1|// Low-level VM code for MIPS CPUs. 2|// Bytecode interpreter, fast functions and helper functions. 3|// Copyright (C) 2005-2017 Mike Pall. See Copyright Notice in luajit.h 4| 5|.arch mips 6|.section code_op, code_sub 7| 8|.actionlist build_actionlist 9|.globals GLOB_ 10|.globalnames globnames 11|.externnames extnames 12| 13|// Note: The ragged indentation of the instructions is intentional. 14|// The starting columns indicate data dependencies. 15| 16|//----------------------------------------------------------------------- 17| 18|// Fixed register assignments for the interpreter. 19|// Don't use: r0 = 0, r26/r27 = reserved, r28 = gp, r29 = sp, r31 = ra 20| 21|// The following must be C callee-save (but BASE is often refetched). 22|.define BASE, r16 // Base of current Lua stack frame. 23|.define KBASE, r17 // Constants of current Lua function. 24|.define PC, r18 // Next PC. 25|.define DISPATCH, r19 // Opcode dispatch table. 26|.define LREG, r20 // Register holding lua_State (also in SAVE_L). 27|.define MULTRES, r21 // Size of multi-result: (nresults+1)*8. 28|// NYI: r22 currently unused. 29| 30|.define JGL, r30 // On-trace: global_State + 32768. 31| 32|// Constants for type-comparisons, stores and conversions. C callee-save. 33|.define TISNIL, r30 34|.define TOBIT, f30 // 2^52 + 2^51. 35| 36|// The following temporaries are not saved across C calls, except for RA. 37|.define RA, r23 // Callee-save. 38|.define RB, r8 39|.define RC, r9 40|.define RD, r10 41|.define INS, r11 42| 43|.define AT, r1 // Assembler temporary. 44|.define TMP0, r12 45|.define TMP1, r13 46|.define TMP2, r14 47|.define TMP3, r15 48| 49|// Calling conventions. 50|.define CFUNCADDR, r25 51|.define CARG1, r4 52|.define CARG2, r5 53|.define CARG3, r6 54|.define CARG4, r7 55| 56|.define CRET1, r2 57|.define CRET2, r3 58| 59|.define FARG1, f12 60|.define FARG2, f14 61| 62|.define FRET1, f0 63|.define FRET2, f2 64| 65|// Stack layout while in interpreter. Must match with lj_frame.h. 66|.define CFRAME_SPACE, 112 // Delta for sp. 67| 68|.define SAVE_ERRF, 124(sp) // 32 bit C frame info. 69|.define SAVE_NRES, 120(sp) 70|.define SAVE_CFRAME, 116(sp) 71|.define SAVE_L, 112(sp) 72|//----- 8 byte aligned, ^^^^ 16 byte register save area, owned by interpreter. 73|.define SAVE_GPR_, 72 // .. 72+10*4: 32 bit GPR saves. 74|.define SAVE_FPR_, 24 // .. 24+6*8: 64 bit FPR saves. 75|.define SAVE_PC, 20(sp) 76|.define ARG5, 16(sp) 77|.define CSAVE_4, 12(sp) 78|.define CSAVE_3, 8(sp) 79|.define CSAVE_2, 4(sp) 80|.define CSAVE_1, 0(sp) 81|//----- 8 byte aligned, ^^^^ 16 byte register save area, owned by callee. 82| 83|.define ARG5_OFS, 16 84|.define SAVE_MULTRES, ARG5 85| 86|.macro saveregs 87| addiu sp, sp, -CFRAME_SPACE 88| sw ra, SAVE_GPR_+9*4(sp) 89| sw r30, SAVE_GPR_+8*4(sp) 90| sdc1 f30, SAVE_FPR_+5*8(sp) 91| sw r23, SAVE_GPR_+7*4(sp) 92| sw r22, SAVE_GPR_+6*4(sp) 93| sdc1 f28, SAVE_FPR_+4*8(sp) 94| sw r21, SAVE_GPR_+5*4(sp) 95| sw r20, SAVE_GPR_+4*4(sp) 96| sdc1 f26, SAVE_FPR_+3*8(sp) 97| sw r19, SAVE_GPR_+3*4(sp) 98| sw r18, SAVE_GPR_+2*4(sp) 99| sdc1 f24, SAVE_FPR_+2*8(sp) 100| sw r17, SAVE_GPR_+1*4(sp) 101| sw r16, SAVE_GPR_+0*4(sp) 102| sdc1 f22, SAVE_FPR_+1*8(sp) 103| sdc1 f20, SAVE_FPR_+0*8(sp) 104|.endmacro 105| 106|.macro restoreregs_ret 107| lw ra, SAVE_GPR_+9*4(sp) 108| lw r30, SAVE_GPR_+8*4(sp) 109| ldc1 f30, SAVE_FPR_+5*8(sp) 110| lw r23, SAVE_GPR_+7*4(sp) 111| lw r22, SAVE_GPR_+6*4(sp) 112| ldc1 f28, SAVE_FPR_+4*8(sp) 113| lw r21, SAVE_GPR_+5*4(sp) 114| lw r20, SAVE_GPR_+4*4(sp) 115| ldc1 f26, SAVE_FPR_+3*8(sp) 116| lw r19, SAVE_GPR_+3*4(sp) 117| lw r18, SAVE_GPR_+2*4(sp) 118| ldc1 f24, SAVE_FPR_+2*8(sp) 119| lw r17, SAVE_GPR_+1*4(sp) 120| lw r16, SAVE_GPR_+0*4(sp) 121| ldc1 f22, SAVE_FPR_+1*8(sp) 122| ldc1 f20, SAVE_FPR_+0*8(sp) 123| jr ra 124| addiu sp, sp, CFRAME_SPACE 125|.endmacro 126| 127|// Type definitions. Some of these are only used for documentation. 128|.type L, lua_State, LREG 129|.type GL, global_State 130|.type TVALUE, TValue 131|.type GCOBJ, GCobj 132|.type STR, GCstr 133|.type TAB, GCtab 134|.type LFUNC, GCfuncL 135|.type CFUNC, GCfuncC 136|.type PROTO, GCproto 137|.type UPVAL, GCupval 138|.type NODE, Node 139|.type NARGS8, int 140|.type TRACE, GCtrace 141| 142|//----------------------------------------------------------------------- 143| 144|// Trap for not-yet-implemented parts. 145|.macro NYI; .long 0xf0f0f0f0; .endmacro 146| 147|// Macros to mark delay slots. 148|.macro ., a; a; .endmacro 149|.macro ., a,b; a,b; .endmacro 150|.macro ., a,b,c; a,b,c; .endmacro 151| 152|//----------------------------------------------------------------------- 153| 154|// Endian-specific defines. 155|.define FRAME_PC, LJ_ENDIAN_SELECT(-4,-8) 156|.define FRAME_FUNC, LJ_ENDIAN_SELECT(-8,-4) 157|.define HI, LJ_ENDIAN_SELECT(4,0) 158|.define LO, LJ_ENDIAN_SELECT(0,4) 159|.define OFS_RD, LJ_ENDIAN_SELECT(2,0) 160|.define OFS_RA, LJ_ENDIAN_SELECT(1,2) 161|.define OFS_OP, LJ_ENDIAN_SELECT(0,3) 162| 163|// Instruction decode. 164|.macro decode_OP1, dst, ins; andi dst, ins, 0xff; .endmacro 165|.macro decode_OP4a, dst, ins; andi dst, ins, 0xff; .endmacro 166|.macro decode_OP4b, dst; sll dst, dst, 2; .endmacro 167|.macro decode_RC4a, dst, ins; srl dst, ins, 14; .endmacro 168|.macro decode_RC4b, dst; andi dst, dst, 0x3fc; .endmacro 169|.macro decode_RD4b, dst; sll dst, dst, 2; .endmacro 170|.macro decode_RA8a, dst, ins; srl dst, ins, 5; .endmacro 171|.macro decode_RA8b, dst; andi dst, dst, 0x7f8; .endmacro 172|.macro decode_RB8a, dst, ins; srl dst, ins, 21; .endmacro 173|.macro decode_RB8b, dst; andi dst, dst, 0x7f8; .endmacro 174|.macro decode_RD8a, dst, ins; srl dst, ins, 16; .endmacro 175|.macro decode_RD8b, dst; sll dst, dst, 3; .endmacro 176|.macro decode_RDtoRC8, dst, src; andi dst, src, 0x7f8; .endmacro 177| 178|// Instruction fetch. 179|.macro ins_NEXT1 180| lw INS, 0(PC) 181| addiu PC, PC, 4 182|.endmacro 183|// Instruction decode+dispatch. 184|.macro ins_NEXT2 185| decode_OP4a TMP1, INS 186| decode_OP4b TMP1 187| addu TMP0, DISPATCH, TMP1 188| decode_RD8a RD, INS 189| lw AT, 0(TMP0) 190| decode_RA8a RA, INS 191| decode_RD8b RD 192| jr AT 193| decode_RA8b RA 194|.endmacro 195|.macro ins_NEXT 196| ins_NEXT1 197| ins_NEXT2 198|.endmacro 199| 200|// Instruction footer. 201|.if 1 202| // Replicated dispatch. Less unpredictable branches, but higher I-Cache use. 203| .define ins_next, ins_NEXT 204| .define ins_next_, ins_NEXT 205| .define ins_next1, ins_NEXT1 206| .define ins_next2, ins_NEXT2 207|.else 208| // Common dispatch. Lower I-Cache use, only one (very) unpredictable branch. 209| // Affects only certain kinds of benchmarks (and only with -j off). 210| .macro ins_next 211| b ->ins_next 212| .endmacro 213| .macro ins_next1 214| .endmacro 215| .macro ins_next2 216| b ->ins_next 217| .endmacro 218| .macro ins_next_ 219| ->ins_next: 220| ins_NEXT 221| .endmacro 222|.endif 223| 224|// Call decode and dispatch. 225|.macro ins_callt 226| // BASE = new base, RB = LFUNC/CFUNC, RC = nargs*8, FRAME_PC(BASE) = PC 227| lw PC, LFUNC:RB->pc 228| lw INS, 0(PC) 229| addiu PC, PC, 4 230| decode_OP4a TMP1, INS 231| decode_RA8a RA, INS 232| decode_OP4b TMP1 233| decode_RA8b RA 234| addu TMP0, DISPATCH, TMP1 235| lw TMP0, 0(TMP0) 236| jr TMP0 237| addu RA, RA, BASE 238|.endmacro 239| 240|.macro ins_call 241| // BASE = new base, RB = LFUNC/CFUNC, RC = nargs*8, PC = caller PC 242| sw PC, FRAME_PC(BASE) 243| ins_callt 244|.endmacro 245| 246|//----------------------------------------------------------------------- 247| 248|.macro branch_RD 249| srl TMP0, RD, 1 250| lui AT, (-(BCBIAS_J*4 >> 16) & 65535) 251| addu TMP0, TMP0, AT 252| addu PC, PC, TMP0 253|.endmacro 254| 255|// Assumes DISPATCH is relative to GL. 256#define DISPATCH_GL(field) (GG_DISP2G + (int)offsetof(global_State, field)) 257#define DISPATCH_J(field) (GG_DISP2J + (int)offsetof(jit_State, field)) 258#define GG_DISP2GOT (GG_OFS(got) - GG_OFS(dispatch)) 259#define DISPATCH_GOT(name) (GG_DISP2GOT + 4*LJ_GOT_##name) 260| 261#define PC2PROTO(field) ((int)offsetof(GCproto, field)-(int)sizeof(GCproto)) 262| 263|.macro load_got, func 264| lw CFUNCADDR, DISPATCH_GOT(func)(DISPATCH) 265|.endmacro 266|// Much faster. Sadly, there's no easy way to force the required code layout. 267|// .macro call_intern, func; bal extern func; .endmacro 268|.macro call_intern, func; jalr CFUNCADDR; .endmacro 269|.macro call_extern; jalr CFUNCADDR; .endmacro 270|.macro jmp_extern; jr CFUNCADDR; .endmacro 271| 272|.macro hotcheck, delta, target 273| srl TMP1, PC, 1 274| andi TMP1, TMP1, 126 275| addu TMP1, TMP1, DISPATCH 276| lhu TMP2, GG_DISP2HOT(TMP1) 277| addiu TMP2, TMP2, -delta 278| bltz TMP2, target 279|. sh TMP2, GG_DISP2HOT(TMP1) 280|.endmacro 281| 282|.macro hotloop 283| hotcheck HOTCOUNT_LOOP, ->vm_hotloop 284|.endmacro 285| 286|.macro hotcall 287| hotcheck HOTCOUNT_CALL, ->vm_hotcall 288|.endmacro 289| 290|// Set current VM state. Uses TMP0. 291|.macro li_vmstate, st; li TMP0, ~LJ_VMST_..st; .endmacro 292|.macro st_vmstate; sw TMP0, DISPATCH_GL(vmstate)(DISPATCH); .endmacro 293| 294|// Move table write barrier back. Overwrites mark and tmp. 295|.macro barrierback, tab, mark, tmp, target 296| lw tmp, DISPATCH_GL(gc.grayagain)(DISPATCH) 297| andi mark, mark, ~LJ_GC_BLACK & 255 // black2gray(tab) 298| sw tab, DISPATCH_GL(gc.grayagain)(DISPATCH) 299| sb mark, tab->marked 300| b target 301|. sw tmp, tab->gclist 302|.endmacro 303| 304|//----------------------------------------------------------------------- 305 306/* Generate subroutines used by opcodes and other parts of the VM. */ 307/* The .code_sub section should be last to help static branch prediction. */ 308static void build_subroutines(BuildCtx *ctx) 309{ 310 |.code_sub 311 | 312 |//----------------------------------------------------------------------- 313 |//-- Return handling ---------------------------------------------------- 314 |//----------------------------------------------------------------------- 315 | 316 |->vm_returnp: 317 | // See vm_return. Also: TMP2 = previous base. 318 | andi AT, PC, FRAME_P 319 | beqz AT, ->cont_dispatch 320 |. li TMP1, LJ_TTRUE 321 | 322 | // Return from pcall or xpcall fast func. 323 | lw PC, FRAME_PC(TMP2) // Fetch PC of previous frame. 324 | move BASE, TMP2 // Restore caller base. 325 | // Prepending may overwrite the pcall frame, so do it at the end. 326 | sw TMP1, FRAME_PC(RA) // Prepend true to results. 327 | addiu RA, RA, -8 328 | 329 |->vm_returnc: 330 | addiu RD, RD, 8 // RD = (nresults+1)*8. 331 | andi TMP0, PC, FRAME_TYPE 332 | beqz RD, ->vm_unwind_c_eh 333 |. li CRET1, LUA_YIELD 334 | beqz TMP0, ->BC_RET_Z // Handle regular return to Lua. 335 |. move MULTRES, RD 336 | 337 |->vm_return: 338 | // BASE = base, RA = resultptr, RD/MULTRES = (nresults+1)*8, PC = return 339 | // TMP0 = PC & FRAME_TYPE 340 | li TMP2, -8 341 | xori AT, TMP0, FRAME_C 342 | and TMP2, PC, TMP2 343 | bnez AT, ->vm_returnp 344 | subu TMP2, BASE, TMP2 // TMP2 = previous base. 345 | 346 | addiu TMP1, RD, -8 347 | sw TMP2, L->base 348 | li_vmstate C 349 | lw TMP2, SAVE_NRES 350 | addiu BASE, BASE, -8 351 | st_vmstate 352 | beqz TMP1, >2 353 |. sll TMP2, TMP2, 3 354 |1: 355 | addiu TMP1, TMP1, -8 356 | ldc1 f0, 0(RA) 357 | addiu RA, RA, 8 358 | sdc1 f0, 0(BASE) 359 | bnez TMP1, <1 360 |. addiu BASE, BASE, 8 361 | 362 |2: 363 | bne TMP2, RD, >6 364 |3: 365 |. sw BASE, L->top // Store new top. 366 | 367 |->vm_leave_cp: 368 | lw TMP0, SAVE_CFRAME // Restore previous C frame. 369 | move CRET1, r0 // Ok return status for vm_pcall. 370 | sw TMP0, L->cframe 371 | 372 |->vm_leave_unw: 373 | restoreregs_ret 374 | 375 |6: 376 | lw TMP1, L->maxstack 377 | slt AT, TMP2, RD 378 | bnez AT, >7 // Less results wanted? 379 | // More results wanted. Check stack size and fill up results with nil. 380 |. slt AT, BASE, TMP1 381 | beqz AT, >8 382 |. nop 383 | sw TISNIL, HI(BASE) 384 | addiu RD, RD, 8 385 | b <2 386 |. addiu BASE, BASE, 8 387 | 388 |7: // Less results wanted. 389 | subu TMP0, RD, TMP2 390 | subu TMP0, BASE, TMP0 // Either keep top or shrink it. 391 | b <3 392 |. movn BASE, TMP0, TMP2 // LUA_MULTRET+1 case? 393 | 394 |8: // Corner case: need to grow stack for filling up results. 395 | // This can happen if: 396 | // - A C function grows the stack (a lot). 397 | // - The GC shrinks the stack in between. 398 | // - A return back from a lua_call() with (high) nresults adjustment. 399 | load_got lj_state_growstack 400 | move MULTRES, RD 401 | srl CARG2, TMP2, 3 402 | call_intern lj_state_growstack // (lua_State *L, int n) 403 |. move CARG1, L 404 | lw TMP2, SAVE_NRES 405 | lw BASE, L->top // Need the (realloced) L->top in BASE. 406 | move RD, MULTRES 407 | b <2 408 |. sll TMP2, TMP2, 3 409 | 410 |->vm_unwind_c: // Unwind C stack, return from vm_pcall. 411 | // (void *cframe, int errcode) 412 | move sp, CARG1 413 | move CRET1, CARG2 414 |->vm_unwind_c_eh: // Landing pad for external unwinder. 415 | lw L, SAVE_L 416 | li TMP0, ~LJ_VMST_C 417 | lw GL:TMP1, L->glref 418 | b ->vm_leave_unw 419 |. sw TMP0, GL:TMP1->vmstate 420 | 421 |->vm_unwind_ff: // Unwind C stack, return from ff pcall. 422 | // (void *cframe) 423 | li AT, -4 424 | and sp, CARG1, AT 425 |->vm_unwind_ff_eh: // Landing pad for external unwinder. 426 | lw L, SAVE_L 427 | lui TMP3, 0x59c0 // TOBIT = 2^52 + 2^51 (float). 428 | li TISNIL, LJ_TNIL 429 | lw BASE, L->base 430 | lw DISPATCH, L->glref // Setup pointer to dispatch table. 431 | mtc1 TMP3, TOBIT 432 | li TMP1, LJ_TFALSE 433 | li_vmstate INTERP 434 | lw PC, FRAME_PC(BASE) // Fetch PC of previous frame. 435 | cvt.d.s TOBIT, TOBIT 436 | addiu RA, BASE, -8 // Results start at BASE-8. 437 | addiu DISPATCH, DISPATCH, GG_G2DISP 438 | sw TMP1, HI(RA) // Prepend false to error message. 439 | st_vmstate 440 | b ->vm_returnc 441 |. li RD, 16 // 2 results: false + error message. 442 | 443 |//----------------------------------------------------------------------- 444 |//-- Grow stack for calls ----------------------------------------------- 445 |//----------------------------------------------------------------------- 446 | 447 |->vm_growstack_c: // Grow stack for C function. 448 | b >2 449 |. li CARG2, LUA_MINSTACK 450 | 451 |->vm_growstack_l: // Grow stack for Lua function. 452 | // BASE = new base, RA = BASE+framesize*8, RC = nargs*8, PC = first PC 453 | addu RC, BASE, RC 454 | subu RA, RA, BASE 455 | sw BASE, L->base 456 | addiu PC, PC, 4 // Must point after first instruction. 