Mercurial

comparison third_party/luajit/src/vm_arm.dasc @ 186:8cf4ec5e2191 hg-web

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Fixed merge conflict.
author MrJuneJune <me@mrjunejune.com>
date Fri, 23 Jan 2026 22:38:59 -0800
parents 94705b5986b3
children
comparison
equal deleted inserted replaced
176:fed99fc04e12 186:8cf4ec5e2191
1 |// Low-level VM code for ARM CPUs.
2 |// Bytecode interpreter, fast functions and helper functions.
3 |// Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
4 |
5 |.arch arm
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 |
20 |// The following must be C callee-save.
21 |.define MASKR8, r4 // 255*8 constant for fast bytecode decoding.
22 |.define KBASE, r5 // Constants of current Lua function.
23 |.define PC, r6 // Next PC.
24 |.define DISPATCH, r7 // Opcode dispatch table.
25 |.define LREG, r8 // Register holding lua_State (also in SAVE_L).
26 |
27 |// C callee-save in EABI, but often refetched. Temporary in iOS 3.0+.
28 |.define BASE, r9 // Base of current Lua stack frame.
29 |
30 |// The following temporaries are not saved across C calls, except for RA/RC.
31 |.define RA, r10 // Callee-save.
32 |.define RC, r11 // Callee-save.
33 |.define RB, r12
34 |.define OP, r12 // Overlaps RB, must not be lr.
35 |.define INS, lr
36 |
37 |// Calling conventions. Also used as temporaries.
38 |.define CARG1, r0
39 |.define CARG2, r1
40 |.define CARG3, r2
41 |.define CARG4, r3
42 |.define CARG12, r0 // For 1st soft-fp double.
43 |.define CARG34, r2 // For 2nd soft-fp double.
44 |
45 |.define CRET1, r0
46 |.define CRET2, r1
47 |
48 |// Stack layout while in interpreter. Must match with lj_frame.h.
49 |.define SAVE_R4, [sp, #28]
50 |.define CFRAME_SPACE, #28
51 |.define SAVE_ERRF, [sp, #24]
52 |.define SAVE_NRES, [sp, #20]
53 |.define SAVE_CFRAME, [sp, #16]
54 |.define SAVE_L, [sp, #12]
55 |.define SAVE_PC, [sp, #8]
56 |.define SAVE_MULTRES, [sp, #4]
57 |.define ARG5, [sp]
58 |
59 |.define TMPDhi, [sp, #4]
60 |.define TMPDlo, [sp]
61 |.define TMPD, [sp]
62 |.define TMPDp, sp
63 |
64 |.if FPU
65 |.macro saveregs
66 | push {r5, r6, r7, r8, r9, r10, r11, lr}
67 | vpush {d8-d15}
68 | sub sp, sp, CFRAME_SPACE+4
69 | str r4, SAVE_R4
70 |.endmacro
71 |.macro restoreregs_ret
72 | ldr r4, SAVE_R4
73 | add sp, sp, CFRAME_SPACE+4
74 | vpop {d8-d15}
75 | pop {r5, r6, r7, r8, r9, r10, r11, pc}
76 |.endmacro
77 |.else
78 |.macro saveregs
79 | push {r4, r5, r6, r7, r8, r9, r10, r11, lr}
80 | sub sp, sp, CFRAME_SPACE
81 |.endmacro
82 |.macro restoreregs_ret
83 | add sp, sp, CFRAME_SPACE
84 | pop {r4, r5, r6, r7, r8, r9, r10, r11, pc}
85 |.endmacro
86 |.endif
87 |
88 |// Type definitions. Some of these are only used for documentation.
89 |.type L, lua_State, LREG
90 |.type GL, global_State
91 |.type TVALUE, TValue
92 |.type GCOBJ, GCobj
93 |.type STR, GCstr
94 |.type TAB, GCtab
95 |.type LFUNC, GCfuncL
96 |.type CFUNC, GCfuncC
97 |.type PROTO, GCproto
98 |.type UPVAL, GCupval
99 |.type NODE, Node
100 |.type NARGS8, int
101 |.type TRACE, GCtrace
102 |.type SBUF, SBuf
103 |
104 |//-----------------------------------------------------------------------
105 |
106 |// Trap for not-yet-implemented parts.
107 |.macro NYI; ud; .endmacro
108 |
109 |//-----------------------------------------------------------------------
110 |
111 |// Access to frame relative to BASE.
112 |.define FRAME_FUNC, #-8
113 |.define FRAME_PC, #-4
114 |
115 |.macro decode_RA8, dst, ins; and dst, MASKR8, ins, lsr #5; .endmacro
116 |.macro decode_RB8, dst, ins; and dst, MASKR8, ins, lsr #21; .endmacro
117 |.macro decode_RC8, dst, ins; and dst, MASKR8, ins, lsr #13; .endmacro
118 |.macro decode_RD, dst, ins; lsr dst, ins, #16; .endmacro
119 |.macro decode_OP, dst, ins; and dst, ins, #255; .endmacro
120 |
121 |// Instruction fetch.
122 |.macro ins_NEXT1
123 | ldrb OP, [PC]
124 |.endmacro
125 |.macro ins_NEXT2
126 | ldr INS, [PC], #4
127 |.endmacro
128 |// Instruction decode+dispatch.
129 |.macro ins_NEXT3
130 | ldr OP, [DISPATCH, OP, lsl #2]
131 | decode_RA8 RA, INS
132 | decode_RD RC, INS
133 | bx OP
134 |.endmacro
135 |.macro ins_NEXT
136 | ins_NEXT1
137 | ins_NEXT2
138 | ins_NEXT3
139 |.endmacro
140 |
141 |// Instruction footer.
142 |.if 1
143 | // Replicated dispatch. Less unpredictable branches, but higher I-Cache use.
144 | .define ins_next, ins_NEXT
145 | .define ins_next_, ins_NEXT
146 | .define ins_next1, ins_NEXT1
147 | .define ins_next2, ins_NEXT2
148 | .define ins_next3, ins_NEXT3
149 |.else
150 | // Common dispatch. Lower I-Cache use, only one (very) unpredictable branch.
151 | // Affects only certain kinds of benchmarks (and only with -j off).
152 | .macro ins_next
153 | b ->ins_next
154 | .endmacro
155 | .macro ins_next1
156 | .endmacro
157 | .macro ins_next2
158 | .endmacro
159 | .macro ins_next3
160 | b ->ins_next
161 | .endmacro
162 | .macro ins_next_
163 | ->ins_next:
164 | ins_NEXT
165 | .endmacro
166 |.endif
167 |
168 |// Avoid register name substitution for field name.
169 #define field_pc pc
170 |
171 |// Call decode and dispatch.
172 |.macro ins_callt
173 | // BASE = new base, CARG3 = LFUNC/CFUNC, RC = nargs*8, FRAME_PC(BASE) = PC
174 | ldr PC, LFUNC:CARG3->field_pc
175 | ldrb OP, [PC] // STALL: load PC. early PC.
176 | ldr INS, [PC], #4
177 | ldr OP, [DISPATCH, OP, lsl #2] // STALL: load OP. early OP.
178 | decode_RA8 RA, INS
179 | add RA, RA, BASE
180 | bx OP
181 |.endmacro
182 |
183 |.macro ins_call
184 | // BASE = new base, CARG3 = LFUNC/CFUNC, RC = nargs*8, PC = caller PC
185 | str PC, [BASE, FRAME_PC]
186 | ins_callt // STALL: locked PC.
187 |.endmacro
188 |
189 |//-----------------------------------------------------------------------
190 |
191 |// Macros to test operand types.
192 |.macro checktp, reg, tp; cmn reg, #-tp; .endmacro
193 |.macro checktpeq, reg, tp; cmneq reg, #-tp; .endmacro
194 |.macro checktpne, reg, tp; cmnne reg, #-tp; .endmacro
195 |.macro checkstr, reg, target; checktp reg, LJ_TSTR; bne target; .endmacro
196 |.macro checktab, reg, target; checktp reg, LJ_TTAB; bne target; .endmacro
197 |.macro checkfunc, reg, target; checktp reg, LJ_TFUNC; bne target; .endmacro
198 |
199 |// Assumes DISPATCH is relative to GL.
200 #define DISPATCH_GL(field) (GG_DISP2G + (int)offsetof(global_State, field))
201 #define DISPATCH_J(field) (GG_DISP2J + (int)offsetof(jit_State, field))
202 |
203 #define PC2PROTO(field) ((int)offsetof(GCproto, field)-(int)sizeof(GCproto))
204 |
205 |.macro hotcheck, delta
206 | lsr CARG1, PC, #1
207 | and CARG1, CARG1, #126
208 | sub CARG1, CARG1, #-GG_DISP2HOT
209 | ldrh CARG2, [DISPATCH, CARG1]
210 | subs CARG2, CARG2, #delta
211 | strh CARG2, [DISPATCH, CARG1]
212 |.endmacro
213 |
214 |.macro hotloop
215 | hotcheck HOTCOUNT_LOOP
216 | blo ->vm_hotloop
217 |.endmacro
218 |
219 |.macro hotcall
220 | hotcheck HOTCOUNT_CALL
221 | blo ->vm_hotcall
222 |.endmacro
223 |
224 |// Set current VM state.
225 |.macro mv_vmstate, reg, st; mvn reg, #LJ_VMST_..st; .endmacro
226 |.macro st_vmstate, reg; str reg, [DISPATCH, #DISPATCH_GL(vmstate)]; .endmacro
227 |
228 |// Move table write barrier back. Overwrites mark and tmp.
229 |.macro barrierback, tab, mark, tmp
230 | ldr tmp, [DISPATCH, #DISPATCH_GL(gc.grayagain)]
231 | bic mark, mark, #LJ_GC_BLACK // black2gray(tab)
232 | str tab, [DISPATCH, #DISPATCH_GL(gc.grayagain)]
233 | strb mark, tab->marked
234 | str tmp, tab->gclist
235 |.endmacro
236 |
237 |.macro .IOS, a, b
238 |.if IOS
239 | a, b
240 |.endif
241 |.endmacro
242 |
243 |//-----------------------------------------------------------------------
244
245 #if !LJ_DUALNUM
246 #error "Only dual-number mode supported for ARM target"
247 #endif
248
249 /* Generate subroutines used by opcodes and other parts of the VM. */
250 /* The .code_sub section should be last to help static branch prediction. */
251 static void build_subroutines(BuildCtx *ctx)
252 {
253 |.code_sub
254 |
255 |//-----------------------------------------------------------------------
256 |//-- Return handling ----------------------------------------------------
257 |//-----------------------------------------------------------------------
258 |
259 |->vm_returnp:
260 | // See vm_return. Also: RB = previous base.
261 | tst PC, #FRAME_P
262 | beq ->cont_dispatch
263 |
264 | // Return from pcall or xpcall fast func.
265 | ldr PC, [RB, FRAME_PC] // Fetch PC of previous frame.
266 | mvn CARG2, #~LJ_TTRUE
267 | mov BASE, RB
268 | // Prepending may overwrite the pcall frame, so do it at the end.
269 | str CARG2, [RA, FRAME_PC] // Prepend true to results.
270 | sub RA, RA, #8
271 |
272 |->vm_returnc:
273 | adds RC, RC, #8 // RC = (nresults+1)*8.
274 | mov CRET1, #LUA_YIELD
275 | beq ->vm_unwind_c_eh
276 | str RC, SAVE_MULTRES
277 | ands CARG1, PC, #FRAME_TYPE
278 | beq ->BC_RET_Z // Handle regular return to Lua.
279 |
280 |->vm_return:
281 | // BASE = base, RA = resultptr, RC/MULTRES = (nresults+1)*8, PC = return
282 | // CARG1 = PC & FRAME_TYPE
283 | bic RB, PC, #FRAME_TYPEP
284 | cmp CARG1, #FRAME_C
285 | sub RB, BASE, RB // RB = previous base.
286 | bne ->vm_returnp
287 |
288 | str RB, L->base
289 | ldr KBASE, SAVE_NRES
290 | mv_vmstate CARG4, C
291 | sub BASE, BASE, #8
292 | subs CARG3, RC, #8
293 | lsl KBASE, KBASE, #3 // KBASE = (nresults_wanted+1)*8
294 | st_vmstate CARG4
295 | beq >2
296 |1:
297 | subs CARG3, CARG3, #8
298 | ldrd CARG12, [RA], #8
299 | strd CARG12, [BASE], #8
300 | bne <1
301 |2:
302 | cmp KBASE, RC // More/less results wanted?
303 | bne >6
304 |3:
305 | str BASE, L->top // Store new top.
306 |
307 |->vm_leave_cp:
308 | ldr RC, SAVE_CFRAME // Restore previous C frame.
309 | mov CRET1, #0 // Ok return status for vm_pcall.
310 | str RC, L->cframe
311 |
312 |->vm_leave_unw:
313 | restoreregs_ret
314 |
315 |6:
316 | blt >7 // Less results wanted?
317 | // More results wanted. Check stack size and fill up results with nil.
318 | ldr CARG3, L->maxstack
319 | mvn CARG2, #~LJ_TNIL
320 | cmp BASE, CARG3
321 | bhs >8
322 | str CARG2, [BASE, #4]
323 | add RC, RC, #8
324 | add BASE, BASE, #8
325 | b <2
326 |
327 |7: // Less results wanted.
328 | sub CARG1, RC, KBASE
329 | cmp KBASE, #0 // LUA_MULTRET+1 case?
330 | subne BASE, BASE, CARG1 // Either keep top or shrink it.
331 | b <3
332 |
333 |8: // Corner case: need to grow stack for filling up results.
334 | // This can happen if:
335 | // - A C function grows the stack (a lot).
336 | // - The GC shrinks the stack in between.
337 | // - A return back from a lua_call() with (high) nresults adjustment.
338 | str BASE, L->top // Save current top held in BASE (yes).
339 | lsr CARG2, KBASE, #3
340 | mov CARG1, L
341 | bl extern lj_state_growstack // (lua_State *L, int n)
342 | ldr BASE, L->top // Need the (realloced) L->top in BASE.
343 | b <2
344 |
345 |->vm_unwind_c: // Unwind C stack, return from vm_pcall.
346 | // (void *cframe, int errcode)
347 | mov sp, CARG1
348 | mov CRET1, CARG2
349 |->vm_unwind_c_eh: // Landing pad for external unwinder.
350 | ldr L, SAVE_L
351 | mv_vmstate CARG4, C
352 | ldr GL:CARG3, L->glref
353 | str CARG4, GL:CARG3->vmstate
354 | b ->vm_leave_unw
355 |
356 |->vm_unwind_ff: // Unwind C stack, return from ff pcall.
357 | // (void *cframe)
358 | bic CARG1, CARG1, #~CFRAME_RAWMASK // Use two steps: bic sp is deprecated.
359 | mov sp, CARG1
360 |->vm_unwind_ff_eh: // Landing pad for external unwinder.
361 | ldr L, SAVE_L
362 | mov MASKR8, #255
363 | mov RC, #16 // 2 results: false + error message.
364 | lsl MASKR8, MASKR8, #3 // MASKR8 = 255*8.
365 | ldr BASE, L->base
366 | ldr DISPATCH, L->glref // Setup pointer to dispatch table.
367 | mvn CARG1, #~LJ_TFALSE
368 | sub RA, BASE, #8 // Results start at BASE-8.
369 | ldr PC, [BASE, FRAME_PC] // Fetch PC of previous frame.
370 | add DISPATCH, DISPATCH, #GG_G2DISP
371 | mv_vmstate CARG2, INTERP
372 | str CARG1, [BASE, #-4] // Prepend false to error message.
373 | st_vmstate CARG2
374 | b ->vm_returnc
375 |
376 |->vm_unwind_ext: // Complete external unwind.
377 #if !LJ_NO_UNWIND
378 | push {r0, r1, r2, lr}
379 | bl extern _Unwind_Complete
380 | ldr r0, [sp]
381 | bl extern _Unwind_DeleteException
382 | pop {r0, r1, r2, lr}
383 | mov r0, r1
384 | bx r2
385 #endif
386 |
387 |//-----------------------------------------------------------------------
388 |//-- Grow stack for calls -----------------------------------------------
389 |//-----------------------------------------------------------------------
390 |
391 |->vm_growstack_c: // Grow stack for C function.
392 | // CARG1 = L
393 | mov CARG2, #LUA_MINSTACK
394 | b >2
395 |
396 |->vm_growstack_l: // Grow stack for Lua function.
397 | // BASE = new base, RA = BASE+framesize*8, RC = nargs*8, PC = first PC
398 | add RC, BASE, RC
399 | sub RA, RA, BASE
400 | mov CARG1, L
401 | str BASE, L->base
402 | add PC, PC, #4 // Must point after first instruction.
403 | str RC, L->top
404 | lsr CARG2, RA, #3
405 |2:
406 | // L->base = new base, L->top = top
407 | str PC, SAVE_PC
408 | bl extern lj_state_growstack // (lua_State *L, int n)
409 | ldr BASE, L->base
410 | ldr RC, L->top
411 | ldr LFUNC:CARG3, [BASE, FRAME_FUNC]
412 | sub NARGS8:RC, RC, BASE
413 | // BASE = new base, RB = LFUNC/CFUNC, RC = nargs*8, FRAME_PC(BASE) = PC
414 | ins_callt // Just retry the call.
415 |
416 |//-----------------------------------------------------------------------
417 |//-- Entry points into the assembler VM ---------------------------------
418 |//-----------------------------------------------------------------------
419 |
420 |->vm_resume: // Setup C frame and resume thread.
421 | // (lua_State *L, TValue *base, int nres1 = 0, ptrdiff_t ef = 0)
422 | saveregs
423 | mov L, CARG1
424 | ldr DISPATCH, L:CARG1->glref // Setup pointer to dispatch table.
425 | mov BASE, CARG2
426 | add DISPATCH, DISPATCH, #GG_G2DISP
427 | str L, SAVE_L
428 | mov PC, #FRAME_CP
429 | str CARG3, SAVE_NRES
430 | add CARG2, sp, #CFRAME_RESUME
431 | ldrb CARG1, L->status
432 | str CARG3, SAVE_ERRF
433 | str L, SAVE_PC // Any value outside of bytecode is ok.
434 | str CARG3, SAVE_CFRAME
435 | cmp CARG1, #0
436 | str CARG2, L->cframe
437 | beq >3
438 |
439 | // Resume after yield (like a return).
440 | str L, [DISPATCH, #DISPATCH_GL(cur_L)]
441 | mov RA, BASE
442 | ldr BASE, L->base
443 | ldr CARG1, L->top
444 | mov MASKR8, #255
445 | strb CARG3, L->status
446 | sub RC, CARG1, BASE
447 | ldr PC, [BASE, FRAME_PC]
448 | lsl MASKR8, MASKR8, #3 // MASKR8 = 255*8.
449 | mv_vmstate CARG2, INTERP
450 | add RC, RC, #8
451 | ands CARG1, PC, #FRAME_TYPE
452 | st_vmstate CARG2
453 | str RC, SAVE_MULTRES
454 | beq ->BC_RET_Z
455 | b ->vm_return
456 |
457 |->vm_pcall: // Setup protected C frame and enter VM.
458 | // (lua_State *L, TValue *base, int nres1, ptrdiff_t ef)
459 | saveregs
460 | mov PC, #FRAME_CP
461 | str CARG4, SAVE_ERRF
462 | b >1
463 |
464 |->vm_call: // Setup C frame and enter VM.
465 | // (lua_State *L, TValue *base, int nres1)
466 | saveregs
467 | mov PC, #FRAME_C
468 |
469 |1: // Entry point for vm_pcall above (PC = ftype).
470 | ldr RC, L:CARG1->cframe
471 | str CARG3, SAVE_NRES
472 | mov L, CARG1
473 | str CARG1, SAVE_L
474 | ldr DISPATCH, L->glref // Setup pointer to dispatch table.
475 | mov BASE, CARG2
476 | str CARG1, SAVE_PC // Any value outside of bytecode is ok.
477 | str RC, SAVE_CFRAME
478 | add DISPATCH, DISPATCH, #GG_G2DISP
479 | str sp, L->cframe // Add our C frame to cframe chain.
480 |
481 |3: // Entry point for vm_cpcall/vm_resume (BASE = base, PC = ftype).
482 | str L, [DISPATCH, #DISPATCH_GL(cur_L)]
483 | ldr RB, L->base // RB = old base (for vmeta_call).
484 | ldr CARG1, L->top
485 | mov MASKR8, #255
486 | add PC, PC, BASE
487 | lsl MASKR8, MASKR8, #3 // MASKR8 = 255*8.
488 | sub PC, PC, RB // PC = frame delta + frame type
489 | mv_vmstate CARG2, INTERP
490 | sub NARGS8:RC, CARG1, BASE
491 | st_vmstate CARG2
492 |
493 |->vm_call_dispatch:
494 | // RB = old base, BASE = new base, RC = nargs*8, PC = caller PC
495 | ldrd CARG34, [BASE, FRAME_FUNC]
496 | checkfunc CARG4, ->vmeta_call
497 |
498 |->vm_call_dispatch_f:
499 | ins_call
500 | // BASE = new base, CARG3 = func, RC = nargs*8, PC = caller PC
501 |
502 |->vm_cpcall: // Setup protected C frame, call C.
503 | // (lua_State *L, lua_CFunction func, void *ud, lua_CPFunction cp)
504 | saveregs
505 | mov L, CARG1
506 | ldr RA, L:CARG1->stack
507 | str CARG1, SAVE_L
508 | ldr DISPATCH, L->glref // Setup pointer to dispatch table.
509 | ldr RB, L->top
510 | str CARG1, SAVE_PC // Any value outside of bytecode is ok.
511 | ldr RC, L->cframe
512 | add DISPATCH, DISPATCH, #GG_G2DISP
513 | sub RA, RA, RB // Compute -savestack(L, L->top).
514 | mov RB, #0
515 | str RA, SAVE_NRES // Neg. delta means cframe w/o frame.
516 | str RB, SAVE_ERRF // No error function.
517 | str RC, SAVE_CFRAME
518 | str sp, L->cframe // Add our C frame to cframe chain.
519 | str L, [DISPATCH, #DISPATCH_GL(cur_L)]
520 | blx CARG4 // (lua_State *L, lua_CFunction func, void *ud)
521 | movs BASE, CRET1
522 | mov PC, #FRAME_CP
523 | bne <3 // Else continue with the call.
524 | b ->vm_leave_cp // No base? Just remove C frame.
525 |
526 |//-----------------------------------------------------------------------
527 |//-- Metamethod handling ------------------------------------------------
528 |//-----------------------------------------------------------------------
529 |
530 |//-- Continuation dispatch ----------------------------------------------
531 |
532 |->cont_dispatch:
533 | // BASE = meta base, RA = resultptr, RC = (nresults+1)*8
534 | ldr LFUNC:CARG3, [RB, FRAME_FUNC]
535 | ldr CARG1, [BASE, #-16] // Get continuation.
536 | mov CARG4, BASE
537 | mov BASE, RB // Restore caller BASE.
538 |.if FFI
539 | cmp CARG1, #1
540 |.endif
541 | ldr PC, [CARG4, #-12] // Restore PC from [cont|PC].
542 | mvn INS, #~LJ_TNIL
543 | add CARG2, RA, RC
544 | str INS, [CARG2, #-4] // Ensure one valid arg.
545 |.if FFI
546 | bls >1
547 |.endif
548 | ldr CARG3, LFUNC:CARG3->field_pc
549 | ldr KBASE, [CARG3, #PC2PROTO(k)]
550 | // BASE = base, RA = resultptr, CARG4 = meta base
551 | bx CARG1
552 |
553 |.if FFI
554 |1:
555 | beq ->cont_ffi_callback // cont = 1: return from FFI callback.
556 | // cont = 0: tailcall from C function.
