, TMP2, 8
3493 |3:
3494 | or TMP0, TMP0, AT
3495 | beqz TMP0, >5
3496 |. nop
3497 |4:
3498 | ins_callt
3499 |
3500 |5: // Tailcall to a fast function with a Lua frame below.
3501 | lw INS, -4(TMP1)
3502 | decode_RA8a RA, INS
3503 | decode_RA8b RA
3504 | subu TMP1, BASE, RA
3505 | lw LFUNC:TMP1, -8+FRAME_FUNC(TMP1)
3506 | lw TMP1, LFUNC:TMP1->pc
3507 | b <4
3508 |. lw KBASE, PC2PROTO(k)(TMP1) // Need to prepare KBASE.
3509 |
3510 |7: // Tailcall from a vararg function.
3511 | andi AT, TMP2, FRAME_TYPEP
3512 | bnez AT, <1 // Vararg frame below?
3513 |. subu TMP2, BASE, TMP2 // Relocate BASE down.
3514 | move BASE, TMP2
3515 | lw TMP1, FRAME_PC(TMP2)
3516 | b <1
3517 |. andi TMP0, TMP1, FRAME_TYPE
3518 break;
3519
3520 case BC_ITERC:
3521 | // RA = base*8, (RB = (nresults+1)*8, RC = (nargs+1)*8 ((2+1)*8))
3522 | move TMP2, BASE
3523 | addu BASE, BASE, RA
3524 | li AT, LJ_TFUNC
3525 | lw TMP1, -24+HI(BASE)
3526 | lw LFUNC:RB, -24+LO(BASE)
3527 | ldc1 f2, -8(BASE)
3528 | ldc1 f0, -16(BASE)
3529 | sw TMP1, HI(BASE) // Copy callable.
3530 | sw LFUNC:RB, LO(BASE)
3531 | sdc1 f2, 16(BASE) // Copy control var.
3532 | sdc1 f0, 8(BASE) // Copy state.
3533 | addiu BASE, BASE, 8
3534 | bne TMP1, AT, ->vmeta_call
3535 |. li NARGS8:RC, 16 // Iterators get 2 arguments.
3536 | ins_call
3537 break;
3538
3539 case BC_ITERN:
3540 | // RA = base*8, (RB = (nresults+1)*8, RC = (nargs+1)*8 (2+1)*8)
3541#if LJ_HASJIT
3542 | // NYI: add hotloop, record BC_ITERN.
3543#endif
3544 | addu RA, BASE, RA
3545 | lw TAB:RB, -16+LO(RA)
3546 | lw RC, -8+LO(RA) // Get index from control var.
3547 | lw TMP0, TAB:RB->asize
3548 | lw TMP1, TAB:RB->array
3549 | addiu PC, PC, 4
3550 |1: // Traverse array part.
3551 | sltu AT, RC, TMP0
3552 | beqz AT, >5 // Index points after array part?
3553 |. sll TMP3, RC, 3
3554 | addu TMP3, TMP1, TMP3
3555 | lw TMP2, HI(TMP3)
3556 | ldc1 f0, 0(TMP3)
3557 | mtc1 RC, f2
3558 | lhu RD, -4+OFS_RD(PC)
3559 | beq TMP2, TISNIL, <1 // Skip holes in array part.
3560 |. addiu RC, RC, 1
3561 | cvt.d.w f2, f2
3562 | lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535)
3563 | sdc1 f0, 8(RA)
3564 | decode_RD4b RD
3565 | addu RD, RD, TMP3
3566 | sw RC, -8+LO(RA) // Update control var.
3567 | addu PC, PC, RD
3568 | sdc1 f2, 0(RA)
3569 |3:
3570 | ins_next
3571 |
3572 |5: // Traverse hash part.
3573 | lw TMP1, TAB:RB->hmask
3574 | subu RC, RC, TMP0
3575 | lw TMP2, TAB:RB->node
3576 |6:
3577 | sltu AT, TMP1, RC // End of iteration? Branch to ITERL+1.