457 | sw RC, L->top 458 | srl CARG2, RA, 3 459 |2: 460 | // L->base = new base, L->top = top 461 | load_got lj_state_growstack 462 | sw PC, SAVE_PC 463 | call_intern lj_state_growstack // (lua_State *L, int n) 464 |. move CARG1, L 465 | lw BASE, L->base 466 | lw RC, L->top 467 | lw LFUNC:RB, FRAME_FUNC(BASE) 468 | subu RC, RC, BASE 469 | // BASE = new base, RB = LFUNC/CFUNC, RC = nargs*8, FRAME_PC(BASE) = PC 470 | ins_callt // Just retry the call. 471 | 472 |//----------------------------------------------------------------------- 473 |//-- Entry points into the assembler VM --------------------------------- 474 |//----------------------------------------------------------------------- 475 | 476 |->vm_resume: // Setup C frame and resume thread. 477 | // (lua_State *L, TValue *base, int nres1 = 0, ptrdiff_t ef = 0) 478 | saveregs 479 | move L, CARG1 480 | lw DISPATCH, L->glref // Setup pointer to dispatch table. 481 | move BASE, CARG2 482 | lbu TMP1, L->status 483 | sw L, SAVE_L 484 | li PC, FRAME_CP 485 | addiu TMP0, sp, CFRAME_RESUME 486 | addiu DISPATCH, DISPATCH, GG_G2DISP 487 | sw r0, SAVE_NRES 488 | sw r0, SAVE_ERRF 489 | sw TMP0, L->cframe 490 | sw r0, SAVE_CFRAME 491 | beqz TMP1, >3 492 |. sw CARG1, SAVE_PC // Any value outside of bytecode is ok. 493 | 494 | // Resume after yield (like a return). 495 | move RA, BASE 496 | lw BASE, L->base 497 | lw TMP1, L->top 498 | lw PC, FRAME_PC(BASE) 499 | lui TMP3, 0x59c0 // TOBIT = 2^52 + 2^51 (float). 500 | subu RD, TMP1, BASE 501 | mtc1 TMP3, TOBIT 502 | sb r0, L->status 503 | cvt.d.s TOBIT, TOBIT 504 | li_vmstate INTERP 505 | addiu RD, RD, 8 506 | st_vmstate 507 | move MULTRES, RD 508 | andi TMP0, PC, FRAME_TYPE 509 | beqz TMP0, ->BC_RET_Z 510 |. li TISNIL, LJ_TNIL 511 | b ->vm_return 512 |. nop 513 | 514 |->vm_pcall: // Setup protected C frame and enter VM. 515 | // (lua_State *L, TValue *base, int nres1, ptrdiff_t ef) 516 | saveregs 517 | sw CARG4, SAVE_ERRF 518 | b >1 519 |. li PC, FRAME_CP 520 | 521 |->vm_call: // Setup C frame and enter VM. 522 | // (lua_State *L, TValue *base, int nres1) 523 | saveregs 524 | li PC, FRAME_C 525 | 526 |1: // Entry point for vm_pcall above (PC = ftype). 527 | lw TMP1, L:CARG1->cframe 528 | sw CARG3, SAVE_NRES 529 | move L, CARG1 530 | sw CARG1, SAVE_L 531 | move BASE, CARG2 532 | sw sp, L->cframe // Add our C frame to cframe chain. 533 | lw DISPATCH, L->glref // Setup pointer to dispatch table. 534 | sw CARG1, SAVE_PC // Any value outside of bytecode is ok. 535 | sw TMP1, SAVE_CFRAME 536 | addiu DISPATCH, DISPATCH, GG_G2DISP 537 | 538 |3: // Entry point for vm_cpcall/vm_resume (BASE = base, PC = ftype). 539 | lw TMP2, L->base // TMP2 = old base (used in vmeta_call). 540 | lui TMP3, 0x59c0 // TOBIT = 2^52 + 2^51 (float). 541 | lw TMP1, L->top 542 | mtc1 TMP3, TOBIT 543 | addu PC, PC, BASE 544 | subu NARGS8:RC, TMP1, BASE 545 | subu PC, PC, TMP2 // PC = frame delta + frame type 546 | cvt.d.s TOBIT, TOBIT 547 | li_vmstate INTERP 548 | li TISNIL, LJ_TNIL 549 | st_vmstate 550 | 551 |->vm_call_dispatch: 552 | // TMP2 = old base, BASE = new base, RC = nargs*8, PC = caller PC 553 | lw TMP0, FRAME_PC(BASE) 554 | li AT, LJ_TFUNC 555 | bne TMP0, AT, ->vmeta_call 556 |. lw LFUNC:RB, FRAME_FUNC(BASE) 557 | 558 |->vm_call_dispatch_f: 559 | ins_call 560 | // BASE = new base, RB = func, RC = nargs*8, PC = caller PC 561 | 562 |->vm_cpcall: // Setup protected C frame, call C. 563 | // (lua_State *L, lua_CFunction func, void *ud, lua_CPFunction cp) 564 | saveregs 565 | move L, CARG1 566 | lw TMP0, L:CARG1->stack 567 | sw CARG1, SAVE_L 568 | lw TMP1, L->top 569 | sw CARG1, SAVE_PC // Any value outside of bytecode is ok. 570 | subu TMP0, TMP0, TMP1 // Compute -savestack(L, L->top). 571 | lw TMP1, L->cframe 572 | sw sp, L->cframe // Add our C frame to cframe chain. 573 | sw TMP0, SAVE_NRES // Neg. delta means cframe w/o frame. 574 | sw r0, SAVE_ERRF // No error function. 575 | move CFUNCADDR, CARG4 576 | jalr CARG4 // (lua_State *L, lua_CFunction func, void *ud) 577 |. sw TMP1, SAVE_CFRAME 578 | move BASE, CRET1 579 | lw DISPATCH, L->glref // Setup pointer to dispatch table. 580 | li PC, FRAME_CP 581 | bnez CRET1, <3 // Else continue with the call. 582 |. addiu DISPATCH, DISPATCH, GG_G2DISP 583 | b ->vm_leave_cp // No base? Just remove C frame. 584 |. nop 585 | 586 |//----------------------------------------------------------------------- 587 |//-- Metamethod handling ------------------------------------------------ 588 |//----------------------------------------------------------------------- 589 | 590 |// The lj_meta_* functions (except for lj_meta_cat) don't reallocate the 591 |// stack, so BASE doesn't need to be reloaded across these calls. 592 | 593 |//-- Continuation dispatch ---------------------------------------------- 594 | 595 |->cont_dispatch: 596 | // BASE = meta base, RA = resultptr, RD = (nresults+1)*8 597 | lw TMP0, -16+LO(BASE) // Continuation. 598 | move RB, BASE 599 | move BASE, TMP2 // Restore caller BASE. 600 | lw LFUNC:TMP1, FRAME_FUNC(TMP2) 601 |.if FFI 602 | sltiu AT, TMP0, 2 603 |.endif 604 | lw PC, -16+HI(RB) // Restore PC from [cont|PC]. 605 | addu TMP2, RA, RD 606 | lw TMP1, LFUNC:TMP1->pc 607 |.if FFI 608 | bnez AT, >1 609 |.endif 610 |. sw TISNIL, -8+HI(TMP2) // Ensure one valid arg. 611 | // BASE = base, RA = resultptr, RB = meta base 612 | jr TMP0 // Jump to continuation. 613 |. lw KBASE, PC2PROTO(k)(TMP1) 614 | 615 |.if FFI 616 |1: 617 | bnez TMP0, ->cont_ffi_callback // cont = 1: return from FFI callback. 618 | // cont = 0: tailcall from C function. 619 |. addiu TMP1, RB, -16 620 | b ->vm_call_tail 621 |. subu RC, TMP1, BASE 622 |.endif 623 | 624 |->cont_cat: // RA = resultptr, RB = meta base 625 | lw INS, -4(PC) 626 | addiu CARG2, RB, -16 627 | ldc1 f0, 0(RA) 628 | decode_RB8a MULTRES, INS 629 | decode_RA8a RA, INS 630 | decode_RB8b MULTRES 631 | decode_RA8b RA 632 | addu TMP1, BASE, MULTRES 633 | sw BASE, L->base 634 | subu CARG3, CARG2, TMP1 635 | bne TMP1, CARG2, ->BC_CAT_Z 636 |. sdc1 f0, 0(CARG2) 637 | addu RA, BASE, RA 638 | b ->cont_nop 639 |. sdc1 f0, 0(RA) 640 | 641 |//-- Table indexing metamethods ----------------------------------------- 642 | 643 |->vmeta_tgets1: 644 | addiu CARG3, DISPATCH, DISPATCH_GL(tmptv) 645 | li TMP0, LJ_TSTR 646 | sw STR:RC, LO(CARG3) 647 | b >1 648 |. sw TMP0, HI(CARG3) 649 | 650 |->vmeta_tgets: 651 | addiu CARG2, DISPATCH, DISPATCH_GL(tmptv) 652 | li TMP0, LJ_TTAB 653 | sw TAB:RB, LO(CARG2) 654 | addiu CARG3, DISPATCH, DISPATCH_GL(tmptv2) 655 | sw TMP0, HI(CARG2) 656 | li TMP1, LJ_TSTR 657 | sw STR:RC, LO(CARG3) 658 | b >1 659 |. sw TMP1, HI(CARG3) 660 | 661 |->vmeta_tgetb: // TMP0 = index 662 | mtc1 TMP0, f0 663 | cvt.d.w f0, f0 664 | addiu CARG3, DISPATCH, DISPATCH_GL(tmptv) 665 | sdc1 f0, 0(CARG3) 666 | 667 |->vmeta_tgetv: 668 |1: 669 | load_got lj_meta_tget 670 | sw BASE, L->base 671 | sw PC, SAVE_PC 672 | call_intern lj_meta_tget // (lua_State *L, TValue *o, TValue *k) 673 |. move CARG1, L 674 | // Returns TValue * (finished) or NULL (metamethod). 675 | beqz CRET1, >3 676 |. addiu TMP1, BASE, -FRAME_CONT 677 | ldc1 f0, 0(CRET1) 678 | ins_next1 679 | sdc1 f0, 0(RA) 680 | ins_next2 681 | 682 |3: // Call __index metamethod. 683 | // BASE = base, L->top = new base, stack = cont/func/t/k 684 | lw BASE, L->top 685 | sw PC, -16+HI(BASE) // [cont|PC] 686 | subu PC, BASE, TMP1 687 | lw LFUNC:RB, FRAME_FUNC(BASE) // Guaranteed to be a function here. 688 | b ->vm_call_dispatch_f 689 |. li NARGS8:RC, 16 // 2 args for func(t, k). 690 | 691 |//----------------------------------------------------------------------- 692 | 693 |->vmeta_tsets1: 694 | addiu CARG3, DISPATCH, DISPATCH_GL(tmptv) 695 | li TMP0, LJ_TSTR 696 | sw STR:RC, LO(CARG3) 697 | b >1 698 |. sw TMP0, HI(CARG3) 699 | 700 |->vmeta_tsets: 701 | addiu CARG2, DISPATCH, DISPATCH_GL(tmptv) 702 | li TMP0, LJ_TTAB 703 | sw TAB:RB, LO(CARG2) 704 | addiu CARG3, DISPATCH, DISPATCH_GL(tmptv2) 705 | sw TMP0, HI(CARG2) 706 | li TMP1, LJ_TSTR 707 | sw STR:RC, LO(CARG3) 708 | b >1 709 |. sw TMP1, HI(CARG3) 710 | 711 |->vmeta_tsetb: // TMP0 = index 712 | mtc1 TMP0, f0 713 | cvt.d.w f0, f0 714 | addiu CARG3, DISPATCH, DISPATCH_GL(tmptv) 715 | sdc1 f0, 0(CARG3) 716 | 717 |->vmeta_tsetv: 718 |1: 719 | load_got lj_meta_tset 720 | sw BASE, L->base 721 | sw PC, SAVE_PC 722 | call_intern lj_meta_tset // (lua_State *L, TValue *o, TValue *k) 723 |. move CARG1, L 724 | // Returns TValue * (finished) or NULL (metamethod). 725 | beqz CRET1, >3 726 |. ldc1 f0, 0(RA) 727 | // NOBARRIER: lj_meta_tset ensures the table is not black. 728 | ins_next1 729 | sdc1 f0, 0(CRET1) 730 | ins_next2 731 | 732 |3: // Call __newindex metamethod. 733 | // BASE = base, L->top = new base, stack = cont/func/t/k/(v) 734 | addiu TMP1, BASE, -FRAME_CONT 735 | lw BASE, L->top 736 | sw PC, -16+HI(BASE) // [cont|PC] 737 | subu PC, BASE, TMP1 738 | lw LFUNC:RB, FRAME_FUNC(BASE) // Guaranteed to be a function here. 739 | sdc1 f0, 16(BASE) // Copy value to third argument. 740 | b ->vm_call_dispatch_f 741 |. li NARGS8:RC, 24 // 3 args for func(t, k, v) 742 | 743 |//-- Comparison metamethods --------------------------------------------- 744 | 745 |->vmeta_comp: 746 | // CARG2, CARG3 are already set by BC_ISLT/BC_ISGE/BC_ISLE/BC_ISGT. 747 | load_got lj_meta_comp 748 | addiu PC, PC, -4 749 | sw BASE, L->base 750 | sw PC, SAVE_PC 751 | decode_OP1 CARG4, INS 752 | call_intern lj_meta_comp // (lua_State *L, TValue *o1, *o2, int op) 753 |. move CARG1, L 754 | // Returns 0/1 or TValue * (metamethod). 755 |3: 756 | sltiu AT, CRET1, 2 757 | beqz AT, ->vmeta_binop 758 | negu TMP2, CRET1 759 |4: 760 | lhu RD, OFS_RD(PC) 761 | addiu PC, PC, 4 762 | lui TMP1, (-(BCBIAS_J*4 >> 16) & 65535) 763 | sll RD, RD, 2 764 | addu RD, RD, TMP1 765 | and RD, RD, TMP2 766 | addu PC, PC, RD 767 |->cont_nop: 768 | ins_next 769 | 770 |->cont_ra: // RA = resultptr 771 | lbu TMP1, -4+OFS_RA(PC) 772 | ldc1 f0, 0(RA) 773 | sll TMP1, TMP1, 3 774 | addu TMP1, BASE, TMP1 775 | b ->cont_nop 776 |. sdc1 f0, 0(TMP1) 777 | 778 |->cont_condt: // RA = resultptr 779 | lw TMP0, HI(RA) 780 | sltiu AT, TMP0, LJ_TISTRUECOND 781 | b <4 782 |. negu TMP2, AT // Branch if result is true. 783 | 784 |->cont_condf: // RA = resultptr 785 | lw TMP0, HI(RA) 786 | sltiu AT, TMP0, LJ_TISTRUECOND 787 | b <4 788 |. addiu TMP2, AT, -1 // Branch if result is false. 789 | 790 |->vmeta_equal: 791 | // CARG2, CARG3, CARG4 are already set by BC_ISEQV/BC_ISNEV. 792 | load_got lj_meta_equal 793 | addiu PC, PC, -4 794 | sw BASE, L->base 795 | sw PC, SAVE_PC 796 | call_intern lj_meta_equal // (lua_State *L, GCobj *o1, *o2, int ne) 797 |. move CARG1, L 798 | // Returns 0/1 or TValue * (metamethod). 799 | b <3 800 |. nop 801 | 802 |->vmeta_equal_cd: 803 |.if FFI 804 | load_got lj_meta_equal_cd 805 | move CARG2, INS 806 | addiu PC, PC, -4 807 | sw BASE, L->base 808 | sw PC, SAVE_PC 809 | call_intern lj_meta_equal_cd // (lua_State *L, BCIns op) 810 |. move CARG1, L 811 | // Returns 0/1 or TValue * (metamethod). 812 | b <3 813 |. nop 814 |.endif 815 | 816 |//-- Arithmetic metamethods --------------------------------------------- 817 | 818 |->vmeta_unm: 819 | move CARG4, CARG3 820 | 821 |->vmeta_arith: 822 | load_got lj_meta_arith 823 | decode_OP1 TMP0, INS 824 | sw BASE, L->base 825 | sw PC, SAVE_PC 826 | move CARG2, RA 827 | sw TMP0, ARG5 828 | call_intern lj_meta_arith // (lua_State *L, TValue *ra,*rb,*rc, BCReg op) 829 |. move CARG1, L 830 | // Returns NULL (finished) or TValue * (metamethod). 831 | beqz CRET1, ->cont_nop 832 |. nop 833 | 834 | // Call metamethod for binary op. 835 |->vmeta_binop: 836 | // BASE = old base, CRET1 = new base, stack = cont/func/o1/o2 837 | subu TMP1, CRET1, BASE 838 | sw PC, -16+HI(CRET1) // [cont|PC] 839 | move TMP2, BASE 840 | addiu PC, TMP1, FRAME_CONT 841 | move BASE, CRET1 842 | b ->vm_call_dispatch 843 |. li NARGS8:RC, 16 // 2 args for func(o1, o2). 844 | 845 |->vmeta_len: 846 | // CARG2 already set by BC_LEN. 847#if LJ_52 848 | move MULTRES, CARG1 849#endif 850 | load_got lj_meta_len 851 | sw BASE, L->base 852 | sw PC, SAVE_PC 853 | call_intern lj_meta_len // (lua_State *L, TValue *o) 854 |. move CARG1, L 855 | // Returns NULL (retry) or TValue * (metamethod base). 856#if LJ_52 857 | bnez CRET1, ->vmeta_binop // Binop call for compatibility. 858 |. nop 859 | b ->BC_LEN_Z 860 |. move CARG1, MULTRES 861#else 862 | b ->vmeta_binop // Binop call for compatibility. 863 |. nop 864#endif 865 | 866 |//-- Call metamethod ---------------------------------------------------- 867 | 868 |->vmeta_call: // Resolve and call __call metamethod. 869 | // TMP2 = old base, BASE = new base, RC = nargs*8 870 | load_got lj_meta_call 871 | sw TMP2, L->base // This is the callers base! 872 | addiu CARG2, BASE, -8 873 | sw PC, SAVE_PC 874 | addu CARG3, BASE, RC 875 | move MULTRES, NARGS8:RC 876 | call_intern lj_meta_call // (lua_State *L, TValue *func, TValue *top) 877 |. move CARG1, L 878 | lw LFUNC:RB, FRAME_FUNC(BASE) // Guaranteed to be a function here. 879 | addiu NARGS8:RC, MULTRES, 8 // Got one more argument now. 880 | ins_call 881 | 882 |->vmeta_callt: // Resolve __call for BC_CALLT. 883 | // BASE = old base, RA = new base, RC = nargs*8 884 | load_got lj_meta_call 885 | sw BASE, L->base 886 | addiu CARG2, RA, -8 887 | sw PC, SAVE_PC 888 | addu CARG3, RA, RC 889 | move MULTRES, NARGS8:RC 890 | call_intern lj_meta_call // (lua_State *L, TValue *func, TValue *top) 891 |. move CARG1, L 892 | lw TMP1, FRAME_PC(BASE) 893 | lw LFUNC:RB, FRAME_FUNC(RA) // Guaranteed to be a function here. 894 | b ->BC_CALLT_Z 895 |. addiu NARGS8:RC, MULTRES, 8 // Got one more argument now. 896 | 897 |//-- Argument coercion for 'for' statement ------------------------------ 898 | 899 |->vmeta_for: 900 | load_got lj_meta_for 901 | sw BASE, L->base 902 | move CARG2, RA 903 | sw PC, SAVE_PC 904 | move MULTRES, INS 905 | call_intern lj_meta_for // (lua_State *L, TValue *base) 906 |. move CARG1, L 907 |.if JIT 908 | decode_OP1 TMP0, MULTRES 909 | li AT, BC_JFORI 910 |.endif 911 | decode_RA8a RA, MULTRES 912 | decode_RD8a RD, MULTRES 913 | decode_RA8b RA 914 |.if JIT 915 | beq TMP0, AT, =>BC_JFORI 916 |. decode_RD8b RD 917 | b =>BC_FORI 918 |. nop 919 |.else 920 | b =>BC_FORI 921 |. decode_RD8b RD 922 |.endif 923 | 924 |//----------------------------------------------------------------------- 925 |//-- Fast functions ----------------------------------------------------- 926 |//----------------------------------------------------------------------- 927 | 928 |.macro .ffunc, name 929 |->ff_ .. name: 930 |.endmacro 931 | 932 |.macro .ffunc_1, name 933 |->ff_ .. name: 934 | beqz NARGS8:RC, ->fff_fallback 935 |. lw CARG3, HI(BASE) 936 | lw CARG1, LO(BASE) 937 |.endmacro 938 | 939 |.macro .ffunc_2, name 940 |->ff_ .. name: 941 | sltiu AT, NARGS8:RC, 16 942 | lw CARG3, HI(BASE) 943 | bnez AT, ->fff_fallback 944 |. lw CARG4, 8+HI(BASE) 945 | lw CARG1, LO(BASE) 946 | lw CARG2, 8+LO(BASE) 947 |.endmacro 948 | 949 |.macro .ffunc_n, name // Caveat: has delay slot! 950 |->ff_ .. name: 951 | lw CARG3, HI(BASE) 952 | beqz NARGS8:RC, ->fff_fallback 953 |. ldc1 FARG1, 0(BASE) 954 | sltiu AT, CARG3, LJ_TISNUM 955 | beqz AT, ->fff_fallback 956 |.endmacro 957 | 958 |.macro .ffunc_nn, name // Caveat: has delay slot! 959 |->ff_ .. name: 960 | sltiu AT, NARGS8:RC, 16 961 | lw CARG3, HI(BASE) 962 | bnez AT, ->fff_fallback 963 |. lw CARG4, 8+HI(BASE) 964 | ldc1 FARG1, 0(BASE) 965 | ldc1 FARG2, 8(BASE) 966 | sltiu TMP0, CARG3, LJ_TISNUM 967 | sltiu TMP1, CARG4, LJ_TISNUM 968 | and TMP0, TMP0, TMP1 969 | beqz TMP0, ->fff_fallback 970 |.endmacro 971 | 972 |// Inlined GC threshold check. Caveat: uses TMP0 and TMP1 and has delay slot! 973 |.macro ffgccheck 974 | lw TMP0, DISPATCH_GL(gc.total)(DISPATCH) 975 | lw TMP1, DISPATCH_GL(gc.threshold)(DISPATCH) 976 | subu AT, TMP0, TMP1 977 | bgezal AT, ->fff_gcstep 978 |.endmacro 979 | 980 |//-- Base library: checks ----------------------------------------------- 981 | 982 |.ffunc_1 assert 983 | sltiu AT, CARG3, LJ_TISTRUECOND 984 | beqz AT, ->fff_fallback 985 |. addiu RA, BASE, -8 986 | lw PC, FRAME_PC(BASE) 987 | addiu RD, NARGS8:RC, 8 // Compute (nresults+1)*8. 