557 | sub CARG4, CARG4, #16
558 | sub RC, CARG4, BASE
559 | b ->vm_call_tail
560 |.endif
561 |
562 |->cont_cat: // RA = resultptr, CARG4 = meta base
563 | ldr INS, [PC, #-4]
564 | sub CARG2, CARG4, #16
565 | ldrd CARG34, [RA]
566 | str BASE, L->base
567 | decode_RB8 RC, INS
568 | decode_RA8 RA, INS
569 | add CARG1, BASE, RC
570 | subs CARG1, CARG2, CARG1
571 | strdne CARG34, [CARG2]
572 | movne CARG3, CARG1
573 | bne ->BC_CAT_Z
574 | strd CARG34, [BASE, RA]
575 | b ->cont_nop
576 |
577 |//-- Table indexing metamethods -----------------------------------------
578 |
579 |->vmeta_tgets1:
580 | add CARG2, BASE, RB
581 | b >2
582 |
583 |->vmeta_tgets:
584 | sub CARG2, DISPATCH, #-DISPATCH_GL(tmptv)
585 | mvn CARG4, #~LJ_TTAB
586 | str TAB:RB, [CARG2]
587 | str CARG4, [CARG2, #4]
588 |2:
589 | mvn CARG4, #~LJ_TSTR
590 | str STR:RC, TMPDlo
591 | str CARG4, TMPDhi
592 | mov CARG3, TMPDp
593 | b >1
594 |
595 |->vmeta_tgetb: // RC = index
596 | decode_RB8 RB, INS
597 | str RC, TMPDlo
598 | mvn CARG4, #~LJ_TISNUM
599 | add CARG2, BASE, RB
600 | str CARG4, TMPDhi
601 | mov CARG3, TMPDp
602 | b >1
603 |
604 |->vmeta_tgetv:
605 | add CARG2, BASE, RB
606 | add CARG3, BASE, RC
607 |1:
608 | str BASE, L->base
609 | mov CARG1, L
610 | str PC, SAVE_PC
611 | bl extern lj_meta_tget // (lua_State *L, TValue *o, TValue *k)
612 | // Returns TValue * (finished) or NULL (metamethod).
613 | .IOS ldr BASE, L->base
614 | cmp CRET1, #0
615 | beq >3
616 | ldrd CARG34, [CRET1]
617 | ins_next1
618 | ins_next2
619 | strd CARG34, [BASE, RA]
620 | ins_next3
621 |
622 |3: // Call __index metamethod.
623 | // BASE = base, L->top = new base, stack = cont/func/t/k
624 | rsb CARG1, BASE, #FRAME_CONT
625 | ldr BASE, L->top
626 | mov NARGS8:RC, #16 // 2 args for func(t, k).
627 | str PC, [BASE, #-12] // [cont|PC]
628 | add PC, CARG1, BASE
629 | ldr LFUNC:CARG3, [BASE, FRAME_FUNC] // Guaranteed to be a function here.
630 | b ->vm_call_dispatch_f
631 |
632 |->vmeta_tgetr:
633 | .IOS mov RC, BASE
634 | bl extern lj_tab_getinth // (GCtab *t, int32_t key)
635 | // Returns cTValue * or NULL.
636 | .IOS mov BASE, RC
637 | cmp CRET1, #0
638 | ldrdne CARG12, [CRET1]
639 | mvneq CARG2, #~LJ_TNIL
640 | b ->BC_TGETR_Z
641 |
642 |//-----------------------------------------------------------------------
643 |
644 |->vmeta_tsets1:
645 | add CARG2, BASE, RB
646 | b >2
647 |
648 |->vmeta_tsets:
649 | sub CARG2, DISPATCH, #-DISPATCH_GL(tmptv)
650 | mvn CARG4, #~LJ_TTAB
651 | str TAB:RB, [CARG2]
652 | str CARG4, [CARG2, #4]
653 |2:
654 | mvn CARG4, #~LJ_TSTR
655 | str STR:RC, TMPDlo
656 | str CARG4, TMPDhi
657 | mov CARG3, TMPDp
658 | b >1
659 |
660 |->vmeta_tsetb: // RC = index
661 | decode_RB8 RB, INS
662 | str RC, TMPDlo
663 | mvn CARG4, #~LJ_TISNUM
664 | add CARG2, BASE, RB
665 | str CARG4, TMPDhi
666 | mov CARG3, TMPDp
667 | b >1
668 |
669 |->vmeta_tsetv:
670 | add CARG2, BASE, RB
671 | add CARG3, BASE, RC
672 |1:
673 | str BASE, L->base
674 | mov CARG1, L
675 | str PC, SAVE_PC
676 | bl extern lj_meta_tset // (lua_State *L, TValue *o, TValue *k)
677 | // Returns TValue * (finished) or NULL (metamethod).
678 | .IOS ldr BASE, L->base
679 | cmp CRET1, #0
680 | ldrd CARG34, [BASE, RA]
681 | beq >3
682 | ins_next1
683 | // NOBARRIER: lj_meta_tset ensures the table is not black.
684 | strd CARG34, [CRET1]
685 | ins_next2
686 | ins_next3
687 |
688 |3: // Call __newindex metamethod.
689 | // BASE = base, L->top = new base, stack = cont/func/t/k/(v)
690 | rsb CARG1, BASE, #FRAME_CONT
691 | ldr BASE, L->top
692 | mov NARGS8:RC, #24 // 3 args for func(t, k, v).
693 | strd CARG34, [BASE, #16] // Copy value to third argument.
694 | str PC, [BASE, #-12] // [cont|PC]
695 | add PC, CARG1, BASE
696 | ldr LFUNC:CARG3, [BASE, FRAME_FUNC] // Guaranteed to be a function here.
697 | b ->vm_call_dispatch_f
698 |
699 |->vmeta_tsetr:
700 | str BASE, L->base
701 | .IOS mov RC, BASE
702 | mov CARG1, L
703 | str PC, SAVE_PC
704 | bl extern lj_tab_setinth // (lua_State *L, GCtab *t, int32_t key)
705 | // Returns TValue *.
706 | .IOS mov BASE, RC
707 | b ->BC_TSETR_Z
708 |
709 |//-- Comparison metamethods ---------------------------------------------
710 |
711 |->vmeta_comp:
712 | mov CARG1, L
713 | sub PC, PC, #4
714 | mov CARG2, RA
715 | str BASE, L->base
716 | mov CARG3, RC
717 | str PC, SAVE_PC
718 | decode_OP CARG4, INS
719 | bl extern lj_meta_comp // (lua_State *L, TValue *o1, *o2, int op)
720 | // Returns 0/1 or TValue * (metamethod).
721 |3:
722 | .IOS ldr BASE, L->base
723 | cmp CRET1, #1
724 | bhi ->vmeta_binop
725 |4:
726 | ldrh RB, [PC, #2]
727 | add PC, PC, #4
728 | add RB, PC, RB, lsl #2
729 | subhs PC, RB, #0x20000
730 |->cont_nop:
731 | ins_next
732 |
733 |->cont_ra: // RA = resultptr
734 | ldr INS, [PC, #-4]
735 | ldrd CARG12, [RA]
736 | decode_RA8 CARG3, INS
737 | strd CARG12, [BASE, CARG3]
738 | b ->cont_nop
739 |
740 |->cont_condt: // RA = resultptr
741 | ldr CARG2, [RA, #4]
742 | mvn CARG1, #~LJ_TTRUE
743 | cmp CARG1, CARG2 // Branch if result is true.
744 | b <4
745 |
746 |->cont_condf: // RA = resultptr
747 | ldr CARG2, [RA, #4]
748 | checktp CARG2, LJ_TFALSE // Branch if result is false.
749 | b <4
750 |
751 |->vmeta_equal:
752 | // CARG2, CARG3, CARG4 are already set by BC_ISEQV/BC_ISNEV.
753 | sub PC, PC, #4
754 | str BASE, L->base
755 | mov CARG1, L
756 | str PC, SAVE_PC
757 | bl extern lj_meta_equal // (lua_State *L, GCobj *o1, *o2, int ne)
758 | // Returns 0/1 or TValue * (metamethod).
759 | b <3
760 |
761 |->vmeta_equal_cd:
762 |.if FFI
763 | sub PC, PC, #4
764 | str BASE, L->base
765 | mov CARG1, L
766 | mov CARG2, INS
767 | str PC, SAVE_PC
768 | bl extern lj_meta_equal_cd // (lua_State *L, BCIns op)
769 | // Returns 0/1 or TValue * (metamethod).
770 | b <3
771 |.endif
772 |
773 |->vmeta_istype:
774 | sub PC, PC, #4
775 | str BASE, L->base
776 | mov CARG1, L
777 | lsr CARG2, RA, #3
778 | mov CARG3, RC
779 | str PC, SAVE_PC
780 | bl extern lj_meta_istype // (lua_State *L, BCReg ra, BCReg tp)
781 | .IOS ldr BASE, L->base
782 | b ->cont_nop
783 |
784 |//-- Arithmetic metamethods ---------------------------------------------
785 |
786 |->vmeta_arith_vn:
787 | decode_RB8 RB, INS
788 | decode_RC8 RC, INS
789 | add CARG3, BASE, RB
790 | add CARG4, KBASE, RC
791 | b >1
792 |
793 |->vmeta_arith_nv:
794 | decode_RB8 RB, INS
795 | decode_RC8 RC, INS
796 | add CARG4, BASE, RB
797 | add CARG3, KBASE, RC
798 | b >1
799 |
800 |->vmeta_unm:
801 | ldr INS, [PC, #-8]
802 | sub PC, PC, #4
803 | add CARG3, BASE, RC
804 | add CARG4, BASE, RC
805 | b >1
806 |
807 |->vmeta_arith_vv:
808 | decode_RB8 RB, INS
809 | decode_RC8 RC, INS
810 | add CARG3, BASE, RB
811 | add CARG4, BASE, RC
812 |1:
813 | decode_OP OP, INS
814 | add CARG2, BASE, RA
815 | str BASE, L->base
816 | mov CARG1, L
817 | str PC, SAVE_PC
818 | str OP, ARG5
819 | bl extern lj_meta_arith // (lua_State *L, TValue *ra,*rb,*rc, BCReg op)
820 | // Returns NULL (finished) or TValue * (metamethod).
821 | .IOS ldr BASE, L->base
822 | cmp CRET1, #0
823 | beq ->cont_nop
824 |
825 | // Call metamethod for binary op.
826 |->vmeta_binop:
827 | // BASE = old base, CRET1 = new base, stack = cont/func/o1/o2
828 | sub CARG2, CRET1, BASE
829 | str PC, [CRET1, #-12] // [cont|PC]
830 | add PC, CARG2, #FRAME_CONT
831 | mov BASE, CRET1
832 | mov NARGS8:RC, #16 // 2 args for func(o1, o2).
833 | b ->vm_call_dispatch
834 |
835 |->vmeta_len:
836 | add CARG2, BASE, RC
837 | str BASE, L->base
838 | mov CARG1, L
839 | str PC, SAVE_PC
840 | bl extern lj_meta_len // (lua_State *L, TValue *o)
841 | // Returns NULL (retry) or TValue * (metamethod base).
842 | .IOS ldr BASE, L->base
843 #if LJ_52
844 | cmp CRET1, #0
845 | bne ->vmeta_binop // Binop call for compatibility.
846 | ldr TAB:CARG1, [BASE, RC]
847 | b ->BC_LEN_Z
848 #else
849 | b ->vmeta_binop // Binop call for compatibility.
850 #endif
851 |
852 |//-- Call metamethod ----------------------------------------------------
853 |
854 |->vmeta_call: // Resolve and call __call metamethod.
855 | // RB = old base, BASE = new base, RC = nargs*8
856 | mov CARG1, L
857 | str RB, L->base // This is the callers base!
858 | sub CARG2, BASE, #8
859 | str PC, SAVE_PC
860 | add CARG3, BASE, NARGS8:RC
861 | .IOS mov RA, BASE
862 | bl extern lj_meta_call // (lua_State *L, TValue *func, TValue *top)
863 | .IOS mov BASE, RA
864 | ldr LFUNC:CARG3, [BASE, FRAME_FUNC] // Guaranteed to be a function here.
865 | add NARGS8:RC, NARGS8:RC, #8 // Got one more argument now.
866 | ins_call
867 |
868 |->vmeta_callt: // Resolve __call for BC_CALLT.
869 | // BASE = old base, RA = new base, RC = nargs*8
870 | mov CARG1, L
871 | str BASE, L->base
872 | sub CARG2, RA, #8
873 | str PC, SAVE_PC
874 | add CARG3, RA, NARGS8:RC
875 | bl extern lj_meta_call // (lua_State *L, TValue *func, TValue *top)
876 | .IOS ldr BASE, L->base
877 | ldr LFUNC:CARG3, [RA, FRAME_FUNC] // Guaranteed to be a function here.
878 | ldr PC, [BASE, FRAME_PC]
879 | add NARGS8:RC, NARGS8:RC, #8 // Got one more argument now.
880 | b ->BC_CALLT2_Z
881 |
882 |//-- Argument coercion for 'for' statement ------------------------------
883 |
884 |->vmeta_for:
885 | mov CARG1, L
886 | str BASE, L->base
887 | mov CARG2, RA
888 | str PC, SAVE_PC
889 | bl extern lj_meta_for // (lua_State *L, TValue *base)
890 | .IOS ldr BASE, L->base
891 |.if JIT
892 | ldrb OP, [PC, #-4]
893 |.endif
894 | ldr INS, [PC, #-4]
895 |.if JIT
896 | cmp OP, #BC_JFORI
897 |.endif
898 | decode_RA8 RA, INS
899 | decode_RD RC, INS
900 |.if JIT
901 | beq =>BC_JFORI
902 |.endif
903 | b =>BC_FORI
904 |
905 |//-----------------------------------------------------------------------
906 |//-- Fast functions -----------------------------------------------------
907 |//-----------------------------------------------------------------------
908 |
909 |.macro .ffunc, name
910 |->ff_ .. name:
911 |.endmacro
912 |
913 |.macro .ffunc_1, name
914 |->ff_ .. name:
915 | ldrd CARG12, [BASE]
916 | cmp NARGS8:RC, #8
917 | blo ->fff_fallback
918 |.endmacro
919 |
920 |.macro .ffunc_2, name
921 |->ff_ .. name:
922 | ldrd CARG12, [BASE]
923 | ldrd CARG34, [BASE, #8]
924 | cmp NARGS8:RC, #16
925 | blo ->fff_fallback
926 |.endmacro
927 |
928 |.macro .ffunc_n, name
929 | .ffunc_1 name
930 | checktp CARG2, LJ_TISNUM
931 | bhs ->fff_fallback
932 |.endmacro
933 |
934 |.macro .ffunc_nn, name
935 | .ffunc_2 name
936 | checktp CARG2, LJ_TISNUM
937 | cmnlo CARG4, #-LJ_TISNUM
938 | bhs ->fff_fallback
939 |.endmacro
940 |
941 |.macro .ffunc_d, name
942 | .ffunc name
943 | ldr CARG2, [BASE, #4]
944 | cmp NARGS8:RC, #8
945 | vldr d0, [BASE]
946 | blo ->fff_fallback
947 | checktp CARG2, LJ_TISNUM
948 | bhs ->fff_fallback
949 |.endmacro
950 |
951 |.macro .ffunc_dd, name
952 | .ffunc name
953 | ldr CARG2, [BASE, #4]
954 | ldr CARG4, [BASE, #12]
955 | cmp NARGS8:RC, #16
956 | vldr d0, [BASE]
957 | vldr d1, [BASE, #8]
958 | blo ->fff_fallback
959 | checktp CARG2, LJ_TISNUM
960 | cmnlo CARG4, #-LJ_TISNUM
961 | bhs ->fff_fallback
962 |.endmacro
963 |
964 |// Inlined GC threshold check. Caveat: uses CARG1 and CARG2.
965 |.macro ffgccheck
966 | ldr CARG1, [DISPATCH, #DISPATCH_GL(gc.total)]
967 | ldr CARG2, [DISPATCH, #DISPATCH_GL(gc.threshold)]
968 | cmp CARG1, CARG2
969 | blge ->fff_gcstep
970 |.endmacro
971 |
972 |//-- Base library: checks -----------------------------------------------
973 |
974 |.ffunc_1 assert
975 | checktp CARG2, LJ_TTRUE
976 | bhi ->fff_fallback
977 | ldr PC, [BASE, FRAME_PC]
978 | strd CARG12, [BASE, #-8]
979 | mov RB, BASE
980 | subs RA, NARGS8:RC, #8
981 | add RC, NARGS8:RC, #8 // Compute (nresults+1)*8.
982 | beq ->fff_res // Done if exactly 1 argument.
983 |1:
984 | ldrd CARG12, [RB, #8]
985 | subs RA, RA, #8
986 | strd CARG12, [RB], #8
987 | bne <1
988 | b ->fff_res
989 |
990 |.ffunc type
991 | ldr CARG2, [BASE, #4]
992 | cmp NARGS8:RC, #8
993 | blo ->fff_fallback
994 | checktp CARG2, LJ_TISNUM
995 | mvnlo CARG2, #~LJ_TISNUM
996 | rsb CARG4, CARG2, #(int)(offsetof(GCfuncC, upvalue)>>3)-1
997 | lsl CARG4, CARG4, #3
998 | ldrd CARG12, [CFUNC:CARG3, CARG4]
999 | b ->fff_restv
1000 |
1001 |//-- Base library: getters and setters ---------------------------------
1002 |
1003 |.ffunc_1 getmetatable
1004 | checktp CARG2, LJ_TTAB
1005 | cmnne CARG2, #-LJ_TUDATA
1006 | bne >6
1007 |1: // Field metatable must be at same offset for GCtab and GCudata!
1008 | ldr TAB:RB, TAB:CARG1->metatable
1009 |2:
1010 | mvn CARG2, #~LJ_TNIL
1011 | ldr STR:RC, [DISPATCH, #DISPATCH_GL(gcroot[GCROOT_MMNAME+MM_metatable])]
1012 | cmp TAB:RB, #0
1013 | beq ->fff_restv
1014 | ldr CARG3, TAB:RB->hmask
1015 | ldr CARG4, STR:RC->sid
1016 | ldr NODE:INS, TAB:RB->node
1017 | and CARG3, CARG3, CARG4 // idx = str->sid & tab->hmask
1018 | add CARG3, CARG3, CARG3, lsl #1
1019 | add NODE:INS, NODE:INS, CARG3, lsl #3 // node = tab->node + idx*3*8
1020 |3: // Rearranged logic, because we expect _not_ to find the key.
1021 | ldrd CARG34, NODE:INS->key // STALL: early NODE:INS.
1022 | ldrd CARG12, NODE:INS->val
1023 | ldr NODE:INS, NODE:INS->next
1024 | checktp CARG4, LJ_TSTR
1025 | cmpeq CARG3, STR:RC
1026 | beq >5
1027 | cmp NODE:INS, #0
1028 | bne <3
1029 |4:
1030 | mov CARG1, RB // Use metatable as default result.
1031 | mvn CARG2, #~LJ_TTAB
1032 | b ->fff_restv
1033 |5:
1034 | checktp CARG2, LJ_TNIL
1035 | bne ->fff_restv
1036 | b <4
1037 |
1038 |6:
1039 | checktp CARG2, LJ_TISNUM
1040 | mvnhs CARG2, CARG2
1041 | movlo CARG2, #~LJ_TISNUM
1042 | add CARG4, DISPATCH, CARG2, lsl #2
1043 | ldr TAB:RB, [CARG4, #DISPATCH_GL(gcroot[GCROOT_BASEMT])]
1044 | b <2
1045 |
1046 |.ffunc_2 setmetatable
1047 | // Fast path: no mt for table yet and not clearing the mt.
1048 | checktp CARG2, LJ_TTAB
1049 | ldreq TAB:RB, TAB:CARG1->metatable
1050 | checktpeq CARG4, LJ_TTAB
1051 | ldrbeq CARG4, TAB:CARG1->marked
1052 | cmpeq TAB:RB, #0
1053 | bne ->fff_fallback
1054 | tst CARG4, #LJ_GC_BLACK // isblack(table)
1055 | str TAB:CARG3, TAB:CARG1->metatable
1056 | beq ->fff_restv
1057 | barrierback TAB:CARG1, CARG4, CARG3
1058 | b ->fff_restv
1059 |
1060 |.ffunc rawget
1061 | ldrd CARG34, [BASE]
1062 | cmp NARGS8:RC, #16
1063 | blo ->fff_fallback
1064 | mov CARG2, CARG3
1065 | checktab CARG4, ->fff_fallback
1066 | mov CARG1, L
1067 | add CARG3, BASE, #8
1068 | .IOS mov RA, BASE
1069 | bl extern lj_tab_get // (lua_State *L, GCtab *t, cTValue *key)
1070 | // Returns cTValue *.
1071 | .IOS mov BASE, RA
1072 | ldrd CARG12, [CRET1]
1073 | b ->fff_restv
1074 |
1075 |//-- Base library: conversions ------------------------------------------
1076 |
1077 |.ffunc tonumber
1078 | // Only handles the number case inline (without a base argument).
1079 | ldrd CARG12, [BASE]
1080 | cmp NARGS8:RC, #8
1081 | bne ->fff_fallback
1082 | checktp CARG2, LJ_TISNUM
1083 | bls ->fff_restv
1084 | b ->fff_fallback
1085 |
1086 |.ffunc_1 tostring
1087 | // Only handles the string or number case inline.
1088 | checktp CARG2, LJ_TSTR
1089 | // A __tostring method in the string base metatable is ignored.
1090 | beq ->fff_restv
1091 | // Handle numbers inline, unless a number base metatable is present.
1092 | ldr CARG4, [DISPATCH, #DISPATCH_GL(gcroot[GCROOT_BASEMT_NUM])]
1093 | str BASE, L->base
1094 | checktp CARG2, LJ_TISNUM
1095 | cmpls CARG4, #0
1096 | str PC, SAVE_PC // Redundant (but a defined value).
1097 | bhi ->fff_fallback
1098 | ffgccheck
1099 | mov CARG1, L
1100 | mov CARG2, BASE
1101 | bl extern lj_strfmt_number // (lua_State *L, cTValue *o)
1102 | // Returns GCstr *.
1103 | ldr BASE, L->base
1104 | mvn CARG2, #~LJ_TSTR
1105 | b ->fff_restv
1106 |
1107 |//-- Base library: iterators -------------------------------------------
1108 |
1109 |.ffunc_1 next
1110 | mvn CARG4, #~LJ_TNIL
1111 | checktab CARG2, ->fff_fallback
1112 | strd CARG34, [BASE, NARGS8:RC] // Set missing 2nd arg to nil.
1113 | ldr PC, [BASE, FRAME_PC]
1114 | add CARG2, BASE, #8
1115 | sub CARG3, BASE, #8
1116 | bl extern lj_tab_next // (GCtab *t, cTValue *key, TValue *o)
1117 | // Returns 1=found, 0=end, -1=error.
1118 | .IOS ldr BASE, L->base
1119 | cmp CRET1, #0
1120 | mov RC, #(2+1)*8
1121 | bgt ->fff_res // Found key/value.
1122 | bmi ->fff_fallback // Invalid key.
1123 | // End of traversal: return nil.
1124 | mvn CRET2, #~LJ_TNIL
1125 | b ->fff_restv
1126 |
1127 |.ffunc_1 pairs
1128 | checktab CARG2, ->fff_fallback
1129 #if LJ_52
1130 | ldr TAB:RB, TAB:CARG1->metatable
1131 #endif
1132 | ldrd CFUNC:CARG34, CFUNC:CARG3->upvalue[0]
1133 | ldr PC, [BASE, FRAME_PC]
1134 #if LJ_52
1135 | cmp TAB:RB, #0
1136 | bne ->fff_fallback
1137 #endif
1138 | mvn CARG2, #~LJ_TNIL
1139 | mov RC, #(3+1)*8
1140 | strd CFUNC:CARG34, [BASE, #-8]
1141 | str CARG2, [BASE, #12]
1142 | b ->fff_res
1143 |
1144 |.ffunc_2 ipairs_aux
1145 | checktp CARG2, LJ_TTAB
1146 | checktpeq CARG4, LJ_TISNUM
1147 | bne ->fff_fallback
1148 | ldr RB, TAB:CARG1->asize
1149 | ldr RC, TAB:CARG1->array
1150 | add CARG3, CARG3, #1
1151 | ldr PC, [BASE, FRAME_PC]
1152 | cmp CARG3, RB
1153 | add RC, RC, CARG3, lsl #3
1154 | strd CARG34, [BASE, #-8]
1155 | ldrdlo CARG12, [RC]
1156 | mov RC, #(0+1)*8
1157 | bhs >2 // Not in array part?
1158 |1:
1159 | checktp CARG2, LJ_TNIL
1160 | movne RC, #(2+1)*8
1161 | strdne CARG12, [BASE]
1162 | b ->fff_res
1163 |2: // Check for empty hash part first. Otherwise call C function.