3578 | bnez AT, <3
3579 |. sll TMP3, RC, 5
3580 | sll RB, RC, 3
3581 | subu TMP3, TMP3, RB
3582 | addu NODE:TMP3, TMP3, TMP2
3583 | lw RB, HI(NODE:TMP3)
3584 | ldc1 f0, 0(NODE:TMP3)
3585 | lhu RD, -4+OFS_RD(PC)
3586 | beq RB, TISNIL, <6 // Skip holes in hash part.
3587 |. addiu RC, RC, 1
3588 | ldc1 f2, NODE:TMP3->key
3589 | lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535)
3590 | sdc1 f0, 8(RA)
3591 | addu RC, RC, TMP0
3592 | decode_RD4b RD
3593 | addu RD, RD, TMP3
3594 | sdc1 f2, 0(RA)
3595 | addu PC, PC, RD
3596 | b <3
3597 |. sw RC, -8+LO(RA) // Update control var.
3598 break;
3599
3600 case BC_ISNEXT:
3601 | // RA = base*8, RD = target (points to ITERN)
3602 | addu RA, BASE, RA
3603 | lw TMP0, -24+HI(RA)
3604 | lw CFUNC:TMP1, -24+LO(RA)
3605 | lw TMP2, -16+HI(RA)
3606 | lw TMP3, -8+HI(RA)
3607 | li AT, LJ_TFUNC
3608 | bne TMP0, AT, >5
3609 |. addiu TMP2, TMP2, -LJ_TTAB
3610 | lbu TMP1, CFUNC:TMP1->ffid
3611 | addiu TMP3, TMP3, -LJ_TNIL
3612 | srl TMP0, RD, 1
3613 | or TMP2, TMP2, TMP3
3614 | addiu TMP1, TMP1, -FF_next_N
3615 | addu TMP0, PC, TMP0
3616 | or TMP1, TMP1, TMP2
3617 | bnez TMP1, >5
3618 |. lui TMP2, (-(BCBIAS_J*4 >> 16) & 65535)
3619 | addu PC, TMP0, TMP2
3620 | sw r0, -8+LO(RA) // Initialize control var.
3621 |1:
3622 | ins_next
3623 |5: // Despecialize bytecode if any of the checks fail.
3624 | li TMP3, BC_JMP
3625 | li TMP1, BC_ITERC
3626 | sb TMP3, -4+OFS_OP(PC)
3627 | addu PC, TMP0, TMP2
3628 | b <1
3629 |. sb TMP1, OFS_OP(PC)
3630 break;
3631
3632 case BC_VARG:
3633 | // RA = base*8, RB = (nresults+1)*8, RC = numparams*8
3634 | lw TMP0, FRAME_PC(BASE)
3635 | decode_RDtoRC8 RC, RD
3636 | decode_RB8a RB, INS
3637 | addu RC, BASE, RC
3638 | decode_RB8b RB
3639 | addu RA, BASE, RA
3640 | addiu RC, RC, FRAME_VARG
3641 | addu TMP2, RA, RB
3642 | addiu TMP3, BASE, -8 // TMP3 = vtop
3643 | subu RC, RC, TMP0 // RC = vbase
3644 | // Note: RC may now be even _above_ BASE if nargs was < numparams.
3645 | beqz RB, >5 // Copy all varargs?
3646 |. subu TMP1, TMP3, RC
3647 | addiu TMP2, TMP2, -16
3648 |1: // Copy vararg slots to destination slots.
3649 | lw CARG1, HI(RC)
3650 | sltu AT, RC, TMP3
3651 | lw CARG2, LO(RC)
3652 | addiu RC, RC, 8
3653 | movz CARG1, TISNIL, AT
3654 | sw CARG1, HI(RA)
3655 | sw CARG2, LO(RA)
3656 | sltu AT, RA, TMP2
3657 | bnez AT, <1
3658 |. addiu RA, RA, 8
3659 |3:
3660 | ins_next
3661 |
3662 |5: // Copy all varargs.
3663 | lw TMP0, L->maxstack
3664 | blez TMP1, <3 // No vararg slots?