988 | addu TMP2, RA, NARGS8:RC 989 | sw CARG3, HI(RA) 990 | addiu TMP1, BASE, 8 991 | beq BASE, TMP2, ->fff_res // Done if exactly 1 argument. 992 |. sw CARG1, LO(RA) 993 |1: 994 | ldc1 f0, 0(TMP1) 995 | sdc1 f0, -8(TMP1) 996 | bne TMP1, TMP2, <1 997 |. addiu TMP1, TMP1, 8 998 | b ->fff_res 999 |. nop 1000 | 1001 |.ffunc type 1002 | lw CARG3, HI(BASE) 1003 | li TMP1, LJ_TISNUM 1004 | beqz NARGS8:RC, ->fff_fallback 1005 |. sltiu TMP0, CARG3, LJ_TISNUM 1006 | movz TMP1, CARG3, TMP0 1007 | not TMP1, TMP1 1008 | sll TMP1, TMP1, 3 1009 | addu TMP1, CFUNC:RB, TMP1 1010 | b ->fff_resn 1011 |. ldc1 FRET1, CFUNC:TMP1->upvalue 1012 | 1013 |//-- Base library: getters and setters --------------------------------- 1014 | 1015 |.ffunc_1 getmetatable 1016 | li AT, LJ_TTAB 1017 | bne CARG3, AT, >6 1018 |. li AT, LJ_TUDATA 1019 |1: // Field metatable must be at same offset for GCtab and GCudata! 1020 | lw TAB:CARG1, TAB:CARG1->metatable 1021 |2: 1022 | lw STR:RC, DISPATCH_GL(gcroot[GCROOT_MMNAME+MM_metatable])(DISPATCH) 1023 | beqz TAB:CARG1, ->fff_restv 1024 |. li CARG3, LJ_TNIL 1025 | lw TMP0, TAB:CARG1->hmask 1026 | li CARG3, LJ_TTAB // Use metatable as default result. 1027 | lw TMP1, STR:RC->hash 1028 | lw NODE:TMP2, TAB:CARG1->node 1029 | and TMP1, TMP1, TMP0 // idx = str->hash & tab->hmask 1030 | sll TMP0, TMP1, 5 1031 | sll TMP1, TMP1, 3 1032 | subu TMP1, TMP0, TMP1 1033 | addu NODE:TMP2, NODE:TMP2, TMP1 // node = tab->node + (idx*32-idx*8) 1034 | li AT, LJ_TSTR 1035 |3: // Rearranged logic, because we expect _not_ to find the key. 1036 | lw CARG4, offsetof(Node, key)+HI(NODE:TMP2) 1037 | lw TMP0, offsetof(Node, key)+LO(NODE:TMP2) 1038 | lw NODE:TMP3, NODE:TMP2->next 1039 | bne CARG4, AT, >4 1040 |. lw CARG2, offsetof(Node, val)+HI(NODE:TMP2) 1041 | beq TMP0, STR:RC, >5 1042 |. lw TMP1, offsetof(Node, val)+LO(NODE:TMP2) 1043 |4: 1044 | beqz NODE:TMP3, ->fff_restv // Not found, keep default result. 1045 |. move NODE:TMP2, NODE:TMP3 1046 | b <3 1047 |. nop 1048 |5: 1049 | beq CARG2, TISNIL, ->fff_restv // Ditto for nil value. 1050 |. nop 1051 | move CARG3, CARG2 // Return value of mt.__metatable. 1052 | b ->fff_restv 1053 |. move CARG1, TMP1 1054 | 1055 |6: 1056 | beq CARG3, AT, <1 1057 |. sltiu TMP0, CARG3, LJ_TISNUM 1058 | li TMP1, LJ_TISNUM 1059 | movz TMP1, CARG3, TMP0 1060 | not TMP1, TMP1 1061 | sll TMP1, TMP1, 2 1062 | addu TMP1, DISPATCH, TMP1 1063 | b <2 1064 |. lw TAB:CARG1, DISPATCH_GL(gcroot[GCROOT_BASEMT])(TMP1) 1065 | 1066 |.ffunc_2 setmetatable 1067 | // Fast path: no mt for table yet and not clearing the mt. 1068 | li AT, LJ_TTAB 1069 | bne CARG3, AT, ->fff_fallback 1070 |. addiu CARG4, CARG4, -LJ_TTAB 1071 | lw TAB:TMP1, TAB:CARG1->metatable 1072 | lbu TMP3, TAB:CARG1->marked 1073 | or AT, CARG4, TAB:TMP1 1074 | bnez AT, ->fff_fallback 1075 |. andi AT, TMP3, LJ_GC_BLACK // isblack(table) 1076 | beqz AT, ->fff_restv 1077 |. sw TAB:CARG2, TAB:CARG1->metatable 1078 | barrierback TAB:CARG1, TMP3, TMP0, ->fff_restv 1079 | 1080 |.ffunc rawget 1081 | lw CARG4, HI(BASE) 1082 | sltiu AT, NARGS8:RC, 16 1083 | lw TAB:CARG2, LO(BASE) 1084 | load_got lj_tab_get 1085 | addiu CARG4, CARG4, -LJ_TTAB 1086 | or AT, AT, CARG4 1087 | bnez AT, ->fff_fallback 1088 | addiu CARG3, BASE, 8 1089 | call_intern lj_tab_get // (lua_State *L, GCtab *t, cTValue *key) 1090 |. move CARG1, L 1091 | // Returns cTValue *. 1092 | b ->fff_resn 1093 |. ldc1 FRET1, 0(CRET1) 1094 | 1095 |//-- Base library: conversions ------------------------------------------ 1096 | 1097 |.ffunc tonumber 1098 | // Only handles the number case inline (without a base argument). 1099 | lw CARG1, HI(BASE) 1100 | xori AT, NARGS8:RC, 8 1101 | sltiu CARG1, CARG1, LJ_TISNUM 1102 | movn CARG1, r0, AT 1103 | beqz CARG1, ->fff_fallback // Exactly one number argument. 1104 |. ldc1 FRET1, 0(BASE) 1105 | b ->fff_resn 1106 |. nop 1107 | 1108 |.ffunc_1 tostring 1109 | // Only handles the string or number case inline. 1110 | li AT, LJ_TSTR 1111 | // A __tostring method in the string base metatable is ignored. 1112 | beq CARG3, AT, ->fff_restv // String key? 1113 | // Handle numbers inline, unless a number base metatable is present. 1114 |. lw TMP1, DISPATCH_GL(gcroot[GCROOT_BASEMT_NUM])(DISPATCH) 1115 | sltiu TMP0, CARG3, LJ_TISNUM 1116 | sltiu TMP1, TMP1, 1 1117 | and TMP0, TMP0, TMP1 1118 | beqz TMP0, ->fff_fallback 1119 |. sw BASE, L->base // Add frame since C call can throw. 1120 | ffgccheck 1121 |. sw PC, SAVE_PC // Redundant (but a defined value). 1122 | load_got lj_str_fromnum 1123 | move CARG1, L 1124 | call_intern lj_str_fromnum // (lua_State *L, lua_Number *np) 1125 |. move CARG2, BASE 1126 | // Returns GCstr *. 1127 | li CARG3, LJ_TSTR 1128 | b ->fff_restv 1129 |. move CARG1, CRET1 1130 | 1131 |//-- Base library: iterators ------------------------------------------- 1132 | 1133 |.ffunc next 1134 | lw CARG1, HI(BASE) 1135 | lw TAB:CARG2, LO(BASE) 1136 | beqz NARGS8:RC, ->fff_fallback 1137 |. addu TMP2, BASE, NARGS8:RC 1138 | li AT, LJ_TTAB 1139 | sw TISNIL, HI(TMP2) // Set missing 2nd arg to nil. 1140 | bne CARG1, AT, ->fff_fallback 1141 |. lw PC, FRAME_PC(BASE) 1142 | load_got lj_tab_next 1143 | sw BASE, L->base // Add frame since C call can throw. 1144 | sw BASE, L->top // Dummy frame length is ok. 1145 | addiu CARG3, BASE, 8 1146 | sw PC, SAVE_PC 1147 | call_intern lj_tab_next // (lua_State *L, GCtab *t, TValue *key) 1148 |. move CARG1, L 1149 | // Returns 0 at end of traversal. 1150 | beqz CRET1, ->fff_restv // End of traversal: return nil. 1151 |. li CARG3, LJ_TNIL 1152 | ldc1 f0, 8(BASE) // Copy key and value to results. 1153 | addiu RA, BASE, -8 1154 | ldc1 f2, 16(BASE) 1155 | li RD, (2+1)*8 1156 | sdc1 f0, 0(RA) 1157 | b ->fff_res 1158 |. sdc1 f2, 8(RA) 1159 | 1160 |.ffunc_1 pairs 1161 | li AT, LJ_TTAB 1162 | bne CARG3, AT, ->fff_fallback 1163 |. lw PC, FRAME_PC(BASE) 1164#if LJ_52 1165 | lw TAB:TMP2, TAB:CARG1->metatable 1166 | ldc1 f0, CFUNC:RB->upvalue[0] 1167 | bnez TAB:TMP2, ->fff_fallback 1168#else 1169 | ldc1 f0, CFUNC:RB->upvalue[0] 1170#endif 1171 |. addiu RA, BASE, -8 1172 | sw TISNIL, 8+HI(BASE) 1173 | li RD, (3+1)*8 1174 | b ->fff_res 1175 |. sdc1 f0, 0(RA) 1176 | 1177 |.ffunc ipairs_aux 1178 | sltiu AT, NARGS8:RC, 16 1179 | lw CARG3, HI(BASE) 1180 | lw TAB:CARG1, LO(BASE) 1181 | lw CARG4, 8+HI(BASE) 1182 | bnez AT, ->fff_fallback 1183 |. ldc1 FARG2, 8(BASE) 1184 | addiu CARG3, CARG3, -LJ_TTAB 1185 | sltiu AT, CARG4, LJ_TISNUM 1186 | li TMP0, 1 1187 | movn AT, r0, CARG3 1188 | mtc1 TMP0, FARG1 1189 | beqz AT, ->fff_fallback 1190 |. lw PC, FRAME_PC(BASE) 1191 | cvt.w.d FRET1, FARG2 1192 | cvt.d.w FARG1, FARG1 1193 | lw TMP0, TAB:CARG1->asize 1194 | lw TMP1, TAB:CARG1->array 1195 | mfc1 TMP2, FRET1 1196 | addiu RA, BASE, -8 1197 | add.d FARG2, FARG2, FARG1 1198 | addiu TMP2, TMP2, 1 1199 | sltu AT, TMP2, TMP0 1200 | sll TMP3, TMP2, 3 1201 | addu TMP3, TMP1, TMP3 1202 | beqz AT, >2 // Not in array part? 1203 |. sdc1 FARG2, 0(RA) 1204 | lw TMP2, HI(TMP3) 1205 | ldc1 f0, 0(TMP3) 1206 |1: 1207 | beq TMP2, TISNIL, ->fff_res // End of iteration, return 0 results. 1208 |. li RD, (0+1)*8 1209 | li RD, (2+1)*8 1210 | b ->fff_res 1211 |. sdc1 f0, 8(RA) 1212 |2: // Check for empty hash part first. Otherwise call C function. 1213 | lw TMP0, TAB:CARG1->hmask 1214 | load_got lj_tab_getinth 1215 | beqz TMP0, ->fff_res 1216 |. li RD, (0+1)*8 1217 | call_intern lj_tab_getinth // (GCtab *t, int32_t key) 1218 |. move CARG2, TMP2 1219 | // Returns cTValue * or NULL. 1220 | beqz CRET1, ->fff_res 1221 |. li RD, (0+1)*8 1222 | lw TMP2, HI(CRET1) 1223 | b <1 1224 |. ldc1 f0, 0(CRET1) 1225 | 1226 |.ffunc_1 ipairs 1227 | li AT, LJ_TTAB 1228 | bne CARG3, AT, ->fff_fallback 1229 |. lw PC, FRAME_PC(BASE) 1230#if LJ_52 1231 | lw TAB:TMP2, TAB:CARG1->metatable 1232 | ldc1 f0, CFUNC:RB->upvalue[0] 1233 | bnez TAB:TMP2, ->fff_fallback 1234#else 1235 | ldc1 f0, CFUNC:RB->upvalue[0] 1236#endif 1237 |. addiu RA, BASE, -8 1238 | sw r0, 8+HI(BASE) 1239 | sw r0, 8+LO(BASE) 1240 | li RD, (3+1)*8 1241 | b ->fff_res 1242 |. sdc1 f0, 0(RA) 1243 | 1244 |//-- Base library: catch errors ---------------------------------------- 1245 | 1246 |.ffunc pcall 1247 | lbu TMP3, DISPATCH_GL(hookmask)(DISPATCH) 1248 | beqz NARGS8:RC, ->fff_fallback 1249 | move TMP2, BASE 1250 | addiu BASE, BASE, 8 1251 | // Remember active hook before pcall. 1252 | srl TMP3, TMP3, HOOK_ACTIVE_SHIFT 1253 | andi TMP3, TMP3, 1 1254 | addiu PC, TMP3, 8+FRAME_PCALL 1255 | b ->vm_call_dispatch 1256 |. addiu NARGS8:RC, NARGS8:RC, -8 1257 | 1258 |.ffunc xpcall 1259 | sltiu AT, NARGS8:RC, 16 1260 | lw CARG4, 8+HI(BASE) 1261 | bnez AT, ->fff_fallback 1262 |. ldc1 FARG2, 8(BASE) 1263 | ldc1 FARG1, 0(BASE) 1264 | lbu TMP1, DISPATCH_GL(hookmask)(DISPATCH) 1265 | li AT, LJ_TFUNC 1266 | move TMP2, BASE 1267 | bne CARG4, AT, ->fff_fallback // Traceback must be a function. 1268 | addiu BASE, BASE, 16 1269 | // Remember active hook before pcall. 1270 | srl TMP3, TMP3, HOOK_ACTIVE_SHIFT 1271 | sdc1 FARG2, 0(TMP2) // Swap function and traceback. 1272 | andi TMP3, TMP3, 1 1273 | sdc1 FARG1, 8(TMP2) 1274 | addiu PC, TMP3, 16+FRAME_PCALL 1275 | b ->vm_call_dispatch 1276 |. addiu NARGS8:RC, NARGS8:RC, -16 1277 | 1278 |//-- Coroutine library -------------------------------------------------- 1279 | 1280 |.macro coroutine_resume_wrap, resume 1281 |.if resume 1282 |.ffunc_1 coroutine_resume 1283 | li AT, LJ_TTHREAD 1284 | bne CARG3, AT, ->fff_fallback 1285 |.else 1286 |.ffunc coroutine_wrap_aux 1287 | lw L:CARG1, CFUNC:RB->upvalue[0].gcr 1288 |.endif 1289 | lbu TMP0, L:CARG1->status 1290 | lw TMP1, L:CARG1->cframe 1291 | lw CARG2, L:CARG1->top 1292 | lw TMP2, L:CARG1->base 1293 | addiu TMP3, TMP0, -LUA_YIELD 1294 | bgtz TMP3, ->fff_fallback // st > LUA_YIELD? 1295 |. xor TMP2, TMP2, CARG2 1296 | bnez TMP1, ->fff_fallback // cframe != 0? 1297 |. or AT, TMP2, TMP0 1298 | lw TMP0, L:CARG1->maxstack 1299 | beqz AT, ->fff_fallback // base == top && st == 0? 1300 |. lw PC, FRAME_PC(BASE) 1301 | addu TMP2, CARG2, NARGS8:RC 1302 | sltu AT, TMP0, TMP2 1303 | bnez AT, ->fff_fallback // Stack overflow? 1304 |. sw PC, SAVE_PC 1305 | sw BASE, L->base 1306 |1: 1307 |.if resume 1308 | addiu BASE, BASE, 8 // Keep resumed thread in stack for GC. 1309 | addiu NARGS8:RC, NARGS8:RC, -8 1310 | addiu TMP2, TMP2, -8 1311 |.endif 1312 | sw TMP2, L:CARG1->top 1313 | addu TMP1, BASE, NARGS8:RC 1314 | move CARG3, CARG2 1315 | sw BASE, L->top 1316 |2: // Move args to coroutine. 1317 | ldc1 f0, 0(BASE) 1318 | sltu AT, BASE, TMP1 1319 | beqz AT, >3 1320 |. addiu BASE, BASE, 8 1321 | sdc1 f0, 0(CARG3) 1322 | b <2 1323 |. addiu CARG3, CARG3, 8 1324 |3: 1325 | bal ->vm_resume // (lua_State *L, TValue *base, 0, 0) 1326 |. move L:RA, L:CARG1 1327 | // Returns thread status. 1328 |4: 1329 | lw TMP2, L:RA->base 1330 | sltiu AT, CRET1, LUA_YIELD+1 1331 | lw TMP3, L:RA->top 1332 | li_vmstate INTERP 1333 | lw BASE, L->base 1334 | st_vmstate 1335 | beqz AT, >8 1336 |. subu RD, TMP3, TMP2 1337 | lw TMP0, L->maxstack 1338 | beqz RD, >6 // No results? 1339 |. addu TMP1, BASE, RD 1340 | sltu AT, TMP0, TMP1 1341 | bnez AT, >9 // Need to grow stack? 1342 |. addu TMP3, TMP2, RD 1343 | sw TMP2, L:RA->top // Clear coroutine stack. 1344 | move TMP1, BASE 1345 |5: // Move results from coroutine. 1346 | ldc1 f0, 0(TMP2) 1347 | addiu TMP2, TMP2, 8 1348 | sltu AT, TMP2, TMP3 1349 | sdc1 f0, 0(TMP1) 1350 | bnez AT, <5 1351 |. addiu TMP1, TMP1, 8 1352 |6: 1353 | andi TMP0, PC, FRAME_TYPE 1354 |.if resume 1355 | li TMP1, LJ_TTRUE 1356 | addiu RA, BASE, -8 1357 | sw TMP1, -8+HI(BASE) // Prepend true to results. 1358 | addiu RD, RD, 16 1359 |.else 1360 | move RA, BASE 1361 | addiu RD, RD, 8 1362 |.endif 1363 |7: 1364 | sw PC, SAVE_PC 1365 | beqz TMP0, ->BC_RET_Z 1366 |. move MULTRES, RD 1367 | b ->vm_return 1368 |. nop 1369 | 1370 |8: // Coroutine returned with error (at co->top-1). 1371 |.if resume 1372 | addiu TMP3, TMP3, -8 1373 | li TMP1, LJ_TFALSE 1374 | ldc1 f0, 0(TMP3) 1375 | sw TMP3, L:RA->top // Remove error from coroutine stack. 1376 | li RD, (2+1)*8 1377 | sw TMP1, -8+HI(BASE) // Prepend false to results. 1378 | addiu RA, BASE, -8 1379 | sdc1 f0, 0(BASE) // Copy error message. 1380 | b <7 1381 |. andi TMP0, PC, FRAME_TYPE 1382 |.else 1383 | load_got lj_ffh_coroutine_wrap_err 1384 | move CARG2, L:RA 1385 | call_intern lj_ffh_coroutine_wrap_err // (lua_State *L, lua_State *co) 1386 |. move CARG1, L 1387 |.endif 1388 | 1389 |9: // Handle stack expansion on return from yield. 1390 | load_got lj_state_growstack 1391 | srl CARG2, RD, 3 1392 | call_intern lj_state_growstack // (lua_State *L, int n) 1393 |. move CARG1, L 1394 | b <4 1395 |. li CRET1, 0 1396 |.endmacro 1397 | 1398 | coroutine_resume_wrap 1 // coroutine.resume 1399 | coroutine_resume_wrap 0 // coroutine.wrap 1400 | 1401 |.ffunc coroutine_yield 1402 | lw TMP0, L->cframe 1403 | addu TMP1, BASE, NARGS8:RC 1404 | sw BASE, L->base 1405 | andi TMP0, TMP0, CFRAME_RESUME 1406 | sw TMP1, L->top 1407 | beqz TMP0, ->fff_fallback 1408 |. li CRET1, LUA_YIELD 1409 | sw r0, L->cframe 1410 | b ->vm_leave_unw 1411 |. sb CRET1, L->status 1412 | 1413 |//-- Math library ------------------------------------------------------- 1414 | 1415 |.ffunc_n math_abs 1416 |. abs.d FRET1, FARG1 1417 |->fff_resn: 1418 | lw PC, FRAME_PC(BASE) 1419 | addiu RA, BASE, -8 1420 | b ->fff_res1 1421 |. sdc1 FRET1, -8(BASE) 1422 | 1423 |->fff_restv: 1424 | // CARG3/CARG1 = TValue result. 1425 | lw PC, FRAME_PC(BASE) 1426 | sw CARG3, -8+HI(BASE) 1427 | addiu RA, BASE, -8 1428 | sw CARG1, -8+LO(BASE) 1429 |->fff_res1: 1430 | // RA = results, PC = return. 1431 | li RD, (1+1)*8 1432 |->fff_res: 1433 | // RA = results, RD = (nresults+1)*8, PC = return. 