1164 | ldr RB, TAB:CARG1->hmask
1165 | mov CARG2, CARG3
1166 | cmp RB, #0
1167 | beq ->fff_res
1168 | .IOS mov RA, BASE
1169 | bl extern lj_tab_getinth // (GCtab *t, int32_t key)
1170 | // Returns cTValue * or NULL.
1171 | .IOS mov BASE, RA
1172 | cmp CRET1, #0
1173 | beq ->fff_res
1174 | ldrd CARG12, [CRET1]
1175 | b <1
1176 |
1177 |.ffunc_1 ipairs
1178 | checktab CARG2, ->fff_fallback
1179 #if LJ_52
1180 | ldr TAB:RB, TAB:CARG1->metatable
1181 #endif
1182 | ldrd CFUNC:CARG34, CFUNC:CARG3->upvalue[0]
1183 | ldr PC, [BASE, FRAME_PC]
1184 #if LJ_52
1185 | cmp TAB:RB, #0
1186 | bne ->fff_fallback
1187 #endif
1188 | mov CARG1, #0
1189 | mvn CARG2, #~LJ_TISNUM
1190 | mov RC, #(3+1)*8
1191 | strd CFUNC:CARG34, [BASE, #-8]
1192 | strd CARG12, [BASE, #8]
1193 | b ->fff_res
1194 |
1195 |//-- Base library: catch errors ----------------------------------------
1196 |
1197 |.ffunc pcall
1198 | ldrb RA, [DISPATCH, #DISPATCH_GL(hookmask)]
1199 | cmp NARGS8:RC, #8
1200 | blo ->fff_fallback
1201 | tst RA, #HOOK_ACTIVE // Remember active hook before pcall.
1202 | mov RB, BASE
1203 | add BASE, BASE, #8
1204 | moveq PC, #8+FRAME_PCALL
1205 | movne PC, #8+FRAME_PCALLH
1206 | sub NARGS8:RC, NARGS8:RC, #8
1207 | b ->vm_call_dispatch
1208 |
1209 |.ffunc_2 xpcall
1210 | ldrb RA, [DISPATCH, #DISPATCH_GL(hookmask)]
1211 | checkfunc CARG4, ->fff_fallback // Traceback must be a function.
1212 | mov RB, BASE
1213 | strd CARG12, [BASE, #8] // Swap function and traceback.
1214 | strd CARG34, [BASE]
1215 | tst RA, #HOOK_ACTIVE // Remember active hook before pcall.
1216 | add BASE, BASE, #16
1217 | moveq PC, #16+FRAME_PCALL
1218 | movne PC, #16+FRAME_PCALLH
1219 | sub NARGS8:RC, NARGS8:RC, #16
1220 | b ->vm_call_dispatch
1221 |
1222 |//-- Coroutine library --------------------------------------------------
1223 |
1224 |.macro coroutine_resume_wrap, resume
1225 |.if resume
1226 |.ffunc_1 coroutine_resume
1227 | checktp CARG2, LJ_TTHREAD
1228 | bne ->fff_fallback
1229 |.else
1230 |.ffunc coroutine_wrap_aux
1231 | ldr L:CARG1, CFUNC:CARG3->upvalue[0].gcr
1232 |.endif
1233 | ldr PC, [BASE, FRAME_PC]
1234 | str BASE, L->base
1235 | ldr CARG2, L:CARG1->top
1236 | ldrb RA, L:CARG1->status
1237 | ldr RB, L:CARG1->base
1238 | add CARG3, CARG2, NARGS8:RC
1239 | add CARG4, CARG2, RA
1240 | str PC, SAVE_PC
1241 | cmp CARG4, RB
1242 | beq ->fff_fallback
1243 | ldr CARG4, L:CARG1->maxstack
1244 | ldr RB, L:CARG1->cframe
1245 | cmp RA, #LUA_YIELD
1246 | cmpls CARG3, CARG4
1247 | cmpls RB, #0
1248 | bhi ->fff_fallback
1249 |1:
1250 |.if resume
1251 | sub CARG3, CARG3, #8 // Keep resumed thread in stack for GC.
1252 | add BASE, BASE, #8
1253 | sub NARGS8:RC, NARGS8:RC, #8
1254 |.endif
1255 | str CARG3, L:CARG1->top
1256 | str BASE, L->top
1257 |2: // Move args to coroutine.
1258 | ldrd CARG34, [BASE, RB]
1259 | cmp RB, NARGS8:RC
1260 | strdne CARG34, [CARG2, RB]
1261 | add RB, RB, #8
1262 | bne <2
1263 |
1264 | mov CARG3, #0
1265 | mov L:RA, L:CARG1
1266 | mov CARG4, #0
1267 | bl ->vm_resume // (lua_State *L, TValue *base, 0, 0)
1268 | // Returns thread status.
1269 |4:
1270 | ldr CARG3, L:RA->base
1271 | mv_vmstate CARG2, INTERP
1272 | ldr CARG4, L:RA->top
1273 | cmp CRET1, #LUA_YIELD
1274 | ldr BASE, L->base
1275 | str L, [DISPATCH, #DISPATCH_GL(cur_L)]
1276 | st_vmstate CARG2
1277 | bhi >8
1278 | subs RC, CARG4, CARG3
1279 | ldr CARG1, L->maxstack
1280 | add CARG2, BASE, RC
1281 | beq >6 // No results?
1282 | cmp CARG2, CARG1
1283 | mov RB, #0
1284 | bhi >9 // Need to grow stack?
1285 |
1286 | sub CARG4, RC, #8
1287 | str CARG3, L:RA->top // Clear coroutine stack.
1288 |5: // Move results from coroutine.
1289 | ldrd CARG12, [CARG3, RB]
1290 | cmp RB, CARG4
1291 | strd CARG12, [BASE, RB]
1292 | add RB, RB, #8
1293 | bne <5
1294 |6:
1295 |.if resume
1296 | mvn CARG3, #~LJ_TTRUE
1297 | add RC, RC, #16
1298 |7:
1299 | str CARG3, [BASE, #-4] // Prepend true/false to results.
1300 | sub RA, BASE, #8
1301 |.else
1302 | mov RA, BASE
1303 | add RC, RC, #8
1304 |.endif
1305 | ands CARG1, PC, #FRAME_TYPE
1306 | str PC, SAVE_PC
1307 | str RC, SAVE_MULTRES
1308 | beq ->BC_RET_Z
1309 | b ->vm_return
1310 |
1311 |8: // Coroutine returned with error (at co->top-1).
1312 |.if resume
1313 | ldrd CARG12, [CARG4, #-8]!
1314 | mvn CARG3, #~LJ_TFALSE
1315 | mov RC, #(2+1)*8
1316 | str CARG4, L:RA->top // Remove error from coroutine stack.
1317 | strd CARG12, [BASE] // Copy error message.
1318 | b <7
1319 |.else
1320 | mov CARG1, L
1321 | mov CARG2, L:RA
1322 | bl extern lj_ffh_coroutine_wrap_err // (lua_State *L, lua_State *co)
1323 | // Never returns.
1324 |.endif
1325 |
1326 |9: // Handle stack expansion on return from yield.
1327 | mov CARG1, L
1328 | lsr CARG2, RC, #3
1329 | bl extern lj_state_growstack // (lua_State *L, int n)
1330 | mov CRET1, #0
1331 | b <4
1332 |.endmacro
1333 |
1334 | coroutine_resume_wrap 1 // coroutine.resume
1335 | coroutine_resume_wrap 0 // coroutine.wrap
1336 |
1337 |.ffunc coroutine_yield
1338 | ldr CARG1, L->cframe
1339 | add CARG2, BASE, NARGS8:RC
1340 | str BASE, L->base
1341 | tst CARG1, #CFRAME_RESUME
1342 | str CARG2, L->top
1343 | mov CRET1, #LUA_YIELD
1344 | mov CARG3, #0
1345 | beq ->fff_fallback
1346 | str CARG3, L->cframe
1347 | strb CRET1, L->status
1348 | b ->vm_leave_unw
1349 |
1350 |//-- Math library -------------------------------------------------------
1351 |
1352 |.macro math_round, func
1353 | .ffunc_1 math_ .. func
1354 | checktp CARG2, LJ_TISNUM
1355 | beq ->fff_restv
1356 | bhi ->fff_fallback
1357 | // Round FP value and normalize result.
1358 | lsl CARG3, CARG2, #1
1359 | adds RB, CARG3, #0x00200000
1360 | bpl >2 // |x| < 1?
1361 | mvn CARG4, #0x3e0
1362 | subs RB, CARG4, RB, asr #21
1363 | lsl CARG4, CARG2, #11
1364 | lsl CARG3, CARG1, #11
1365 | orr CARG4, CARG4, #0x80000000
1366 | rsb INS, RB, #32
1367 | orr CARG4, CARG4, CARG1, lsr #21
1368 | bls >3 // |x| >= 2^31?
1369 | orr CARG3, CARG3, CARG4, lsl INS
1370 | lsr CARG1, CARG4, RB
1371 |.if "func" == "floor"
1372 | tst CARG3, CARG2, asr #31
1373 | addne CARG1, CARG1, #1
1374 |.else
1375 | bics CARG3, CARG3, CARG2, asr #31
1376 | addsne CARG1, CARG1, #1
1377 | ldrdvs CARG12, >9
1378 | bvs ->fff_restv
1379 |.endif
1380 | cmp CARG2, #0
1381 | rsblt CARG1, CARG1, #0
1382 |1:
1383 | mvn CARG2, #~LJ_TISNUM
1384 | b ->fff_restv
1385 |
1386 |2: // |x| < 1
1387 | bcs ->fff_restv // |x| is not finite.
1388 | orr CARG3, CARG3, CARG1 // ztest = abs(hi) | lo
1389 |.if "func" == "floor"
1390 | tst CARG3, CARG2, asr #31 // return (ztest & sign) == 0 ? 0 : -1
1391 | moveq CARG1, #0
1392 | mvnne CARG1, #0
1393 |.else
1394 | bics CARG3, CARG3, CARG2, asr #31 // return (ztest & ~sign) == 0 ? 0 : 1
1395 | moveq CARG1, #0
1396 | movne CARG1, #1
1397 |.endif
1398 | mvn CARG2, #~LJ_TISNUM
1399 | b ->fff_restv
1400 |
1401 |3: // |x| >= 2^31. Check for x == -(2^31).
1402 | cmpeq CARG4, #0x80000000
1403 |.if "func" == "floor"
1404 | cmpeq CARG3, #0
1405 |.endif
1406 | bne >4
1407 | cmp CARG2, #0
1408 | movmi CARG1, #0x80000000
1409 | bmi <1
1410 |4:
1411 | bl ->vm_..func.._sf
1412 | b ->fff_restv
1413 |.endmacro
1414 |
1415 | math_round floor
1416 | math_round ceil
1417 |
1418 |.align 8
1419 |9:
1420 | .long 0x00000000, 0x41e00000 // 2^31.
1421 |
1422 |.ffunc_1 math_abs
1423 | checktp CARG2, LJ_TISNUM
1424 | bhi ->fff_fallback
1425 | bicne CARG2, CARG2, #0x80000000
1426 | bne ->fff_restv
1427 | cmp CARG1, #0
1428 | rsbslt CARG1, CARG1, #0
1429 | ldrdvs CARG12, <9
1430 | // Fallthrough.
1431 |
1432 |->fff_restv:
1433 | // CARG12 = TValue result.
1434 | ldr PC, [BASE, FRAME_PC]
1435 | strd CARG12, [BASE, #-8]
1436 |->fff_res1:
1437 | // PC = return.
1438 | mov RC, #(1+1)*8
1439 |->fff_res:
1440 | // RC = (nresults+1)*8, PC = return.
1441 | ands CARG1, PC, #FRAME_TYPE
1442 | ldreq INS, [PC, #-4]
1443 | str RC, SAVE_MULTRES
1444 | sub RA, BASE, #8
1445 | bne ->vm_return
1446 | decode_RB8 RB, INS
1447 |5:
1448 | cmp RB, RC // More results expected?
1449 | bhi >6
1450 | decode_RA8 CARG1, INS
1451 | ins_next1
1452 | ins_next2
1453 | // Adjust BASE. KBASE is assumed to be set for the calling frame.
1454 | sub BASE, RA, CARG1
1455 | ins_next3
1456 |
1457 |6: // Fill up results with nil.
1458 | add CARG2, RA, RC
1459 | mvn CARG1, #~LJ_TNIL
1460 | add RC, RC, #8
1461 | str CARG1, [CARG2, #-4]
1462 | b <5
1463 |
1464 |.macro math_extern, func
1465 |.if HFABI
1466 | .ffunc_d math_ .. func
1467 |.else
1468 | .ffunc_n math_ .. func
1469 |.endif
1470 | .IOS mov RA, BASE
1471 | bl extern func
1472 | .IOS mov BASE, RA
1473 |.if HFABI
1474 | b ->fff_resd
1475 |.else
1476 | b ->fff_restv
1477 |.endif
1478 |.endmacro
1479 |
1480 |.macro math_extern2, func
1481 |.if HFABI
1482 | .ffunc_dd math_ .. func
1483 |.else
1484 | .ffunc_nn math_ .. func
1485 |.endif
1486 | .IOS mov RA, BASE
1487 | bl extern func
1488 | .IOS mov BASE, RA
1489 |.if HFABI
1490 | b ->fff_resd
1491 |.else
1492 | b ->fff_restv
1493 |.endif
1494 |.endmacro
1495 |
1496 |.if FPU
1497 | .ffunc_d math_sqrt
1498 | vsqrt.f64 d0, d0
1499 |->fff_resd:
1500 | ldr PC, [BASE, FRAME_PC]
1501 | vstr d0, [BASE, #-8]
1502 | b ->fff_res1
1503 |.else
1504 | math_extern sqrt
1505 |.endif
1506 |
1507 |.ffunc math_log
1508 |.if HFABI
1509 | ldr CARG2, [BASE, #4]
1510 | cmp NARGS8:RC, #8 // Need exactly 1 argument.
1511 | vldr d0, [BASE]
1512 | bne ->fff_fallback
1513 |.else
1514 | ldrd CARG12, [BASE]
1515 | cmp NARGS8:RC, #8 // Need exactly 1 argument.
1516 | bne ->fff_fallback
1517 |.endif
1518 | checktp CARG2, LJ_TISNUM
1519 | bhs ->fff_fallback
1520 | .IOS mov RA, BASE
1521 | bl extern log
1522 | .IOS mov BASE, RA
1523 |.if HFABI
1524 | b ->fff_resd
1525 |.else
1526 | b ->fff_restv
1527 |.endif
1528 |
1529 | math_extern log10
1530 | math_extern exp
1531 | math_extern sin
1532 | math_extern cos
1533 | math_extern tan
1534 | math_extern asin
1535 | math_extern acos
1536 | math_extern atan
1537 | math_extern sinh
1538 | math_extern cosh
1539 | math_extern tanh
1540 | math_extern2 pow
1541 | math_extern2 atan2
1542 | math_extern2 fmod
1543 |
1544 |.if HFABI
1545 | .ffunc math_ldexp
1546 | ldr CARG4, [BASE, #4]
1547 | ldrd CARG12, [BASE, #8]
1548 | cmp NARGS8:RC, #16
1549 | blo ->fff_fallback
1550 | vldr d0, [BASE]
1551 | checktp CARG4, LJ_TISNUM
1552 | bhs ->fff_fallback
1553 | checktp CARG2, LJ_TISNUM
1554 | bne ->fff_fallback
1555 | .IOS mov RA, BASE
1556 | bl extern ldexp // (double x, int exp)
1557 | .IOS mov BASE, RA
1558 | b ->fff_resd
1559 |.else
1560 |.ffunc_2 math_ldexp
1561 | checktp CARG2, LJ_TISNUM
1562 | bhs ->fff_fallback
1563 | checktp CARG4, LJ_TISNUM
1564 | bne ->fff_fallback
1565 | .IOS mov RA, BASE
1566 | bl extern ldexp // (double x, int exp)
1567 | .IOS mov BASE, RA
1568 | b ->fff_restv
1569 |.endif
1570 |
1571 |.if HFABI
1572 |.ffunc_d math_frexp
1573 | mov CARG1, sp
1574 | .IOS mov RA, BASE
1575 | bl extern frexp
1576 | .IOS mov BASE, RA
1577 | ldr CARG3, [sp]
1578 | mvn CARG4, #~LJ_TISNUM
1579 | ldr PC, [BASE, FRAME_PC]
1580 | vstr d0, [BASE, #-8]
1581 | mov RC, #(2+1)*8
1582 | strd CARG34, [BASE]
1583 | b ->fff_res
1584 |.else
1585 |.ffunc_n math_frexp
1586 | mov CARG3, sp
1587 | .IOS mov RA, BASE
1588 | bl extern frexp
1589 | .IOS mov BASE, RA
1590 | ldr CARG3, [sp]
1591 | mvn CARG4, #~LJ_TISNUM
1592 | ldr PC, [BASE, FRAME_PC]
1593 | strd CARG12, [BASE, #-8]
1594 | mov RC, #(2+1)*8
1595 | strd CARG34, [BASE]
1596 | b ->fff_res
1597 |.endif
1598 |
1599 |.if HFABI
1600 |.ffunc_d math_modf
1601 | sub CARG1, BASE, #8
1602 | ldr PC, [BASE, FRAME_PC]
1603 | .IOS mov RA, BASE
1604 | bl extern modf
1605 | .IOS mov BASE, RA
1606 | mov RC, #(2+1)*8
1607 | vstr d0, [BASE]
1608 | b ->fff_res
1609 |.else
1610 |.ffunc_n math_modf
1611 | sub CARG3, BASE, #8
1612 | ldr PC, [BASE, FRAME_PC]
1613 | .IOS mov RA, BASE
1614 | bl extern modf
1615 | .IOS mov BASE, RA
1616 | mov RC, #(2+1)*8
1617 | strd CARG12, [BASE]
1618 | b ->fff_res
1619 |.endif
1620 |
1621 |.macro math_minmax, name, cond, fcond
1622 |.if FPU
1623 | .ffunc_1 name
1624 | add RB, BASE, RC
1625 | checktp CARG2, LJ_TISNUM
1626 | add RA, BASE, #8
1627 | bne >4
1628 |1: // Handle integers.
1629 | ldrd CARG34, [RA]
1630 | cmp RA, RB
1631 | bhs ->fff_restv
1632 | checktp CARG4, LJ_TISNUM
1633 | bne >3
1634 | cmp CARG1, CARG3
1635 | add RA, RA, #8
1636 | mov..cond CARG1, CARG3
1637 | b <1
1638 |3: // Convert intermediate result to number and continue below.
1639 | vmov s4, CARG1
1640 | bhi ->fff_fallback
1641 | vldr d1, [RA]
1642 | vcvt.f64.s32 d0, s4
1643 | b >6
1644 |
1645 |4:
1646 | vldr d0, [BASE]
1647 | bhi ->fff_fallback
1648 |5: // Handle numbers.
1649 | ldrd CARG34, [RA]
1650 | vldr d1, [RA]
1651 | cmp RA, RB
1652 | bhs ->fff_resd
1653 | checktp CARG4, LJ_TISNUM
1654 | bhs >7
1655 |6:
1656 | vcmp.f64 d0, d1
1657 | vmrs
1658 | add RA, RA, #8
1659 | vmov..fcond.f64 d0, d1
1660 | b <5
1661 |7: // Convert integer to number and continue above.
1662 | vmov s4, CARG3
1663 | bhi ->fff_fallback
1664 | vcvt.f64.s32 d1, s4
1665 | b <6
1666 |
1667 |.else
1668 |
1669 | .ffunc_1 name
1670 | checktp CARG2, LJ_TISNUM
1671 | mov RA, #8
1672 | bne >4
1673 |1: // Handle integers.
1674 | ldrd CARG34, [BASE, RA]
1675 | cmp RA, RC
1676 | bhs ->fff_restv
1677 | checktp CARG4, LJ_TISNUM
1678 | bne >3
1679 | cmp CARG1, CARG3
1680 | add RA, RA, #8
1681 | mov..cond CARG1, CARG3
1682 | b <1
1683 |3: // Convert intermediate result to number and continue below.
1684 | bhi ->fff_fallback
1685 | bl extern __aeabi_i2d
1686 | ldrd CARG34, [BASE, RA]
1687 | b >6
1688 |
1689 |4:
1690 | bhi ->fff_fallback
1691 |5: // Handle numbers.
1692 | ldrd CARG34, [BASE, RA]
1693 | cmp RA, RC
1694 | bhs ->fff_restv
1695 | checktp CARG4, LJ_TISNUM
1696 | bhs >7
1697 |6:
1698 | bl extern __aeabi_cdcmple
1699 | add RA, RA, #8
1700 | mov..fcond CARG1, CARG3
1701 | mov..fcond CARG2, CARG4
1702 | b <5
1703 |7: // Convert integer to number and continue above.
1704 | bhi ->fff_fallback
1705 | strd CARG12, TMPD
1706 | mov CARG1, CARG3
1707 | bl extern __aeabi_i2d
1708 | ldrd CARG34, TMPD
1709 | b <6
1710 |.endif
1711 |.endmacro
1712 |
1713 | math_minmax math_min, gt, pl
1714 | math_minmax math_max, lt, le
1715 |
1716 |//-- String library -----------------------------------------------------
1717 |
1718 |.ffunc string_byte // Only handle the 1-arg case here.
1719 | ldrd CARG12, [BASE]
1720 | ldr PC, [BASE, FRAME_PC]
1721 | cmp NARGS8:RC, #8
1722 | checktpeq CARG2, LJ_TSTR // Need exactly 1 argument.
1723 | bne ->fff_fallback
1724 | ldr CARG3, STR:CARG1->len
1725 | ldrb CARG1, STR:CARG1[1] // Access is always ok (NUL at end).
1726 | mvn CARG2, #~LJ_TISNUM
1727 | cmp CARG3, #0
1728 | moveq RC, #(0+1)*8
1729 | movne RC, #(1+1)*8
1730 | strd CARG12, [BASE, #-8]
1731 | b ->fff_res
1732 |
1733 |.ffunc string_char // Only handle the 1-arg case here.
1734 | ffgccheck
1735 | ldrd CARG12, [BASE]
1736 | ldr PC, [BASE, FRAME_PC]
1737 | cmp NARGS8:RC, #8 // Need exactly 1 argument.
1738 | checktpeq CARG2, LJ_TISNUM
1739 | bicseq CARG4, CARG1, #255
1740 | mov CARG3, #1
1741 | bne ->fff_fallback
1742 | str CARG1, TMPD
1743 | mov CARG2, TMPDp // Points to stack. Little-endian.
1744 |->fff_newstr:
1745 | // CARG2 = str, CARG3 = len.
1746 | str BASE, L->base
1747 | mov CARG1, L
1748 | str PC, SAVE_PC
1749 | bl extern lj_str_new // (lua_State *L, char *str, size_t l)
1750 |->fff_resstr:
1751 | // Returns GCstr *.
1752 | ldr BASE, L->base
1753 | mvn CARG2, #~LJ_TSTR
1754 | b ->fff_restv
1755 |
1756 |.ffunc string_sub
1757 | ffgccheck
1758 | ldrd CARG12, [BASE]
1759 | ldrd CARG34, [BASE, #16]
1760 | cmp NARGS8:RC, #16
1761 | mvn RB, #0
1762 | beq >1
1763 | blo ->fff_fallback
1764 | checktp CARG4, LJ_TISNUM
1765 | mov RB, CARG3
1766 | bne ->fff_fallback
1767 |1:
1768 | ldrd CARG34, [BASE, #8]
1769 | checktp CARG2, LJ_TSTR
1770 | ldreq CARG2, STR:CARG1->len
1771 | checktpeq CARG4, LJ_TISNUM
1772 | bne ->fff_fallback
1773 | // CARG1 = str, CARG2 = str->len, CARG3 = start, RB = end
1774 | add CARG4, CARG2, #1
1775 | cmp CARG3, #0 // if (start < 0) start += len+1
1776 | addlt CARG3, CARG3, CARG4
1777 | cmp CARG3, #1 // if (start < 1) start = 1
1778 | movlt CARG3, #1
1779 | cmp RB, #0 // if (end < 0) end += len+1
1780 | addlt RB, RB, CARG4
1781 | bic RB, RB, RB, asr #31 // if (end < 0) end = 0
1782 | cmp RB, CARG2 // if (end > len) end = len
1783 | add CARG1, STR:CARG1, #sizeof(GCstr)-1
1784 | movgt RB, CARG2
1785 | add CARG2, CARG1, CARG3
1786 | subs CARG3, RB, CARG3 // len = end - start
1787 | add CARG3, CARG3, #1 // len += 1
1788 | bge ->fff_newstr
1789 |->fff_emptystr:
1790 | sub STR:CARG1, DISPATCH, #-DISPATCH_GL(strempty)
1791 | mvn CARG2, #~LJ_TSTR
1792 | b ->fff_restv
1793 |
1794 |.macro ffstring_op, name
1795 | .ffunc string_ .. name
1796 | ffgccheck
1797 | ldr CARG3, [BASE, #4]
1798 | cmp NARGS8:RC, #8
1799 | ldr STR:CARG2, [BASE]
1800 | blo ->fff_fallback
1801 | sub SBUF:CARG1, DISPATCH, #-DISPATCH_GL(tmpbuf)
1802 | checkstr CARG3, ->fff_fallback
1803 | ldr CARG4, SBUF:CARG1->b
1804 | str BASE, L->base
1805 | str PC, SAVE_PC
1806 | str L, SBUF:CARG1->L
1807 | str CARG4, SBUF:CARG1->w
1808 | bl extern lj_buf_putstr_ .. name
1809 | bl extern lj_buf_tostr
1810 | b ->fff_resstr
1811 |.endmacro
1812 |
1813 |ffstring_op reverse
1814 |ffstring_op lower
1815 |ffstring_op upper
1816 |
1817 |//-- Bit library --------------------------------------------------------
1818 |
1819 |// FP number to bit conversion for soft-float. Clobbers r0-r3.