3665 |. li MULTRES, 8 // MULTRES = (0+1)*8
3666 | addu TMP2, RA, TMP1
3667 | sltu AT, TMP0, TMP2
3668 | bnez AT, >7
3669 |. addiu MULTRES, TMP1, 8
3670 |6:
3671 | ldc1 f0, 0(RC)
3672 | addiu RC, RC, 8
3673 | sdc1 f0, 0(RA)
3674 | sltu AT, RC, TMP3
3675 | bnez AT, <6 // More vararg slots?
3676 |. addiu RA, RA, 8
3677 | b <3
3678 |. nop
3679 |
3680 |7: // Grow stack for varargs.
3681 | load_got lj_state_growstack
3682 | sw RA, L->top
3683 | subu RA, RA, BASE
3684 | sw BASE, L->base
3685 | subu BASE, RC, BASE // Need delta, because BASE may change.
3686 | sw PC, SAVE_PC
3687 | srl CARG2, TMP1, 3
3688 | call_intern lj_state_growstack // (lua_State *L, int n)
3689 |. move CARG1, L
3690 | move RC, BASE
3691 | lw BASE, L->base
3692 | addu RA, BASE, RA
3693 | addu RC, BASE, RC
3694 | b <6
3695 |. addiu TMP3, BASE, -8
3696 break;
3697
3698 /* -- Returns ----------------------------------------------------------- */
3699
3700 case BC_RETM:
3701 | // RA = results*8, RD = extra_nresults*8
3702 | addu RD, RD, MULTRES // MULTRES >= 8, so RD >= 8.
3703 | // Fall through. Assumes BC_RET follows.
3704 break;
3705
3706 case BC_RET:
3707 | // RA = results*8, RD = (nresults+1)*8
3708 | lw PC, FRAME_PC(BASE)
3709 | addu RA, BASE, RA
3710 | move MULTRES, RD
3711 |1:
3712 | andi TMP0, PC, FRAME_TYPE
3713 | bnez TMP0, ->BC_RETV_Z
3714 |. xori TMP1, PC, FRAME_VARG
3715 |
3716 |->BC_RET_Z:
3717 | // BASE = base, RA = resultptr, RD = (nresults+1)*8, PC = return
3718 | lw INS, -4(PC)
3719 | addiu TMP2, BASE, -8
3720 | addiu RC, RD, -8
3721 | decode_RA8a TMP0, INS
3722 | decode_RB8a RB, INS
3723 | decode_RA8b TMP0
3724 | decode_RB8b RB
3725 | addu TMP3, TMP2, RB
3726 | beqz RC, >3
3727 |. subu BASE, TMP2, TMP0
3728 |2:
3729 | ldc1 f0, 0(RA)
3730 | addiu RA, RA, 8
3731 | addiu RC, RC, -8
3732 | sdc1 f0, 0(TMP2)
3733 | bnez RC, <2
3734 |. addiu TMP2, TMP2, 8
3735 |3:
3736 | addiu TMP3, TMP3, -8
3737 |5:
3738 | sltu AT, TMP2, TMP3
3739 | bnez AT, >6
3740 |. lw LFUNC:TMP1, FRAME_FUNC(BASE)
3741 | ins_next1
3742 | lw TMP1, LFUNC:TMP1->pc
3743 | lw KBASE, PC2PROTO(k)(TMP1)
3744 | ins_next2
3745 |
3746 |6: // Fill up results with nil.
3747 | sw TISNIL, HI(TMP2)
3748 | b <5
3749 |. addiu TMP2, TMP2, 8
3750 |
3751 |->BC_RETV_Z: // Non-standard return case.
3752 | andi TMP2, TMP1, FRAME_TYPEP
3753 | bnez TMP2, ->vm_return
3754 |. nop
3755 | // Return from vararg function: relocate BASE down.