1434 | andi TMP0, PC, FRAME_TYPE 1435 | bnez TMP0, ->vm_return 1436 |. move MULTRES, RD 1437 | lw INS, -4(PC) 1438 | decode_RB8a RB, INS 1439 | decode_RB8b RB 1440 |5: 1441 | sltu AT, RD, RB 1442 | bnez AT, >6 // More results expected? 1443 |. decode_RA8a TMP0, INS 1444 | decode_RA8b TMP0 1445 | ins_next1 1446 | // Adjust BASE. KBASE is assumed to be set for the calling frame. 1447 | subu BASE, RA, TMP0 1448 | ins_next2 1449 | 1450 |6: // Fill up results with nil. 1451 | addu TMP1, RA, RD 1452 | addiu RD, RD, 8 1453 | b <5 1454 |. sw TISNIL, -8+HI(TMP1) 1455 | 1456 |.macro math_extern, func 1457 |->ff_math_ .. func: 1458 | lw CARG3, HI(BASE) 1459 | beqz NARGS8:RC, ->fff_fallback 1460 |. load_got func 1461 | sltiu AT, CARG3, LJ_TISNUM 1462 | beqz AT, ->fff_fallback 1463 |. nop 1464 | call_extern 1465 |. ldc1 FARG1, 0(BASE) 1466 | b ->fff_resn 1467 |. nop 1468 |.endmacro 1469 | 1470 |.macro math_extern2, func 1471 | .ffunc_nn math_ .. func 1472 |. load_got func 1473 | call_extern 1474 |. nop 1475 | b ->fff_resn 1476 |. nop 1477 |.endmacro 1478 | 1479 |.macro math_round, func 1480 | .ffunc_n math_ .. func 1481 |. nop 1482 | bal ->vm_ .. func 1483 |. nop 1484 | b ->fff_resn 1485 |. nop 1486 |.endmacro 1487 | 1488 | math_round floor 1489 | math_round ceil 1490 | 1491 |.ffunc math_log 1492 | lw CARG3, HI(BASE) 1493 | li AT, 8 1494 | bne NARGS8:RC, AT, ->fff_fallback // Exactly 1 argument. 1495 |. load_got log 1496 | sltiu AT, CARG3, LJ_TISNUM 1497 | beqz AT, ->fff_fallback 1498 |. nop 1499 | call_extern 1500 |. ldc1 FARG1, 0(BASE) 1501 | b ->fff_resn 1502 |. nop 1503 | 1504 | math_extern log10 1505 | math_extern exp 1506 | math_extern sin 1507 | math_extern cos 1508 | math_extern tan 1509 | math_extern asin 1510 | math_extern acos 1511 | math_extern atan 1512 | math_extern sinh 1513 | math_extern cosh 1514 | math_extern tanh 1515 | math_extern2 pow 1516 | math_extern2 atan2 1517 | math_extern2 fmod 1518 | 1519 |.ffunc_n math_sqrt 1520 |. sqrt.d FRET1, FARG1 1521 | b ->fff_resn 1522 |. nop 1523 | 1524 |->ff_math_deg: 1525 |.ffunc_n math_rad 1526 |. ldc1 FARG2, CFUNC:RB->upvalue[0] 1527 | b ->fff_resn 1528 |. mul.d FRET1, FARG1, FARG2 1529 | 1530 |.ffunc_nn math_ldexp 1531 | cvt.w.d FARG2, FARG2 1532 | load_got ldexp 1533 | mfc1 CARG3, FARG2 1534 | call_extern 1535 |. nop 1536 | b ->fff_resn 1537 |. nop 1538 | 1539 |.ffunc_n math_frexp 1540 | load_got frexp 1541 | lw PC, FRAME_PC(BASE) 1542 | call_extern 1543 |. addiu CARG3, DISPATCH, DISPATCH_GL(tmptv) 1544 | lw TMP1, DISPATCH_GL(tmptv)(DISPATCH) 1545 | addiu RA, BASE, -8 1546 | mtc1 TMP1, FARG2 1547 | sdc1 FRET1, 0(RA) 1548 | cvt.d.w FARG2, FARG2 1549 | sdc1 FARG2, 8(RA) 1550 | b ->fff_res 1551 |. li RD, (2+1)*8 1552 | 1553 |.ffunc_n math_modf 1554 | load_got modf 1555 | lw PC, FRAME_PC(BASE) 1556 | call_extern 1557 |. addiu CARG3, BASE, -8 1558 | addiu RA, BASE, -8 1559 | sdc1 FRET1, 0(BASE) 1560 | b ->fff_res 1561 |. li RD, (2+1)*8 1562 | 1563 |.macro math_minmax, name, ismax 1564 |->ff_ .. name: 1565 | lw CARG3, HI(BASE) 1566 | beqz NARGS8:RC, ->fff_fallback 1567 |. ldc1 FRET1, 0(BASE) 1568 | sltiu AT, CARG3, LJ_TISNUM 1569 | beqz AT, ->fff_fallback 1570 |. addu TMP2, BASE, NARGS8:RC 1571 | addiu TMP1, BASE, 8 1572 | beq TMP1, TMP2, ->fff_resn 1573 |1: 1574 |. lw CARG3, HI(TMP1) 1575 | ldc1 FARG1, 0(TMP1) 1576 | addiu TMP1, TMP1, 8 1577 | sltiu AT, CARG3, LJ_TISNUM 1578 | beqz AT, ->fff_fallback 1579 |.if ismax 1580 |. c.olt.d FARG1, FRET1 1581 |.else 1582 |. c.olt.d FRET1, FARG1 1583 |.endif 1584 | bne TMP1, TMP2, <1 1585 |. movf.d FRET1, FARG1 1586 | b ->fff_resn 1587 |. nop 1588 |.endmacro 1589 | 1590 | math_minmax math_min, 0 1591 | math_minmax math_max, 1 1592 | 1593 |//-- String library ----------------------------------------------------- 1594 | 1595 |.ffunc_1 string_len 1596 | li AT, LJ_TSTR 1597 | bne CARG3, AT, ->fff_fallback 1598 |. nop 1599 | b ->fff_resi 1600 |. lw CRET1, STR:CARG1->len 1601 | 1602 |.ffunc string_byte // Only handle the 1-arg case here. 1603 | lw CARG3, HI(BASE) 1604 | lw STR:CARG1, LO(BASE) 1605 | xori AT, NARGS8:RC, 8 1606 | addiu CARG3, CARG3, -LJ_TSTR 1607 | or AT, AT, CARG3 1608 | bnez AT, ->fff_fallback // Need exactly 1 string argument. 1609 |. nop 1610 | lw TMP0, STR:CARG1->len 1611 | lbu TMP1, STR:CARG1[1] // Access is always ok (NUL at end). 1612 | addiu RA, BASE, -8 1613 | sltu RD, r0, TMP0 1614 | mtc1 TMP1, f0 1615 | addiu RD, RD, 1 1616 | cvt.d.w f0, f0 1617 | lw PC, FRAME_PC(BASE) 1618 | sll RD, RD, 3 // RD = ((str->len != 0)+1)*8 1619 | b ->fff_res 1620 |. sdc1 f0, 0(RA) 1621 | 1622 |.ffunc string_char // Only handle the 1-arg case here. 1623 | ffgccheck 1624 |. nop 1625 | lw CARG3, HI(BASE) 1626 | ldc1 FARG1, 0(BASE) 1627 | li AT, 8 1628 | bne NARGS8:RC, AT, ->fff_fallback // Exactly 1 argument. 1629 |. sltiu AT, CARG3, LJ_TISNUM 1630 | beqz AT, ->fff_fallback 1631 |. li CARG3, 1 1632 | cvt.w.d FARG1, FARG1 1633 | addiu CARG2, sp, ARG5_OFS 1634 | sltiu AT, TMP0, 256 1635 | mfc1 TMP0, FARG1 1636 | beqz AT, ->fff_fallback 1637 |. sw TMP0, ARG5 1638 |->fff_newstr: 1639 | load_got lj_str_new 1640 | sw BASE, L->base 1641 | sw PC, SAVE_PC 1642 | call_intern lj_str_new // (lua_State *L, char *str, size_t l) 1643 |. move CARG1, L 1644 | // Returns GCstr *. 1645 | lw BASE, L->base 1646 | move CARG1, CRET1 1647 | b ->fff_restv 1648 |. li CARG3, LJ_TSTR 1649 | 1650 |.ffunc string_sub 1651 | ffgccheck 1652 |. nop 1653 | addiu AT, NARGS8:RC, -16 1654 | lw CARG3, 16+HI(BASE) 1655 | ldc1 f0, 16(BASE) 1656 | lw TMP0, HI(BASE) 1657 | lw STR:CARG1, LO(BASE) 1658 | bltz AT, ->fff_fallback 1659 | lw CARG2, 8+HI(BASE) 1660 | ldc1 f2, 8(BASE) 1661 | beqz AT, >1 1662 |. li CARG4, -1 1663 | cvt.w.d f0, f0 1664 | sltiu AT, CARG3, LJ_TISNUM 1665 | beqz AT, ->fff_fallback 1666 |. mfc1 CARG4, f0 1667 |1: 1668 | sltiu AT, CARG2, LJ_TISNUM 1669 | beqz AT, ->fff_fallback 1670 |. li AT, LJ_TSTR 1671 | cvt.w.d f2, f2 1672 | bne TMP0, AT, ->fff_fallback 1673 |. lw CARG2, STR:CARG1->len 1674 | mfc1 CARG3, f2 1675 | // STR:CARG1 = str, CARG2 = str->len, CARG3 = start, CARG4 = end 1676 | slt AT, CARG4, r0 1677 | addiu TMP0, CARG2, 1 1678 | addu TMP1, CARG4, TMP0 1679 | slt TMP3, CARG3, r0 1680 | movn CARG4, TMP1, AT // if (end < 0) end += len+1 1681 | addu TMP1, CARG3, TMP0 1682 | movn CARG3, TMP1, TMP3 // if (start < 0) start += len+1 1683 | li TMP2, 1 1684 | slt AT, CARG4, r0 1685 | slt TMP3, r0, CARG3 1686 | movn CARG4, r0, AT // if (end < 0) end = 0 1687 | movz CARG3, TMP2, TMP3 // if (start < 1) start = 1 1688 | slt AT, CARG2, CARG4 1689 | movn CARG4, CARG2, AT // if (end > len) end = len 1690 | addu CARG2, STR:CARG1, CARG3 1691 | subu CARG3, CARG4, CARG3 // len = end - start 1692 | addiu CARG2, CARG2, sizeof(GCstr)-1 1693 | bgez CARG3, ->fff_newstr 1694 |. addiu CARG3, CARG3, 1 // len++ 1695 |->fff_emptystr: // Return empty string. 1696 | addiu STR:CARG1, DISPATCH, DISPATCH_GL(strempty) 1697 | b ->fff_restv 1698 |. li CARG3, LJ_TSTR 1699 | 1700 |.ffunc string_rep // Only handle the 1-char case inline. 1701 | ffgccheck 1702 |. nop 1703 | lw TMP0, HI(BASE) 1704 | addiu AT, NARGS8:RC, -16 // Exactly 2 arguments. 1705 | lw CARG4, 8+HI(BASE) 1706 | lw STR:CARG1, LO(BASE) 1707 | addiu TMP0, TMP0, -LJ_TSTR 1708 | ldc1 f0, 8(BASE) 1709 | or AT, AT, TMP0 1710 | bnez AT, ->fff_fallback 1711 |. sltiu AT, CARG4, LJ_TISNUM 1712 | cvt.w.d f0, f0 1713 | beqz AT, ->fff_fallback 1714 |. lw TMP0, STR:CARG1->len 1715 | mfc1 CARG3, f0 1716 | lw TMP1, DISPATCH_GL(tmpbuf.sz)(DISPATCH) 1717 | li AT, 1 1718 | blez CARG3, ->fff_emptystr // Count <= 0? 1719 |. sltu AT, AT, TMP0 1720 | beqz TMP0, ->fff_emptystr // Zero length string? 1721 |. sltu TMP0, TMP1, CARG3 1722 | or AT, AT, TMP0 1723 | lw CARG2, DISPATCH_GL(tmpbuf.buf)(DISPATCH) 1724 | bnez AT, ->fff_fallback // Fallback for > 1-char strings. 1725 |. lbu TMP0, STR:CARG1[1] 1726 | addu TMP2, CARG2, CARG3 1727 |1: // Fill buffer with char. Yes, this is suboptimal code (do you care?). 1728 | addiu TMP2, TMP2, -1 1729 | sltu AT, CARG2, TMP2 1730 | bnez AT, <1 1731 |. sb TMP0, 0(TMP2) 1732 | b ->fff_newstr 1733 |. nop 1734 | 1735 |.ffunc string_reverse 1736 | ffgccheck 1737 |. nop 1738 | lw CARG3, HI(BASE) 1739 | lw STR:CARG1, LO(BASE) 1740 | beqz NARGS8:RC, ->fff_fallback 1741 |. li AT, LJ_TSTR 1742 | bne CARG3, AT, ->fff_fallback 1743 |. lw TMP1, DISPATCH_GL(tmpbuf.sz)(DISPATCH) 1744 | lw CARG3, STR:CARG1->len 1745 | addiu CARG1, STR:CARG1, #STR 1746 | lw CARG2, DISPATCH_GL(tmpbuf.buf)(DISPATCH) 1747 | sltu AT, TMP1, CARG3 1748 | bnez AT, ->fff_fallback 1749 |. addu TMP3, CARG1, CARG3 1750 | addu CARG4, CARG2, CARG3 1751 |1: // Reverse string copy. 1752 | lbu TMP1, 0(CARG1) 1753 | sltu AT, CARG1, TMP3 1754 | beqz AT, ->fff_newstr 1755 |. addiu CARG1, CARG1, 1 1756 | addiu CARG4, CARG4, -1 1757 | b <1 1758 | sb TMP1, 0(CARG4) 1759 | 1760 |.macro ffstring_case, name, lo 1761 | .ffunc name 1762 | ffgccheck 1763 |. nop 1764 | lw CARG3, HI(BASE) 1765 | lw STR:CARG1, LO(BASE) 1766 | beqz NARGS8:RC, ->fff_fallback 1767 |. li AT, LJ_TSTR 1768 | bne CARG3, AT, ->fff_fallback 1769 |. lw TMP1, DISPATCH_GL(tmpbuf.sz)(DISPATCH) 1770 | lw CARG3, STR:CARG1->len 1771 | addiu CARG1, STR:CARG1, #STR 1772 | lw CARG2, DISPATCH_GL(tmpbuf.buf)(DISPATCH) 1773 | sltu AT, TMP1, CARG3 1774 | bnez AT, ->fff_fallback 1775 |. addu TMP3, CARG1, CARG3 1776 | move CARG4, CARG2 1777 |1: // ASCII case conversion. 1778 | lbu TMP1, 0(CARG1) 1779 | sltu AT, CARG1, TMP3 1780 | beqz AT, ->fff_newstr 1781 |. addiu TMP0, TMP1, -lo 1782 | xori TMP2, TMP1, 0x20 1783 | sltiu AT, TMP0, 26 1784 | movn TMP1, TMP2, AT 1785 | addiu CARG1, CARG1, 1 1786 | sb TMP1, 0(CARG4) 1787 | b <1 1788 |. addiu CARG4, CARG4, 1 1789 |.endmacro 1790 | 1791 |ffstring_case string_lower, 65 1792 |ffstring_case string_upper, 97 1793 | 1794 |//-- Table library ------------------------------------------------------ 1795 | 1796 |.ffunc_1 table_getn 1797 | li AT, LJ_TTAB 1798 | bne CARG3, AT, ->fff_fallback 1799 |. load_got lj_tab_len 1800 | call_intern lj_tab_len // (GCtab *t) 1801 |. nop 1802 | // Returns uint32_t (but less than 2^31). 1803 | b ->fff_resi 1804 |. nop 1805 | 1806 |//-- Bit library -------------------------------------------------------- 1807 | 1808 |.macro .ffunc_bit, name 1809 | .ffunc_n bit_..name 1810 |. add.d FARG1, FARG1, TOBIT 1811 | mfc1 CRET1, FARG1 1812 |.endmacro 1813 | 1814 |.macro .ffunc_bit_op, name, ins 1815 | .ffunc_bit name 1816 | addiu TMP1, BASE, 8 1817 | addu TMP2, BASE, NARGS8:RC 1818 |1: 1819 | lw CARG4, HI(TMP1) 1820 | beq TMP1, TMP2, ->fff_resi 1821 |. ldc1 FARG1, 0(TMP1) 1822 | sltiu AT, CARG4, LJ_TISNUM 1823 | beqz AT, ->fff_fallback 1824 | add.d FARG1, FARG1, TOBIT 1825 | mfc1 CARG2, FARG1 1826 | ins CRET1, CRET1, CARG2 1827 | b <1 1828 |. addiu TMP1, TMP1, 8 1829 |.endmacro 1830 | 1831 |.ffunc_bit_op band, and 1832 |.ffunc_bit_op bor, or 1833 |.ffunc_bit_op bxor, xor 1834 | 1835 |.ffunc_bit bswap 1836 | srl TMP0, CRET1, 24 1837 | srl TMP2, CRET1, 8 1838 | sll TMP1, CRET1, 24 1839 | andi TMP2, TMP2, 0xff00 1840 | or TMP0, TMP0, TMP1 1841 | andi CRET1, CRET1, 0xff00 1842 | or TMP0, TMP0, TMP2 1843 | sll CRET1, CRET1, 8 1844 | b ->fff_resi 1845 |. or CRET1, TMP0, CRET1 1846 | 1847 |.ffunc_bit bnot 1848 | b ->fff_resi 1849 |. not CRET1, CRET1 1850 | 1851 |.macro .ffunc_bit_sh, name, ins, shmod 1852 | .ffunc_nn bit_..name 1853 |. add.d FARG1, FARG1, TOBIT 1854 | add.d FARG2, FARG2, TOBIT 1855 | mfc1 CARG1, FARG1 1856 | mfc1 CARG2, FARG2 1857 |.if shmod == 1 1858 | li AT, 32 1859 | subu TMP0, AT, CARG2 1860 | sllv CARG2, CARG1, CARG2 1861 | srlv CARG1, CARG1, TMP0 1862 |.elif shmod == 2 1863 | li AT, 32 1864 | subu TMP0, AT, CARG2 1865 | srlv CARG2, CARG1, CARG2 1866 | sllv CARG1, CARG1, TMP0 1867 |.endif 1868 | b ->fff_resi 1869 |. ins CRET1, CARG1, CARG2 1870 |.endmacro 1871 | 1872 |.ffunc_bit_sh lshift, sllv, 0 1873 |.ffunc_bit_sh rshift, srlv, 0 1874 |.ffunc_bit_sh arshift, srav, 0 1875 |// Can't use rotrv, since it's only in MIPS32R2. 1876 |.ffunc_bit_sh rol, or, 1 1877 |.ffunc_bit_sh ror, or, 2 1878 | 1879 |.ffunc_bit tobit 1880 |->fff_resi: 1881 | mtc1 CRET1, FRET1 1882 | b ->fff_resn 1883 |. cvt.d.w FRET1, FRET1 1884 | 1885 |//----------------------------------------------------------------------- 1886 | 1887 |->fff_fallback: // Call fast function fallback handler. 1888 | // BASE = new base, RB = CFUNC, RC = nargs*8 1889 | lw TMP3, CFUNC:RB->f 1890 | addu TMP1, BASE, NARGS8:RC 1891 | lw PC, FRAME_PC(BASE) // Fallback may overwrite PC. 1892 | addiu TMP0, TMP1, 8*LUA_MINSTACK 1893 | lw TMP2, L->maxstack 1894 | sw PC, SAVE_PC // Redundant (but a defined value). 1895 | sltu AT, TMP2, TMP0 1896 | sw BASE, L->base 1897 | sw TMP1, L->top 1898 | bnez AT, >5 // Need to grow stack. 1899 |. move CFUNCADDR, TMP3 1900 | jalr TMP3 // (lua_State *L) 1901 |. move CARG1, L 1902 | // Either throws an error, or recovers and returns -1, 0 or nresults+1. 1903 | lw BASE, L->base 1904 | sll RD, CRET1, 3 1905 | bgtz CRET1, ->fff_res // Returned nresults+1? 1906 |. addiu RA, BASE, -8 1907 |1: // Returned 0 or -1: retry fast path. 1908 | lw TMP0, L->top 1909 | lw LFUNC:RB, FRAME_FUNC(BASE) 1910 | bnez CRET1, ->vm_call_tail // Returned -1? 1911 |. subu NARGS8:RC, TMP0, BASE 1912 | ins_callt // Returned 0: retry fast path. 1913 | 1914 |// Reconstruct previous base for vmeta_call during tailcall. 1915 |->vm_call_tail: 1916 | andi TMP0, PC, FRAME_TYPE 1917 | li AT, -4 1918 | bnez TMP0, >3 1919 |. and TMP1, PC, AT 1920 | lbu TMP1, OFS_RA(PC) 1921 | sll TMP1, TMP1, 3 1922 | addiu TMP1, TMP1, 8 1923 |3: 1924 | b ->vm_call_dispatch // Resolve again for tailcall. 1925 |. subu TMP2, BASE, TMP1 1926 | 1927 |5: // Grow stack for fallback handler. 1928 | load_got lj_state_growstack 1929 | li CARG2, LUA_MINSTACK 1930 | call_intern lj_state_growstack // (lua_State *L, int n) 1931 |. move CARG1, L 1932 | lw BASE, L->base 1933 | b <1 1934 |. li CRET1, 0 // Force retry. 1935 | 1936 |->fff_gcstep: // Call GC step function. 1937 | // BASE = new base, RC = nargs*8 1938 | move MULTRES, ra 1939 | load_got lj_gc_step 1940 | sw BASE, L->base 1941 | addu TMP0, BASE, NARGS8:RC 1942 | sw PC, SAVE_PC // Redundant (but a defined value). 1943 | sw TMP0, L->top 1944 | call_intern lj_gc_step // (lua_State *L) 1945 |. move CARG1, L 1946 | lw BASE, L->base 1947 | move ra, MULTRES 1948 | lw TMP0, L->top 1949 | lw CFUNC:RB, FRAME_FUNC(BASE) 1950 | jr ra 1951 |. subu NARGS8:RC, TMP0, BASE 1952 | 1953 |//----------------------------------------------------------------------- 1954 |//-- Special dispatch targets ------------------------------------------- 1955 |//----------------------------------------------------------------------- 1956 | 1957 |->vm_record: // Dispatch target for recording phase. 