1820 |->vm_tobit_fb:
1821 | bhi ->fff_fallback
1822 |->vm_tobit:
1823 | lsl RB, CARG2, #1
1824 | adds RB, RB, #0x00200000
1825 | movpl CARG1, #0 // |x| < 1?
1826 | bxpl lr
1827 | mvn CARG4, #0x3e0
1828 | subs RB, CARG4, RB, asr #21
1829 | bmi >1 // |x| >= 2^32?
1830 | lsl CARG4, CARG2, #11
1831 | orr CARG4, CARG4, #0x80000000
1832 | orr CARG4, CARG4, CARG1, lsr #21
1833 | cmp CARG2, #0
1834 | lsr CARG1, CARG4, RB
1835 | rsblt CARG1, CARG1, #0
1836 | bx lr
1837 |1:
1838 | add RB, RB, #21
1839 | lsr CARG4, CARG1, RB
1840 | rsb RB, RB, #20
1841 | lsl CARG1, CARG2, #12
1842 | cmp CARG2, #0
1843 | orr CARG1, CARG4, CARG1, lsl RB
1844 | rsblt CARG1, CARG1, #0
1845 | bx lr
1846 |
1847 |.macro .ffunc_bit, name
1848 | .ffunc_1 bit_..name
1849 | checktp CARG2, LJ_TISNUM
1850 | blne ->vm_tobit_fb
1851 |.endmacro
1852 |
1853 |.ffunc_bit tobit
1854 | mvn CARG2, #~LJ_TISNUM
1855 | b ->fff_restv
1856 |
1857 |.macro .ffunc_bit_op, name, ins
1858 | .ffunc_bit name
1859 | mov CARG3, CARG1
1860 | mov RA, #8
1861 |1:
1862 | ldrd CARG12, [BASE, RA]
1863 | cmp RA, NARGS8:RC
1864 | add RA, RA, #8
1865 | bge >2
1866 | checktp CARG2, LJ_TISNUM
1867 | blne ->vm_tobit_fb
1868 | ins CARG3, CARG3, CARG1
1869 | b <1
1870 |.endmacro
1871 |
1872 |.ffunc_bit_op band, and
1873 |.ffunc_bit_op bor, orr
1874 |.ffunc_bit_op bxor, eor
1875 |
1876 |2:
1877 | mvn CARG4, #~LJ_TISNUM
1878 | ldr PC, [BASE, FRAME_PC]
1879 | strd CARG34, [BASE, #-8]
1880 | b ->fff_res1
1881 |
1882 |.ffunc_bit bswap
1883 | eor CARG3, CARG1, CARG1, ror #16
1884 | bic CARG3, CARG3, #0x00ff0000
1885 | ror CARG1, CARG1, #8
1886 | mvn CARG2, #~LJ_TISNUM
1887 | eor CARG1, CARG1, CARG3, lsr #8
1888 | b ->fff_restv
1889 |
1890 |.ffunc_bit bnot
1891 | mvn CARG1, CARG1
1892 | mvn CARG2, #~LJ_TISNUM
1893 | b ->fff_restv
1894 |
1895 |.macro .ffunc_bit_sh, name, ins, shmod
1896 | .ffunc bit_..name
1897 | ldrd CARG12, [BASE, #8]
1898 | cmp NARGS8:RC, #16
1899 | blo ->fff_fallback
1900 | checktp CARG2, LJ_TISNUM
1901 | blne ->vm_tobit_fb
1902 |.if shmod == 0
1903 | and RA, CARG1, #31
1904 |.else
1905 | rsb RA, CARG1, #0
1906 |.endif
1907 | ldrd CARG12, [BASE]
1908 | checktp CARG2, LJ_TISNUM
1909 | blne ->vm_tobit_fb
1910 | ins CARG1, CARG1, RA
1911 | mvn CARG2, #~LJ_TISNUM
1912 | b ->fff_restv
1913 |.endmacro
1914 |
1915 |.ffunc_bit_sh lshift, lsl, 0
1916 |.ffunc_bit_sh rshift, lsr, 0
1917 |.ffunc_bit_sh arshift, asr, 0
1918 |.ffunc_bit_sh rol, ror, 1
1919 |.ffunc_bit_sh ror, ror, 0
1920 |
1921 |//-----------------------------------------------------------------------
1922 |
1923 |->fff_fallback: // Call fast function fallback handler.
1924 | // BASE = new base, RC = nargs*8
1925 | ldr CARG3, [BASE, FRAME_FUNC]
1926 | ldr CARG2, L->maxstack
1927 | add CARG1, BASE, NARGS8:RC
1928 | ldr PC, [BASE, FRAME_PC] // Fallback may overwrite PC.
1929 | str CARG1, L->top
1930 | ldr CARG3, CFUNC:CARG3->f
1931 | str BASE, L->base
1932 | add CARG1, CARG1, #8*LUA_MINSTACK
1933 | str PC, SAVE_PC // Redundant (but a defined value).
1934 | cmp CARG1, CARG2
1935 | mov CARG1, L
1936 | bhi >5 // Need to grow stack.
1937 | blx CARG3 // (lua_State *L)
1938 | // Either throws an error, or recovers and returns -1, 0 or nresults+1.
1939 | ldr BASE, L->base
1940 | cmp CRET1, #0
1941 | lsl RC, CRET1, #3
1942 | sub RA, BASE, #8
1943 | bgt ->fff_res // Returned nresults+1?
1944 |1: // Returned 0 or -1: retry fast path.
1945 | ldr CARG1, L->top
1946 | ldr LFUNC:CARG3, [BASE, FRAME_FUNC]
1947 | sub NARGS8:RC, CARG1, BASE
1948 | bne ->vm_call_tail // Returned -1?
1949 | ins_callt // Returned 0: retry fast path.
1950 |
1951 |// Reconstruct previous base for vmeta_call during tailcall.
1952 |->vm_call_tail:
1953 | ands CARG1, PC, #FRAME_TYPE
1954 | bic CARG2, PC, #FRAME_TYPEP
1955 | ldreq INS, [PC, #-4]
1956 | andeq CARG2, MASKR8, INS, lsr #5 // Conditional decode_RA8.
1957 | addeq CARG2, CARG2, #8
1958 | sub RB, BASE, CARG2
1959 | b ->vm_call_dispatch // Resolve again for tailcall.
1960 |
1961 |5: // Grow stack for fallback handler.
1962 | mov CARG2, #LUA_MINSTACK
1963 | bl extern lj_state_growstack // (lua_State *L, int n)
1964 | ldr BASE, L->base
1965 | cmp CARG1, CARG1 // Set zero-flag to force retry.
1966 | b <1
1967 |
1968 |->fff_gcstep: // Call GC step function.
1969 | // BASE = new base, RC = nargs*8
1970 | mov RA, lr
1971 | str BASE, L->base
1972 | add CARG2, BASE, NARGS8:RC
1973 | str PC, SAVE_PC // Redundant (but a defined value).
1974 | str CARG2, L->top
1975 | mov CARG1, L
1976 | bl extern lj_gc_step // (lua_State *L)
1977 | ldr BASE, L->base
1978 | mov lr, RA // Help return address predictor.
1979 | ldr CFUNC:CARG3, [BASE, FRAME_FUNC]
1980 | bx lr
1981 |
1982 |//-----------------------------------------------------------------------
1983 |//-- Special dispatch targets -------------------------------------------
1984 |//-----------------------------------------------------------------------
1985 |
1986 |->vm_record: // Dispatch target for recording phase.
1987 |.if JIT
1988 | ldrb CARG1, [DISPATCH, #DISPATCH_GL(hookmask)]
1989 | tst CARG1, #HOOK_VMEVENT // No recording while in vmevent.
1990 | bne >5
1991 | // Decrement the hookcount for consistency, but always do the call.
1992 | ldr CARG2, [DISPATCH, #DISPATCH_GL(hookcount)]
1993 | tst CARG1, #HOOK_ACTIVE
1994 | bne >1
1995 | sub CARG2, CARG2, #1
1996 | tst CARG1, #LUA_MASKLINE|LUA_MASKCOUNT
1997 | strne CARG2, [DISPATCH, #DISPATCH_GL(hookcount)]
1998 | b >1
1999 |.endif
2000 |
2001 |->vm_rethook: // Dispatch target for return hooks.
2002 | ldrb CARG1, [DISPATCH, #DISPATCH_GL(hookmask)]
2003 | tst CARG1, #HOOK_ACTIVE // Hook already active?
2004 | beq >1
2005 |5: // Re-dispatch to static ins.
2006 | decode_OP OP, INS
2007 | add OP, DISPATCH, OP, lsl #2
2008 | ldr pc, [OP, #GG_DISP2STATIC]
2009 |
2010 |->vm_inshook: // Dispatch target for instr/line hooks.
2011 | ldrb CARG1, [DISPATCH, #DISPATCH_GL(hookmask)]
2012 | ldr CARG2, [DISPATCH, #DISPATCH_GL(hookcount)]
2013 | tst CARG1, #HOOK_ACTIVE // Hook already active?
2014 | bne <5
2015 | tst CARG1, #LUA_MASKLINE|LUA_MASKCOUNT
2016 | beq <5
2017 | subs CARG2, CARG2, #1
2018 | str CARG2, [DISPATCH, #DISPATCH_GL(hookcount)]
2019 | beq >1
2020 | tst CARG1, #LUA_MASKLINE
2021 | beq <5
2022 |1:
2023 | mov CARG1, L
2024 | str BASE, L->base
2025 | mov CARG2, PC
2026 | // SAVE_PC must hold the _previous_ PC. The callee updates it with PC.
2027 | bl extern lj_dispatch_ins // (lua_State *L, const BCIns *pc)
2028 |3:
2029 | ldr BASE, L->base
2030 |4: // Re-dispatch to static ins.
2031 | ldrb OP, [PC, #-4]
2032 | ldr INS, [PC, #-4]
2033 | add OP, DISPATCH, OP, lsl #2
2034 | ldr OP, [OP, #GG_DISP2STATIC]
2035 | decode_RA8 RA, INS
2036 | decode_RD RC, INS
2037 | bx OP
2038 |
2039 |->cont_hook: // Continue from hook yield.
2040 | ldr CARG1, [CARG4, #-24]
2041 | add PC, PC, #4
2042 | str CARG1, SAVE_MULTRES // Restore MULTRES for *M ins.
2043 | b <4
2044 |
2045 |->vm_hotloop: // Hot loop counter underflow.
2046 |.if JIT
2047 | ldr LFUNC:CARG3, [BASE, FRAME_FUNC] // Same as curr_topL(L).
2048 | sub CARG1, DISPATCH, #-GG_DISP2J
2049 | str PC, SAVE_PC
2050 | ldr CARG3, LFUNC:CARG3->field_pc
2051 | mov CARG2, PC
2052 | str L, [DISPATCH, #DISPATCH_J(L)]
2053 | ldrb CARG3, [CARG3, #PC2PROTO(framesize)]
2054 | str BASE, L->base
2055 | add CARG3, BASE, CARG3, lsl #3
2056 | str CARG3, L->top
2057 | bl extern lj_trace_hot // (jit_State *J, const BCIns *pc)
2058 | b <3
2059 |.endif
2060 |
2061 |->vm_callhook: // Dispatch target for call hooks.
2062 | mov CARG2, PC
2063 |.if JIT
2064 | b >1
2065 |.endif
2066 |
2067 |->vm_hotcall: // Hot call counter underflow.
2068 |.if JIT
2069 | orr CARG2, PC, #1
2070 |1:
2071 |.endif
2072 | add CARG4, BASE, RC
2073 | str PC, SAVE_PC
2074 | mov CARG1, L
2075 | str BASE, L->base
2076 | sub RA, RA, BASE
2077 | str CARG4, L->top
2078 | bl extern lj_dispatch_call // (lua_State *L, const BCIns *pc)
2079 | // Returns ASMFunction.
2080 | ldr BASE, L->base
2081 | ldr CARG4, L->top
2082 | mov CARG2, #0
2083 | add RA, BASE, RA
2084 | sub NARGS8:RC, CARG4, BASE
2085 | str CARG2, SAVE_PC // Invalidate for subsequent line hook.
2086 | ldr LFUNC:CARG3, [BASE, FRAME_FUNC]
2087 | ldr INS, [PC, #-4]
2088 | bx CRET1
2089 |
2090 |->cont_stitch: // Trace stitching.
2091 |.if JIT
2092 | // RA = resultptr, CARG4 = meta base
2093 | ldr RB, SAVE_MULTRES
2094 | ldr INS, [PC, #-4]
2095 | ldr TRACE:CARG3, [CARG4, #-24] // Save previous trace.
2096 | subs RB, RB, #8
2097 | decode_RA8 RC, INS // Call base.
2098 | beq >2
2099 |1: // Move results down.
2100 | ldrd CARG12, [RA]
2101 | add RA, RA, #8
2102 | subs RB, RB, #8
2103 | strd CARG12, [BASE, RC]
2104 | add RC, RC, #8
2105 | bne <1
2106 |2:
2107 | decode_RA8 RA, INS
2108 | decode_RB8 RB, INS
2109 | add RA, RA, RB
2110 |3:
2111 | cmp RA, RC
2112 | mvn CARG2, #~LJ_TNIL
2113 | bhi >9 // More results wanted?
2114 |
2115 | ldrh RA, TRACE:CARG3->traceno
2116 | ldrh RC, TRACE:CARG3->link
2117 | cmp RC, RA
2118 | beq ->cont_nop // Blacklisted.
2119 | cmp RC, #0
2120 | bne =>BC_JLOOP // Jump to stitched trace.
2121 |
2122 | // Stitch a new trace to the previous trace.
2123 | str RA, [DISPATCH, #DISPATCH_J(exitno)]
2124 | str L, [DISPATCH, #DISPATCH_J(L)]
2125 | str BASE, L->base
2126 | sub CARG1, DISPATCH, #-GG_DISP2J
2127 | mov CARG2, PC
2128 | bl extern lj_dispatch_stitch // (jit_State *J, const BCIns *pc)
2129 | ldr BASE, L->base
2130 | b ->cont_nop
2131 |
2132 |9: // Fill up results with nil.
2133 | strd CARG12, [BASE, RC]
2134 | add RC, RC, #8
2135 | b <3
2136 |.endif
2137 |
2138 |->vm_profhook: // Dispatch target for profiler hook.
2139 #if LJ_HASPROFILE
2140 | mov CARG1, L
2141 | str BASE, L->base
2142 | mov CARG2, PC
2143 | bl extern lj_dispatch_profile // (lua_State *L, const BCIns *pc)
2144 | // HOOK_PROFILE is off again, so re-dispatch to dynamic instruction.
2145 | ldr BASE, L->base
2146 | sub PC, PC, #4
2147 | b ->cont_nop
2148 #endif
2149 |
2150 |//-----------------------------------------------------------------------
2151 |//-- Trace exit handler -------------------------------------------------
2152 |//-----------------------------------------------------------------------
2153 |
2154 |->vm_exit_handler:
2155 |.if JIT
2156 | sub sp, sp, #12
2157 | push {r0,r1,r2,r3,r4,r5,r6,r7,r8,r9,r10,r11,r12}
2158 | ldr CARG1, [sp, #64] // Load original value of lr.
2159 | ldr DISPATCH, [lr] // Load DISPATCH.
2160 | add CARG3, sp, #64 // Recompute original value of sp.
2161 | mv_vmstate CARG4, EXIT
2162 | str CARG3, [sp, #52] // Store sp in RID_SP
2163 | st_vmstate CARG4
2164 | ldr CARG2, [CARG1, #-4]! // Get exit instruction.
2165 | str CARG1, [sp, #56] // Store exit pc in RID_LR and RID_PC.
2166 | str CARG1, [sp, #60]
2167 |.if FPU
2168 | vpush {d0-d15}
2169 |.endif
2170 | lsl CARG2, CARG2, #8
2171 | add CARG1, CARG1, CARG2, asr #6
2172 | ldr CARG2, [lr, #4] // Load exit stub group offset.
2173 | sub CARG1, CARG1, lr
2174 | ldr L, [DISPATCH, #DISPATCH_GL(cur_L)]
2175 | add CARG1, CARG2, CARG1, lsr #2 // Compute exit number.
2176 | ldr BASE, [DISPATCH, #DISPATCH_GL(jit_base)]
2177 | str CARG1, [DISPATCH, #DISPATCH_J(exitno)]
2178 | mov CARG4, #0
2179 | str BASE, L->base
2180 | str L, [DISPATCH, #DISPATCH_J(L)]
2181 | str CARG4, [DISPATCH, #DISPATCH_GL(jit_base)]
2182 | sub CARG1, DISPATCH, #-GG_DISP2J
2183 | mov CARG2, sp
2184 | bl extern lj_trace_exit // (jit_State *J, ExitState *ex)
2185 | // Returns MULTRES (unscaled) or negated error code.
2186 | ldr CARG2, L->cframe
2187 | ldr BASE, L->base
2188 | bic CARG2, CARG2, #~CFRAME_RAWMASK // Use two steps: bic sp is deprecated.
2189 | mov sp, CARG2
2190 | ldr PC, SAVE_PC // Get SAVE_PC.
2191 | str L, SAVE_L // Set SAVE_L (on-trace resume/yield).
2192 | b >1
2193 |.endif
2194 |->vm_exit_interp:
2195 | // CARG1 = MULTRES or negated error code, BASE, PC and DISPATCH set.
2196 |.if JIT
2197 | ldr L, SAVE_L
2198 |1:
2199 | cmn CARG1, #LUA_ERRERR
2200 | bhs >9 // Check for error from exit.
2201 | lsl RC, CARG1, #3
2202 | ldr LFUNC:CARG2, [BASE, FRAME_FUNC]
2203 | str RC, SAVE_MULTRES
2204 | mov CARG3, #0
2205 | str BASE, L->base
2206 | ldr CARG2, LFUNC:CARG2->field_pc
2207 | str CARG3, [DISPATCH, #DISPATCH_GL(jit_base)]
2208 | mv_vmstate CARG4, INTERP
2209 | ldr KBASE, [CARG2, #PC2PROTO(k)]
2210 | // Modified copy of ins_next which handles function header dispatch, too.
2211 | ldrb OP, [PC]
2212 | mov MASKR8, #255
2213 | ldr INS, [PC], #4
2214 | lsl MASKR8, MASKR8, #3 // MASKR8 = 255*8.
2215 | st_vmstate CARG4
2216 | cmn CARG1, #17 // Static dispatch?
2217 | beq >5
2218 | cmp OP, #BC_FUNCC+2 // Fast function?
2219 | bhs >4
2220 |2:
2221 | cmp OP, #BC_FUNCF // Function header?
2222 | ldr OP, [DISPATCH, OP, lsl #2]
2223 | decode_RA8 RA, INS
2224 | lsrlo RC, INS, #16 // No: Decode operands A*8 and D.
2225 | subhs RC, RC, #8
2226 | addhs RA, RA, BASE // Yes: RA = BASE+framesize*8, RC = nargs*8
2227 | ldrhs CARG3, [BASE, FRAME_FUNC]
2228 | bx OP
2229 |
2230 |4: // Check frame below fast function.
2231 | ldr CARG1, [BASE, FRAME_PC]
2232 | ands CARG2, CARG1, #FRAME_TYPE
2233 | bne <2 // Trace stitching continuation?
2234 | // Otherwise set KBASE for Lua function below fast function.
2235 | ldr CARG3, [CARG1, #-4]
2236 | decode_RA8 CARG1, CARG3
2237 | sub CARG2, BASE, CARG1
2238 | ldr LFUNC:CARG3, [CARG2, #-16]
2239 | ldr CARG3, LFUNC:CARG3->field_pc
2240 | ldr KBASE, [CARG3, #PC2PROTO(k)]
2241 | b <2
2242 |
2243 |5: // Dispatch to static entry of original ins replaced by BC_JLOOP.
2244 | ldr CARG1, [DISPATCH, #DISPATCH_J(trace)]
2245 | decode_RD RC, INS
2246 | ldr TRACE:CARG1, [CARG1, RC, lsl #2]
2247 | ldr INS, TRACE:CARG1->startins
2248 | decode_OP OP, INS
2249 | decode_RA8 RA, INS
2250 | add OP, DISPATCH, OP, lsl #2
2251 | decode_RD RC, INS
2252 | ldr pc, [OP, #GG_DISP2STATIC]
2253 |
2254 |9: // Rethrow error from the right C frame.
2255 | rsb CARG2, CARG1, #0
2256 | mov CARG1, L
2257 | bl extern lj_err_trace // (lua_State *L, int errcode)
2258 |.endif
2259 |
2260 |//-----------------------------------------------------------------------
2261 |//-- Math helper functions ----------------------------------------------
2262 |//-----------------------------------------------------------------------
2263 |
2264 |// FP value rounding. Called from JIT code.
2265 |//
2266 |// double lj_vm_floor/ceil/trunc(double x);
2267 |.macro vm_round, func, hf
2268 |.if hf == 1
2269 | vmov CARG1, CARG2, d0
2270 |.endif
2271 | lsl CARG3, CARG2, #1
2272 | adds RB, CARG3, #0x00200000
2273 | bpl >2 // |x| < 1?
2274 | mvn CARG4, #0x3cc
2275 | subs RB, CARG4, RB, asr #21 // 2^0: RB = 51, 2^51: RB = 0.
2276 | bxlo lr // |x| >= 2^52: done.
2277 | mvn CARG4, #1
2278 | bic CARG3, CARG1, CARG4, lsl RB // ztest = lo & ~lomask
2279 | and CARG1, CARG1, CARG4, lsl RB // lo &= lomask
2280 | subs RB, RB, #32
2281 | bicpl CARG4, CARG2, CARG4, lsl RB // |x| <= 2^20: ztest |= hi & ~himask
2282 | orrpl CARG3, CARG3, CARG4
2283 | mvnpl CARG4, #1
2284 | andpl CARG2, CARG2, CARG4, lsl RB // |x| <= 2^20: hi &= himask
2285 |.if "func" == "floor"
2286 | tst CARG3, CARG2, asr #31 // iszero = ((ztest & signmask) == 0)
2287 |.else
2288 | bics CARG3, CARG3, CARG2, asr #31 // iszero = ((ztest & ~signmask) == 0)
2289 |.endif
2290 |.if hf == 1
2291 | vmoveq d0, CARG1, CARG2
2292 |.endif
2293 | bxeq lr // iszero: done.
2294 | mvn CARG4, #1
2295 | cmp RB, #0
2296 | lslpl CARG3, CARG4, RB
2297 | mvnmi CARG3, #0
2298 | add RB, RB, #32
2299 | subs CARG1, CARG1, CARG4, lsl RB // lo = lo-lomask
2300 | sbc CARG2, CARG2, CARG3 // hi = hi-himask+carry
2301 |.if hf == 1
2302 | vmov d0, CARG1, CARG2
2303 |.endif
2304 | bx lr
2305 |
2306 |2: // |x| < 1:
2307 | bxcs lr // |x| is not finite.