3756 | subu BASE, BASE, TMP1
3757 | b <1
3758 |. lw PC, FRAME_PC(BASE)
3759 break;
3760
3761 case BC_RET0: case BC_RET1:
3762 | // RA = results*8, RD = (nresults+1)*8
3763 | lw PC, FRAME_PC(BASE)
3764 | addu RA, BASE, RA
3765 | move MULTRES, RD
3766 | andi TMP0, PC, FRAME_TYPE
3767 | bnez TMP0, ->BC_RETV_Z
3768 |. xori TMP1, PC, FRAME_VARG
3769 |
3770 | lw INS, -4(PC)
3771 | addiu TMP2, BASE, -8
3772 if (op == BC_RET1) {
3773 | ldc1 f0, 0(RA)
3774 }
3775 | decode_RB8a RB, INS
3776 | decode_RA8a RA, INS
3777 | decode_RB8b RB
3778 | decode_RA8b RA
3779 if (op == BC_RET1) {
3780 | sdc1 f0, 0(TMP2)
3781 }
3782 | subu BASE, TMP2, RA
3783 |5:
3784 | sltu AT, RD, RB
3785 | bnez AT, >6
3786 |. lw LFUNC:TMP1, FRAME_FUNC(BASE)
3787 | ins_next1
3788 | lw TMP1, LFUNC:TMP1->pc
3789 | lw KBASE, PC2PROTO(k)(TMP1)
3790 | ins_next2
3791 |
3792 |6: // Fill up results with nil.
3793 | addiu TMP2, TMP2, 8
3794 | addiu RD, RD, 8
3795 | b <5
3796 if (op == BC_RET1) {
3797 |. sw TISNIL, HI(TMP2)
3798 } else {
3799 |. sw TISNIL, -8+HI(TMP2)
3800 }
3801 break;
3802
3803 /* -- Loops and branches ------------------------------------------------ */
3804
3805 case BC_FORL:
3806#if LJ_HASJIT
3807 | hotloop
3808#endif
3809 | // Fall through. Assumes BC_IFORL follows.
3810 break;
3811
3812 case BC_JFORI:
3813 case BC_JFORL:
3814#if !LJ_HASJIT
3815 break;
3816#endif
3817 case BC_FORI:
3818 case BC_IFORL:
3819 | // RA = base*8, RD = target (after end of loop or start of loop)
3820 vk = (op == BC_IFORL || op == BC_JFORL);
3821 | addu RA, BASE, RA
3822 if (vk) {
3823 | ldc1 f0, FORL_IDX*8(RA)
3824 | ldc1 f4, FORL_STEP*8(RA)
3825 | ldc1 f2, FORL_STOP*8(RA)
3826 | lw TMP3, FORL_STEP*8+HI(RA)
3827 | add.d f0, f0, f4
3828 | sdc1 f0, FORL_IDX*8(RA)
3829 } else {
3830 | lw TMP1, FORL_IDX*8+HI(RA)
3831 | lw TMP3, FORL_STEP*8+HI(RA)
3832 | lw TMP2, FORL_STOP*8+HI(RA)
3833 | sltiu TMP1, TMP1, LJ_TISNUM
3834 | sltiu TMP0, TMP3, LJ_TISNUM
3835 | sltiu TMP2, TMP2, LJ_TISNUM
3836 | and TMP1, TMP1, TMP0
3837 | and TMP1, TMP1, TMP2
3838 | ldc1 f0, FORL_IDX*8(RA)
3839 | beqz TMP1, ->vmeta_for
3840 |. ldc1 f2, FORL_STOP*8(RA)
3841 }
3842 if (op != BC_JFORL) {
3843 | srl RD, RD, 1
3844 | lui TMP0, (-(BCBIAS_J*4 >> 16) & 65535)
3845 }
3846 | c.le.d 0, f0, f2
3847 | c.le.d 1, f2, f0
3848 | sdc1 f0, FORL_EXT*8(RA)
3849 if (op == BC_JFORI) {
3850 | li TMP1, 1
3851 | li TMP2, 1
3852 | addu TMP0, RD, TMP0
3853 | slt TMP3, TMP3, r0
3854 | movf TMP1, r0, 0
3855 | addu PC, PC, TMP0
3856 | movf TMP2, r0, 1
3857 | lhu RD, -4+OFS_RD(PC)
3858 | movn TMP1, TMP2, TMP3
3859 | bnez TMP1, =>BC_JLOOP
3860 |. decode_RD8b RD
3861 } else if (op == BC_JFORL) {
3862 | li TMP1, 1
3863 | li TMP2, 1
3864 | slt TMP3, TMP3, r0
3865 | movf TMP1, r0, 0
3866 | movf TMP2, r0, 1
3867 | movn TMP1, TMP2, TMP3
3868 | bnez TMP1, =>BC_JLOOP
3869 |. nop
3870 } else {
3871 | addu TMP1, RD, TMP0
3872 | slt TMP3, TMP3, r0
3873 | move TMP2, TMP1
3874 if (op == BC_FORI) {
3875 | movt TMP1, r0, 0
3876 | movt TMP2, r0, 1
3877 } else {
3878 | movf TMP1, r0, 0
3879 | movf TMP2, r0, 1
3880 }
3881 | movn TMP1, TMP2, TMP3
3882 | addu PC, PC, TMP1
3883 }
3884 | ins_next
3885 break;
3886
3887 case BC_ITERL:
3888#if LJ_HASJIT
3889 | hotloop
3890#endif
3891 | // Fall through. Assumes BC_IITERL follows.