1958 |.if JIT 1959 | lbu TMP3, DISPATCH_GL(hookmask)(DISPATCH) 1960 | andi AT, TMP3, HOOK_VMEVENT // No recording while in vmevent. 1961 | bnez AT, >5 1962 | // Decrement the hookcount for consistency, but always do the call. 1963 |. lw TMP2, DISPATCH_GL(hookcount)(DISPATCH) 1964 | andi AT, TMP3, HOOK_ACTIVE 1965 | bnez AT, >1 1966 |. addiu TMP2, TMP2, -1 1967 | andi AT, TMP3, LUA_MASKLINE|LUA_MASKCOUNT 1968 | beqz AT, >1 1969 |. nop 1970 | b >1 1971 |. sw TMP2, DISPATCH_GL(hookcount)(DISPATCH) 1972 |.endif 1973 | 1974 |->vm_rethook: // Dispatch target for return hooks. 1975 | lbu TMP3, DISPATCH_GL(hookmask)(DISPATCH) 1976 | andi AT, TMP3, HOOK_ACTIVE // Hook already active? 1977 | beqz AT, >1 1978 |5: // Re-dispatch to static ins. 1979 |. lw AT, GG_DISP2STATIC(TMP0) // Assumes TMP0 holds DISPATCH+OP*4. 1980 | jr AT 1981 |. nop 1982 | 1983 |->vm_inshook: // Dispatch target for instr/line hooks. 1984 | lbu TMP3, DISPATCH_GL(hookmask)(DISPATCH) 1985 | lw TMP2, DISPATCH_GL(hookcount)(DISPATCH) 1986 | andi AT, TMP3, HOOK_ACTIVE // Hook already active? 1987 | bnez AT, <5 1988 |. andi AT, TMP3, LUA_MASKLINE|LUA_MASKCOUNT 1989 | beqz AT, <5 1990 |. addiu TMP2, TMP2, -1 1991 | beqz TMP2, >1 1992 |. sw TMP2, DISPATCH_GL(hookcount)(DISPATCH) 1993 | andi AT, TMP3, LUA_MASKLINE 1994 | beqz AT, <5 1995 |1: 1996 |. load_got lj_dispatch_ins 1997 | sw MULTRES, SAVE_MULTRES 1998 | move CARG2, PC 1999 | sw BASE, L->base 2000 | // SAVE_PC must hold the _previous_ PC. The callee updates it with PC. 2001 | call_intern lj_dispatch_ins // (lua_State *L, const BCIns *pc) 2002 |. move CARG1, L 2003 |3: 2004 | lw BASE, L->base 2005 |4: // Re-dispatch to static ins. 2006 | lw INS, -4(PC) 2007 | decode_OP4a TMP1, INS 2008 | decode_OP4b TMP1 2009 | addu TMP0, DISPATCH, TMP1 2010 | decode_RD8a RD, INS 2011 | lw AT, GG_DISP2STATIC(TMP0) 2012 | decode_RA8a RA, INS 2013 | decode_RD8b RD 2014 | jr AT 2015 | decode_RA8b RA 2016 | 2017 |->cont_hook: // Continue from hook yield. 2018 | addiu PC, PC, 4 2019 | b <4 2020 |. lw MULTRES, -24+LO(RB) // Restore MULTRES for *M ins. 2021 | 2022 |->vm_hotloop: // Hot loop counter underflow. 2023 |.if JIT 2024 | lw LFUNC:TMP1, FRAME_FUNC(BASE) 2025 | addiu CARG1, DISPATCH, GG_DISP2J 2026 | sw PC, SAVE_PC 2027 | lw TMP1, LFUNC:TMP1->pc 2028 | move CARG2, PC 2029 | sw L, DISPATCH_J(L)(DISPATCH) 2030 | lbu TMP1, PC2PROTO(framesize)(TMP1) 2031 | load_got lj_trace_hot 2032 | sw BASE, L->base 2033 | sll TMP1, TMP1, 3 2034 | addu TMP1, BASE, TMP1 2035 | call_intern lj_trace_hot // (jit_State *J, const BCIns *pc) 2036 |. sw TMP1, L->top 2037 | b <3 2038 |. nop 2039 |.endif 2040 | 2041 |->vm_callhook: // Dispatch target for call hooks. 2042 |.if JIT 2043 | b >1 2044 |.endif 2045 |. move CARG2, PC 2046 | 2047 |->vm_hotcall: // Hot call counter underflow. 2048 |.if JIT 2049 | ori CARG2, PC, 1 2050 |1: 2051 |.endif 2052 | load_got lj_dispatch_call 2053 | addu TMP0, BASE, RC 2054 | sw PC, SAVE_PC 2055 | sw BASE, L->base 2056 | subu RA, RA, BASE 2057 | sw TMP0, L->top 2058 | call_intern lj_dispatch_call // (lua_State *L, const BCIns *pc) 2059 |. move CARG1, L 2060 | // Returns ASMFunction. 2061 | lw BASE, L->base 2062 | lw TMP0, L->top 2063 | sw r0, SAVE_PC // Invalidate for subsequent line hook. 2064 | subu NARGS8:RC, TMP0, BASE 2065 | addu RA, BASE, RA 2066 | lw LFUNC:RB, FRAME_FUNC(BASE) 2067 | jr CRET1 2068 |. lw INS, -4(PC) 2069 | 2070 |//----------------------------------------------------------------------- 2071 |//-- Trace exit handler ------------------------------------------------- 2072 |//----------------------------------------------------------------------- 2073 | 2074 |.macro savex_, a, b 2075 | sdc1 f..a, 16+a*8(sp) 2076 | sw r..a, 16+32*8+a*4(sp) 2077 | sw r..b, 16+32*8+b*4(sp) 2078 |.endmacro 2079 | 2080 |->vm_exit_handler: 2081 |.if JIT 2082 | addiu sp, sp, -(16+32*8+32*4) 2083 | savex_ 0, 1 2084 | savex_ 2, 3 2085 | savex_ 4, 5 2086 | savex_ 6, 7 2087 | savex_ 8, 9 2088 | savex_ 10, 11 2089 | savex_ 12, 13 2090 | savex_ 14, 15 2091 | savex_ 16, 17 2092 | savex_ 18, 19 2093 | savex_ 20, 21 2094 | savex_ 22, 23 2095 | savex_ 24, 25 2096 | savex_ 26, 27 2097 | sdc1 f28, 16+28*8(sp) 2098 | sw r28, 16+32*8+28*4(sp) 2099 | sdc1 f30, 16+30*8(sp) 2100 | sw r30, 16+32*8+30*4(sp) 2101 | sw r0, 16+32*8+31*4(sp) // Clear RID_TMP. 2102 | li_vmstate EXIT 2103 | addiu TMP2, sp, 16+32*8+32*4 // Recompute original value of sp. 2104 | addiu DISPATCH, JGL, -GG_DISP2G-32768 2105 | lw TMP1, 0(TMP2) // Load exit number. 2106 | st_vmstate 2107 | sw TMP2, 16+32*8+29*4(sp) // Store sp in RID_SP. 2108 | lw L, DISPATCH_GL(jit_L)(DISPATCH) 2109 | lw BASE, DISPATCH_GL(jit_base)(DISPATCH) 2110 | load_got lj_trace_exit 2111 | sw L, DISPATCH_J(L)(DISPATCH) 2112 | sw ra, DISPATCH_J(parent)(DISPATCH) // Store trace number. 2113 | sw TMP1, DISPATCH_J(exitno)(DISPATCH) // Store exit number. 2114 | addiu CARG1, DISPATCH, GG_DISP2J 2115 | sw BASE, L->base 2116 | call_intern lj_trace_exit // (jit_State *J, ExitState *ex) 2117 |. addiu CARG2, sp, 16 2118 | // Returns MULTRES (unscaled) or negated error code. 2119 | lw TMP1, L->cframe 2120 | li AT, -4 2121 | lw BASE, L->base 2122 | and sp, TMP1, AT 2123 | lw PC, SAVE_PC // Get SAVE_PC. 2124 | b >1 2125 |. sw L, SAVE_L // Set SAVE_L (on-trace resume/yield). 2126 |.endif 2127 |->vm_exit_interp: 2128 |.if JIT 2129 | // CRET1 = MULTRES or negated error code, BASE, PC and JGL set. 2130 | lw L, SAVE_L 2131 | addiu DISPATCH, JGL, -GG_DISP2G-32768 2132 |1: 2133 | bltz CRET1, >3 // Check for error from exit. 2134 |. lw LFUNC:TMP1, FRAME_FUNC(BASE) 2135 | lui TMP3, 0x59c0 // TOBIT = 2^52 + 2^51 (float). 2136 | sll MULTRES, CRET1, 3 2137 | li TISNIL, LJ_TNIL 2138 | sw MULTRES, SAVE_MULTRES 2139 | mtc1 TMP3, TOBIT 2140 | lw TMP1, LFUNC:TMP1->pc 2141 | sw r0, DISPATCH_GL(jit_L)(DISPATCH) 2142 | lw KBASE, PC2PROTO(k)(TMP1) 2143 | cvt.d.s TOBIT, TOBIT 2144 | // Modified copy of ins_next which handles function header dispatch, too. 2145 | lw INS, 0(PC) 2146 | addiu PC, PC, 4 2147 | // Assumes TISNIL == ~LJ_VMST_INTERP == -1 2148 | sw TISNIL, DISPATCH_GL(vmstate)(DISPATCH) 2149 | decode_OP4a TMP1, INS 2150 | decode_OP4b TMP1 2151 | sltiu TMP2, TMP1, BC_FUNCF*4 // Function header? 2152 | addu TMP0, DISPATCH, TMP1 2153 | decode_RD8a RD, INS 2154 | lw AT, 0(TMP0) 2155 | decode_RA8a RA, INS 2156 | beqz TMP2, >2 2157 |. decode_RA8b RA 2158 | jr AT 2159 |. decode_RD8b RD 2160 |2: 2161 | addiu RC, MULTRES, -8 2162 | jr AT 2163 |. addu RA, RA, BASE 2164 | 2165 |3: // Rethrow error from the right C frame. 2166 | load_got lj_err_throw 2167 | negu CARG2, CRET1 2168 | call_intern lj_err_throw // (lua_State *L, int errcode) 2169 |. move CARG1, L 2170 |.endif 2171 | 2172 |//----------------------------------------------------------------------- 2173 |//-- Math helper functions ---------------------------------------------- 2174 |//----------------------------------------------------------------------- 2175 | 2176 |// Modifies AT, TMP0, FRET1, FRET2, f4. Keeps all others incl. FARG1. 2177 |.macro vm_round, func 2178 | lui TMP0, 0x4330 // Hiword of 2^52 (double). 2179 | mtc1 r0, f4 2180 | mtc1 TMP0, f5 2181 | abs.d FRET2, FARG1 // |x| 2182 | mfc1 AT, f13 2183 | c.olt.d 0, FRET2, f4 2184 | add.d FRET1, FRET2, f4 // (|x| + 2^52) - 2^52 2185 | bc1f 0, >1 // Truncate only if |x| < 2^52. 2186 |. sub.d FRET1, FRET1, f4 2187 | slt AT, AT, r0 2188 |.if "func" == "ceil" 2189 | lui TMP0, 0xbff0 // Hiword of -1 (double). Preserves -0. 2190 |.else 2191 | lui TMP0, 0x3ff0 // Hiword of +1 (double). 2192 |.endif 2193 |.if "func" == "trunc" 2194 | mtc1 TMP0, f5 2195 | c.olt.d 0, FRET2, FRET1 // |x| < result? 2196 | sub.d FRET2, FRET1, f4 2197 | movt.d FRET1, FRET2, 0 // If yes, subtract +1. 2198 | neg.d FRET2, FRET1 2199 | jr ra 2200 |. movn.d FRET1, FRET2, AT // Merge sign bit back in. 2201 |.else 2202 | neg.d FRET2, FRET1 2203 | mtc1 TMP0, f5 2204 | movn.d FRET1, FRET2, AT // Merge sign bit back in. 2205 |.if "func" == "ceil" 2206 | c.olt.d 0, FRET1, FARG1 // x > result? 2207 |.else 2208 | c.olt.d 0, FARG1, FRET1 // x < result? 2209 |.endif 2210 | sub.d FRET2, FRET1, f4 // If yes, subtract +-1. 2211 | jr ra 2212 |. movt.d FRET1, FRET2, 0 2213 |.endif 2214 |1: 2215 | jr ra 2216 |. mov.d FRET1, FARG1 2217 |.endmacro 2218 | 2219 |->vm_floor: 2220 | vm_round floor 2221 |->vm_ceil: 2222 | vm_round ceil 2223 |->vm_trunc: 2224 |.if JIT 2225 | vm_round trunc 2226 |.endif 2227 | 2228 |//----------------------------------------------------------------------- 2229 |//-- Miscellaneous functions -------------------------------------------- 2230 |//----------------------------------------------------------------------- 2231 | 2232 |//----------------------------------------------------------------------- 2233 |//-- FFI helper functions ----------------------------------------------- 2234 |//----------------------------------------------------------------------- 2235 | 2236 |// Handler for callback functions. Callback slot number in r1, g in r2. 2237 |->vm_ffi_callback: 2238 |.if FFI 2239 |.type CTSTATE, CTState, PC 2240 | saveregs 2241 | lw CTSTATE, GL:r2->ctype_state 2242 | addiu DISPATCH, r2, GG_G2DISP 2243 | load_got lj_ccallback_enter 2244 | sw r1, CTSTATE->cb.slot 2245 | sw CARG1, CTSTATE->cb.gpr[0] 2246 | sw CARG2, CTSTATE->cb.gpr[1] 2247 | sdc1 FARG1, CTSTATE->cb.fpr[0] 2248 | sw CARG3, CTSTATE->cb.gpr[2] 2249 | sw CARG4, CTSTATE->cb.gpr[3] 2250 | sdc1 FARG2, CTSTATE->cb.fpr[1] 2251 | addiu TMP0, sp, CFRAME_SPACE+16 2252 | sw TMP0, CTSTATE->cb.stack 2253 | sw r0, SAVE_PC // Any value outside of bytecode is ok. 2254 | move CARG2, sp 2255 | call_intern lj_ccallback_enter // (CTState *cts, void *cf) 2256 |. move CARG1, CTSTATE 2257 | // Returns lua_State *. 2258 | lw BASE, L:CRET1->base 2259 | lw RC, L:CRET1->top 2260 | move L, CRET1 2261 | lui TMP3, 0x59c0 // TOBIT = 2^52 + 2^51 (float). 2262 | lw LFUNC:RB, FRAME_FUNC(BASE) 2263 | mtc1 TMP3, TOBIT 2264 | li_vmstate INTERP 2265 | li TISNIL, LJ_TNIL 2266 | subu RC, RC, BASE 2267 | st_vmstate 2268 | cvt.d.s TOBIT, TOBIT 2269 | ins_callt 2270 |.endif 2271 | 2272 |->cont_ffi_callback: // Return from FFI callback. 2273 |.if FFI 2274 | load_got lj_ccallback_leave 2275 | lw CTSTATE, DISPATCH_GL(ctype_state)(DISPATCH) 2276 | sw BASE, L->base 2277 | sw RB, L->top 2278 | sw L, CTSTATE->L 2279 | move CARG2, RA 2280 | call_intern lj_ccallback_leave // (CTState *cts, TValue *o) 2281 |. move CARG1, CTSTATE 2282 | lw CRET1, CTSTATE->cb.gpr[0] 2283 | ldc1 FRET1, CTSTATE->cb.fpr[0] 2284 | lw CRET2, CTSTATE->cb.gpr[1] 2285 | b ->vm_leave_unw 2286 |. ldc1 FRET2, CTSTATE->cb.fpr[1] 2287 |.endif 2288 | 2289 |->vm_ffi_call: // Call C function via FFI. 2290 | // Caveat: needs special frame unwinding, see below. 2291 |.if FFI 2292 | .type CCSTATE, CCallState, CARG1 2293 | lw TMP1, CCSTATE->spadj 2294 | lbu CARG2, CCSTATE->nsp 2295 | move TMP2, sp 2296 | subu sp, sp, TMP1 2297 | sw ra, -4(TMP2) 2298 | sll CARG2, CARG2, 2 2299 | sw r16, -8(TMP2) 2300 | sw CCSTATE, -12(TMP2) 2301 | move r16, TMP2 2302 | addiu TMP1, CCSTATE, offsetof(CCallState, stack) 2303 | addiu TMP2, sp, 16 2304 | beqz CARG2, >2 2305 |. addu TMP3, TMP1, CARG2 2306 |1: 2307 | lw TMP0, 0(TMP1) 2308 | addiu TMP1, TMP1, 4 2309 | sltu AT, TMP1, TMP3 2310 | sw TMP0, 0(TMP2) 2311 | bnez AT, <1 2312 |. addiu TMP2, TMP2, 4 2313 |2: 2314 | lw CFUNCADDR, CCSTATE->func 2315 | lw CARG2, CCSTATE->gpr[1] 2316 | lw CARG3, CCSTATE->gpr[2] 2317 | lw CARG4, CCSTATE->gpr[3] 2318 | ldc1 FARG1, CCSTATE->fpr[0] 2319 | ldc1 FARG2, CCSTATE->fpr[1] 2320 | jalr CFUNCADDR 2321 |. lw CARG1, CCSTATE->gpr[0] // Do this last, since CCSTATE is CARG1. 2322 | lw CCSTATE:TMP1, -12(r16) 2323 | lw TMP2, -8(r16) 2324 | lw ra, -4(r16) 2325 | sw CRET1, CCSTATE:TMP1->gpr[0] 2326 | sw CRET2, CCSTATE:TMP1->gpr[1] 2327 | sdc1 FRET1, CCSTATE:TMP1->fpr[0] 2328 | sdc1 FRET2, CCSTATE:TMP1->fpr[1] 2329 | move sp, r16 2330 | jr ra 2331 |. move r16, TMP2 2332 |.endif 2333 |// Note: vm_ffi_call must be the last function in this object file! 2334 | 2335 |//----------------------------------------------------------------------- 2336} 2337 2338/* Generate the code for a single instruction. */ 2339static void build_ins(BuildCtx *ctx, BCOp op, int defop) 2340{ 2341 int vk = 0; 2342 |=>defop: 2343 2344 switch (op) { 2345 2346 /* -- Comparison ops ---------------------------------------------------- */ 2347 2348 /* Remember: all ops branch for a true comparison, fall through otherwise. */ 2349 2350 case BC_ISLT: case BC_ISGE: case BC_ISLE: case BC_ISGT: 2351 | // RA = src1*8, RD = src2*8, JMP with RD = target 2352 | addu CARG2, BASE, RA 2353 | addu CARG3, BASE, RD 2354 | lw TMP0, HI(CARG2) 2355 | lw TMP1, HI(CARG3) 2356 | ldc1 f0, 0(CARG2) 2357 | ldc1 f2, 0(CARG3) 2358 | sltiu TMP0, TMP0, LJ_TISNUM 2359 | sltiu TMP1, TMP1, LJ_TISNUM 2360 | lhu TMP2, OFS_RD(PC) 2361 | and TMP0, TMP0, TMP1 2362 | addiu PC, PC, 4 2363 | beqz TMP0, ->vmeta_comp 2364 |. lui TMP1, (-(BCBIAS_J*4 >> 16) & 65535) 2365 | decode_RD4b TMP2 2366 | addu TMP2, TMP2, TMP1 2367 if (op == BC_ISLT || op == BC_ISGE) { 2368 | c.olt.d f0, f2 2369 } else { 2370 | c.ole.d f0, f2 2371 } 2372 if (op == BC_ISLT || op == BC_ISLE) { 2373 | movf TMP2, r0 2374 } else { 2375 | movt TMP2, r0 2376 } 2377 | addu PC, PC, TMP2 2378 |1: 2379 | ins_next 2380 break; 2381 2382 case BC_ISEQV: case BC_ISNEV: 2383 vk = op == BC_ISEQV; 2384 | // RA = src1*8, RD = src2*8, JMP with RD = target 2385 | addu RA, BASE, RA 2386 | addiu PC, PC, 4 2387 | lw TMP0, HI(RA) 2388 | ldc1 f0, 0(RA) 2389 | addu RD, BASE, RD 2390 | lhu TMP2, -4+OFS_RD(PC) 2391 | lw TMP1, HI(RD) 2392 | ldc1 f2, 0(RD) 2393 | lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535) 2394 | sltiu AT, TMP0, LJ_TISNUM 2395 | sltiu CARG1, TMP1, LJ_TISNUM 2396 | decode_RD4b TMP2 2397 | and AT, AT, CARG1 2398 | beqz AT, >5 2399 |. addu TMP2, TMP2, TMP3 2400 | c.eq.d f0, f2 2401 if (vk) { 2402 | movf TMP2, r0 2403 } else { 2404 | movt TMP2, r0 2405 } 2406 |1: 2407 | addu PC, PC, TMP2 2408 | ins_next 2409 |5: // Either or both types are not numbers. 2410 | lw CARG2, LO(RA) 2411 | lw CARG3, LO(RD) 2412 |.if FFI 2413 | li TMP3, LJ_TCDATA 2414 | beq TMP0, TMP3, ->vmeta_equal_cd 2415 |.endif 2416 |. sltiu AT, TMP0, LJ_TISPRI // Not a primitive? 2417 |.if FFI 2418 | beq TMP1, TMP3, ->vmeta_equal_cd 2419 |.endif 2420 |. xor TMP3, CARG2, CARG3 // Same tv? 2421 | xor TMP1, TMP1, TMP0 // Same type? 2422 | sltiu CARG1, TMP0, LJ_TISTABUD+1 // Table or userdata? 2423 | movz TMP3, r0, AT // Ignore tv if primitive. 2424 | movn CARG1, r0, TMP1 // Tab/ud and same type? 2425 | or AT, TMP1, TMP3 // Same type && (pri||same tv). 