2308 | orr CARG3, CARG3, CARG1 // ztest = (2*hi) | lo
2309 |.if "func" == "floor"
2310 | tst CARG3, CARG2, asr #31 // iszero = ((ztest & signmask) == 0)
2311 |.else
2312 | bics CARG3, CARG3, CARG2, asr #31 // iszero = ((ztest & ~signmask) == 0)
2313 |.endif
2314 | mov CARG1, #0 // lo = 0
2315 | and CARG2, CARG2, #0x80000000
2316 | ldrne CARG4, <9 // hi = sign(x) | (iszero ? 0.0 : 1.0)
2317 | orrne CARG2, CARG2, CARG4
2318 |.if hf == 1
2319 | vmov d0, CARG1, CARG2
2320 |.endif
2321 | bx lr
2322 |.endmacro
2323 |
2324 |9:
2325 | .long 0x3ff00000 // hiword(+1.0)
2326 |
2327 |->vm_floor:
2328 |.if HFABI
2329 | vm_round floor, 1
2330 |.endif
2331 |->vm_floor_sf:
2332 | vm_round floor, 0
2333 |
2334 |->vm_ceil:
2335 |.if HFABI
2336 | vm_round ceil, 1
2337 |.endif
2338 |->vm_ceil_sf:
2339 | vm_round ceil, 0
2340 |
2341 |.macro vm_trunc, hf
2342 |.if JIT
2343 |.if hf == 1
2344 | vmov CARG1, CARG2, d0
2345 |.endif
2346 | lsl CARG3, CARG2, #1
2347 | adds RB, CARG3, #0x00200000
2348 | andpl CARG2, CARG2, #0x80000000 // |x| < 1? hi = sign(x), lo = 0.
2349 | movpl CARG1, #0
2350 |.if hf == 1
2351 | vmovpl d0, CARG1, CARG2
2352 |.endif
2353 | bxpl lr
2354 | mvn CARG4, #0x3cc
2355 | subs RB, CARG4, RB, asr #21 // 2^0: RB = 51, 2^51: RB = 0.
2356 | bxlo lr // |x| >= 2^52: already done.
2357 | mvn CARG4, #1
2358 | and CARG1, CARG1, CARG4, lsl RB // lo &= lomask
2359 | subs RB, RB, #32
2360 | andpl CARG2, CARG2, CARG4, lsl RB // |x| <= 2^20: hi &= himask
2361 |.if hf == 1
2362 | vmov d0, CARG1, CARG2
2363 |.endif
2364 | bx lr
2365 |.endif
2366 |.endmacro
2367 |
2368 |->vm_trunc:
2369 |.if HFABI
2370 | vm_trunc 1
2371 |.endif
2372 |->vm_trunc_sf:
2373 | vm_trunc 0
2374 |
2375 | // double lj_vm_mod(double dividend, double divisor);
2376 |->vm_mod:
2377 |.if FPU
2378 | // Special calling convention. Also, RC (r11) is not preserved.
2379 | vdiv.f64 d0, d6, d7
2380 | mov RC, lr
2381 | vmov CARG1, CARG2, d0
2382 | bl ->vm_floor_sf
2383 | vmov d0, CARG1, CARG2
2384 | vmul.f64 d0, d0, d7
2385 | mov lr, RC
2386 | vsub.f64 d6, d6, d0
2387 | bx lr
2388 |.else
2389 | push {r0, r1, r2, r3, r4, lr}
2390 | bl extern __aeabi_ddiv
2391 | bl ->vm_floor_sf
2392 | ldrd CARG34, [sp, #8]
2393 | bl extern __aeabi_dmul
2394 | ldrd CARG34, [sp]
2395 | eor CARG2, CARG2, #0x80000000
2396 | bl extern __aeabi_dadd
2397 | add sp, sp, #20
2398 | pop {pc}
2399 |.endif
2400 |
2401 | // int lj_vm_modi(int dividend, int divisor);
2402 |->vm_modi:
2403 | ands RB, CARG1, #0x80000000
2404 | rsbmi CARG1, CARG1, #0 // a = |dividend|
2405 | eor RB, RB, CARG2, asr #1 // Keep signdiff and sign(divisor).
2406 | cmp CARG2, #0
2407 | rsbmi CARG2, CARG2, #0 // b = |divisor|
2408 | subs CARG4, CARG2, #1
2409 | cmpne CARG1, CARG2
2410 | moveq CARG1, #0 // if (b == 1 || a == b) a = 0
2411 | tsthi CARG2, CARG4
2412 | andeq CARG1, CARG1, CARG4 // else if ((b & (b-1)) == 0) a &= b-1
2413 | bls >1
2414 | // Use repeated subtraction to get the remainder.
2415 | clz CARG3, CARG1
2416 | clz CARG4, CARG2
2417 | sub CARG4, CARG4, CARG3
2418 | rsbs CARG3, CARG4, #31 // entry = (31-(clz(b)-clz(a)))*8
2419 | addne pc, pc, CARG3, lsl #3 // Duff's device.
2420 | nop
2421 {
2422 int i;
2423 for (i = 31; i >= 0; i--) {
2424 | cmp CARG1, CARG2, lsl #i
2425 | subhs CARG1, CARG1, CARG2, lsl #i
2426 }
2427 }
2428 |1:
2429 | cmp CARG1, #0
2430 | cmpne RB, #0
2431 | submi CARG1, CARG1, CARG2 // if (y != 0 && signdiff) y = y - b
2432 | eors CARG2, CARG1, RB, lsl #1
2433 | rsbmi CARG1, CARG1, #0 // if (sign(divisor) != sign(y)) y = -y
2434 | bx lr
2435 |
2436 |//-----------------------------------------------------------------------
2437 |//-- Miscellaneous functions --------------------------------------------
2438 |//-----------------------------------------------------------------------
2439 |
2440 |.define NEXT_TAB, TAB:CARG1
2441 |.define NEXT_RES, CARG1
2442 |.define NEXT_IDX, CARG2
2443 |.define NEXT_TMP0, CARG3
2444 |.define NEXT_TMP1, CARG4
2445 |.define NEXT_LIM, r12
2446 |.define NEXT_RES_PTR, sp
2447 |.define NEXT_RES_VAL, [sp]
2448 |.define NEXT_RES_KEY_I, [sp, #8]
2449 |.define NEXT_RES_KEY_IT, [sp, #12]
2450 |
2451 |// TValue *lj_vm_next(GCtab *t, uint32_t idx)
2452 |// Next idx returned in CRET2.
2453 |->vm_next:
2454 |.if JIT
2455 | ldr NEXT_TMP0, NEXT_TAB->array
2456 | ldr NEXT_LIM, NEXT_TAB->asize
2457 | add NEXT_TMP0, NEXT_TMP0, NEXT_IDX, lsl #3
2458 |1: // Traverse array part.
2459 | subs NEXT_TMP1, NEXT_IDX, NEXT_LIM
2460 | bhs >5
2461 | ldr NEXT_TMP1, [NEXT_TMP0, #4]
2462 | str NEXT_IDX, NEXT_RES_KEY_I
2463 | add NEXT_TMP0, NEXT_TMP0, #8
2464 | add NEXT_IDX, NEXT_IDX, #1
2465 | checktp NEXT_TMP1, LJ_TNIL
2466 | beq <1 // Skip holes in array part.
2467 | ldr NEXT_TMP0, [NEXT_TMP0, #-8]
2468 | mov NEXT_RES, NEXT_RES_PTR
2469 | strd NEXT_TMP0, NEXT_RES_VAL // Stores NEXT_TMP1, too.
2470 | mvn NEXT_TMP0, #~LJ_TISNUM
2471 | str NEXT_TMP0, NEXT_RES_KEY_IT
2472 | bx lr
2473 |
2474 |5: // Traverse hash part.
2475 | ldr NEXT_TMP0, NEXT_TAB->hmask
2476 | ldr NODE:NEXT_RES, NEXT_TAB->node
2477 | add NEXT_TMP1, NEXT_TMP1, NEXT_TMP1, lsl #1
2478 | add NEXT_LIM, NEXT_LIM, NEXT_TMP0
2479 | add NODE:NEXT_RES, NODE:NEXT_RES, NEXT_TMP1, lsl #3
2480 |6:
2481 | cmp NEXT_IDX, NEXT_LIM
2482 | bhi >9
2483 | ldr NEXT_TMP1, NODE:NEXT_RES->val.it
2484 | checktp NEXT_TMP1, LJ_TNIL
2485 | add NEXT_IDX, NEXT_IDX, #1
2486 | bxne lr
2487 | // Skip holes in hash part.
2488 | add NEXT_RES, NEXT_RES, #sizeof(Node)
2489 | b <6
2490 |
2491 |9: // End of iteration. Set the key to nil (not the value).
2492 | mvn NEXT_TMP0, #0
2493 | mov NEXT_RES, NEXT_RES_PTR
2494 | str NEXT_TMP0, NEXT_RES_KEY_IT
2495 | bx lr
2496 |.endif
2497 |
2498 |//-----------------------------------------------------------------------
2499 |//-- FFI helper functions -----------------------------------------------
2500 |//-----------------------------------------------------------------------
2501 |
2502 |// Handler for callback functions.
2503 |// Saveregs already performed. Callback slot number in [sp], g in r12.
2504 |->vm_ffi_callback:
2505 |.if FFI
2506 |.type CTSTATE, CTState, PC
2507 | ldr CTSTATE, GL:r12->ctype_state
2508 | add DISPATCH, r12, #GG_G2DISP
2509 |.if FPU
2510 | str r4, SAVE_R4
2511 | add r4, sp, CFRAME_SPACE+4+8*8
2512 | vstmdb r4!, {d8-d15}
2513 |.endif
2514 |.if HFABI
2515 | add r12, CTSTATE, #offsetof(CTState, cb.fpr[8])
2516 |.endif
2517 | strd CARG34, CTSTATE->cb.gpr[2]
2518 | strd CARG12, CTSTATE->cb.gpr[0]
2519 |.if HFABI
2520 | vstmdb r12!, {d0-d7}
2521 |.endif
2522 | ldr CARG4, [sp]
2523 | add CARG3, sp, #CFRAME_SIZE
2524 | mov CARG1, CTSTATE
2525 | lsr CARG4, CARG4, #3
2526 | str CARG3, CTSTATE->cb.stack
2527 | mov CARG2, sp
2528 | str CARG4, CTSTATE->cb.slot
2529 | str CTSTATE, SAVE_PC // Any value outside of bytecode is ok.
2530 | bl extern lj_ccallback_enter // (CTState *cts, void *cf)
2531 | // Returns lua_State *.
2532 | ldr BASE, L:CRET1->base
2533 | mv_vmstate CARG2, INTERP
2534 | ldr RC, L:CRET1->top
2535 | mov MASKR8, #255
2536 | ldr LFUNC:CARG3, [BASE, FRAME_FUNC]
2537 | mov L, CRET1
2538 | sub RC, RC, BASE
2539 | lsl MASKR8, MASKR8, #3 // MASKR8 = 255*8.
2540 | st_vmstate CARG2
2541 | ins_callt
2542 |.endif
2543 |
2544 |->cont_ffi_callback: // Return from FFI callback.
2545 |.if FFI
2546 | ldr CTSTATE, [DISPATCH, #DISPATCH_GL(ctype_state)]
2547 | str BASE, L->base
2548 | str CARG4, L->top
2549 | str L, CTSTATE->L
2550 | mov CARG1, CTSTATE
2551 | mov CARG2, RA
2552 | bl extern lj_ccallback_leave // (CTState *cts, TValue *o)
2553 | ldrd CARG12, CTSTATE->cb.gpr[0]
2554 |.if HFABI
2555 | vldr d0, CTSTATE->cb.fpr[0]
2556 |.endif
2557 | b ->vm_leave_unw
2558 |.endif
2559 |
2560 |->vm_ffi_call: // Call C function via FFI.
2561 | // Caveat: needs special frame unwinding, see below.
2562 |.if FFI
2563 | .type CCSTATE, CCallState, r4
2564 | push {CCSTATE, r5, r11, lr}
2565 | mov CCSTATE, CARG1
2566 | ldr CARG1, CCSTATE:CARG1->spadj
2567 | ldrb CARG2, CCSTATE->nsp
2568 | add CARG3, CCSTATE, #offsetof(CCallState, stack)
2569 |.if HFABI
2570 | add RB, CCSTATE, #offsetof(CCallState, fpr[0])
2571 |.endif
2572 | mov r11, sp
2573 | sub sp, sp, CARG1 // Readjust stack.
2574 | subs CARG2, CARG2, #1
2575 |.if HFABI
2576 | vldm RB, {d0-d7}
2577 |.endif
2578 | ldr RB, CCSTATE->func
2579 | bmi >2
2580 |1: // Copy stack slots.
2581 | ldr CARG4, [CARG3, CARG2, lsl #2]
2582 | str CARG4, [sp, CARG2, lsl #2]
2583 | subs CARG2, CARG2, #1
2584 | bpl <1
2585 |2:
2586 | ldrd CARG12, CCSTATE->gpr[0]
2587 | ldrd CARG34, CCSTATE->gpr[2]
2588 | blx RB
2589 | mov sp, r11
2590 |.if HFABI
2591 | add r12, CCSTATE, #offsetof(CCallState, fpr[4])
2592 |.endif
2593 | strd CRET1, CCSTATE->gpr[0]
2594 |.if HFABI
2595 | vstmdb r12!, {d0-d3}
2596 |.endif
2597 | pop {CCSTATE, r5, r11, pc}
2598 |.endif
2599 |// Note: vm_ffi_call must be the last function in this object file!
2600 |
2601 |//-----------------------------------------------------------------------
2602 }
2603
2604 /* Generate the code for a single instruction. */
2605 static void build_ins(BuildCtx *ctx, BCOp op, int defop)
2606 {
2607 int vk = 0;
2608 |=>defop:
2609
2610 switch (op) {
2611
2612 /* -- Comparison ops ---------------------------------------------------- */
2613
2614 /* Remember: all ops branch for a true comparison, fall through otherwise. */
2615
2616 case BC_ISLT: case BC_ISGE: case BC_ISLE: case BC_ISGT:
2617 | // RA = src1*8, RC = src2, JMP with RC = target
2618 | lsl RC, RC, #3
2619 | ldrd CARG12, [RA, BASE]!
2620 | ldrh RB, [PC, #2]
2621 | ldrd CARG34, [RC, BASE]!
2622 | add PC, PC, #4
2623 | add RB, PC, RB, lsl #2
2624 | checktp CARG2, LJ_TISNUM
2625 | bne >3
2626 | checktp CARG4, LJ_TISNUM
2627 | bne >4
2628 | cmp CARG1, CARG3
2629 if (op == BC_ISLT) {
2630 | sublt PC, RB, #0x20000
2631 } else if (op == BC_ISGE) {
2632 | subge PC, RB, #0x20000
2633 } else if (op == BC_ISLE) {
2634 | suble PC, RB, #0x20000
2635 } else {
2636 | subgt PC, RB, #0x20000
2637 }
2638 |1:
2639 | ins_next
2640 |
2641 |3: // CARG12 is not an integer.
2642 |.if FPU
2643 | vldr d0, [RA]
2644 | bhi ->vmeta_comp
2645 | // d0 is a number.
2646 | checktp CARG4, LJ_TISNUM
2647 | vldr d1, [RC]
2648 | blo >5
2649 | bhi ->vmeta_comp
2650 | // d0 is a number, CARG3 is an integer.
2651 | vmov s4, CARG3
2652 | vcvt.f64.s32 d1, s4
2653 | b >5
2654 |4: // CARG1 is an integer, CARG34 is not an integer.
2655 | vldr d1, [RC]
2656 | bhi ->vmeta_comp
2657 | // CARG1 is an integer, d1 is a number.
2658 | vmov s4, CARG1
2659 | vcvt.f64.s32 d0, s4
2660 |5: // d0 and d1 are numbers.
2661 | vcmp.f64 d0, d1
2662 | vmrs
2663 | // To preserve NaN semantics GE/GT branch on unordered, but LT/LE don't.
2664 if (op == BC_ISLT) {
2665 | sublo PC, RB, #0x20000
2666 } else if (op == BC_ISGE) {
2667 | subhs PC, RB, #0x20000
2668 } else if (op == BC_ISLE) {
2669 | subls PC, RB, #0x20000
2670 } else {
2671 | subhi PC, RB, #0x20000
2672 }
2673 | b <1
2674 |.else
2675 | bhi ->vmeta_comp
2676 | // CARG12 is a number.
2677 | checktp CARG4, LJ_TISNUM
2678 | movlo RA, RB // Save RB.
2679 | blo >5
2680 | bhi ->vmeta_comp
2681 | // CARG12 is a number, CARG3 is an integer.
2682 | mov CARG1, CARG3
2683 | mov RC, RA
2684 | mov RA, RB // Save RB.
2685 | bl extern __aeabi_i2d
2686 | mov CARG3, CARG1
2687 | mov CARG4, CARG2
2688 | ldrd CARG12, [RC] // Restore first operand.
2689 | b >5
2690 |4: // CARG1 is an integer, CARG34 is not an integer.
2691 | bhi ->vmeta_comp
2692 | // CARG1 is an integer, CARG34 is a number.
2693 | mov RA, RB // Save RB.
2694 | bl extern __aeabi_i2d
2695 | ldrd CARG34, [RC] // Restore second operand.
2696 |5: // CARG12 and CARG34 are numbers.
2697 | bl extern __aeabi_cdcmple
2698 | // To preserve NaN semantics GE/GT branch on unordered, but LT/LE don't.
2699 if (op == BC_ISLT) {
2700 | sublo PC, RA, #0x20000
2701 } else if (op == BC_ISGE) {
2702 | subhs PC, RA, #0x20000
2703 } else if (op == BC_ISLE) {
2704 | subls PC, RA, #0x20000
2705 } else {
2706 | subhi PC, RA, #0x20000
2707 }
2708 | b <1
2709 |.endif
2710 break;
2711
2712 case BC_ISEQV: case BC_ISNEV:
2713 vk = op == BC_ISEQV;
2714 | // RA = src1*8, RC = src2, JMP with RC = target
2715 | lsl RC, RC, #3
2716 | ldrd CARG12, [RA, BASE]!
2717 | ldrh RB, [PC, #2]
2718 | ldrd CARG34, [RC, BASE]!
2719 | add PC, PC, #4
2720 | add RB, PC, RB, lsl #2
2721 | checktp CARG2, LJ_TISNUM
2722 | cmnls CARG4, #-LJ_TISNUM
2723 if (vk) {
2724 | bls ->BC_ISEQN_Z
2725 } else {
2726 | bls ->BC_ISNEN_Z
2727 }
2728 | // Either or both types are not numbers.
2729 |.if FFI
2730 | checktp CARG2, LJ_TCDATA
2731 | checktpne CARG4, LJ_TCDATA
2732 | beq ->vmeta_equal_cd
2733 |.endif
2734 | cmp CARG2, CARG4 // Compare types.
2735 | bne >2 // Not the same type?
2736 | checktp CARG2, LJ_TISPRI
2737 | bhs >1 // Same type and primitive type?
2738 |
2739 | // Same types and not a primitive type. Compare GCobj or pvalue.
2740 | cmp CARG1, CARG3
2741 if (vk) {
2742 | bne >3 // Different GCobjs or pvalues?
2743 |1: // Branch if same.
2744 | sub PC, RB, #0x20000
2745 |2: // Different.
2746 | ins_next
2747 |3:
2748 | checktp CARG2, LJ_TISTABUD
2749 | bhi <2 // Different objects and not table/ud?
2750 } else {
2751 | beq >1 // Same GCobjs or pvalues?
2752 | checktp CARG2, LJ_TISTABUD
2753 | bhi >2 // Different objects and not table/ud?
2754 }
2755 | // Different tables or userdatas. Need to check __eq metamethod.
2756 | // Field metatable must be at same offset for GCtab and GCudata!
2757 | ldr TAB:RA, TAB:CARG1->metatable
2758 | cmp TAB:RA, #0
2759 if (vk) {
2760 | beq <2 // No metatable?
2761 } else {
2762 | beq >2 // No metatable?
2763 }
2764 | ldrb RA, TAB:RA->nomm
2765 | mov CARG4, #1-vk // ne = 0 or 1.
2766 | mov CARG2, CARG1
2767 | tst RA, #1<<MM_eq
2768 | beq ->vmeta_equal // 'no __eq' flag not set?
2769 if (vk) {
2770 | b <2
2771 } else {
2772 |2: // Branch if different.
2773 | sub PC, RB, #0x20000
2774 |1: // Same.
2775 | ins_next
2776 }
2777 break;
2778
2779 case BC_ISEQS: case BC_ISNES:
2780 vk = op == BC_ISEQS;
2781 | // RA = src*8, RC = str_const (~), JMP with RC = target
2782 | mvn RC, RC
2783 | ldrd CARG12, [BASE, RA]
2784 | ldrh RB, [PC, #2]
2785 | ldr STR:CARG3, [KBASE, RC, lsl #2]
2786 | add PC, PC, #4
2787 | add RB, PC, RB, lsl #2
2788 | checktp CARG2, LJ_TSTR
2789 |.if FFI
2790 | bne >7
2791 | cmp CARG1, CARG3
2792 |.else
2793 | cmpeq CARG1, CARG3
2794 |.endif
2795 if (vk) {
2796 | subeq PC, RB, #0x20000
2797 |1:
2798 } else {
2799 |1:
2800 | subne PC, RB, #0x20000
2801 }
2802 | ins_next
2803 |
2804 |.if FFI
2805 |7:
2806 | checktp CARG2, LJ_TCDATA
2807 | bne <1
2808 | b ->vmeta_equal_cd
2809 |.endif
2810 break;
2811
2812 case BC_ISEQN: case BC_ISNEN:
2813 vk = op == BC_ISEQN;
2814 | // RA = src*8, RC = num_const (~), JMP with RC = target
2815 | lsl RC, RC, #3
2816 | ldrd CARG12, [RA, BASE]!
2817 | ldrh RB, [PC, #2]
2818 | ldrd CARG34, [RC, KBASE]!
2819 | add PC, PC, #4
2820 | add RB, PC, RB, lsl #2
2821 if (vk) {
2822 |->BC_ISEQN_Z:
2823 } else {
2824 |->BC_ISNEN_Z:
2825 }
2826 | checktp CARG2, LJ_TISNUM
2827 | bne >3
2828 | checktp CARG4, LJ_TISNUM
2829 | bne >4
2830 | cmp CARG1, CARG3
2831 if (vk) {
2832 | subeq PC, RB, #0x20000
2833 |1:
2834 } else {
2835 |1:
2836 | subne PC, RB, #0x20000
2837 }
2838 |2:
2839 | ins_next
2840 |
2841 |3: // CARG12 is not an integer.
2842 |.if FFI
2843 | bhi >7
2844 |.else
2845 if (!vk) {
2846 | subhi PC, RB, #0x20000
2847 }
2848 | bhi <2
2849 |.endif
2850 |.if FPU
2851 | checktp CARG4, LJ_TISNUM
2852 | vmov s4, CARG3
2853 | vldr d0, [RA]
2854 | vldrlo d1, [RC]
2855 | vcvths.f64.s32 d1, s4
2856 | b >5
2857 |4: // CARG1 is an integer, d1 is a number.
2858 | vmov s4, CARG1
2859 | vldr d1, [RC]
2860 | vcvt.f64.s32 d0, s4
2861 |5: // d0 and d1 are numbers.
2862 | vcmp.f64 d0, d1
2863 | vmrs
2864 if (vk) {
2865 | subeq PC, RB, #0x20000
2866 } else {
2867 | subne PC, RB, #0x20000
2868 }
2869 | b <2
2870 |.else
2871 | // CARG12 is a number.
2872 | checktp CARG4, LJ_TISNUM
2873 | movlo RA, RB // Save RB.
2874 | blo >5
2875 | // CARG12 is a number, CARG3 is an integer.
2876 | mov CARG1, CARG3
2877 | mov RC, RA
2878 |4: // CARG1 is an integer, CARG34 is a number.
2879 | mov RA, RB // Save RB.
2880 | bl extern __aeabi_i2d
2881 | ldrd CARG34, [RC] // Restore other operand.
2882 |5: // CARG12 and CARG34 are numbers.