3892 break;
3893
3894 case BC_JITERL:
3895#if !LJ_HASJIT
3896 break;
3897#endif
3898 case BC_IITERL:
3899 | // RA = base*8, RD = target
3900 | addu RA, BASE, RA
3901 | lw TMP1, HI(RA)
3902 | beq TMP1, TISNIL, >1 // Stop if iterator returned nil.
3903 |. lw TMP2, LO(RA)
3904 if (op == BC_JITERL) {
3905 | sw TMP1, -8+HI(RA)
3906 | b =>BC_JLOOP
3907 |. sw TMP2, -8+LO(RA)
3908 } else {
3909 | branch_RD // Otherwise save control var + branch.
3910 | sw TMP1, -8+HI(RA)
3911 | sw TMP2, -8+LO(RA)
3912 }
3913 |1:
3914 | ins_next
3915 break;
3916
3917 case BC_LOOP:
3918 | // RA = base*8, RD = target (loop extent)
3919 | // Note: RA/RD is only used by trace recorder to determine scope/extent
3920 | // This opcode does NOT jump, it's only purpose is to detect a hot loop.
3921#if LJ_HASJIT
3922 | hotloop
3923#endif
3924 | // Fall through. Assumes BC_ILOOP follows.
3925 break;
3926
3927 case BC_ILOOP:
3928 | // RA = base*8, RD = target (loop extent)
3929 | ins_next
3930 break;
3931
3932 case BC_JLOOP:
3933#if LJ_HASJIT
3934 | // RA = base*8 (ignored), RD = traceno*8
3935 | lw TMP1, DISPATCH_J(trace)(DISPATCH)
3936 | srl RD, RD, 1
3937 | li AT, 0
3938 | addu TMP1, TMP1, RD
3939 | // Traces on MIPS don't store the trace number, so use 0.
3940 | sw AT, DISPATCH_GL(vmstate)(DISPATCH)
3941 | lw TRACE:TMP2, 0(TMP1)
3942 | sw BASE, DISPATCH_GL(jit_base)(DISPATCH)
3943 | sw L, DISPATCH_GL(jit_L)(DISPATCH)
3944 | lw TMP2, TRACE:TMP2->mcode
3945 | jr TMP2
3946 |. addiu JGL, DISPATCH, GG_DISP2G+32768
3947#endif
3948 break;
3949
3950 case BC_JMP:
3951 | // RA = base*8 (only used by trace recorder), RD = target
3952 | branch_RD
3953 | ins_next
3954 break;
3955
3956 /* -- Function headers -------------------------------------------------- */
3957
3958 case BC_FUNCF:
3959#if LJ_HASJIT
3960 | hotcall
3961#endif
3962 case BC_FUNCV: /* NYI: compiled vararg functions. */
3963 | // Fall through. Assumes BC_IFUNCF/BC_IFUNCV follow.