2426 | movz CARG1, r0, AT 2427 | beqz CARG1, <1 // Done if not tab/ud or not same type or same tv. 2428 if (vk) { 2429 |. movn TMP2, r0, AT 2430 } else { 2431 |. movz TMP2, r0, AT 2432 } 2433 | // Different tables or userdatas. Need to check __eq metamethod. 2434 | // Field metatable must be at same offset for GCtab and GCudata! 2435 | lw TAB:TMP1, TAB:CARG2->metatable 2436 | beqz TAB:TMP1, <1 // No metatable? 2437 |. nop 2438 | lbu TMP1, TAB:TMP1->nomm 2439 | andi TMP1, TMP1, 1<<MM_eq 2440 | bnez TMP1, <1 // Or 'no __eq' flag set? 2441 |. nop 2442 | b ->vmeta_equal // Handle __eq metamethod. 2443 |. li CARG4, 1-vk // ne = 0 or 1. 2444 break; 2445 2446 case BC_ISEQS: case BC_ISNES: 2447 vk = op == BC_ISEQS; 2448 | // RA = src*8, RD = str_const*8 (~), JMP with RD = target 2449 | addu RA, BASE, RA 2450 | addiu PC, PC, 4 2451 | lw TMP0, HI(RA) 2452 | srl RD, RD, 1 2453 | lw STR:TMP3, LO(RA) 2454 | subu RD, KBASE, RD 2455 | lhu TMP2, -4+OFS_RD(PC) 2456 |.if FFI 2457 | li AT, LJ_TCDATA 2458 | beq TMP0, AT, ->vmeta_equal_cd 2459 |.endif 2460 |. lw STR:TMP1, -4(RD) // KBASE-4-str_const*4 2461 | addiu TMP0, TMP0, -LJ_TSTR 2462 | decode_RD4b TMP2 2463 | xor TMP1, STR:TMP1, STR:TMP3 2464 | or TMP0, TMP0, TMP1 2465 | lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535) 2466 | addu TMP2, TMP2, TMP3 2467 if (vk) { 2468 | movn TMP2, r0, TMP0 2469 } else { 2470 | movz TMP2, r0, TMP0 2471 } 2472 | addu PC, PC, TMP2 2473 | ins_next 2474 break; 2475 2476 case BC_ISEQN: case BC_ISNEN: 2477 vk = op == BC_ISEQN; 2478 | // RA = src*8, RD = num_const*8, JMP with RD = target 2479 | addu RA, BASE, RA 2480 | addiu PC, PC, 4 2481 | lw TMP0, HI(RA) 2482 | ldc1 f0, 0(RA) 2483 | addu RD, KBASE, RD 2484 | lhu TMP2, -4+OFS_RD(PC) 2485 | ldc1 f2, 0(RD) 2486 | lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535) 2487 | sltiu AT, TMP0, LJ_TISNUM 2488 | decode_RD4b TMP2 2489 |.if FFI 2490 | beqz AT, >5 2491 |.else 2492 | beqz AT, >1 2493 |.endif 2494 |. addu TMP2, TMP2, TMP3 2495 | c.eq.d f0, f2 2496 if (vk) { 2497 | movf TMP2, r0 2498 | addu PC, PC, TMP2 2499 |1: 2500 } else { 2501 | movt TMP2, r0 2502 |1: 2503 | addu PC, PC, TMP2 2504 } 2505 | ins_next 2506 |.if FFI 2507 |5: 2508 | li AT, LJ_TCDATA 2509 | beq TMP0, AT, ->vmeta_equal_cd 2510 |. nop 2511 | b <1 2512 |. nop 2513 |.endif 2514 break; 2515 2516 case BC_ISEQP: case BC_ISNEP: 2517 vk = op == BC_ISEQP; 2518 | // RA = src*8, RD = primitive_type*8 (~), JMP with RD = target 2519 | addu RA, BASE, RA 2520 | srl TMP1, RD, 3 2521 | lw TMP0, HI(RA) 2522 | lhu TMP2, OFS_RD(PC) 2523 | not TMP1, TMP1 2524 | addiu PC, PC, 4 2525 |.if FFI 2526 | li AT, LJ_TCDATA 2527 | beq TMP0, AT, ->vmeta_equal_cd 2528 |.endif 2529 |. xor TMP0, TMP0, TMP1 2530 | decode_RD4b TMP2 2531 | lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535) 2532 | addu TMP2, TMP2, TMP3 2533 if (vk) { 2534 | movn TMP2, r0, TMP0 2535 } else { 2536 | movz TMP2, r0, TMP0 2537 } 2538 | addu PC, PC, TMP2 2539 | ins_next 2540 break; 2541 2542 /* -- Unary test and copy ops ------------------------------------------- */ 2543 2544 case BC_ISTC: case BC_ISFC: case BC_IST: case BC_ISF: 2545 | // RA = dst*8 or unused, RD = src*8, JMP with RD = target 2546 | addu RD, BASE, RD 2547 | lhu TMP2, OFS_RD(PC) 2548 | lw TMP0, HI(RD) 2549 | addiu PC, PC, 4 2550 if (op == BC_IST || op == BC_ISF) { 2551 | sltiu TMP0, TMP0, LJ_TISTRUECOND 2552 | decode_RD4b TMP2 2553 | lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535) 2554 | addu TMP2, TMP2, TMP3 2555 if (op == BC_IST) { 2556 | movz TMP2, r0, TMP0 2557 } else { 2558 | movn TMP2, r0, TMP0 2559 } 2560 | addu PC, PC, TMP2 2561 } else { 2562 | sltiu TMP0, TMP0, LJ_TISTRUECOND 2563 | ldc1 f0, 0(RD) 2564 if (op == BC_ISTC) { 2565 | beqz TMP0, >1 2566 } else { 2567 | bnez TMP0, >1 2568 } 2569 |. addu RA, BASE, RA 2570 | decode_RD4b TMP2 2571 | lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535) 2572 | addu TMP2, TMP2, TMP3 2573 | sdc1 f0, 0(RA) 2574 | addu PC, PC, TMP2 2575 |1: 2576 } 2577 | ins_next 2578 break; 2579 2580 /* -- Unary ops --------------------------------------------------------- */ 2581 2582 case BC_MOV: 2583 | // RA = dst*8, RD = src*8 2584 | addu RD, BASE, RD 2585 | addu RA, BASE, RA 2586 | ldc1 f0, 0(RD) 2587 | ins_next1 2588 | sdc1 f0, 0(RA) 2589 | ins_next2 2590 break; 2591 case BC_NOT: 2592 | // RA = dst*8, RD = src*8 2593 | addu RD, BASE, RD 2594 | addu RA, BASE, RA 2595 | lw TMP0, HI(RD) 2596 | li TMP1, LJ_TFALSE 2597 | sltiu TMP0, TMP0, LJ_TISTRUECOND 2598 | addiu TMP1, TMP0, LJ_TTRUE 2599 | ins_next1 2600 | sw TMP1, HI(RA) 2601 | ins_next2 2602 break; 2603 case BC_UNM: 2604 | // RA = dst*8, RD = src*8 2605 | addu CARG3, BASE, RD 2606 | addu RA, BASE, RA 2607 | lw TMP0, HI(CARG3) 2608 | ldc1 f0, 0(CARG3) 2609 | sltiu AT, TMP0, LJ_TISNUM 2610 | beqz AT, ->vmeta_unm 2611 |. neg.d f0, f0 2612 | ins_next1 2613 | sdc1 f0, 0(RA) 2614 | ins_next2 2615 break; 2616 case BC_LEN: 2617 | // RA = dst*8, RD = src*8 2618 | addu CARG2, BASE, RD 2619 | addu RA, BASE, RA 2620 | lw TMP0, HI(CARG2) 2621 | lw CARG1, LO(CARG2) 2622 | li AT, LJ_TSTR 2623 | bne TMP0, AT, >2 2624 |. li AT, LJ_TTAB 2625 | lw CRET1, STR:CARG1->len 2626 |1: 2627 | mtc1 CRET1, f0 2628 | cvt.d.w f0, f0 2629 | ins_next1 2630 | sdc1 f0, 0(RA) 2631 | ins_next2 2632 |2: 2633 | bne TMP0, AT, ->vmeta_len 2634 |. nop 2635#if LJ_52 2636 | lw TAB:TMP2, TAB:CARG1->metatable 2637 | bnez TAB:TMP2, >9 2638 |. nop 2639 |3: 2640#endif 2641 |->BC_LEN_Z: 2642 | load_got lj_tab_len 2643 | call_intern lj_tab_len // (GCtab *t) 2644 |. nop 2645 | // Returns uint32_t (but less than 2^31). 2646 | b <1 2647 |. nop 2648#if LJ_52 2649 |9: 2650 | lbu TMP0, TAB:TMP2->nomm 2651 | andi TMP0, TMP0, 1<<MM_len 2652 | bnez TMP0, <3 // 'no __len' flag set: done. 2653 |. nop 2654 | b ->vmeta_len 2655 |. nop 2656#endif 2657 break; 2658 2659 /* -- Binary ops -------------------------------------------------------- */ 2660 2661 |.macro ins_arithpre 2662 ||vk = ((int)op - BC_ADDVN) / (BC_ADDNV-BC_ADDVN); 2663 | decode_RB8a RB, INS 2664 | decode_RB8b RB 2665 | decode_RDtoRC8 RC, RD 2666 | // RA = dst*8, RB = src1*8, RC = src2*8 | num_const*8 2667 ||switch (vk) { 2668 ||case 0: 2669 | addu CARG3, BASE, RB 2670 | addu CARG4, KBASE, RC 2671 | lw TMP1, HI(CARG3) 2672 | ldc1 f20, 0(CARG3) 2673 | ldc1 f22, 0(CARG4) 2674 | sltiu AT, TMP1, LJ_TISNUM 2675 || break; 2676 ||case 1: 2677 | addu CARG4, BASE, RB 2678 | addu CARG3, KBASE, RC 2679 | lw TMP1, HI(CARG4) 2680 | ldc1 f22, 0(CARG4) 2681 | ldc1 f20, 0(CARG3) 2682 | sltiu AT, TMP1, LJ_TISNUM 2683 || break; 2684 ||default: 2685 | addu CARG3, BASE, RB 2686 | addu CARG4, BASE, RC 2687 | lw TMP1, HI(CARG3) 2688 | lw TMP2, HI(CARG4) 2689 | ldc1 f20, 0(CARG3) 2690 | ldc1 f22, 0(CARG4) 2691 | sltiu AT, TMP1, LJ_TISNUM 2692 | sltiu TMP0, TMP2, LJ_TISNUM 2693 | and AT, AT, TMP0 2694 || break; 2695 ||} 2696 | beqz AT, ->vmeta_arith 2697 |. addu RA, BASE, RA 2698 |.endmacro 2699 | 2700 |.macro fpmod, a, b, c 2701 |->BC_MODVN_Z: 2702 | bal ->vm_floor // floor(b/c) 2703 |. div.d FARG1, b, c 2704 | mul.d a, FRET1, c 2705 | sub.d a, b, a // b - floor(b/c)*c 2706 |.endmacro 2707 | 2708 |.macro ins_arith, ins 2709 | ins_arithpre 2710 |.if "ins" == "fpmod_" 2711 | b ->BC_MODVN_Z // Avoid 3 copies. It's slow anyway. 2712 |. nop 2713 |.else 2714 | ins f0, f20, f22 2715 | ins_next1 2716 | sdc1 f0, 0(RA) 2717 | ins_next2 2718 |.endif 2719 |.endmacro 2720 2721 case BC_ADDVN: case BC_ADDNV: case BC_ADDVV: 2722 | ins_arith add.d 2723 break; 2724 case BC_SUBVN: case BC_SUBNV: case BC_SUBVV: 2725 | ins_arith sub.d 2726 break; 2727 case BC_MULVN: case BC_MULNV: case BC_MULVV: 2728 | ins_arith mul.d 2729 break; 2730 case BC_DIVVN: case BC_DIVNV: case BC_DIVVV: 2731 | ins_arith div.d 2732 break; 2733 case BC_MODVN: 2734 | ins_arith fpmod 2735 break; 2736 case BC_MODNV: case BC_MODVV: 2737 | ins_arith fpmod_ 2738 break; 2739 case BC_POW: 2740 | decode_RB8a RB, INS 2741 | decode_RB8b RB 2742 | decode_RDtoRC8 RC, RD 2743 | addu CARG3, BASE, RB 2744 | addu CARG4, BASE, RC 2745 | lw TMP1, HI(CARG3) 2746 | lw TMP2, HI(CARG4) 2747 | ldc1 FARG1, 0(CARG3) 2748 | ldc1 FARG2, 0(CARG4) 2749 | sltiu AT, TMP1, LJ_TISNUM 2750 | sltiu TMP0, TMP2, LJ_TISNUM 2751 | and AT, AT, TMP0 2752 | load_got pow 2753 | beqz AT, ->vmeta_arith 2754 |. addu RA, BASE, RA 2755 | call_extern 2756 |. nop 2757 | ins_next1 2758 | sdc1 FRET1, 0(RA) 2759 | ins_next2 2760 break; 2761 2762 case BC_CAT: 2763 | // RA = dst*8, RB = src_start*8, RC = src_end*8 2764 | decode_RB8a RB, INS 2765 | decode_RB8b RB 2766 | decode_RDtoRC8 RC, RD 2767 | subu CARG3, RC, RB 2768 | sw BASE, L->base 2769 | addu CARG2, BASE, RC 2770 | move MULTRES, RB 2771 |->BC_CAT_Z: 2772 | load_got lj_meta_cat 2773 | srl CARG3, CARG3, 3 2774 | sw PC, SAVE_PC 2775 | call_intern lj_meta_cat // (lua_State *L, TValue *top, int left) 2776 |. move CARG1, L 2777 | // Returns NULL (finished) or TValue * (metamethod). 2778 | bnez CRET1, ->vmeta_binop 2779 |. lw BASE, L->base 2780 | addu RB, BASE, MULTRES 2781 | ldc1 f0, 0(RB) 2782 | addu RA, BASE, RA 2783 | ins_next1 2784 | sdc1 f0, 0(RA) // Copy result from RB to RA. 2785 | ins_next2 2786 break; 2787 2788 /* -- Constant ops ------------------------------------------------------ */ 2789 2790 case BC_KSTR: 2791 | // RA = dst*8, RD = str_const*8 (~) 2792 | srl TMP1, RD, 1 2793 | subu TMP1, KBASE, TMP1 2794 | ins_next1 2795 | lw TMP0, -4(TMP1) // KBASE-4-str_const*4 2796 | addu RA, BASE, RA 2797 | li TMP2, LJ_TSTR 2798 | sw TMP0, LO(RA) 2799 | sw TMP2, HI(RA) 2800 | ins_next2 2801 break; 2802 case BC_KCDATA: 2803 |.if FFI 2804 | // RA = dst*8, RD = cdata_const*8 (~) 2805 | srl TMP1, RD, 1 2806 | subu TMP1, KBASE, TMP1 2807 | ins_next1 2808 | lw TMP0, -4(TMP1) // KBASE-4-cdata_const*4 2809 | addu RA, BASE, RA 2810 | li TMP2, LJ_TCDATA 2811 | sw TMP0, LO(RA) 2812 | sw TMP2, HI(RA) 2813 | ins_next2 2814 |.endif 2815 break; 2816 case BC_KSHORT: 2817 | // RA = dst*8, RD = int16_literal*8 2818 | sra RD, INS, 16 2819 | mtc1 RD, f0 2820 | addu RA, BASE, RA 2821 | cvt.d.w f0, f0 2822 | ins_next1 2823 | sdc1 f0, 0(RA) 2824 | ins_next2 2825 break; 2826 case BC_KNUM: 2827 | // RA = dst*8, RD = num_const*8 2828 | addu RD, KBASE, RD 2829 | addu RA, BASE, RA 2830 | ldc1 f0, 0(RD) 2831 | ins_next1 2832 | sdc1 f0, 0(RA) 2833 | ins_next2 2834 break; 2835 case BC_KPRI: 2836 | // RA = dst*8, RD = primitive_type*8 (~) 2837 | srl TMP1, RD, 3 2838 | addu RA, BASE, RA 2839 | not TMP0, TMP1 2840 | ins_next1 2841 | sw TMP0, HI(RA) 2842 | ins_next2 2843 break; 2844 case BC_KNIL: 2845 | // RA = base*8, RD = end*8 2846 | addu RA, BASE, RA 2847 | sw TISNIL, HI(RA) 2848 | addiu RA, RA, 8 2849 | addu RD, BASE, RD 2850 |1: 2851 | sw TISNIL, HI(RA) 2852 | slt AT, RA, RD 2853 | bnez AT, <1 2854 |. addiu RA, RA, 8 2855 | ins_next_ 2856 break; 2857 2858 /* -- Upvalue and function ops ------------------------------------------ */ 2859 2860 case BC_UGET: 2861 | // RA = dst*8, RD = uvnum*8 2862 | lw LFUNC:RB, FRAME_FUNC(BASE) 2863 | srl RD, RD, 1 2864 | addu RD, RD, LFUNC:RB 2865 | lw UPVAL:RB, LFUNC:RD->uvptr 2866 | ins_next1 2867 | lw TMP1, UPVAL:RB->v 2868 | ldc1 f0, 0(TMP1) 2869 | addu RA, BASE, RA 2870 | sdc1 f0, 0(RA) 2871 | ins_next2 2872 break; 2873 case BC_USETV: 2874 | // RA = uvnum*8, RD = src*8 2875 | lw LFUNC:RB, FRAME_FUNC(BASE) 2876 | srl RA, RA, 1 2877 | addu RD, BASE, RD 2878 | addu RA, RA, LFUNC:RB 2879 | ldc1 f0, 0(RD) 2880 | lw UPVAL:RB, LFUNC:RA->uvptr 2881 | lbu TMP3, UPVAL:RB->marked 2882 | lw CARG2, UPVAL:RB->v 2883 | andi TMP3, TMP3, LJ_GC_BLACK // isblack(uv) 2884 | lbu TMP0, UPVAL:RB->closed 2885 | lw TMP2, HI(RD) 2886 | sdc1 f0, 0(CARG2) 2887 | li AT, LJ_GC_BLACK|1 2888 | or TMP3, TMP3, TMP0 2889 | beq TMP3, AT, >2 // Upvalue is closed and black? 2890 |. addiu TMP2, TMP2, -(LJ_TNUMX+1) 2891 |1: 2892 | ins_next 2893 | 2894 |2: // Check if new value is collectable. 2895 | sltiu AT, TMP2, LJ_TISGCV - (LJ_TNUMX+1) 2896 | beqz AT, <1 // tvisgcv(v) 2897 |. lw TMP1, LO(RD) 2898 | lbu TMP3, GCOBJ:TMP1->gch.marked 2899 | andi TMP3, TMP3, LJ_GC_WHITES // iswhite(v) 2900 | beqz TMP3, <1 2901 |. load_got lj_gc_barrieruv 2902 | // Crossed a write barrier. Move the barrier forward. 2903 | call_intern lj_gc_barrieruv // (global_State *g, TValue *tv) 2904 |. addiu CARG1, DISPATCH, GG_DISP2G 2905 | b <1 2906 |. nop 2907 break; 2908 case BC_USETS: 2909 | // RA = uvnum*8, RD = str_const*8 (~) 2910 | lw LFUNC:RB, FRAME_FUNC(BASE) 2911 | srl RA, RA, 1 2912 | srl TMP1, RD, 1 2913 | addu RA, RA, LFUNC:RB 2914 | subu TMP1, KBASE, TMP1 2915 | lw UPVAL:RB, LFUNC:RA->uvptr 2916 | lw STR:TMP1, -4(TMP1) // KBASE-4-str_const*4 2917 | lbu TMP2, UPVAL:RB->marked 2918 | lw CARG2, UPVAL:RB->v 2919 | lbu TMP3, STR:TMP1->marked 2920 | andi AT, TMP2, LJ_GC_BLACK // isblack(uv) 2921 | lbu TMP2, UPVAL:RB->closed 2922 | li TMP0, LJ_TSTR 2923 | sw STR:TMP1, LO(CARG2) 2924 | bnez AT, >2 2925 |. sw TMP0, HI(CARG2) 2926 |1: 2927 | ins_next 2928 | 2929 |2: // Check if string is white and ensure upvalue is closed. 2930 | beqz TMP2, <1 2931 |. andi AT, TMP3, LJ_GC_WHITES // iswhite(str) 2932 | beqz AT, <1 2933 |. load_got lj_gc_barrieruv 2934 | // Crossed a write barrier. Move the barrier forward. 2935 | call_intern lj_gc_barrieruv // (global_State *g, TValue *tv) 2936 |. addiu CARG1, DISPATCH, GG_DISP2G 2937 | b <1 2938 |. nop 2939 break; 2940 case BC_USETN: 2941 | // RA = uvnum*8, RD = num_const*8 2942 | lw LFUNC:RB, FRAME_FUNC(BASE) 2943 | srl RA, RA, 1 2944 | addu RD, KBASE, RD 2945 | addu RA, RA, LFUNC:RB 2946 | ldc1 f0, 0(RD) 2947 | lw UPVAL:RB, LFUNC:RA->uvptr 2948 | ins_next1 2949 | lw TMP1, UPVAL:RB->v 2950 | sdc1 f0, 0(TMP1) 2951 | ins_next2 2952 break; 2953 case BC_USETP: 2954 | // RA = uvnum*8, RD = primitive_type*8 (~) 2955 | lw LFUNC:RB, FRAME_FUNC(BASE) 2956 | srl RA, RA, 1 2957 | srl TMP0, RD, 3 2958 | addu RA, RA, LFUNC:RB 2959 | not TMP0, TMP0 2960 | lw UPVAL:RB, LFUNC:RA->uvptr 2961 | ins_next1 2962 | lw TMP1, UPVAL:RB->v 2963 | sw TMP0, HI(TMP1) 2964 | ins_next2 2965 break; 2966 2967 case BC_UCLO: 2968 | // RA = level*8, RD = target 2969 | lw TMP2, L->openupval 2970 | branch_RD // Do this first since RD is not saved. 2971 | load_got lj_func_closeuv 2972 | sw BASE, L->base 2973 | beqz TMP2, >1 2974 |. move CARG1, L 2975 | call_intern lj_func_closeuv // (lua_State *L, TValue *level) 2976 |. addu CARG2, BASE, RA 2977 | lw BASE, L->base 2978 |1: 2979 | ins_next 2980 break; 2981 2982 case BC_FNEW: 2983 | // RA = dst*8, RD = proto_const*8 (~) (holding function prototype) 2984 | srl TMP1, RD, 1 2985 | load_got lj_func_newL_gc 2986 | subu TMP1, KBASE, TMP1 2987 | lw CARG3, FRAME_FUNC(BASE) 2988 | lw CARG2, -4(TMP1) // KBASE-4-tab_const*4 2989 | sw BASE, L->base 2990 | sw PC, SAVE_PC 2991 | // (lua_State *L, GCproto *pt, GCfuncL *parent) 2992 | call_intern lj_func_newL_gc 2993 |. move CARG1, L 2994 | // Returns GCfuncL *. 