2883 | bl extern __aeabi_cdcmpeq
2884 if (vk) {
2885 | subeq PC, RA, #0x20000
2886 } else {
2887 | subne PC, RA, #0x20000
2888 }
2889 | b <2
2890 |.endif
2891 |
2892 |.if FFI
2893 |7:
2894 | checktp CARG2, LJ_TCDATA
2895 | bne <1
2896 | b ->vmeta_equal_cd
2897 |.endif
2898 break;
2899
2900 case BC_ISEQP: case BC_ISNEP:
2901 vk = op == BC_ISEQP;
2902 | // RA = src*8, RC = primitive_type (~), JMP with RC = target
2903 | ldrd CARG12, [BASE, RA]
2904 | ldrh RB, [PC, #2]
2905 | add PC, PC, #4
2906 | mvn RC, RC
2907 | add RB, PC, RB, lsl #2
2908 |.if FFI
2909 | checktp CARG2, LJ_TCDATA
2910 | beq ->vmeta_equal_cd
2911 |.endif
2912 | cmp CARG2, RC
2913 if (vk) {
2914 | subeq PC, RB, #0x20000
2915 } else {
2916 | subne PC, RB, #0x20000
2917 }
2918 | ins_next
2919 break;
2920
2921 /* -- Unary test and copy ops ------------------------------------------- */
2922
2923 case BC_ISTC: case BC_ISFC: case BC_IST: case BC_ISF:
2924 | // RA = dst*8 or unused, RC = src, JMP with RC = target
2925 | add RC, BASE, RC, lsl #3
2926 | ldrh RB, [PC, #2]
2927 | ldrd CARG12, [RC]
2928 | add PC, PC, #4
2929 | add RB, PC, RB, lsl #2
2930 | checktp CARG2, LJ_TTRUE
2931 if (op == BC_ISTC || op == BC_IST) {
2932 | subls PC, RB, #0x20000
2933 if (op == BC_ISTC) {
2934 | strdls CARG12, [BASE, RA]
2935 }
2936 } else {
2937 | subhi PC, RB, #0x20000
2938 if (op == BC_ISFC) {
2939 | strdhi CARG12, [BASE, RA]
2940 }
2941 }
2942 | ins_next
2943 break;
2944
2945 case BC_ISTYPE:
2946 | // RA = src*8, RC = -type
2947 | ldrd CARG12, [BASE, RA]
2948 | ins_next1
2949 | cmn CARG2, RC
2950 | ins_next2
2951 | bne ->vmeta_istype
2952 | ins_next3
2953 break;
2954 case BC_ISNUM:
2955 | // RA = src*8, RC = -(TISNUM-1)
2956 | ldrd CARG12, [BASE, RA]
2957 | ins_next1
2958 | checktp CARG2, LJ_TISNUM
2959 | ins_next2
2960 | bhs ->vmeta_istype
2961 | ins_next3
2962 break;
2963
2964 /* -- Unary ops --------------------------------------------------------- */
2965
2966 case BC_MOV:
2967 | // RA = dst*8, RC = src
2968 | lsl RC, RC, #3
2969 | ins_next1
2970 | ldrd CARG12, [BASE, RC]
2971 | ins_next2
2972 | strd CARG12, [BASE, RA]
2973 | ins_next3
2974 break;
2975 case BC_NOT:
2976 | // RA = dst*8, RC = src
2977 | add RC, BASE, RC, lsl #3
2978 | ins_next1
2979 | ldr CARG1, [RC, #4]
2980 | add RA, BASE, RA
2981 | ins_next2
2982 | checktp CARG1, LJ_TTRUE
2983 | mvnls CARG2, #~LJ_TFALSE
2984 | mvnhi CARG2, #~LJ_TTRUE
2985 | str CARG2, [RA, #4]
2986 | ins_next3
2987 break;
2988 case BC_UNM:
2989 | // RA = dst*8, RC = src
2990 | lsl RC, RC, #3
2991 | ldrd CARG12, [BASE, RC]
2992 | ins_next1
2993 | ins_next2
2994 | checktp CARG2, LJ_TISNUM
2995 | bhi ->vmeta_unm
2996 | eorne CARG2, CARG2, #0x80000000
2997 | bne >5
2998 | rsbseq CARG1, CARG1, #0
2999 | ldrdvs CARG12, >9
3000 |5:
3001 | strd CARG12, [BASE, RA]
3002 | ins_next3
3003 |
3004 |.align 8
3005 |9:
3006 | .long 0x00000000, 0x41e00000 // 2^31.
3007 break;
3008 case BC_LEN:
3009 | // RA = dst*8, RC = src
3010 | lsl RC, RC, #3
3011 | ldrd CARG12, [BASE, RC]
3012 | checkstr CARG2, >2
3013 | ldr CARG1, STR:CARG1->len
3014 |1:
3015 | mvn CARG2, #~LJ_TISNUM
3016 | ins_next1
3017 | ins_next2
3018 | strd CARG12, [BASE, RA]
3019 | ins_next3
3020 |2:
3021 | checktab CARG2, ->vmeta_len
3022 #if LJ_52
3023 | ldr TAB:CARG3, TAB:CARG1->metatable
3024 | cmp TAB:CARG3, #0
3025 | bne >9
3026 |3:
3027 #endif
3028 |->BC_LEN_Z:
3029 | .IOS mov RC, BASE
3030 | bl extern lj_tab_len // (GCtab *t)
3031 | // Returns uint32_t (but less than 2^31).
3032 | .IOS mov BASE, RC
3033 | b <1
3034 #if LJ_52
3035 |9:
3036 | ldrb CARG4, TAB:CARG3->nomm
3037 | tst CARG4, #1<<MM_len
3038 | bne <3 // 'no __len' flag set: done.
3039 | b ->vmeta_len
3040 #endif
3041 break;
3042
3043 /* -- Binary ops -------------------------------------------------------- */
3044
3045 |.macro ins_arithcheck, cond, ncond, target
3046 ||if (vk == 1) {
3047 | cmn CARG4, #-LJ_TISNUM
3048 | cmn..cond CARG2, #-LJ_TISNUM
3049 ||} else {
3050 | cmn CARG2, #-LJ_TISNUM
3051 | cmn..cond CARG4, #-LJ_TISNUM
3052 ||}
3053 | b..ncond target
3054 |.endmacro
3055 |.macro ins_arithcheck_int, target
3056 | ins_arithcheck eq, ne, target
3057 |.endmacro
3058 |.macro ins_arithcheck_num, target
3059 | ins_arithcheck lo, hs, target
3060 |.endmacro
3061 |
3062 |.macro ins_arithpre
3063 | decode_RB8 RB, INS
3064 | decode_RC8 RC, INS
3065 | // RA = dst*8, RB = src1*8, RC = src2*8 | num_const*8
3066 ||vk = ((int)op - BC_ADDVN) / (BC_ADDNV-BC_ADDVN);
3067 ||switch (vk) {
3068 ||case 0:
3069 | .if FPU
3070 | ldrd CARG12, [RB, BASE]!
3071 | ldrd CARG34, [RC, KBASE]!
3072 | .else
3073 | ldrd CARG12, [BASE, RB]
3074 | ldrd CARG34, [KBASE, RC]
3075 | .endif
3076 || break;
3077 ||case 1:
3078 | .if FPU
3079 | ldrd CARG34, [RB, BASE]!
3080 | ldrd CARG12, [RC, KBASE]!
3081 | .else
3082 | ldrd CARG34, [BASE, RB]
3083 | ldrd CARG12, [KBASE, RC]
3084 | .endif
3085 || break;
3086 ||default:
3087 | .if FPU
3088 | ldrd CARG12, [RB, BASE]!
3089 | ldrd CARG34, [RC, BASE]!
3090 | .else
3091 | ldrd CARG12, [BASE, RB]
3092 | ldrd CARG34, [BASE, RC]
3093 | .endif
3094 || break;
3095 ||}
3096 |.endmacro
3097 |
3098 |.macro ins_arithpre_fpu, reg1, reg2
3099 |.if FPU
3100 ||if (vk == 1) {
3101 | vldr reg2, [RB]
3102 | vldr reg1, [RC]
3103 ||} else {
3104 | vldr reg1, [RB]
3105 | vldr reg2, [RC]
3106 ||}
3107 |.endif
3108 |.endmacro
3109 |
3110 |.macro ins_arithpost_fpu, reg
3111 | ins_next1
3112 | add RA, BASE, RA
3113 | ins_next2
3114 | vstr reg, [RA]
3115 | ins_next3
3116 |.endmacro
3117 |
3118 |.macro ins_arithfallback, ins
3119 ||switch (vk) {
3120 ||case 0:
3121 | ins ->vmeta_arith_vn
3122 || break;
3123 ||case 1:
3124 | ins ->vmeta_arith_nv
3125 || break;
3126 ||default:
3127 | ins ->vmeta_arith_vv
3128 || break;
3129 ||}
3130 |.endmacro
3131 |
3132 |.macro ins_arithdn, intins, fpins, fpcall
3133 | ins_arithpre
3134 |.if "intins" ~= "vm_modi" and not FPU
3135 | ins_next1
3136 |.endif
3137 | ins_arithcheck_int >5
3138 |.if "intins" == "smull"
3139 | smull CARG1, RC, CARG3, CARG1
3140 | cmp RC, CARG1, asr #31
3141 | ins_arithfallback bne
3142 |.elif "intins" == "vm_modi"
3143 | movs CARG2, CARG3
3144 | ins_arithfallback beq
3145 | bl ->vm_modi
3146 | mvn CARG2, #~LJ_TISNUM
3147 |.else
3148 | intins CARG1, CARG1, CARG3
3149 | ins_arithfallback bvs
3150 |.endif
3151 |4:
3152 |.if "intins" == "vm_modi" or FPU
3153 | ins_next1
3154 |.endif
3155 | ins_next2
3156 | strd CARG12, [BASE, RA]
3157 | ins_next3
3158 |5: // FP variant.
3159 | ins_arithpre_fpu d6, d7
3160 | ins_arithfallback ins_arithcheck_num
3161 |.if FPU
3162 |.if "intins" == "vm_modi"
3163 | bl fpcall
3164 |.else
3165 | fpins d6, d6, d7
3166 |.endif
3167 | ins_arithpost_fpu d6
3168 |.else
3169 | bl fpcall
3170 |.if "intins" ~= "vm_modi"
3171 | ins_next1
3172 |.endif
3173 | b <4
3174 |.endif
3175 |.endmacro
3176 |
3177 |.macro ins_arithfp, fpins, fpcall
3178 | ins_arithpre
3179 |.if "fpins" ~= "extern" or HFABI
3180 | ins_arithpre_fpu d0, d1
3181 |.endif
3182 | ins_arithfallback ins_arithcheck_num
3183 |.if "fpins" == "extern"
3184 | .IOS mov RC, BASE
3185 | bl fpcall
3186 | .IOS mov BASE, RC
3187 |.elif FPU
3188 | fpins d0, d0, d1
3189 |.else
3190 | bl fpcall
3191 |.endif
3192 |.if ("fpins" ~= "extern" or HFABI) and FPU
3193 | ins_arithpost_fpu d0
3194 |.else
3195 | ins_next1
3196 | ins_next2
3197 | strd CARG12, [BASE, RA]
3198 | ins_next3
3199 |.endif
3200 |.endmacro
3201
3202 case BC_ADDVN: case BC_ADDNV: case BC_ADDVV:
3203 | ins_arithdn adds, vadd.f64, extern __aeabi_dadd
3204 break;
3205 case BC_SUBVN: case BC_SUBNV: case BC_SUBVV:
3206 | ins_arithdn subs, vsub.f64, extern __aeabi_dsub
3207 break;
3208 case BC_MULVN: case BC_MULNV: case BC_MULVV:
3209 | ins_arithdn smull, vmul.f64, extern __aeabi_dmul
3210 break;
3211 case BC_DIVVN: case BC_DIVNV: case BC_DIVVV:
3212 | ins_arithfp vdiv.f64, extern __aeabi_ddiv
3213 break;
3214 case BC_MODVN: case BC_MODNV: case BC_MODVV:
3215 | ins_arithdn vm_modi, vm_mod, ->vm_mod
3216 break;
3217 case BC_POW:
3218 | // NYI: (partial) integer arithmetic.
3219 | ins_arithfp extern, extern pow
3220 break;
3221
3222 case BC_CAT:
3223 | decode_RB8 RC, INS
3224 | decode_RC8 RB, INS
3225 | // RA = dst*8, RC = src_start*8, RB = src_end*8 (note: RB/RC swapped!)
3226 | sub CARG3, RB, RC
3227 | str BASE, L->base
3228 | add CARG2, BASE, RB
3229 |->BC_CAT_Z:
3230 | // RA = dst*8, RC = src_start*8, CARG2 = top-1
3231 | mov CARG1, L
3232 | str PC, SAVE_PC
3233 | lsr CARG3, CARG3, #3
3234 | bl extern lj_meta_cat // (lua_State *L, TValue *top, int left)
3235 | // Returns NULL (finished) or TValue * (metamethod).
3236 | ldr BASE, L->base
3237 | cmp CRET1, #0
3238 | bne ->vmeta_binop
3239 | ldrd CARG34, [BASE, RC]
3240 | ins_next1
3241 | ins_next2
3242 | strd CARG34, [BASE, RA] // Copy result to RA.
3243 | ins_next3
3244 break;
3245
3246 /* -- Constant ops ------------------------------------------------------ */
3247
3248 case BC_KSTR:
3249 | // RA = dst*8, RC = str_const (~)
3250 | mvn RC, RC
3251 | ins_next1
3252 | ldr CARG1, [KBASE, RC, lsl #2]
3253 | mvn CARG2, #~LJ_TSTR
3254 | ins_next2
3255 | strd CARG12, [BASE, RA]
3256 | ins_next3
3257 break;
3258 case BC_KCDATA:
3259 |.if FFI
3260 | // RA = dst*8, RC = cdata_const (~)
3261 | mvn RC, RC
3262 | ins_next1
3263 | ldr CARG1, [KBASE, RC, lsl #2]
3264 | mvn CARG2, #~LJ_TCDATA
3265 | ins_next2
3266 | strd CARG12, [BASE, RA]
3267 | ins_next3
3268 |.endif
3269 break;
3270 case BC_KSHORT:
3271 | // RA = dst*8, (RC = int16_literal)
3272 | mov CARG1, INS, asr #16 // Refetch sign-extended reg.
3273 | mvn CARG2, #~LJ_TISNUM
3274 | ins_next1
3275 | ins_next2
3276 | strd CARG12, [BASE, RA]
3277 | ins_next3
3278 break;
3279 case BC_KNUM:
3280 | // RA = dst*8, RC = num_const
3281 | lsl RC, RC, #3
3282 | ins_next1
3283 | ldrd CARG12, [KBASE, RC]
3284 | ins_next2
3285 | strd CARG12, [BASE, RA]
3286 | ins_next3
3287 break;
3288 case BC_KPRI:
3289 | // RA = dst*8, RC = primitive_type (~)
3290 | add RA, BASE, RA
3291 | mvn RC, RC
3292 | ins_next1
3293 | ins_next2
3294 | str RC, [RA, #4]
3295 | ins_next3
3296 break;
3297 case BC_KNIL:
3298 | // RA = base*8, RC = end
3299 | add RA, BASE, RA
3300 | add RC, BASE, RC, lsl #3
3301 | mvn CARG1, #~LJ_TNIL
3302 | str CARG1, [RA, #4]
3303 | add RA, RA, #8
3304 |1:
3305 | str CARG1, [RA, #4]
3306 | cmp RA, RC
3307 | add RA, RA, #8
3308 | blt <1
3309 | ins_next_
3310 break;
3311
3312 /* -- Upvalue and function ops ------------------------------------------ */
3313
3314 case BC_UGET:
3315 | // RA = dst*8, RC = uvnum
3316 | ldr LFUNC:CARG2, [BASE, FRAME_FUNC]
3317 | lsl RC, RC, #2
3318 | add RC, RC, #offsetof(GCfuncL, uvptr)
3319 | ldr UPVAL:CARG2, [LFUNC:CARG2, RC]
3320 | ldr CARG2, UPVAL:CARG2->v
3321 | ldrd CARG34, [CARG2]
3322 | ins_next1
3323 | ins_next2
3324 | strd CARG34, [BASE, RA]
3325 | ins_next3
3326 break;
3327 case BC_USETV:
3328 | // RA = uvnum*8, RC = src
3329 | ldr LFUNC:CARG2, [BASE, FRAME_FUNC]
3330 | lsr RA, RA, #1
3331 | add RA, RA, #offsetof(GCfuncL, uvptr)
3332 | lsl RC, RC, #3
3333 | ldr UPVAL:CARG2, [LFUNC:CARG2, RA]
3334 | ldrd CARG34, [BASE, RC]
3335 | ldrb RB, UPVAL:CARG2->marked
3336 | ldrb RC, UPVAL:CARG2->closed
3337 | ldr CARG2, UPVAL:CARG2->v
3338 | tst RB, #LJ_GC_BLACK // isblack(uv)
3339 | add RB, CARG4, #-LJ_TISGCV
3340 | cmpne RC, #0
3341 | strd CARG34, [CARG2]
3342 | bne >2 // Upvalue is closed and black?
3343 |1:
3344 | ins_next
3345 |
3346 |2: // Check if new value is collectable.
3347 | cmn RB, #-(LJ_TNUMX - LJ_TISGCV)
3348 | ldrbhi RC, GCOBJ:CARG3->gch.marked
3349 | bls <1 // tvisgcv(v)
3350 | sub CARG1, DISPATCH, #-GG_DISP2G
3351 | tst RC, #LJ_GC_WHITES
3352 | // Crossed a write barrier. Move the barrier forward.
3353 |.if IOS
3354 | beq <1
3355 | mov RC, BASE
3356 | bl extern lj_gc_barrieruv // (global_State *g, TValue *tv)
3357 | mov BASE, RC
3358 |.else
3359 | blne extern lj_gc_barrieruv // (global_State *g, TValue *tv)
3360 |.endif
3361 | b <1
3362 break;
3363 case BC_USETS:
3364 | // RA = uvnum*8, RC = str_const (~)
3365 | ldr LFUNC:CARG2, [BASE, FRAME_FUNC]
3366 | lsr RA, RA, #1
3367 | add RA, RA, #offsetof(GCfuncL, uvptr)
3368 | mvn RC, RC
3369 | ldr UPVAL:CARG2, [LFUNC:CARG2, RA]
3370 | ldr STR:CARG3, [KBASE, RC, lsl #2]
3371 | ldrb RB, UPVAL:CARG2->marked
3372 | ldrb RC, UPVAL:CARG2->closed
3373 | ldr CARG2, UPVAL:CARG2->v
3374 | mvn CARG4, #~LJ_TSTR
3375 | tst RB, #LJ_GC_BLACK // isblack(uv)
3376 | ldrb RB, STR:CARG3->marked
3377 | strd CARG34, [CARG2]
3378 | bne >2
3379 |1:
3380 | ins_next
3381 |
3382 |2: // Check if string is white and ensure upvalue is closed.
3383 | tst RB, #LJ_GC_WHITES // iswhite(str)
3384 | cmpne RC, #0
3385 | sub CARG1, DISPATCH, #-GG_DISP2G
3386 | // Crossed a write barrier. Move the barrier forward.
3387 |.if IOS
3388 | beq <1
3389 | mov RC, BASE
3390 | bl extern lj_gc_barrieruv // (global_State *g, TValue *tv)
3391 | mov BASE, RC
3392 |.else
3393 | blne extern lj_gc_barrieruv // (global_State *g, TValue *tv)
3394 |.endif
3395 | b <1
3396 break;
3397 case BC_USETN:
3398 | // RA = uvnum*8, RC = num_const
3399 | ldr LFUNC:CARG2, [BASE, FRAME_FUNC]
3400 | lsr RA, RA, #1
3401 | add RA, RA, #offsetof(GCfuncL, uvptr)
3402 | lsl RC, RC, #3
3403 | ldr UPVAL:CARG2, [LFUNC:CARG2, RA]
3404 | ldrd CARG34, [KBASE, RC]
3405 | ldr CARG2, UPVAL:CARG2->v
3406 | ins_next1
3407 | ins_next2
3408 | strd CARG34, [CARG2]
3409 | ins_next3
3410 break;
3411 case BC_USETP:
3412 | // RA = uvnum*8, RC = primitive_type (~)
3413 | ldr LFUNC:CARG2, [BASE, FRAME_FUNC]
3414 | lsr RA, RA, #1
3415 | add RA, RA, #offsetof(GCfuncL, uvptr)
3416 | ldr UPVAL:CARG2, [LFUNC:CARG2, RA]
3417 | mvn RC, RC
3418 | ldr CARG2, UPVAL:CARG2->v
3419 | ins_next1
3420 | ins_next2
3421 | str RC, [CARG2, #4]
3422 | ins_next3
3423 break;
3424
3425 case BC_UCLO:
3426 | // RA = level*8, RC = target
3427 | ldr CARG3, L->openupval
3428 | add RC, PC, RC, lsl #2
3429 | str BASE, L->base
3430 | cmp CARG3, #0
3431 | sub PC, RC, #0x20000
3432 | beq >1
3433 | mov CARG1, L
3434 | add CARG2, BASE, RA
3435 | bl extern lj_func_closeuv // (lua_State *L, TValue *level)
3436 | ldr BASE, L->base
3437 |1:
3438 | ins_next
3439 break;
3440
3441 case BC_FNEW:
3442 | // RA = dst*8, RC = proto_const (~) (holding function prototype)
3443 | mvn RC, RC
3444 | str BASE, L->base
3445 | ldr CARG2, [KBASE, RC, lsl #2]
3446 | str PC, SAVE_PC
3447 | ldr CARG3, [BASE, FRAME_FUNC]
3448 | mov CARG1, L
3449 | // (lua_State *L, GCproto *pt, GCfuncL *parent)
3450 | bl extern lj_func_newL_gc
3451 | // Returns GCfuncL *.
3452 | ldr BASE, L->base
3453 | mvn CARG2, #~LJ_TFUNC
3454 | ins_next1
3455 | ins_next2
3456 | strd CARG12, [BASE, RA]
3457 | ins_next3
3458 break;
3459
3460 /* -- Table ops --------------------------------------------------------- */
3461
3462 case BC_TNEW:
3463 case BC_TDUP:
3464 | // RA = dst*8, RC = (hbits|asize) | tab_const (~)
3465 if (op == BC_TDUP) {
3466 | mvn RC, RC
3467 }
3468 | ldr CARG3, [DISPATCH, #DISPATCH_GL(gc.total)]
3469 | ldr CARG4, [DISPATCH, #DISPATCH_GL(gc.threshold)]
3470 | str BASE, L->base
3471 | str PC, SAVE_PC
3472 | cmp CARG3, CARG4
3473 | mov CARG1, L
3474 | bhs >5
3475 |1:
3476 if (op == BC_TNEW) {
3477 | lsl CARG2, RC, #21
3478 | lsr CARG3, RC, #11
3479 | asr RC, CARG2, #21
3480 | lsr CARG2, CARG2, #21
3481 | cmn RC, #1
3482 | addeq CARG2, CARG2, #2
3483 | bl extern lj_tab_new // (lua_State *L, int32_t asize, uint32_t hbits)
3484 | // Returns GCtab *.
3485 } else {
3486 | ldr CARG2, [KBASE, RC, lsl #2]
3487 | bl extern lj_tab_dup // (lua_State *L, Table *kt)
3488 | // Returns GCtab *.
3489 }
3490 | ldr BASE, L->base
3491 | mvn CARG2, #~LJ_TTAB
3492 | ins_next1
3493 | ins_next2
3494 | strd CARG12, [BASE, RA]
3495 | ins_next3
3496 |5:
3497 | bl extern lj_gc_step_fixtop // (lua_State *L)
3498 | mov CARG1, L
3499 | b <1
3500 break;
3501
3502 case BC_GGET:
3503 | // RA = dst*8, RC = str_const (~)
3504 case BC_GSET:
3505 | // RA = dst*8, RC = str_const (~)
3506 | ldr LFUNC:CARG2, [BASE, FRAME_FUNC]
3507 | mvn RC, RC
3508 | ldr TAB:CARG1, LFUNC:CARG2->env
3509 | ldr STR:RC, [KBASE, RC, lsl #2]
3510 if (op == BC_GGET) {
3511 | b ->BC_TGETS_Z
3512 } else {
3513 | b ->BC_TSETS_Z
3514 }
3515 break;
3516
3517 case BC_TGETV:
3518 | decode_RB8 RB, INS
3519 | decode_RC8 RC, INS
3520 | // RA = dst*8, RB = table*8, RC = key*8
3521 | ldrd TAB:CARG12, [BASE, RB]
3522 | ldrd CARG34, [BASE, RC]
3523 | checktab CARG2, ->vmeta_tgetv // STALL: load CARG12.