3964 break;
3965
3966 case BC_JFUNCF:
3967#if !LJ_HASJIT
3968 break;
3969#endif
3970 case BC_IFUNCF:
3971 | // BASE = new base, RA = BASE+framesize*8, RB = LFUNC, RC = nargs*8
3972 | lw TMP2, L->maxstack
3973 | lbu TMP1, -4+PC2PROTO(numparams)(PC)
3974 | lw KBASE, -4+PC2PROTO(k)(PC)
3975 | sltu AT, TMP2, RA
3976 | bnez AT, ->vm_growstack_l
3977 |. sll TMP1, TMP1, 3
3978 if (op != BC_JFUNCF) {
3979 | ins_next1
3980 }
3981 |2:
3982 | sltu AT, NARGS8:RC, TMP1 // Check for missing parameters.
3983 | bnez AT, >3
3984 |. addu AT, BASE, NARGS8:RC
3985 if (op == BC_JFUNCF) {
3986 | decode_RD8a RD, INS
3987 | b =>BC_JLOOP
3988 |. decode_RD8b RD
3989 } else {
3990 | ins_next2
3991 }
3992 |
3993 |3: // Clear missing parameters.
3994 | sw TISNIL, HI(AT)
3995 | b <2
3996 |. addiu NARGS8:RC, NARGS8:RC, 8
3997 break;
3998
3999 case BC_JFUNCV:
4000#if !LJ_HASJIT
4001 break;
4002#endif
4003 | NYI // NYI: compiled vararg functions
4004 break; /* NYI: compiled vararg functions. */
4005
4006 case BC_IFUNCV:
4007 | // BASE = new base, RA = BASE+framesize*8, RB = LFUNC, RC = nargs*8
4008 | addu TMP1, BASE, RC
4009 | lw TMP2, L->maxstack
4010 | addu TMP0, RA, RC
4011 | sw LFUNC:RB, LO(TMP1) // Store copy of LFUNC.
4012 | addiu TMP3, RC, 8+FRAME_VARG
4013 | sltu AT, TMP0, TMP2
4014 | lw KBASE, -4+PC2PROTO(k)(PC)
4015 | beqz AT, ->vm_growstack_l
4016 |. sw TMP3, HI(TMP1) // Store delta + FRAME_VARG.
4017 | lbu TMP2, -4+PC2PROTO(numparams)(PC)
4018 | move RA, BASE
4019 | move RC, TMP1
4020 | ins_next1
4021 | beqz TMP2, >3
4022 |. addiu BASE, TMP1, 8
4023 |1:
4024 | lw TMP0, HI(RA)
4025 | lw TMP3, LO(RA)
4026 | sltu AT, RA, RC // Less args than parameters?
4027 | move CARG1, TMP0
4028 | movz TMP0, TISNIL, AT // Clear missing parameters.
4029 | movn CARG1, TISNIL, AT // Clear old fixarg slot (help the GC).
4030 | sw TMP3, 8+LO(TMP1)
4031 | addiu TMP2, TMP2, -1
4032 | sw TMP0, 8+HI(TMP1)
4033 | addiu TMP1, TMP1, 8
4034 | sw CARG1, HI(RA)
4035 | bnez TMP2, <1
4036 |. addiu RA, RA, 8
4037 |3:
4038 | ins_next2
4039 break;
4040
4041 case BC_FUNCC:
4042 case BC_FUNCCW:
4043 | // BASE = new base, RA = BASE+framesize*8, RB = CFUNC, RC = nargs*8
4044 if (op == BC_FUNCC) {
4045 | lw CFUNCADDR, CFUNC:RB->f
4046 } else {
4047 | lw CFUNCADDR, DISPATCH_GL(wrapf)(DISPATCH)
4048 }
4049 | addu TMP1, RA, NARGS8:RC
4050 | lw TMP2, L->maxstack
4051 | addu RC, BASE, NARGS8:RC
4052 | sw BASE, L->base
4053 | sltu AT, TMP2, TMP1
4054 | sw RC, L->top
4055 | li_vmstate C
4056 if (op == BC_FUNCCW) {
4057 | lw CARG2, CFUNC:RB->f
4058 }
4059 | bnez AT, ->vm_growstack_c // Need to grow stack.