2995 | lw BASE, L->base 2996 | li TMP0, LJ_TFUNC 2997 | ins_next1 2998 | addu RA, BASE, RA 2999 | sw TMP0, HI(RA) 3000 | sw LFUNC:CRET1, LO(RA) 3001 | ins_next2 3002 break; 3003 3004 /* -- Table ops --------------------------------------------------------- */ 3005 3006 case BC_TNEW: 3007 case BC_TDUP: 3008 | // RA = dst*8, RD = (hbits|asize)*8 | tab_const*8 (~) 3009 | lw TMP0, DISPATCH_GL(gc.total)(DISPATCH) 3010 | lw TMP1, DISPATCH_GL(gc.threshold)(DISPATCH) 3011 | sw BASE, L->base 3012 | sw PC, SAVE_PC 3013 | sltu AT, TMP0, TMP1 3014 | beqz AT, >5 3015 |1: 3016 if (op == BC_TNEW) { 3017 | load_got lj_tab_new 3018 | srl CARG2, RD, 3 3019 | andi CARG2, CARG2, 0x7ff 3020 | li TMP0, 0x801 3021 | addiu AT, CARG2, -0x7ff 3022 | srl CARG3, RD, 14 3023 | movz CARG2, TMP0, AT 3024 | // (lua_State *L, int32_t asize, uint32_t hbits) 3025 | call_intern lj_tab_new 3026 |. move CARG1, L 3027 | // Returns Table *. 3028 } else { 3029 | load_got lj_tab_dup 3030 | srl TMP1, RD, 1 3031 | subu TMP1, KBASE, TMP1 3032 | move CARG1, L 3033 | call_intern lj_tab_dup // (lua_State *L, Table *kt) 3034 |. lw CARG2, -4(TMP1) // KBASE-4-str_const*4 3035 | // Returns Table *. 3036 } 3037 | lw BASE, L->base 3038 | ins_next1 3039 | addu RA, BASE, RA 3040 | li TMP0, LJ_TTAB 3041 | sw TAB:CRET1, LO(RA) 3042 | sw TMP0, HI(RA) 3043 | ins_next2 3044 |5: 3045 | load_got lj_gc_step_fixtop 3046 | move MULTRES, RD 3047 | call_intern lj_gc_step_fixtop // (lua_State *L) 3048 |. move CARG1, L 3049 | b <1 3050 |. move RD, MULTRES 3051 break; 3052 3053 case BC_GGET: 3054 | // RA = dst*8, RD = str_const*8 (~) 3055 case BC_GSET: 3056 | // RA = src*8, RD = str_const*8 (~) 3057 | lw LFUNC:TMP2, FRAME_FUNC(BASE) 3058 | srl TMP1, RD, 1 3059 | subu TMP1, KBASE, TMP1 3060 | lw TAB:RB, LFUNC:TMP2->env 3061 | lw STR:RC, -4(TMP1) // KBASE-4-str_const*4 3062 if (op == BC_GGET) { 3063 | b ->BC_TGETS_Z 3064 } else { 3065 | b ->BC_TSETS_Z 3066 } 3067 |. addu RA, BASE, RA 3068 break; 3069 3070 case BC_TGETV: 3071 | // RA = dst*8, RB = table*8, RC = key*8 3072 | decode_RB8a RB, INS 3073 | decode_RB8b RB 3074 | decode_RDtoRC8 RC, RD 3075 | addu CARG2, BASE, RB 3076 | addu CARG3, BASE, RC 3077 | lw TMP1, HI(CARG2) 3078 | lw TMP2, HI(CARG3) 3079 | lw TAB:RB, LO(CARG2) 3080 | li AT, LJ_TTAB 3081 | ldc1 f0, 0(CARG3) 3082 | bne TMP1, AT, ->vmeta_tgetv 3083 |. addu RA, BASE, RA 3084 | sltiu AT, TMP2, LJ_TISNUM 3085 | beqz AT, >5 3086 |. li AT, LJ_TSTR 3087 | 3088 | // Convert number key to integer, check for integerness and range. 3089 | cvt.w.d f2, f0 3090 | lw TMP0, TAB:RB->asize 3091 | mfc1 TMP2, f2 3092 | cvt.d.w f4, f2 3093 | lw TMP1, TAB:RB->array 3094 | c.eq.d f0, f4 3095 | sltu AT, TMP2, TMP0 3096 | movf AT, r0 3097 | sll TMP2, TMP2, 3 3098 | beqz AT, ->vmeta_tgetv // Integer key and in array part? 3099 |. addu TMP2, TMP1, TMP2 3100 | lw TMP0, HI(TMP2) 3101 | beq TMP0, TISNIL, >2 3102 |. ldc1 f0, 0(TMP2) 3103 |1: 3104 | ins_next1 3105 | sdc1 f0, 0(RA) 3106 | ins_next2 3107 | 3108 |2: // Check for __index if table value is nil. 3109 | lw TAB:TMP2, TAB:RB->metatable 3110 | beqz TAB:TMP2, <1 // No metatable: done. 3111 |. nop 3112 | lbu TMP0, TAB:TMP2->nomm 3113 | andi TMP0, TMP0, 1<<MM_index 3114 | bnez TMP0, <1 // 'no __index' flag set: done. 3115 |. nop 3116 | b ->vmeta_tgetv 3117 |. nop 3118 | 3119 |5: 3120 | bne TMP2, AT, ->vmeta_tgetv 3121 |. lw STR:RC, LO(CARG3) 3122 | b ->BC_TGETS_Z // String key? 3123 |. nop 3124 break; 3125 case BC_TGETS: 3126 | // RA = dst*8, RB = table*8, RC = str_const*4 (~) 3127 | decode_RB8a RB, INS 3128 | decode_RB8b RB 3129 | addu CARG2, BASE, RB 3130 | decode_RC4a RC, INS 3131 | lw TMP0, HI(CARG2) 3132 | decode_RC4b RC 3133 | li AT, LJ_TTAB 3134 | lw TAB:RB, LO(CARG2) 3135 | subu CARG3, KBASE, RC 3136 | lw STR:RC, -4(CARG3) // KBASE-4-str_const*4 3137 | bne TMP0, AT, ->vmeta_tgets1 3138 |. addu RA, BASE, RA 3139 |->BC_TGETS_Z: 3140 | // TAB:RB = GCtab *, STR:RC = GCstr *, RA = dst*8 3141 | lw TMP0, TAB:RB->hmask 3142 | lw TMP1, STR:RC->hash 3143 | lw NODE:TMP2, TAB:RB->node 3144 | and TMP1, TMP1, TMP0 // idx = str->hash & tab->hmask 3145 | sll TMP0, TMP1, 5 3146 | sll TMP1, TMP1, 3 3147 | subu TMP1, TMP0, TMP1 3148 | addu NODE:TMP2, NODE:TMP2, TMP1 // node = tab->node + (idx*32-idx*8) 3149 |1: 3150 | lw CARG1, offsetof(Node, key)+HI(NODE:TMP2) 3151 | lw TMP0, offsetof(Node, key)+LO(NODE:TMP2) 3152 | lw NODE:TMP1, NODE:TMP2->next 3153 | lw CARG2, offsetof(Node, val)+HI(NODE:TMP2) 3154 | addiu CARG1, CARG1, -LJ_TSTR 3155 | xor TMP0, TMP0, STR:RC 3156 | or AT, CARG1, TMP0 3157 | bnez AT, >4 3158 |. lw TAB:TMP3, TAB:RB->metatable 3159 | beq CARG2, TISNIL, >5 // Key found, but nil value? 3160 |. lw CARG1, offsetof(Node, val)+LO(NODE:TMP2) 3161 |3: 3162 | ins_next1 3163 | sw CARG2, HI(RA) 3164 | sw CARG1, LO(RA) 3165 | ins_next2 3166 | 3167 |4: // Follow hash chain. 3168 | bnez NODE:TMP1, <1 3169 |. move NODE:TMP2, NODE:TMP1 3170 | // End of hash chain: key not found, nil result. 3171 | 3172 |5: // Check for __index if table value is nil. 3173 | beqz TAB:TMP3, <3 // No metatable: done. 3174 |. li CARG2, LJ_TNIL 3175 | lbu TMP0, TAB:TMP3->nomm 3176 | andi TMP0, TMP0, 1<<MM_index 3177 | bnez TMP0, <3 // 'no __index' flag set: done. 3178 |. nop 3179 | b ->vmeta_tgets 3180 |. nop 3181 break; 3182 case BC_TGETB: 3183 | // RA = dst*8, RB = table*8, RC = index*8 3184 | decode_RB8a RB, INS 3185 | decode_RB8b RB 3186 | addu CARG2, BASE, RB 3187 | decode_RDtoRC8 RC, RD 3188 | lw CARG1, HI(CARG2) 3189 | li AT, LJ_TTAB 3190 | lw TAB:RB, LO(CARG2) 3191 | addu RA, BASE, RA 3192 | bne CARG1, AT, ->vmeta_tgetb 3193 |. srl TMP0, RC, 3 3194 | lw TMP1, TAB:RB->asize 3195 | lw TMP2, TAB:RB->array 3196 | sltu AT, TMP0, TMP1 3197 | beqz AT, ->vmeta_tgetb 3198 |. addu RC, TMP2, RC 3199 | lw TMP1, HI(RC) 3200 | beq TMP1, TISNIL, >5 3201 |. ldc1 f0, 0(RC) 3202 |1: 3203 | ins_next1 3204 | sdc1 f0, 0(RA) 3205 | ins_next2 3206 | 3207 |5: // Check for __index if table value is nil. 3208 | lw TAB:TMP2, TAB:RB->metatable 3209 | beqz TAB:TMP2, <1 // No metatable: done. 3210 |. nop 3211 | lbu TMP1, TAB:TMP2->nomm 3212 | andi TMP1, TMP1, 1<<MM_index 3213 | bnez TMP1, <1 // 'no __index' flag set: done. 3214 |. nop 3215 | b ->vmeta_tgetb // Caveat: preserve TMP0! 3216 |. nop 3217 break; 3218 3219 case BC_TSETV: 3220 | // RA = src*8, RB = table*8, RC = key*8 3221 | decode_RB8a RB, INS 3222 | decode_RB8b RB 3223 | decode_RDtoRC8 RC, RD 3224 | addu CARG2, BASE, RB 3225 | addu CARG3, BASE, RC 3226 | lw TMP1, HI(CARG2) 3227 | lw TMP2, HI(CARG3) 3228 | lw TAB:RB, LO(CARG2) 3229 | li AT, LJ_TTAB 3230 | ldc1 f0, 0(CARG3) 3231 | bne TMP1, AT, ->vmeta_tsetv 3232 |. addu RA, BASE, RA 3233 | sltiu AT, TMP2, LJ_TISNUM 3234 | beqz AT, >5 3235 |. li AT, LJ_TSTR 3236 | 3237 | // Convert number key to integer, check for integerness and range. 3238 | cvt.w.d f2, f0 3239 | lw TMP0, TAB:RB->asize 3240 | mfc1 TMP2, f2 3241 | cvt.d.w f4, f2 3242 | lw TMP1, TAB:RB->array 3243 | c.eq.d f0, f4 3244 | sltu AT, TMP2, TMP0 3245 | movf AT, r0 3246 | sll TMP2, TMP2, 3 3247 | beqz AT, ->vmeta_tsetv // Integer key and in array part? 3248 |. addu TMP1, TMP1, TMP2 3249 | lbu TMP3, TAB:RB->marked 3250 | lw TMP0, HI(TMP1) 3251 | beq TMP0, TISNIL, >3 3252 |. ldc1 f0, 0(RA) 3253 |1: 3254 | andi AT, TMP3, LJ_GC_BLACK // isblack(table) 3255 | bnez AT, >7 3256 |. sdc1 f0, 0(TMP1) 3257 |2: 3258 | ins_next 3259 | 3260 |3: // Check for __newindex if previous value is nil. 3261 | lw TAB:TMP2, TAB:RB->metatable 3262 | beqz TAB:TMP2, <1 // No metatable: done. 3263 |. nop 3264 | lbu TMP2, TAB:TMP2->nomm 3265 | andi TMP2, TMP2, 1<<MM_newindex 3266 | bnez TMP2, <1 // 'no __newindex' flag set: done. 3267 |. nop 3268 | b ->vmeta_tsetv 3269 |. nop 3270 | 3271 |5: 3272 | bne TMP2, AT, ->vmeta_tsetv 3273 |. lw STR:RC, LO(CARG3) 3274 | b ->BC_TSETS_Z // String key? 3275 |. nop 3276 | 3277 |7: // Possible table write barrier for the value. Skip valiswhite check. 3278 | barrierback TAB:RB, TMP3, TMP0, <2 3279 break; 3280 case BC_TSETS: 3281 | // RA = src*8, RB = table*8, RC = str_const*8 (~) 3282 | decode_RB8a RB, INS 3283 | decode_RB8b RB 3284 | addu CARG2, BASE, RB 3285 | decode_RC4a RC, INS 3286 | lw TMP0, HI(CARG2) 3287 | decode_RC4b RC 3288 | li AT, LJ_TTAB 3289 | subu CARG3, KBASE, RC 3290 | lw TAB:RB, LO(CARG2) 3291 | lw STR:RC, -4(CARG3) // KBASE-4-str_const*4 3292 | bne TMP0, AT, ->vmeta_tsets1 3293 |. addu RA, BASE, RA 3294 |->BC_TSETS_Z: 3295 | // TAB:RB = GCtab *, STR:RC = GCstr *, RA = BASE+src*8 3296 | lw TMP0, TAB:RB->hmask 3297 | lw TMP1, STR:RC->hash 3298 | lw NODE:TMP2, TAB:RB->node 3299 | sb r0, TAB:RB->nomm // Clear metamethod cache. 3300 | and TMP1, TMP1, TMP0 // idx = str->hash & tab->hmask 3301 | sll TMP0, TMP1, 5 3302 | sll TMP1, TMP1, 3 3303 | subu TMP1, TMP0, TMP1 3304 | addu NODE:TMP2, NODE:TMP2, TMP1 // node = tab->node + (idx*32-idx*8) 3305 | ldc1 f20, 0(RA) 3306 |1: 3307 | lw CARG1, offsetof(Node, key)+HI(NODE:TMP2) 3308 | lw TMP0, offsetof(Node, key)+LO(NODE:TMP2) 3309 | li AT, LJ_TSTR 3310 | lw NODE:TMP1, NODE:TMP2->next 3311 | bne CARG1, AT, >5 3312 |. lw CARG2, offsetof(Node, val)+HI(NODE:TMP2) 3313 | bne TMP0, STR:RC, >5 3314 |. lbu TMP3, TAB:RB->marked 3315 | beq CARG2, TISNIL, >4 // Key found, but nil value? 3316 |. lw TAB:TMP0, TAB:RB->metatable 3317 |2: 3318 | andi AT, TMP3, LJ_GC_BLACK // isblack(table) 3319 | bnez AT, >7 3320 |. sdc1 f20, NODE:TMP2->val 3321 |3: 3322 | ins_next 3323 | 3324 |4: // Check for __newindex if previous value is nil. 3325 | beqz TAB:TMP0, <2 // No metatable: done. 3326 |. nop 3327 | lbu TMP0, TAB:TMP0->nomm 3328 | andi TMP0, TMP0, 1<<MM_newindex 3329 | bnez TMP0, <2 // 'no __newindex' flag set: done. 3330 |. nop 3331 | b ->vmeta_tsets 3332 |. nop 3333 | 3334 |5: // Follow hash chain. 3335 | bnez NODE:TMP1, <1 3336 |. move NODE:TMP2, NODE:TMP1 3337 | // End of hash chain: key not found, add a new one 3338 | 3339 | // But check for __newindex first. 3340 | lw TAB:TMP2, TAB:RB->metatable 3341 | beqz TAB:TMP2, >6 // No metatable: continue. 3342 |. addiu CARG3, DISPATCH, DISPATCH_GL(tmptv) 3343 | lbu TMP0, TAB:TMP2->nomm 3344 | andi TMP0, TMP0, 1<<MM_newindex 3345 | beqz TMP0, ->vmeta_tsets // 'no __newindex' flag NOT set: check. 3346 |. li AT, LJ_TSTR 3347 |6: 3348 | load_got lj_tab_newkey 3349 | sw STR:RC, LO(CARG3) 3350 | sw AT, HI(CARG3) 3351 | sw BASE, L->base 3352 | move CARG2, TAB:RB 3353 | sw PC, SAVE_PC 3354 | call_intern lj_tab_newkey // (lua_State *L, GCtab *t, TValue *k 3355 |. move CARG1, L 3356 | // Returns TValue *. 3357 | lw BASE, L->base 3358 | b <3 // No 2nd write barrier needed. 3359 |. sdc1 f20, 0(CRET1) 3360 | 3361 |7: // Possible table write barrier for the value. Skip valiswhite check. 3362 | barrierback TAB:RB, TMP3, TMP0, <3 3363 break; 3364 case BC_TSETB: 3365 | // RA = src*8, RB = table*8, RC = index*8 3366 | decode_RB8a RB, INS 3367 | decode_RB8b RB 3368 | addu CARG2, BASE, RB 3369 | decode_RDtoRC8 RC, RD 3370 | lw CARG1, HI(CARG2) 3371 | li AT, LJ_TTAB 3372 | lw TAB:RB, LO(CARG2) 3373 | addu RA, BASE, RA 3374 | bne CARG1, AT, ->vmeta_tsetb 3375 |. srl TMP0, RC, 3 3376 | lw TMP1, TAB:RB->asize 3377 | lw TMP2, TAB:RB->array 3378 | sltu AT, TMP0, TMP1 3379 | beqz AT, ->vmeta_tsetb 3380 |. addu RC, TMP2, RC 3381 | lw TMP1, HI(RC) 3382 | lbu TMP3, TAB:RB->marked 3383 | beq TMP1, TISNIL, >5 3384 |. ldc1 f0, 0(RA) 3385 |1: 3386 | andi AT, TMP3, LJ_GC_BLACK // isblack(table) 3387 | bnez AT, >7 3388 |. sdc1 f0, 0(RC) 3389 |2: 3390 | ins_next 3391 | 3392 |5: // Check for __newindex if previous value is nil. 3393 | lw TAB:TMP2, TAB:RB->metatable 3394 | beqz TAB:TMP2, <1 // No metatable: done. 3395 |. nop 3396 | lbu TMP1, TAB:TMP2->nomm 3397 | andi TMP1, TMP1, 1<<MM_newindex 3398 | bnez TMP1, <1 // 'no __newindex' flag set: done. 3399 |. nop 3400 | b ->vmeta_tsetb // Caveat: preserve TMP0! 3401 |. nop 3402 | 3403 |7: // Possible table write barrier for the value. Skip valiswhite check. 3404 | barrierback TAB:RB, TMP3, TMP0, <2 3405 break; 3406 3407 case BC_TSETM: 3408 | // RA = base*8 (table at base-1), RD = num_const*8 (start index) 3409 | addu RA, BASE, RA 3410 |1: 3411 | addu TMP3, KBASE, RD 3412 | lw TAB:CARG2, -8+LO(RA) // Guaranteed to be a table. 3413 | addiu TMP0, MULTRES, -8 3414 | lw TMP3, LO(TMP3) // Integer constant is in lo-word. 3415 | beqz TMP0, >4 // Nothing to copy? 3416 |. srl CARG3, TMP0, 3 3417 | addu CARG3, CARG3, TMP3 3418 | lw TMP2, TAB:CARG2->asize 3419 | sll TMP1, TMP3, 3 3420 | lbu TMP3, TAB:CARG2->marked 3421 | lw CARG1, TAB:CARG2->array 3422 | sltu AT, TMP2, CARG3 3423 | bnez AT, >5 3424 |. addu TMP2, RA, TMP0 3425 | addu TMP1, TMP1, CARG1 3426 | andi TMP0, TMP3, LJ_GC_BLACK // isblack(table) 3427 |3: // Copy result slots to table. 3428 | ldc1 f0, 0(RA) 3429 | addiu RA, RA, 8 3430 | sltu AT, RA, TMP2 3431 | sdc1 f0, 0(TMP1) 3432 | bnez AT, <3 3433 |. addiu TMP1, TMP1, 8 3434 | bnez TMP0, >7 3435 |. nop 3436 |4: 3437 | ins_next 3438 | 3439 |5: // Need to resize array part. 3440 | load_got lj_tab_reasize 3441 | sw BASE, L->base 3442 | sw PC, SAVE_PC 3443 | move BASE, RD 3444 | call_intern lj_tab_reasize // (lua_State *L, GCtab *t, int nasize) 3445 |. move CARG1, L 3446 | // Must not reallocate the stack. 3447 | move RD, BASE 3448 | b <1 3449 |. lw BASE, L->base // Reload BASE for lack of a saved register. 3450 | 3451 |7: // Possible table write barrier for any value. Skip valiswhite check. 3452 | barrierback TAB:CARG2, TMP3, TMP0, <4 3453 break; 3454 3455 /* -- Calls and vararg handling ----------------------------------------- */ 3456 3457 case BC_CALLM: 3458 | // RA = base*8, (RB = (nresults+1)*8,) RC = extra_nargs*8 3459 | decode_RDtoRC8 NARGS8:RC, RD 3460 | b ->BC_CALL_Z 3461 |. addu NARGS8:RC, NARGS8:RC, MULTRES 3462 break; 3463 case BC_CALL: 3464 | // RA = base*8, (RB = (nresults+1)*8,) RC = (nargs+1)*8 3465 | decode_RDtoRC8 NARGS8:RC, RD 3466 |->BC_CALL_Z: 3467 | move TMP2, BASE 3468 | addu BASE, BASE, RA 3469 | li AT, LJ_TFUNC 3470 | lw TMP0, HI(BASE) 3471 | lw LFUNC:RB, LO(BASE) 3472 | addiu BASE, BASE, 8 3473 | bne TMP0, AT, ->vmeta_call 3474 |. addiu NARGS8:RC, NARGS8:RC, -8 3475 | ins_call 3476 break; 3477 3478 case BC_CALLMT: 3479 | // RA = base*8, (RB = 0,) RC = extra_nargs*8 3480 | addu NARGS8:RD, NARGS8:RD, MULTRES // BC_CALLT gets RC from RD. 3481 | // Fall through. Assumes BC_CALLT follows. 3482 break; 3483 case BC_CALLT: 3484 | // RA = base*8, (RB = 0,) RC = (nargs+1)*8 3485 | addu RA, BASE, RA 3486 | li AT, LJ_TFUNC 3487 | lw TMP0, HI(RA) 3488 | lw LFUNC:RB, LO(RA) 3489 | move NARGS8:RC, RD 3490 | lw TMP1, FRAME_PC(BASE) 3491 | addiu RA, RA, 8 3492 | bne TMP0, AT, ->vmeta_callt 3493 |. addiu NARGS8:RC, NARGS8:RC, -8 3494 |->BC_CALLT_Z: 3495 | andi TMP0, TMP1, FRAME_TYPE // Caveat: preserve TMP0 until the 'or'. 