3524 | checktp CARG4, LJ_TISNUM // Integer key?
3525 | ldreq CARG4, TAB:CARG1->array
3526 | ldreq CARG2, TAB:CARG1->asize
3527 | bne >9
3528 |
3529 | add CARG4, CARG4, CARG3, lsl #3
3530 | cmp CARG3, CARG2 // In array part?
3531 | ldrdlo CARG34, [CARG4]
3532 | bhs ->vmeta_tgetv
3533 | ins_next1 // Overwrites RB!
3534 | checktp CARG4, LJ_TNIL
3535 | beq >5
3536 |1:
3537 | ins_next2
3538 | strd CARG34, [BASE, RA]
3539 | ins_next3
3540 |
3541 |5: // Check for __index if table value is nil.
3542 | ldr TAB:CARG2, TAB:CARG1->metatable
3543 | cmp TAB:CARG2, #0
3544 | beq <1 // No metatable: done.
3545 | ldrb CARG2, TAB:CARG2->nomm
3546 | tst CARG2, #1<<MM_index
3547 | bne <1 // 'no __index' flag set: done.
3548 | decode_RB8 RB, INS // Restore RB.
3549 | b ->vmeta_tgetv
3550 |
3551 |9:
3552 | checktp CARG4, LJ_TSTR // String key?
3553 | moveq STR:RC, CARG3
3554 | beq ->BC_TGETS_Z
3555 | b ->vmeta_tgetv
3556 break;
3557 case BC_TGETS:
3558 | decode_RB8 RB, INS
3559 | and RC, RC, #255
3560 | // RA = dst*8, RB = table*8, RC = str_const (~)
3561 | ldrd CARG12, [BASE, RB]
3562 | mvn RC, RC
3563 | ldr STR:RC, [KBASE, RC, lsl #2] // STALL: early RC.
3564 | checktab CARG2, ->vmeta_tgets1
3565 |->BC_TGETS_Z:
3566 | // (TAB:RB =) TAB:CARG1 = GCtab *, STR:RC = GCstr *, RA = dst*8
3567 | ldr CARG3, TAB:CARG1->hmask
3568 | ldr CARG4, STR:RC->sid
3569 | ldr NODE:INS, TAB:CARG1->node
3570 | mov TAB:RB, TAB:CARG1
3571 | and CARG3, CARG3, CARG4 // idx = str->sid & tab->hmask
3572 | add CARG3, CARG3, CARG3, lsl #1
3573 | add NODE:INS, NODE:INS, CARG3, lsl #3 // node = tab->node + idx*3*8
3574 |1:
3575 | ldrd CARG12, NODE:INS->key // STALL: early NODE:INS.
3576 | ldrd CARG34, NODE:INS->val
3577 | ldr NODE:INS, NODE:INS->next
3578 | checktp CARG2, LJ_TSTR
3579 | cmpeq CARG1, STR:RC
3580 | bne >4
3581 | checktp CARG4, LJ_TNIL
3582 | beq >5
3583 |3:
3584 | ins_next1
3585 | ins_next2
3586 | strd CARG34, [BASE, RA]
3587 | ins_next3
3588 |
3589 |4: // Follow hash chain.
3590 | cmp NODE:INS, #0
3591 | bne <1
3592 | // End of hash chain: key not found, nil result.
3593 |
3594 |5: // Check for __index if table value is nil.
3595 | ldr TAB:CARG1, TAB:RB->metatable
3596 | mov CARG3, #0 // Optional clear of undef. value (during load stall).
3597 | mvn CARG4, #~LJ_TNIL
3598 | cmp TAB:CARG1, #0
3599 | beq <3 // No metatable: done.
3600 | ldrb CARG2, TAB:CARG1->nomm
3601 | tst CARG2, #1<<MM_index
3602 | bne <3 // 'no __index' flag set: done.
3603 | b ->vmeta_tgets
3604 break;
3605 case BC_TGETB:
3606 | decode_RB8 RB, INS
3607 | and RC, RC, #255
3608 | // RA = dst*8, RB = table*8, RC = index
3609 | ldrd CARG12, [BASE, RB]
3610 | checktab CARG2, ->vmeta_tgetb // STALL: load CARG12.
3611 | ldr CARG3, TAB:CARG1->asize
3612 | ldr CARG4, TAB:CARG1->array
3613 | lsl CARG2, RC, #3
3614 | cmp RC, CARG3
3615 | ldrdlo CARG34, [CARG4, CARG2]
3616 | bhs ->vmeta_tgetb
3617 | ins_next1 // Overwrites RB!
3618 | checktp CARG4, LJ_TNIL
3619 | beq >5
3620 |1:
3621 | ins_next2
3622 | strd CARG34, [BASE, RA]
3623 | ins_next3
3624 |
3625 |5: // Check for __index if table value is nil.
3626 | ldr TAB:CARG2, TAB:CARG1->metatable
3627 | cmp TAB:CARG2, #0
3628 | beq <1 // No metatable: done.
3629 | ldrb CARG2, TAB:CARG2->nomm
3630 | tst CARG2, #1<<MM_index
3631 | bne <1 // 'no __index' flag set: done.
3632 | b ->vmeta_tgetb
3633 break;
3634 case BC_TGETR:
3635 | decode_RB8 RB, INS
3636 | decode_RC8 RC, INS
3637 | // RA = dst*8, RB = table*8, RC = key*8
3638 | ldr TAB:CARG1, [BASE, RB]
3639 | ldr CARG2, [BASE, RC]
3640 | ldr CARG4, TAB:CARG1->array
3641 | ldr CARG3, TAB:CARG1->asize
3642 | add CARG4, CARG4, CARG2, lsl #3
3643 | cmp CARG2, CARG3 // In array part?
3644 | bhs ->vmeta_tgetr
3645 | ldrd CARG12, [CARG4]
3646 |->BC_TGETR_Z:
3647 | ins_next1
3648 | ins_next2
3649 | strd CARG12, [BASE, RA]
3650 | ins_next3
3651 break;
3652
3653 case BC_TSETV:
3654 | decode_RB8 RB, INS
3655 | decode_RC8 RC, INS
3656 | // RA = src*8, RB = table*8, RC = key*8
3657 | ldrd TAB:CARG12, [BASE, RB]
3658 | ldrd CARG34, [BASE, RC]
3659 | checktab CARG2, ->vmeta_tsetv // STALL: load CARG12.
3660 | checktp CARG4, LJ_TISNUM // Integer key?
3661 | ldreq CARG2, TAB:CARG1->array
3662 | ldreq CARG4, TAB:CARG1->asize
3663 | bne >9
3664 |
3665 | add CARG2, CARG2, CARG3, lsl #3
3666 | cmp CARG3, CARG4 // In array part?
3667 | ldrlo INS, [CARG2, #4]
3668 | bhs ->vmeta_tsetv
3669 | ins_next1 // Overwrites RB!
3670 | checktp INS, LJ_TNIL
3671 | ldrb INS, TAB:CARG1->marked
3672 | ldrd CARG34, [BASE, RA]
3673 | beq >5
3674 |1:
3675 | tst INS, #LJ_GC_BLACK // isblack(table)
3676 | strd CARG34, [CARG2]
3677 | bne >7
3678 |2:
3679 | ins_next2
3680 | ins_next3
3681 |
3682 |5: // Check for __newindex if previous value is nil.
3683 | ldr TAB:RA, TAB:CARG1->metatable
3684 | cmp TAB:RA, #0
3685 | beq <1 // No metatable: done.
3686 | ldrb RA, TAB:RA->nomm
3687 | tst RA, #1<<MM_newindex
3688 | bne <1 // 'no __newindex' flag set: done.
3689 | ldr INS, [PC, #-4] // Restore RA and RB.
3690 | decode_RB8 RB, INS
3691 | decode_RA8 RA, INS
3692 | b ->vmeta_tsetv
3693 |
3694 |7: // Possible table write barrier for the value. Skip valiswhite check.
3695 | barrierback TAB:CARG1, INS, CARG3
3696 | b <2
3697 |
3698 |9:
3699 | checktp CARG4, LJ_TSTR // String key?
3700 | moveq STR:RC, CARG3
3701 | beq ->BC_TSETS_Z
3702 | b ->vmeta_tsetv
3703 break;
3704 case BC_TSETS:
3705 | decode_RB8 RB, INS
3706 | and RC, RC, #255
3707 | // RA = src*8, RB = table*8, RC = str_const (~)
3708 | ldrd CARG12, [BASE, RB]
3709 | mvn RC, RC
3710 | ldr STR:RC, [KBASE, RC, lsl #2] // STALL: early RC.
3711 | checktab CARG2, ->vmeta_tsets1
3712 |->BC_TSETS_Z:
3713 | // (TAB:RB =) TAB:CARG1 = GCtab *, STR:RC = GCstr *, RA = dst*8
3714 | ldr CARG3, TAB:CARG1->hmask
3715 | ldr CARG4, STR:RC->sid
3716 | ldr NODE:INS, TAB:CARG1->node
3717 | mov TAB:RB, TAB:CARG1
3718 | and CARG3, CARG3, CARG4 // idx = str->sid & tab->hmask
3719 | add CARG3, CARG3, CARG3, lsl #1
3720 | mov CARG4, #0
3721 | add NODE:INS, NODE:INS, CARG3, lsl #3 // node = tab->node + idx*3*8
3722 | strb CARG4, TAB:RB->nomm // Clear metamethod cache.
3723 |1:
3724 | ldrd CARG12, NODE:INS->key
3725 | ldr CARG4, NODE:INS->val.it
3726 | ldr NODE:CARG3, NODE:INS->next
3727 | checktp CARG2, LJ_TSTR
3728 | cmpeq CARG1, STR:RC
3729 | bne >5
3730 | ldrb CARG2, TAB:RB->marked
3731 | checktp CARG4, LJ_TNIL // Key found, but nil value?
3732 | ldrd CARG34, [BASE, RA]
3733 | beq >4
3734 |2:
3735 | tst CARG2, #LJ_GC_BLACK // isblack(table)
3736 | strd CARG34, NODE:INS->val
3737 | bne >7
3738 |3:
3739 | ins_next
3740 |
3741 |4: // Check for __newindex if previous value is nil.
3742 | ldr TAB:CARG1, TAB:RB->metatable
3743 | cmp TAB:CARG1, #0
3744 | beq <2 // No metatable: done.
3745 | ldrb CARG1, TAB:CARG1->nomm
3746 | tst CARG1, #1<<MM_newindex
3747 | bne <2 // 'no __newindex' flag set: done.
3748 | b ->vmeta_tsets
3749 |
3750 |5: // Follow hash chain.
3751 | movs NODE:INS, NODE:CARG3
3752 | bne <1
3753 | // End of hash chain: key not found, add a new one.
3754 |
3755 | // But check for __newindex first.
3756 | ldr TAB:CARG1, TAB:RB->metatable
3757 | mov CARG3, TMPDp
3758 | str PC, SAVE_PC
3759 | cmp TAB:CARG1, #0 // No metatable: continue.
3760 | str BASE, L->base
3761 | ldrbne CARG2, TAB:CARG1->nomm
3762 | mov CARG1, L
3763 | beq >6
3764 | tst CARG2, #1<<MM_newindex
3765 | beq ->vmeta_tsets // 'no __newindex' flag NOT set: check.
3766 |6:
3767 | mvn CARG4, #~LJ_TSTR
3768 | str STR:RC, TMPDlo
3769 | mov CARG2, TAB:RB
3770 | str CARG4, TMPDhi
3771 | bl extern lj_tab_newkey // (lua_State *L, GCtab *t, TValue *k)
3772 | // Returns TValue *.
3773 | ldr BASE, L->base
3774 | ldrd CARG34, [BASE, RA]
3775 | strd CARG34, [CRET1]
3776 | b <3 // No 2nd write barrier needed.
3777 |
3778 |7: // Possible table write barrier for the value. Skip valiswhite check.
3779 | barrierback TAB:RB, CARG2, CARG3
3780 | b <3
3781 break;
3782 case BC_TSETB:
3783 | decode_RB8 RB, INS
3784 | and RC, RC, #255
3785 | // RA = src*8, RB = table*8, RC = index
3786 | ldrd CARG12, [BASE, RB]
3787 | checktab CARG2, ->vmeta_tsetb // STALL: load CARG12.
3788 | ldr CARG3, TAB:CARG1->asize
3789 | ldr RB, TAB:CARG1->array
3790 | lsl CARG2, RC, #3
3791 | cmp RC, CARG3
3792 | ldrdlo CARG34, [CARG2, RB]!
3793 | bhs ->vmeta_tsetb
3794 | ins_next1 // Overwrites RB!
3795 | checktp CARG4, LJ_TNIL
3796 | ldrb INS, TAB:CARG1->marked
3797 | ldrd CARG34, [BASE, RA]
3798 | beq >5
3799 |1:
3800 | tst INS, #LJ_GC_BLACK // isblack(table)
3801 | strd CARG34, [CARG2]
3802 | bne >7
3803 |2:
3804 | ins_next2
3805 | ins_next3
3806 |
3807 |5: // Check for __newindex if previous value is nil.
3808 | ldr TAB:RA, TAB:CARG1->metatable
3809 | cmp TAB:RA, #0
3810 | beq <1 // No metatable: done.
3811 | ldrb RA, TAB:RA->nomm
3812 | tst RA, #1<<MM_newindex
3813 | bne <1 // 'no __newindex' flag set: done.
3814 | ldr INS, [PC, #-4] // Restore INS.
3815 | decode_RA8 RA, INS
3816 | b ->vmeta_tsetb
3817 |
3818 |7: // Possible table write barrier for the value. Skip valiswhite check.
3819 | barrierback TAB:CARG1, INS, CARG3
3820 | b <2
3821 break;
3822 case BC_TSETR:
3823 | decode_RB8 RB, INS
3824 | decode_RC8 RC, INS
3825 | // RA = src*8, RB = table*8, RC = key*8
3826 | ldr TAB:CARG2, [BASE, RB]
3827 | ldr CARG3, [BASE, RC]
3828 | ldrb INS, TAB:CARG2->marked
3829 | ldr CARG1, TAB:CARG2->array
3830 | ldr CARG4, TAB:CARG2->asize
3831 | tst INS, #LJ_GC_BLACK // isblack(table)
3832 | add CARG1, CARG1, CARG3, lsl #3
3833 | bne >7
3834 |2:
3835 | cmp CARG3, CARG4 // In array part?
3836 | bhs ->vmeta_tsetr
3837 |->BC_TSETR_Z:
3838 | ldrd CARG34, [BASE, RA]
3839 | ins_next1
3840 | ins_next2
3841 | strd CARG34, [CARG1]
3842 | ins_next3
3843 |
3844 |7: // Possible table write barrier for the value. Skip valiswhite check.
3845 | barrierback TAB:CARG2, INS, RB
3846 | b <2
3847 break;
3848
3849 case BC_TSETM:
3850 | // RA = base*8 (table at base-1), RC = num_const (start index)
3851 | add RA, BASE, RA
3852 |1:
3853 | ldr RB, SAVE_MULTRES
3854 | ldr TAB:CARG2, [RA, #-8] // Guaranteed to be a table.
3855 | ldr CARG1, [KBASE, RC, lsl #3] // Integer constant is in lo-word.
3856 | subs RB, RB, #8
3857 | ldr CARG4, TAB:CARG2->asize
3858 | beq >4 // Nothing to copy?
3859 | add CARG3, CARG1, RB, lsr #3
3860 | cmp CARG3, CARG4
3861 | ldr CARG4, TAB:CARG2->array
3862 | add RB, RA, RB
3863 | bhi >5
3864 | add INS, CARG4, CARG1, lsl #3
3865 | ldrb CARG1, TAB:CARG2->marked
3866 |3: // Copy result slots to table.
3867 | ldrd CARG34, [RA], #8
3868 | strd CARG34, [INS], #8
3869 | cmp RA, RB
3870 | blo <3
3871 | tst CARG1, #LJ_GC_BLACK // isblack(table)
3872 | bne >7
3873 |4:
3874 | ins_next
3875 |
3876 |5: // Need to resize array part.
3877 | str BASE, L->base
3878 | mov CARG1, L
3879 | str PC, SAVE_PC
3880 | bl extern lj_tab_reasize // (lua_State *L, GCtab *t, int nasize)
3881 | // Must not reallocate the stack.
3882 | .IOS ldr BASE, L->base
3883 | b <1
3884 |
3885 |7: // Possible table write barrier for any value. Skip valiswhite check.
3886 | barrierback TAB:CARG2, CARG1, CARG3
3887 | b <4
3888 break;
3889
3890 /* -- Calls and vararg handling ----------------------------------------- */
3891
3892 case BC_CALLM:
3893 | // RA = base*8, (RB = nresults+1,) RC = extra_nargs
3894 | ldr CARG1, SAVE_MULTRES
3895 | decode_RC8 NARGS8:RC, INS
3896 | add NARGS8:RC, NARGS8:RC, CARG1
3897 | b ->BC_CALL_Z
3898 break;
3899 case BC_CALL:
3900 | decode_RC8 NARGS8:RC, INS
3901 | // RA = base*8, (RB = nresults+1,) RC = (nargs+1)*8
3902 |->BC_CALL_Z:
3903 | mov RB, BASE // Save old BASE for vmeta_call.
3904 | ldrd CARG34, [BASE, RA]!
3905 | sub NARGS8:RC, NARGS8:RC, #8
3906 | add BASE, BASE, #8
3907 | checkfunc CARG4, ->vmeta_call
3908 | ins_call
3909 break;
3910
3911 case BC_CALLMT:
3912 | // RA = base*8, (RB = 0,) RC = extra_nargs
3913 | ldr CARG1, SAVE_MULTRES
3914 | add NARGS8:RC, CARG1, RC, lsl #3
3915 | b ->BC_CALLT1_Z
3916 break;
3917 case BC_CALLT:
3918 | lsl NARGS8:RC, RC, #3
3919 | // RA = base*8, (RB = 0,) RC = (nargs+1)*8
3920 |->BC_CALLT1_Z:
3921 | ldrd LFUNC:CARG34, [RA, BASE]!
3922 | sub NARGS8:RC, NARGS8:RC, #8
3923 | add RA, RA, #8
3924 | checkfunc CARG4, ->vmeta_callt
3925 | ldr PC, [BASE, FRAME_PC]
3926 |->BC_CALLT2_Z:
3927 | mov RB, #0
3928 | ldrb CARG4, LFUNC:CARG3->ffid
3929 | tst PC, #FRAME_TYPE
3930 | bne >7
3931 |1:
3932 | str LFUNC:CARG3, [BASE, FRAME_FUNC] // Copy function down, but keep PC.
3933 | cmp NARGS8:RC, #0
3934 | beq >3
3935 |2:
3936 | ldrd CARG12, [RA, RB]
3937 | add INS, RB, #8
3938 | cmp INS, NARGS8:RC
3939 | strd CARG12, [BASE, RB]
3940 | mov RB, INS
3941 | bne <2
3942 |3:
3943 | cmp CARG4, #1 // (> FF_C) Calling a fast function?
3944 | bhi >5
3945 |4:
3946 | ins_callt
3947 |
3948 |5: // Tailcall to a fast function with a Lua frame below.
3949 | ldr INS, [PC, #-4]
3950 | decode_RA8 RA, INS
3951 | sub CARG1, BASE, RA
3952 | ldr LFUNC:CARG1, [CARG1, #-16]
3953 | ldr CARG1, LFUNC:CARG1->field_pc
3954 | ldr KBASE, [CARG1, #PC2PROTO(k)]
3955 | b <4
3956 |
3957 |7: // Tailcall from a vararg function.
3958 | eor PC, PC, #FRAME_VARG
3959 | tst PC, #FRAME_TYPEP // Vararg frame below?
3960 | movne CARG4, #0 // Clear ffid if no Lua function below.
3961 | bne <1
3962 | sub BASE, BASE, PC
3963 | ldr PC, [BASE, FRAME_PC]
3964 | tst PC, #FRAME_TYPE
3965 | movne CARG4, #0 // Clear ffid if no Lua function below.
3966 | b <1
3967 break;
3968
3969 case BC_ITERC:
3970 | // RA = base*8, (RB = nresults+1, RC = nargs+1 (2+1))
3971 | add RA, BASE, RA
3972 | mov RB, BASE // Save old BASE for vmeta_call.
3973 | ldrd CARG34, [RA, #-16]
3974 | ldrd CARG12, [RA, #-8]
3975 | add BASE, RA, #8
3976 | strd CARG34, [RA, #8] // Copy state.
3977 | strd CARG12, [RA, #16] // Copy control var.
3978 | // STALL: locked CARG34.
3979 | ldrd LFUNC:CARG34, [RA, #-24]
3980 | mov NARGS8:RC, #16 // Iterators get 2 arguments.
3981 | // STALL: load CARG34.
3982 | strd LFUNC:CARG34, [RA] // Copy callable.
3983 | checkfunc CARG4, ->vmeta_call
3984 | ins_call
3985 break;
3986
3987 case BC_ITERN:
3988 |.if JIT
3989 | hotloop
3990 |.endif
3991 |->vm_IITERN:
3992 | // RA = base*8, (RB = nresults+1, RC = nargs+1 (2+1))
3993 | add RA, BASE, RA
3994 | ldr TAB:RB, [RA, #-16]
3995 | ldr CARG1, [RA, #-8] // Get index from control var.
3996 | ldr INS, TAB:RB->asize
3997 | ldr CARG2, TAB:RB->array
3998 | add PC, PC, #4
3999 |1: // Traverse array part.
4000 | subs RC, CARG1, INS
4001 | add CARG3, CARG2, CARG1, lsl #3
4002 | bhs >5 // Index points after array part?
4003 | ldrd CARG34, [CARG3]
4004 | checktp CARG4, LJ_TNIL
4005 | addeq CARG1, CARG1, #1 // Skip holes in array part.
4006 | beq <1
4007 | ldrh RC, [PC, #-2]
4008 | mvn CARG2, #~LJ_TISNUM
4009 | strd CARG34, [RA, #8]
4010 | add RC, PC, RC, lsl #2
4011 | add RB, CARG1, #1
4012 | strd CARG12, [RA]
4013 | sub PC, RC, #0x20000
4014 | str RB, [RA, #-8] // Update control var.
4015 |3:
4016 | ins_next
4017 |
4018 |5: // Traverse hash part.
4019 | ldr CARG4, TAB:RB->hmask
4020 | ldr NODE:RB, TAB:RB->node
4021 |6:
4022 | add CARG1, RC, RC, lsl #1
4023 | cmp RC, CARG4 // End of iteration? Branch to ITERL+1.
4024 | add NODE:CARG3, NODE:RB, CARG1, lsl #3 // node = tab->node + idx*3*8
4025 | bhi <3
4026 | ldrd CARG12, NODE:CARG3->val
4027 | checktp CARG2, LJ_TNIL
4028 | add RC, RC, #1
4029 | beq <6 // Skip holes in hash part.
4030 | ldrh RB, [PC, #-2]
4031 | add RC, RC, INS
4032 | ldrd CARG34, NODE:CARG3->key
4033 | str RC, [RA, #-8] // Update control var.
4034 | strd CARG12, [RA, #8]
4035 | add RC, PC, RB, lsl #2
4036 | sub PC, RC, #0x20000
4037 | strd CARG34, [RA]
4038 | b <3
4039 break;
4040
4041 case BC_ISNEXT:
4042 | // RA = base*8, RC = target (points to ITERN)
4043 | add RA, BASE, RA
4044 | add RC, PC, RC, lsl #2
4045 | ldrd CFUNC:CARG12, [RA, #-24]
4046 | ldr CARG3, [RA, #-12]
4047 | ldr CARG4, [RA, #-4]
4048 | checktp CARG2, LJ_TFUNC
4049 | ldrbeq CARG1, CFUNC:CARG1->ffid
4050 | checktpeq CARG3, LJ_TTAB
4051 | checktpeq CARG4, LJ_TNIL
4052 | cmpeq CARG1, #FF_next_N
4053 | subeq PC, RC, #0x20000
4054 | bne >5
4055 | ins_next1
4056 | ins_next2
4057 | mov CARG1, #0
4058 | mvn CARG2, #~LJ_KEYINDEX
4059 | strd CARG1, [RA, #-8] // Initialize control var.
4060 |1:
4061 | ins_next3
4062 |5: // Despecialize bytecode if any of the checks fail.