4060 |. move CARG1, L
4061 | jalr CFUNCADDR // (lua_State *L [, lua_CFunction f])
4062 |. st_vmstate
4063 | // Returns nresults.
4064 | lw BASE, L->base
4065 | sll RD, CRET1, 3
4066 | lw TMP1, L->top
4067 | li_vmstate INTERP
4068 | lw PC, FRAME_PC(BASE) // Fetch PC of caller.
4069 | subu RA, TMP1, RD // RA = L->top - nresults*8
4070 | b ->vm_returnc
4071 |. st_vmstate
4072 break;
4073
4074 /* ---------------------------------------------------------------------- */
4075
4076 default:
4077 fprintf(stderr, "Error: undefined opcode BC_%s\n", bc_names[op]);
4078 exit(2);
4079 break;
4080 }
4081}
4082
4083static int build_backend(BuildCtx *ctx)
4084{
4085 int op;
4086
4087 dasm_growpc(Dst, BC__MAX);
4088
4089 build_subroutines(ctx);
4090
4091 |.code_op
4092 for (op = 0; op < BC__MAX; op++)
4093 build_ins(ctx, (BCOp)op, op);
4094
4095 return BC__MAX;
4096}
4097
4098/* Emit pseudo frame-info for all assembler functions. */
4099static void emit_asm_debug(BuildCtx *ctx)
4100{
4101 int fcofs = (int)((uint8_t *)ctx->glob[GLOB_vm_ffi_call] - ctx->code);
4102 int i;
4103 switch (ctx->mode) {
4104 case BUILD_elfasm:
4105 fprintf(ctx->fp, "\t.section .debug_frame,\"\",@progbits\n");
4106 fprintf(ctx->fp,
4107 ".Lframe0:\n"
4108 "\t.4byte .LECIE0-.LSCIE0\n"
4109 ".LSCIE0:\n"
4110 "\t.4byte 0xffffffff\n"
4111 "\t.byte 0x1\n"
4112 "\t.string \"\"\n"
4113 "\t.uleb128 0x1\n"
4114 "\t.sleb128 -4\n"
4115 "\t.byte 31\n"
4116 "\t.byte 0xc\n\t.uleb128 29\n\t.uleb128 0\n"
4117 "\t.align 2\n"
4118 ".LECIE0:\n\n");
4119 fprintf(ctx->fp,
4120 ".LSFDE0:\n"
4121 "\t.4byte .LEFDE0-.LASFDE0\n"
4122 ".LASFDE0:\n"
4123 "\t.4byte .Lframe0\n"
4124 "\t.4byte .Lbegin\n"
4125 "\t.4byte %d\n"
4126 "\t.byte 0xe\n\t.uleb128 %d\n"
4127 "\t.byte 0x9f\n\t.sleb128 1\n"
4128 "\t.byte 0x9e\n\t.sleb128 2\n",
4129 fcofs, CFRAME_SIZE);
4130 for (i = 23; i >= 16; i--)
4131 fprintf(ctx->fp, "\t.byte %d\n\t.uleb128 %d\n", 0x80+i, 26-i);
4132 for (i = 30; i >= 20; i -= 2)
4133 fprintf(ctx->fp, "\t.byte %d\n\t.uleb128 %d\n", 0x80+32+i, 42-i);
4134 fprintf(ctx->fp,
4135 "\t.align 2\n"
4136 ".LEFDE0:\n\n");
4137#if LJ_HASFFI
4138 fprintf(ctx->fp,
4139 ".LSFDE1:\n"
4140 "\t.4byte .LEFDE1-.LASFDE1\n"
4141 ".LASFDE1:\n"
4142 "\t.4byte .Lframe0\n"
4143 "\t.4byte lj_vm_ffi_call\n"
4144 "\t.4byte %d\n"
4145 "\t.byte 0x9f\n\t.uleb128 1\n"