4063 | mov CARG1, #BC_JMP
4064 | mov OP, #BC_ITERC
4065 | strb CARG1, [PC, #-4]
4066 | sub PC, RC, #0x20000
4067 |.if JIT
4068 | ldrb CARG1, [PC]
4069 | cmp CARG1, #BC_ITERN
4070 | bne >6
4071 |.endif
4072 | strb OP, [PC] // Subsumes ins_next1.
4073 | ins_next2
4074 | b <1
4075 |.if JIT
4076 |6: // Unpatch JLOOP.
4077 | ldr CARG1, [DISPATCH, #DISPATCH_J(trace)]
4078 | ldrh CARG2, [PC, #2]
4079 | ldr TRACE:CARG1, [CARG1, CARG2, lsl #2]
4080 | // Subsumes ins_next1 and ins_next2.
4081 | ldr INS, TRACE:CARG1->startins
4082 | bfi INS, OP, #0, #8
4083 | str INS, [PC], #4
4084 | b <1
4085 |.endif
4086 break;
4087
4088 case BC_VARG:
4089 | decode_RB8 RB, INS
4090 | decode_RC8 RC, INS
4091 | // RA = base*8, RB = (nresults+1)*8, RC = numparams*8
4092 | ldr CARG1, [BASE, FRAME_PC]
4093 | add RC, BASE, RC
4094 | add RA, BASE, RA
4095 | add RC, RC, #FRAME_VARG
4096 | add CARG4, RA, RB
4097 | sub CARG3, BASE, #8 // CARG3 = vtop
4098 | sub RC, RC, CARG1 // RC = vbase
4099 | // Note: RC may now be even _above_ BASE if nargs was < numparams.
4100 | cmp RB, #0
4101 | sub CARG1, CARG3, RC
4102 | beq >5 // Copy all varargs?
4103 | sub CARG4, CARG4, #16
4104 |1: // Copy vararg slots to destination slots.
4105 | cmp RC, CARG3
4106 | ldrdlo CARG12, [RC], #8
4107 | mvnhs CARG2, #~LJ_TNIL
4108 | cmp RA, CARG4
4109 | strd CARG12, [RA], #8
4110 | blo <1
4111 |2:
4112 | ins_next
4113 |
4114 |5: // Copy all varargs.
4115 | ldr CARG4, L->maxstack
4116 | cmp CARG1, #0
4117 | movle RB, #8 // MULTRES = (0+1)*8
4118 | addgt RB, CARG1, #8
4119 | add CARG2, RA, CARG1
4120 | str RB, SAVE_MULTRES
4121 | ble <2
4122 | cmp CARG2, CARG4
4123 | bhi >7
4124 |6:
4125 | ldrd CARG12, [RC], #8
4126 | strd CARG12, [RA], #8
4127 | cmp RC, CARG3
4128 | blo <6
4129 | b <2
4130 |
4131 |7: // Grow stack for varargs.
4132 | lsr CARG2, CARG1, #3
4133 | str RA, L->top
4134 | mov CARG1, L
4135 | str BASE, L->base
4136 | sub RC, RC, BASE // Need delta, because BASE may change.
4137 | str PC, SAVE_PC
4138 | sub RA, RA, BASE
4139 | bl extern lj_state_growstack // (lua_State *L, int n)
4140 | ldr BASE, L->base
4141 | add RA, BASE, RA
4142 | add RC, BASE, RC
4143 | sub CARG3, BASE, #8
4144 | b <6
4145 break;
4146
4147 /* -- Returns ----------------------------------------------------------- */
4148
4149 case BC_RETM:
4150 | // RA = results*8, RC = extra results
4151 | ldr CARG1, SAVE_MULTRES
4152 | ldr PC, [BASE, FRAME_PC]
4153 | add RA, BASE, RA
4154 | add RC, CARG1, RC, lsl #3
4155 | b ->BC_RETM_Z
4156 break;
4157
4158 case BC_RET:
4159 | // RA = results*8, RC = nresults+1
4160 | ldr PC, [BASE, FRAME_PC]
4161 | lsl RC, RC, #3
4162 | add RA, BASE, RA
4163 |->BC_RETM_Z:
4164 | str RC, SAVE_MULTRES
4165 |1:
4166 | ands CARG1, PC, #FRAME_TYPE
4167 | eor CARG2, PC, #FRAME_VARG
4168 | bne ->BC_RETV2_Z
4169 |
4170 |->BC_RET_Z:
4171 | // BASE = base, RA = resultptr, RC = (nresults+1)*8, PC = return
4172 | ldr INS, [PC, #-4]
4173 | subs CARG4, RC, #8
4174 | sub CARG3, BASE, #8
4175 | beq >3
4176 |2:
4177 | ldrd CARG12, [RA], #8
4178 | add BASE, BASE, #8
4179 | subs CARG4, CARG4, #8
4180 | strd CARG12, [BASE, #-16]
4181 | bne <2
4182 |3:
4183 | decode_RA8 RA, INS
4184 | sub CARG4, CARG3, RA
4185 | decode_RB8 RB, INS
4186 | ldr LFUNC:CARG1, [CARG4, FRAME_FUNC]
4187 |5:
4188 | cmp RB, RC // More results expected?
4189 | bhi >6
4190 | mov BASE, CARG4
4191 | ldr CARG2, LFUNC:CARG1->field_pc
4192 | ins_next1
4193 | ins_next2
4194 | ldr KBASE, [CARG2, #PC2PROTO(k)]
4195 | ins_next3
4196 |
4197 |6: // Fill up results with nil.
4198 | mvn CARG2, #~LJ_TNIL
4199 | add BASE, BASE, #8
4200 | add RC, RC, #8
4201 | str CARG2, [BASE, #-12]
4202 | b <5
4203 |
4204 |->BC_RETV1_Z: // Non-standard return case.
4205 | add RA, BASE, RA
4206 |->BC_RETV2_Z:
4207 | tst CARG2, #FRAME_TYPEP
4208 | bne ->vm_return
4209 | // Return from vararg function: relocate BASE down.
4210 | sub BASE, BASE, CARG2
4211 | ldr PC, [BASE, FRAME_PC]
4212 | b <1
4213 break;
4214
4215 case BC_RET0: case BC_RET1:
4216 | // RA = results*8, RC = nresults+1
4217 | ldr PC, [BASE, FRAME_PC]
4218 | lsl RC, RC, #3
4219 | str RC, SAVE_MULTRES
4220 | ands CARG1, PC, #FRAME_TYPE
4221 | eor CARG2, PC, #FRAME_VARG
4222 | ldreq INS, [PC, #-4]
4223 | bne ->BC_RETV1_Z
4224 if (op == BC_RET1) {
4225 | ldrd CARG12, [BASE, RA]
4226 }
4227 | sub CARG4, BASE, #8
4228 | decode_RA8 RA, INS
4229 if (op == BC_RET1) {
4230 | strd CARG12, [CARG4]
4231 }
4232 | sub BASE, CARG4, RA
4233 | decode_RB8 RB, INS
4234 | ldr LFUNC:CARG1, [BASE, FRAME_FUNC]
4235 |5:
4236 | cmp RB, RC
4237 | bhi >6
4238 | ldr CARG2, LFUNC:CARG1->field_pc
4239 | ins_next1
4240 | ins_next2
4241 | ldr KBASE, [CARG2, #PC2PROTO(k)]
4242 | ins_next3
4243 |
4244 |6: // Fill up results with nil.
4245 | sub CARG2, CARG4, #4
4246 | mvn CARG3, #~LJ_TNIL
4247 | str CARG3, [CARG2, RC]
4248 | add RC, RC, #8
4249 | b <5
4250 break;
4251
4252 /* -- Loops and branches ------------------------------------------------ */
4253
4254 |.define FOR_IDX, [RA]; .define FOR_TIDX, [RA, #4]
4255 |.define FOR_STOP, [RA, #8]; .define FOR_TSTOP, [RA, #12]
4256 |.define FOR_STEP, [RA, #16]; .define FOR_TSTEP, [RA, #20]
4257 |.define FOR_EXT, [RA, #24]; .define FOR_TEXT, [RA, #28]
4258
4259 case BC_FORL:
4260 |.if JIT
4261 | hotloop
4262 |.endif
4263 | // Fall through. Assumes BC_IFORL follows.
4264 break;
4265
4266 case BC_JFORI:
4267 case BC_JFORL:
4268 #if !LJ_HASJIT
4269 break;
4270 #endif
4271 case BC_FORI:
4272 case BC_IFORL:
4273 | // RA = base*8, RC = target (after end of loop or start of loop)
4274 vk = (op == BC_IFORL || op == BC_JFORL);
4275 | ldrd CARG12, [RA, BASE]!
4276 if (op != BC_JFORL) {
4277 | add RC, PC, RC, lsl #2
4278 }
4279 if (!vk) {
4280 | ldrd CARG34, FOR_STOP
4281 | checktp CARG2, LJ_TISNUM
4282 | ldr RB, FOR_TSTEP
4283 | bne >5
4284 | checktp CARG4, LJ_TISNUM
4285 | ldr CARG4, FOR_STEP
4286 | checktpeq RB, LJ_TISNUM
4287 | bne ->vmeta_for
4288 | cmp CARG4, #0
4289 | blt >4
4290 | cmp CARG1, CARG3
4291 } else {
4292 | ldrd CARG34, FOR_STEP
4293 | checktp CARG2, LJ_TISNUM
4294 | bne >5
4295 | adds CARG1, CARG1, CARG3
4296 | ldr CARG4, FOR_STOP
4297 if (op == BC_IFORL) {
4298 | addvs RC, PC, #0x20000 // Overflow: prevent branch.
4299 } else {
4300 | bvs >2 // Overflow: do not enter mcode.
4301 }
4302 | cmp CARG3, #0
4303 | blt >4
4304 | cmp CARG1, CARG4
4305 }
4306 |1:
4307 if (op == BC_FORI) {
4308 | subgt PC, RC, #0x20000
4309 } else if (op == BC_JFORI) {
4310 | sub PC, RC, #0x20000
4311 | ldrhle RC, [PC, #-2]
4312 } else if (op == BC_IFORL) {
4313 | suble PC, RC, #0x20000
4314 }
4315 if (vk) {
4316 | strd CARG12, FOR_IDX
4317 }
4318 |2:
4319 | ins_next1
4320 | ins_next2
4321 | strd CARG12, FOR_EXT
4322 if (op == BC_JFORI || op == BC_JFORL) {
4323 | ble =>BC_JLOOP
4324 }
4325 |3:
4326 | ins_next3
4327 |
4328 |4: // Invert check for negative step.
4329 if (!vk) {
4330 | cmp CARG3, CARG1
4331 } else {
4332 | cmp CARG4, CARG1
4333 }
4334 | b <1
4335 |
4336 |5: // FP loop.
4337 if (!vk) {
4338 | cmnlo CARG4, #-LJ_TISNUM
4339 | cmnlo RB, #-LJ_TISNUM
4340 | bhs ->vmeta_for
4341 |.if FPU
4342 | vldr d0, FOR_IDX
4343 | vldr d1, FOR_STOP
4344 | cmp RB, #0
4345 | vstr d0, FOR_EXT
4346 |.else
4347 | cmp RB, #0
4348 | strd CARG12, FOR_EXT
4349 | blt >8
4350 |.endif
4351 } else {
4352 |.if FPU
4353 | vldr d0, FOR_IDX
4354 | vldr d2, FOR_STEP
4355 | vldr d1, FOR_STOP
4356 | cmp CARG4, #0
4357 | vadd.f64 d0, d0, d2
4358 |.else
4359 | cmp CARG4, #0
4360 | blt >8
4361 | bl extern __aeabi_dadd
4362 | strd CARG12, FOR_IDX
4363 | ldrd CARG34, FOR_STOP
4364 | strd CARG12, FOR_EXT
4365 |.endif
4366 }
4367 |6:
4368 |.if FPU
4369 | vcmpge.f64 d0, d1
4370 | vcmplt.f64 d1, d0
4371 | vmrs
4372 |.else
4373 | bl extern __aeabi_cdcmple
4374 |.endif
4375 if (vk) {
4376 |.if FPU
4377 | vstr d0, FOR_IDX
4378 | vstr d0, FOR_EXT
4379 |.endif
4380 }
4381 if (op == BC_FORI) {
4382 | subhi PC, RC, #0x20000
4383 } else if (op == BC_JFORI) {
4384 | sub PC, RC, #0x20000
4385 | ldrhls RC, [PC, #-2]
4386 | bls =>BC_JLOOP
4387 } else if (op == BC_IFORL) {
4388 | subls PC, RC, #0x20000
4389 } else {
4390 | bls =>BC_JLOOP
4391 }
4392 | ins_next1
4393 | ins_next2
4394 | b <3
4395 |
4396 |.if not FPU
4397 |8: // Invert check for negative step.
4398 if (vk) {
4399 | bl extern __aeabi_dadd
4400 | strd CARG12, FOR_IDX
4401 | strd CARG12, FOR_EXT
4402 }
4403 | mov CARG3, CARG1
4404 | mov CARG4, CARG2
4405 | ldrd CARG12, FOR_STOP
4406 | b <6
4407 |.endif
4408 break;
4409
4410 case BC_ITERL:
4411 |.if JIT
4412 | hotloop
4413 |.endif
4414 | // Fall through. Assumes BC_IITERL follows.
4415 break;
4416
4417 case BC_JITERL:
4418 #if !LJ_HASJIT
4419 break;
4420 #endif
4421 case BC_IITERL:
4422 | // RA = base*8, RC = target
4423 | ldrd CARG12, [RA, BASE]!
4424 if (op == BC_JITERL) {
4425 | cmn CARG2, #-LJ_TNIL // Stop if iterator returned nil.
4426 | strdne CARG12, [RA, #-8]
4427 | bne =>BC_JLOOP
4428 } else {
4429 | add RC, PC, RC, lsl #2
4430 | // STALL: load CARG12.
4431 | cmn CARG2, #-LJ_TNIL // Stop if iterator returned nil.
4432 | subne PC, RC, #0x20000 // Otherwise save control var + branch.
4433 | strdne CARG12, [RA, #-8]
4434 }
4435 | ins_next
4436 break;
4437
4438 case BC_LOOP:
4439 | // RA = base*8, RC = target (loop extent)
4440 | // Note: RA/RC is only used by trace recorder to determine scope/extent
4441 | // This opcode does NOT jump, it's only purpose is to detect a hot loop.
4442 |.if JIT
4443 | hotloop
4444 |.endif
4445 | // Fall through. Assumes BC_ILOOP follows.
4446 break;
4447
4448 case BC_ILOOP:
4449 | // RA = base*8, RC = target (loop extent)
4450 | ins_next
4451 break;
4452
4453 case BC_JLOOP:
4454 |.if JIT
4455 | // RA = base (ignored), RC = traceno
4456 | ldr CARG1, [DISPATCH, #DISPATCH_J(trace)]
4457 | mov CARG2, #0 // Traces on ARM don't store the trace number, so use 0.
4458 | ldr TRACE:RC, [CARG1, RC, lsl #2]
4459 | st_vmstate CARG2
4460 | ldr RA, TRACE:RC->mcode
4461 | str BASE, [DISPATCH, #DISPATCH_GL(jit_base)]
4462 | str L, [DISPATCH, #DISPATCH_GL(tmpbuf.L)]
4463 | bx RA
4464 |.endif
4465 break;
4466
4467 case BC_JMP:
4468 | // RA = base*8 (only used by trace recorder), RC = target
4469 | add RC, PC, RC, lsl #2
4470 | sub PC, RC, #0x20000
4471 | ins_next
4472 break;
4473
4474 /* -- Function headers -------------------------------------------------- */
4475
4476 case BC_FUNCF:
4477 |.if JIT
4478 | hotcall
4479 |.endif
4480 case BC_FUNCV: /* NYI: compiled vararg functions. */
4481 | // Fall through. Assumes BC_IFUNCF/BC_IFUNCV follow.
4482 break;
4483
4484 case BC_JFUNCF:
4485 #if !LJ_HASJIT
4486 break;
4487 #endif
4488 case BC_IFUNCF:
4489 | // BASE = new base, RA = BASE+framesize*8, CARG3 = LFUNC, RC = nargs*8
4490 | ldr CARG1, L->maxstack
4491 | ldrb CARG2, [PC, #-4+PC2PROTO(numparams)]
4492 | ldr KBASE, [PC, #-4+PC2PROTO(k)]
4493 | cmp RA, CARG1
4494 | bhi ->vm_growstack_l
4495 if (op != BC_JFUNCF) {
4496 | ins_next1
4497 | ins_next2
4498 }
4499 |2:
4500 | cmp NARGS8:RC, CARG2, lsl #3 // Check for missing parameters.
4501 | mvn CARG4, #~LJ_TNIL
4502 | blo >3
4503 if (op == BC_JFUNCF) {
4504 | decode_RD RC, INS
4505 | b =>BC_JLOOP
4506 } else {
4507 | ins_next3
4508 }
4509 |
4510 |3: // Clear missing parameters.
4511 | strd CARG34, [BASE, NARGS8:RC]
4512 | add NARGS8:RC, NARGS8:RC, #8
4513 | b <2
4514 break;
4515
4516 case BC_JFUNCV:
4517 #if !LJ_HASJIT
4518 break;
4519 #endif
4520 | NYI // NYI: compiled vararg functions
4521 break; /* NYI: compiled vararg functions. */
4522
4523 case BC_IFUNCV:
4524 | // BASE = new base, RA = BASE+framesize*8, CARG3 = LFUNC, RC = nargs*8
4525 | ldr CARG1, L->maxstack
4526 | add CARG4, BASE, RC
4527 | add RA, RA, RC
4528 | str LFUNC:CARG3, [CARG4] // Store copy of LFUNC.
4529 | add CARG2, RC, #8+FRAME_VARG
4530 | ldr KBASE, [PC, #-4+PC2PROTO(k)]
4531 | cmp RA, CARG1
4532 | str CARG2, [CARG4, #4] // Store delta + FRAME_VARG.
4533 | bhs ->vm_growstack_l
4534 | ldrb RB, [PC, #-4+PC2PROTO(numparams)]
4535 | mov RA, BASE
4536 | mov RC, CARG4
4537 | cmp RB, #0
4538 | add BASE, CARG4, #8
4539 | beq >3
4540 | mvn CARG3, #~LJ_TNIL
4541 |1:
4542 | cmp RA, RC // Less args than parameters?
4543 | ldrdlo CARG12, [RA], #8
4544 | movhs CARG2, CARG3
4545 | strlo CARG3, [RA, #-4] // Clear old fixarg slot (help the GC).
4546 |2:
4547 | subs RB, RB, #1
4548 | strd CARG12, [CARG4, #8]!
4549 | bne <1
4550 |3:
4551 | ins_next
4552 break;
4553
4554 case BC_FUNCC:
4555 case BC_FUNCCW:
4556 | // BASE = new base, RA = BASE+framesize*8, CARG3 = CFUNC, RC = nargs*8
4557 if (op == BC_FUNCC) {
4558 | ldr CARG4, CFUNC:CARG3->f
4559 } else {
4560 | ldr CARG4, [DISPATCH, #DISPATCH_GL(wrapf)]
4561 }
4562 | add CARG2, RA, NARGS8:RC
4563 | ldr CARG1, L->maxstack
4564 | add RC, BASE, NARGS8:RC
4565 | str BASE, L->base
4566 | cmp CARG2, CARG1
4567 | str RC, L->top
4568 if (op == BC_FUNCCW) {
4569 | ldr CARG2, CFUNC:CARG3->f
4570 }
4571 | mv_vmstate CARG3, C
4572 | mov CARG1, L
4573 | bhi ->vm_growstack_c // Need to grow stack.
4574 | st_vmstate CARG3
4575 | blx CARG4 // (lua_State *L [, lua_CFunction f])
4576 | // Returns nresults.
4577 | ldr BASE, L->base
4578 | mv_vmstate CARG3, INTERP
4579 | ldr CRET2, L->top
4580 | str L, [DISPATCH, #DISPATCH_GL(cur_L)]
4581 | lsl RC, CRET1, #3
4582 | st_vmstate CARG3
4583 | ldr PC, [BASE, FRAME_PC]
4584 | sub RA, CRET2, RC // RA = L->top - nresults*8
4585 | b ->vm_returnc
4586 break;
4587
4588 /* ---------------------------------------------------------------------- */
4589
4590 default:
4591 fprintf(stderr, "Error: undefined opcode BC_%s\n", bc_names[op]);
4592 exit(2);
4593 break;
4594 }
4595 }
4596
4597 static int build_backend(BuildCtx *ctx)
4598 {
4599 int op;
4600
4601 dasm_growpc(Dst, BC__MAX);
4602
4603 build_subroutines(ctx);
4604
4605 |.code_op
4606 for (op = 0; op < BC__MAX; op++)
4607 build_ins(ctx, (BCOp)op, op);
4608
4609 return BC__MAX;
4610 }
4611
4612 /* Emit pseudo frame-info for all assembler functions. */
4613 static void emit_asm_debug(BuildCtx *ctx)
4614 {
4615 int fcofs = (int)((uint8_t *)ctx->glob[GLOB_vm_ffi_call] - ctx->code);
4616 int i;
4617 switch (ctx->mode) {
4618 case BUILD_elfasm:
4619 fprintf(ctx->fp, "\t.section .debug_frame,\"\",%%progbits\n");
4620 fprintf(ctx->fp,
4621 ".Lframe0:\n"
4622 "\t.long .LECIE0-.LSCIE0\n"
4623 ".LSCIE0:\n"
4624 "\t.long 0xffffffff\n"
4625 "\t.byte 0x1\n"
4626 "\t.string \"\"\n"
4627 "\t.uleb128 0x1\n"
4628 "\t.sleb128 -4\n"
4629 "\t.byte 0xe\n" /* Return address is in lr. */
4630 "\t.byte 0xc\n\t.uleb128 0xd\n\t.uleb128 0\n" /* def_cfa sp */
4631 "\t.align 2\n"
4632 ".LECIE0:\n\n");
4633 fprintf(ctx->fp,
4634 ".LSFDE0:\n"
4635 "\t.long .LEFDE0-.LASFDE0\n"
4636 ".LASFDE0:\n"
4637 "\t.long .Lframe0\n"
4638 "\t.long .Lbegin\n"
4639 "\t.long %d\n"
4640 "\t.byte 0xe\n\t.uleb128 %d\n" /* def_cfa_offset */
4641 "\t.byte 0x8e\n\t.uleb128 1\n", /* offset lr */
4642 fcofs, CFRAME_SIZE);
4643 for (i = 11; i >= (LJ_ARCH_HASFPU ? 5 : 4); i--) /* offset r4-r11 */
4644 fprintf(ctx->fp, "\t.byte %d\n\t.uleb128 %d\n", 0x80+i, 2+(11-i));
4645 #if LJ_ARCH_HASFPU
4646 for (i = 15; i >= 8; i--) /* offset d8-d15 */
4647 fprintf(ctx->fp, "\t.byte 5\n\t.uleb128 %d, %d\n",
4648 64+2*i, 10+2*(15-i));
4649 fprintf(ctx->fp, "\t.byte 0x84\n\t.uleb128 %d\n", 25); /* offset r4 */
4650 #endif
4651 fprintf(ctx->fp,
4652 "\t.align 2\n"
4653 ".LEFDE0:\n\n");
4654 #if LJ_HASFFI
4655 fprintf(ctx->fp,
4656 ".LSFDE1:\n"
4657 "\t.long .LEFDE1-.LASFDE1\n"
4658 ".LASFDE1:\n"
4659 "\t.long .Lframe0\n"
4660 "\t.long lj_vm_ffi_call\n"
4661 "\t.long %d\n"
4662 "\t.byte 0xe\n\t.uleb128 16\n" /* def_cfa_offset */
4663 "\t.byte 0x8e\n\t.uleb128 1\n" /* offset lr */
4664 "\t.byte 0x8b\n\t.uleb128 2\n" /* offset r11 */
4665 "\t.byte 0x85\n\t.uleb128 3\n" /* offset r5 */
4666 "\t.byte 0x84\n\t.uleb128 4\n" /* offset r4 */
4667 "\t.byte 0xd\n\t.uleb128 0xb\n" /* def_cfa_register r11 */
4668 "\t.align 2\n"
4669 ".LEFDE1:\n\n", (int)ctx->codesz - fcofs);
4670 #endif
4671 break;
4672 default:
4673 break;
4674 }
4675 }
4676