4146 "\t.byte 0x90\n\t.uleb128 2\n"
4147 "\t.byte 0xd\n\t.uleb128 0x10\n"
4148 "\t.align 2\n"
4149 ".LEFDE1:\n\n", (int)ctx->codesz - fcofs);
4150#endif
4151 fprintf(ctx->fp, "\t.section .eh_frame,\"aw\",@progbits\n");
4152 fprintf(ctx->fp,
4153 "\t.globl lj_err_unwind_dwarf\n"
4154 ".Lframe1:\n"
4155 "\t.4byte .LECIE1-.LSCIE1\n"
4156 ".LSCIE1:\n"
4157 "\t.4byte 0\n"
4158 "\t.byte 0x1\n"
4159 "\t.string \"zPR\"\n"
4160 "\t.uleb128 0x1\n"
4161 "\t.sleb128 -4\n"
4162 "\t.byte 31\n"
4163 "\t.uleb128 6\n" /* augmentation length */
4164 "\t.byte 0\n"
4165 "\t.4byte lj_err_unwind_dwarf\n"
4166 "\t.byte 0\n"
4167 "\t.byte 0xc\n\t.uleb128 29\n\t.uleb128 0\n"
4168 "\t.align 2\n"
4169 ".LECIE1:\n\n");
4170 fprintf(ctx->fp,
4171 ".LSFDE2:\n"
4172 "\t.4byte .LEFDE2-.LASFDE2\n"
4173 ".LASFDE2:\n"
4174 "\t.4byte .LASFDE2-.Lframe1\n"
4175 "\t.4byte .Lbegin\n"
4176 "\t.4byte %d\n"
4177 "\t.uleb128 0\n" /* augmentation length */
4178 "\t.byte 0xe\n\t.uleb128 %d\n"
4179 "\t.byte 0x9f\n\t.sleb128 1\n"
4180 "\t.byte 0x9e\n\t.sleb128 2\n",
4181 fcofs, CFRAME_SIZE);
4182 for (i = 23; i >= 16; i--)
4183 fprintf(ctx->fp, "\t.byte %d\n\t.uleb128 %d\n", 0x80+i, 26-i);
4184 for (i = 30; i >= 20; i -= 2)
4185 fprintf(ctx->fp, "\t.byte %d\n\t.uleb128 %d\n", 0x80+32+i, 42-i);
4186 fprintf(ctx->fp,
4187 "\t.align 2\n"
4188 ".LEFDE2:\n\n");
4189#if LJ_HASFFI
4190 fprintf(ctx->fp,
4191 ".Lframe2:\n"
4192 "\t.4byte .LECIE2-.LSCIE2\n"
4193 ".LSCIE2:\n"
4194 "\t.4byte 0\n"
4195 "\t.byte 0x1\n"
4196 "\t.string \"zR\"\n"
4197 "\t.uleb128 0x1\n"
4198 "\t.sleb128 -4\n"
4199 "\t.byte 31\n"
4200 "\t.uleb128 1\n" /* augmentation length */
4201 "\t.byte 0\n"
4202 "\t.byte 0xc\n\t.uleb128 29\n\t.uleb128 0\n"
4203 "\t.align 2\n"
4204 ".LECIE2:\n\n");
4205 fprintf(ctx->fp,
4206 ".LSFDE3:\n"
4207 "\t.4byte .LEFDE3-.LASFDE3\n"
4208 ".LASFDE3:\n"
4209 "\t.4byte .LASFDE3-.Lframe2\n"
4210 "\t.4byte lj_vm_ffi_call\n"
4211 "\t.4byte %d\n"
4212 "\t.uleb128 0\n" /* augmentation length */
4213 "\t.byte 0x9f\n\t.uleb128 1\n"
4214 "\t.byte 0x90\n\t.uleb128 2\n"
4215 "\t.byte 0xd\n\t.uleb128 0x10\n"
4216 "\t.align 2\n"
4217 ".LEFDE3:\n\n", (int)ctx->codesz - fcofs);
4218#endif
4219 break;
4220 default:
4221 break;
4222 }
4223}
4224
generated by cgit v1.2.3-55-g6feb (git 2.39.1) at 2025-08-11 09:03:46 +0000