3496 | lbu TMP3, LFUNC:RB->ffid 3497 | bnez TMP0, >7 3498 |. xori TMP2, TMP1, FRAME_VARG 3499 |1: 3500 | sw LFUNC:RB, FRAME_FUNC(BASE) // Copy function down, but keep PC. 3501 | sltiu AT, TMP3, 2 // (> FF_C) Calling a fast function? 3502 | move TMP2, BASE 3503 | beqz NARGS8:RC, >3 3504 |. move TMP3, NARGS8:RC 3505 |2: 3506 | ldc1 f0, 0(RA) 3507 | addiu RA, RA, 8 3508 | addiu TMP3, TMP3, -8 3509 | sdc1 f0, 0(TMP2) 3510 | bnez TMP3, <2 3511 |. addiu TMP2, TMP2, 8 3512 |3: 3513 | or TMP0, TMP0, AT 3514 | beqz TMP0, >5 3515 |. nop 3516 |4: 3517 | ins_callt 3518 | 3519 |5: // Tailcall to a fast function with a Lua frame below. 3520 | lw INS, -4(TMP1) 3521 | decode_RA8a RA, INS 3522 | decode_RA8b RA 3523 | subu TMP1, BASE, RA 3524 | lw LFUNC:TMP1, -8+FRAME_FUNC(TMP1) 3525 | lw TMP1, LFUNC:TMP1->pc 3526 | b <4 3527 |. lw KBASE, PC2PROTO(k)(TMP1) // Need to prepare KBASE. 3528 | 3529 |7: // Tailcall from a vararg function. 3530 | andi AT, TMP2, FRAME_TYPEP 3531 | bnez AT, <1 // Vararg frame below? 3532 |. subu TMP2, BASE, TMP2 // Relocate BASE down. 3533 | move BASE, TMP2 3534 | lw TMP1, FRAME_PC(TMP2) 3535 | b <1 3536 |. andi TMP0, TMP1, FRAME_TYPE 3537 break; 3538 3539 case BC_ITERC: 3540 | // RA = base*8, (RB = (nresults+1)*8, RC = (nargs+1)*8 ((2+1)*8)) 3541 | move TMP2, BASE 3542 | addu BASE, BASE, RA 3543 | li AT, LJ_TFUNC 3544 | lw TMP1, -24+HI(BASE) 3545 | lw LFUNC:RB, -24+LO(BASE) 3546 | ldc1 f2, -8(BASE) 3547 | ldc1 f0, -16(BASE) 3548 | sw TMP1, HI(BASE) // Copy callable. 3549 | sw LFUNC:RB, LO(BASE) 3550 | sdc1 f2, 16(BASE) // Copy control var. 3551 | sdc1 f0, 8(BASE) // Copy state. 3552 | addiu BASE, BASE, 8 3553 | bne TMP1, AT, ->vmeta_call 3554 |. li NARGS8:RC, 16 // Iterators get 2 arguments. 3555 | ins_call 3556 break; 3557 3558 case BC_ITERN: 3559 | // RA = base*8, (RB = (nresults+1)*8, RC = (nargs+1)*8 (2+1)*8) 3560 |.if JIT 3561 | // NYI: add hotloop, record BC_ITERN. 3562 |.endif 3563 | addu RA, BASE, RA 3564 | lw TAB:RB, -16+LO(RA) 3565 | lw RC, -8+LO(RA) // Get index from control var. 3566 | lw TMP0, TAB:RB->asize 3567 | lw TMP1, TAB:RB->array 3568 | addiu PC, PC, 4 3569 |1: // Traverse array part. 3570 | sltu AT, RC, TMP0 3571 | beqz AT, >5 // Index points after array part? 3572 |. sll TMP3, RC, 3 3573 | addu TMP3, TMP1, TMP3 3574 | lw TMP2, HI(TMP3) 3575 | ldc1 f0, 0(TMP3) 3576 | mtc1 RC, f2 3577 | lhu RD, -4+OFS_RD(PC) 3578 | beq TMP2, TISNIL, <1 // Skip holes in array part. 3579 |. addiu RC, RC, 1 3580 | cvt.d.w f2, f2 3581 | lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535) 3582 | sdc1 f0, 8(RA) 3583 | decode_RD4b RD 3584 | addu RD, RD, TMP3 3585 | sw RC, -8+LO(RA) // Update control var. 3586 | addu PC, PC, RD 3587 | sdc1 f2, 0(RA) 3588 |3: 3589 | ins_next 3590 | 3591 |5: // Traverse hash part. 3592 | lw TMP1, TAB:RB->hmask 3593 | subu RC, RC, TMP0 3594 | lw TMP2, TAB:RB->node 3595 |6: 3596 | sltu AT, TMP1, RC // End of iteration? Branch to ITERL+1. 3597 | bnez AT, <3 3598 |. sll TMP3, RC, 5 3599 | sll RB, RC, 3 3600 | subu TMP3, TMP3, RB 3601 | addu NODE:TMP3, TMP3, TMP2 3602 | lw RB, HI(NODE:TMP3) 3603 | ldc1 f0, 0(NODE:TMP3) 3604 | lhu RD, -4+OFS_RD(PC) 3605 | beq RB, TISNIL, <6 // Skip holes in hash part. 3606 |. addiu RC, RC, 1 3607 | ldc1 f2, NODE:TMP3->key 3608 | lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535) 3609 | sdc1 f0, 8(RA) 3610 | addu RC, RC, TMP0 3611 | decode_RD4b RD 3612 | addu RD, RD, TMP3 3613 | sdc1 f2, 0(RA) 3614 | addu PC, PC, RD 3615 | b <3 3616 |. sw RC, -8+LO(RA) // Update control var. 3617 break; 3618 3619 case BC_ISNEXT: 3620 | // RA = base*8, RD = target (points to ITERN) 3621 | addu RA, BASE, RA 3622 | srl TMP0, RD, 1 3623 | lw CARG1, -24+HI(RA) 3624 | lw CFUNC:CARG2, -24+LO(RA) 3625 | addu TMP0, PC, TMP0 3626 | lw CARG3, -16+HI(RA) 3627 | lw CARG4, -8+HI(RA) 3628 | li AT, LJ_TFUNC 3629 | bne CARG1, AT, >5 3630 |. lui TMP2, (-(BCBIAS_J*4 >> 16) & 65535) 3631 | lbu CARG2, CFUNC:CARG2->ffid 3632 | addiu CARG3, CARG3, -LJ_TTAB 3633 | addiu CARG4, CARG4, -LJ_TNIL 3634 | or CARG3, CARG3, CARG4 3635 | addiu CARG2, CARG2, -FF_next_N 3636 | or CARG2, CARG2, CARG3 3637 | bnez CARG2, >5 3638 |. lui TMP1, 0xfffe 3639 | addu PC, TMP0, TMP2 3640 | ori TMP1, TMP1, 0x7fff 3641 | sw r0, -8+LO(RA) // Initialize control var. 3642 | sw TMP1, -8+HI(RA) 3643 |1: 3644 | ins_next 3645 |5: // Despecialize bytecode if any of the checks fail. 3646 | li TMP3, BC_JMP 3647 | li TMP1, BC_ITERC 3648 | sb TMP3, -4+OFS_OP(PC) 3649 | addu PC, TMP0, TMP2 3650 | b <1 3651 |. sb TMP1, OFS_OP(PC) 3652 break; 3653 3654 case BC_VARG: 3655 | // RA = base*8, RB = (nresults+1)*8, RC = numparams*8 3656 | lw TMP0, FRAME_PC(BASE) 3657 | decode_RDtoRC8 RC, RD 3658 | decode_RB8a RB, INS 3659 | addu RC, BASE, RC 3660 | decode_RB8b RB 3661 | addu RA, BASE, RA 3662 | addiu RC, RC, FRAME_VARG 3663 | addu TMP2, RA, RB 3664 | addiu TMP3, BASE, -8 // TMP3 = vtop 3665 | subu RC, RC, TMP0 // RC = vbase 3666 | // Note: RC may now be even _above_ BASE if nargs was < numparams. 3667 | beqz RB, >5 // Copy all varargs? 3668 |. subu TMP1, TMP3, RC 3669 | addiu TMP2, TMP2, -16 3670 |1: // Copy vararg slots to destination slots. 3671 | lw CARG1, HI(RC) 3672 | sltu AT, RC, TMP3 3673 | lw CARG2, LO(RC) 3674 | addiu RC, RC, 8 3675 | movz CARG1, TISNIL, AT 3676 | sw CARG1, HI(RA) 3677 | sw CARG2, LO(RA) 3678 | sltu AT, RA, TMP2 3679 | bnez AT, <1 3680 |. addiu RA, RA, 8 3681 |3: 3682 | ins_next 3683 | 3684 |5: // Copy all varargs. 3685 | lw TMP0, L->maxstack 3686 | blez TMP1, <3 // No vararg slots? 3687 |. li MULTRES, 8 // MULTRES = (0+1)*8 3688 | addu TMP2, RA, TMP1 3689 | sltu AT, TMP0, TMP2 3690 | bnez AT, >7 3691 |. addiu MULTRES, TMP1, 8 3692 |6: 3693 | ldc1 f0, 0(RC) 3694 | addiu RC, RC, 8 3695 | sdc1 f0, 0(RA) 3696 | sltu AT, RC, TMP3 3697 | bnez AT, <6 // More vararg slots? 3698 |. addiu RA, RA, 8 3699 | b <3 3700 |. nop 3701 | 3702 |7: // Grow stack for varargs. 3703 | load_got lj_state_growstack 3704 | sw RA, L->top 3705 | subu RA, RA, BASE 3706 | sw BASE, L->base 3707 | subu BASE, RC, BASE // Need delta, because BASE may change. 3708 | sw PC, SAVE_PC 3709 | srl CARG2, TMP1, 3 3710 | call_intern lj_state_growstack // (lua_State *L, int n) 3711 |. move CARG1, L 3712 | move RC, BASE 3713 | lw BASE, L->base 3714 | addu RA, BASE, RA 3715 | addu RC, BASE, RC 3716 | b <6 3717 |. addiu TMP3, BASE, -8 3718 break; 3719 3720 /* -- Returns ----------------------------------------------------------- */ 3721 3722 case BC_RETM: 3723 | // RA = results*8, RD = extra_nresults*8 3724 | addu RD, RD, MULTRES // MULTRES >= 8, so RD >= 8. 3725 | // Fall through. Assumes BC_RET follows. 3726 break; 3727 3728 case BC_RET: 3729 | // RA = results*8, RD = (nresults+1)*8 3730 | lw PC, FRAME_PC(BASE) 3731 | addu RA, BASE, RA 3732 | move MULTRES, RD 3733 |1: 3734 | andi TMP0, PC, FRAME_TYPE 3735 | bnez TMP0, ->BC_RETV_Z 3736 |. xori TMP1, PC, FRAME_VARG 3737 | 3738 |->BC_RET_Z: 3739 | // BASE = base, RA = resultptr, RD = (nresults+1)*8, PC = return 3740 | lw INS, -4(PC) 3741 | addiu TMP2, BASE, -8 3742 | addiu RC, RD, -8 3743 | decode_RA8a TMP0, INS 3744 | decode_RB8a RB, INS 3745 | decode_RA8b TMP0 3746 | decode_RB8b RB 3747 | addu TMP3, TMP2, RB 3748 | beqz RC, >3 3749 |. subu BASE, TMP2, TMP0 3750 |2: 3751 | ldc1 f0, 0(RA) 3752 | addiu RA, RA, 8 3753 | addiu RC, RC, -8 3754 | sdc1 f0, 0(TMP2) 3755 | bnez RC, <2 3756 |. addiu TMP2, TMP2, 8 3757 |3: 3758 | addiu TMP3, TMP3, -8 3759 |5: 3760 | sltu AT, TMP2, TMP3 3761 | bnez AT, >6 3762 |. lw LFUNC:TMP1, FRAME_FUNC(BASE) 3763 | ins_next1 3764 | lw TMP1, LFUNC:TMP1->pc 3765 | lw KBASE, PC2PROTO(k)(TMP1) 3766 | ins_next2 3767 | 3768 |6: // Fill up results with nil. 3769 | sw TISNIL, HI(TMP2) 3770 | b <5 3771 |. addiu TMP2, TMP2, 8 3772 | 3773 |->BC_RETV_Z: // Non-standard return case. 3774 | andi TMP2, TMP1, FRAME_TYPEP 3775 | bnez TMP2, ->vm_return 3776 |. nop 3777 | // Return from vararg function: relocate BASE down. 3778 | subu BASE, BASE, TMP1 3779 | b <1 3780 |. lw PC, FRAME_PC(BASE) 3781 break; 3782 3783 case BC_RET0: case BC_RET1: 3784 | // RA = results*8, RD = (nresults+1)*8 3785 | lw PC, FRAME_PC(BASE) 3786 | addu RA, BASE, RA 3787 | move MULTRES, RD 3788 | andi TMP0, PC, FRAME_TYPE 3789 | bnez TMP0, ->BC_RETV_Z 3790 |. xori TMP1, PC, FRAME_VARG 3791 | 3792 | lw INS, -4(PC) 3793 | addiu TMP2, BASE, -8 3794 if (op == BC_RET1) { 3795 | ldc1 f0, 0(RA) 3796 } 3797 | decode_RB8a RB, INS 3798 | decode_RA8a RA, INS 3799 | decode_RB8b RB 3800 | decode_RA8b RA 3801 if (op == BC_RET1) { 3802 | sdc1 f0, 0(TMP2) 3803 } 3804 | subu BASE, TMP2, RA 3805 |5: 3806 | sltu AT, RD, RB 3807 | bnez AT, >6 3808 |. lw LFUNC:TMP1, FRAME_FUNC(BASE) 3809 | ins_next1 3810 | lw TMP1, LFUNC:TMP1->pc 3811 | lw KBASE, PC2PROTO(k)(TMP1) 3812 | ins_next2 3813 | 3814 |6: // Fill up results with nil. 3815 | addiu TMP2, TMP2, 8 3816 | addiu RD, RD, 8 3817 | b <5 3818 if (op == BC_RET1) { 3819 |. sw TISNIL, HI(TMP2) 3820 } else { 3821 |. sw TISNIL, -8+HI(TMP2) 3822 } 3823 break; 3824 3825 /* -- Loops and branches ------------------------------------------------ */ 3826 3827 case BC_FORL: 3828 |.if JIT 3829 | hotloop 3830 |.endif 3831 | // Fall through. Assumes BC_IFORL follows. 3832 break; 3833 3834 case BC_JFORI: 3835 case BC_JFORL: 3836#if !LJ_HASJIT 3837 break; 3838#endif 3839 case BC_FORI: 3840 case BC_IFORL: 3841 | // RA = base*8, RD = target (after end of loop or start of loop) 3842 vk = (op == BC_IFORL || op == BC_JFORL); 3843 | addu RA, BASE, RA 3844 if (vk) { 3845 | ldc1 f0, FORL_IDX*8(RA) 3846 | ldc1 f4, FORL_STEP*8(RA) 3847 | ldc1 f2, FORL_STOP*8(RA) 3848 | lw TMP3, FORL_STEP*8+HI(RA) 3849 | add.d f0, f0, f4 3850 | sdc1 f0, FORL_IDX*8(RA) 3851 } else { 3852 | lw TMP1, FORL_IDX*8+HI(RA) 3853 | lw TMP3, FORL_STEP*8+HI(RA) 3854 | lw TMP2, FORL_STOP*8+HI(RA) 3855 | sltiu TMP1, TMP1, LJ_TISNUM 3856 | sltiu TMP0, TMP3, LJ_TISNUM 3857 | sltiu TMP2, TMP2, LJ_TISNUM 3858 | and TMP1, TMP1, TMP0 3859 | and TMP1, TMP1, TMP2 3860 | ldc1 f0, FORL_IDX*8(RA) 3861 | beqz TMP1, ->vmeta_for 3862 |. ldc1 f2, FORL_STOP*8(RA) 3863 } 3864 if (op != BC_JFORL) { 3865 | srl RD, RD, 1 3866 | lui TMP0, (-(BCBIAS_J*4 >> 16) & 65535) 3867 } 3868 | c.le.d 0, f0, f2 3869 | c.le.d 1, f2, f0 3870 | sdc1 f0, FORL_EXT*8(RA) 3871 if (op == BC_JFORI) { 3872 | li TMP1, 1 3873 | li TMP2, 1 3874 | addu TMP0, RD, TMP0 3875 | slt TMP3, TMP3, r0 3876 | movf TMP1, r0, 0 3877 | addu PC, PC, TMP0 3878 | movf TMP2, r0, 1 3879 | lhu RD, -4+OFS_RD(PC) 3880 | movn TMP1, TMP2, TMP3 3881 | bnez TMP1, =>BC_JLOOP 3882 |. decode_RD8b RD 3883 } else if (op == BC_JFORL) { 3884 | li TMP1, 1 3885 | li TMP2, 1 3886 | slt TMP3, TMP3, r0 3887 | movf TMP1, r0, 0 3888 | movf TMP2, r0, 1 3889 | movn TMP1, TMP2, TMP3 3890 | bnez TMP1, =>BC_JLOOP 3891 |. nop 3892 } else { 3893 | addu TMP1, RD, TMP0 3894 | slt TMP3, TMP3, r0 3895 | move TMP2, TMP1 3896 if (op == BC_FORI) { 3897 | movt TMP1, r0, 0 3898 | movt TMP2, r0, 1 3899 } else { 3900 | movf TMP1, r0, 0 3901 | movf TMP2, r0, 1 3902 } 3903 | movn TMP1, TMP2, TMP3 3904 | addu PC, PC, TMP1 3905 } 3906 | ins_next 3907 break; 3908 3909 case BC_ITERL: 3910 |.if JIT 3911 | hotloop 3912 |.endif 3913 | // Fall through. Assumes BC_IITERL follows. 3914 break; 3915 3916 case BC_JITERL: 3917#if !LJ_HASJIT 3918 break; 3919#endif 3920 case BC_IITERL: 3921 | // RA = base*8, RD = target 3922 | addu RA, BASE, RA 3923 | lw TMP1, HI(RA) 3924 | beq TMP1, TISNIL, >1 // Stop if iterator returned nil. 3925 |. lw TMP2, LO(RA) 3926 if (op == BC_JITERL) { 3927 | sw TMP1, -8+HI(RA) 3928 | b =>BC_JLOOP 3929 |. sw TMP2, -8+LO(RA) 3930 } else { 3931 | branch_RD // Otherwise save control var + branch. 3932 | sw TMP1, -8+HI(RA) 3933 | sw TMP2, -8+LO(RA) 3934 } 3935 |1: 3936 | ins_next 3937 break; 3938 3939 case BC_LOOP: 3940 | // RA = base*8, RD = target (loop extent) 3941 | // Note: RA/RD is only used by trace recorder to determine scope/extent 3942 | // This opcode does NOT jump, it's only purpose is to detect a hot loop. 3943 |.if JIT 3944 | hotloop 3945 |.endif 3946 | // Fall through. Assumes BC_ILOOP follows. 3947 break; 3948 3949 case BC_ILOOP: 3950 | // RA = base*8, RD = target (loop extent) 3951 | ins_next 3952 break; 3953 3954 case BC_JLOOP: 3955 |.if JIT 3956 | // RA = base*8 (ignored), RD = traceno*8 3957 | lw TMP1, DISPATCH_J(trace)(DISPATCH) 3958 | srl RD, RD, 1 3959 | li AT, 0 3960 | addu TMP1, TMP1, RD 3961 | // Traces on MIPS don't store the trace number, so use 0. 3962 | sw AT, DISPATCH_GL(vmstate)(DISPATCH) 3963 | lw TRACE:TMP2, 0(TMP1) 3964 | sw BASE, DISPATCH_GL(jit_base)(DISPATCH) 3965 | sw L, DISPATCH_GL(jit_L)(DISPATCH) 3966 | lw TMP2, TRACE:TMP2->mcode 3967 | jr TMP2 3968 |. addiu JGL, DISPATCH, GG_DISP2G+32768 3969 |.endif 3970 break; 3971 3972 case BC_JMP: 3973 | // RA = base*8 (only used by trace recorder), RD = target 3974 | branch_RD 3975 | ins_next 3976 break; 3977 3978 /* -- Function headers -------------------------------------------------- */ 3979 3980 case BC_FUNCF: 3981 |.if JIT 3982 | hotcall 3983 |.endif 3984 case BC_FUNCV: /* NYI: compiled vararg functions. */ 3985 | // Fall through. Assumes BC_IFUNCF/BC_IFUNCV follow. 3986 break; 3987 3988 case BC_JFUNCF: 3989#if !LJ_HASJIT 3990 break; 3991#endif 3992 case BC_IFUNCF: 3993 | // BASE = new base, RA = BASE+framesize*8, RB = LFUNC, RC = nargs*8 3994 | lw TMP2, L->maxstack 3995 | lbu TMP1, -4+PC2PROTO(numparams)(PC) 3996 | lw KBASE, -4+PC2PROTO(k)(PC) 3997 | sltu AT, TMP2, RA 3998 | bnez AT, ->vm_growstack_l 3999 |. sll TMP1, TMP1, 3 4000 if (op != BC_JFUNCF) { 4001 | ins_next1 4002 } 4003 |2: 4004 | sltu AT, NARGS8:RC, TMP1 // Check for missing parameters. 4005 | bnez AT, >3 4006 |. addu AT, BASE, NARGS8:RC 4007 if (op == BC_JFUNCF) { 4008 | decode_RD8a RD, INS 4009 | b =>BC_JLOOP 4010 |. decode_RD8b RD 4011 } else { 4012 | ins_next2 4013 } 4014 | 4015 |3: // Clear missing parameters. 4016 | sw TISNIL, HI(AT) 4017 | b <2 4018 |. addiu NARGS8:RC, NARGS8:RC, 8 4019 break; 4020 4021 case BC_JFUNCV: 4022#if !LJ_HASJIT 4023 break; 4024#endif 4025 | NYI // NYI: compiled vararg functions 4026 break; /* NYI: compiled vararg functions. */ 4027 4028 case BC_IFUNCV: 4029 | // BASE = new base, RA = BASE+framesize*8, RB = LFUNC, RC = nargs*8 4030 | addu TMP1, BASE, RC 4031 | lw TMP2, L->maxstack 4032 | addu TMP0, RA, RC 4033 | sw LFUNC:RB, LO(TMP1) // Store copy of LFUNC. 4034 | addiu TMP3, RC, 8+FRAME_VARG 4035 | sltu AT, TMP0, TMP2 4036 | lw KBASE, -4+PC2PROTO(k)(PC) 4037 | beqz AT, ->vm_growstack_l 4038 |. sw TMP3, HI(TMP1) // Store delta + FRAME_VARG. 4039 | lbu TMP2, -4+PC2PROTO(numparams)(PC) 4040 | move RA, BASE 4041 | move RC, TMP1 4042 | ins_next1 4043 | beqz TMP2, >3 4044 |. addiu BASE, TMP1, 8 4045 |1: 4046 | lw TMP0, HI(RA) 4047 | lw TMP3, LO(RA) 4048 | sltu AT, RA, RC // Less args than parameters? 4049 | move CARG1, TMP0 4050 | movz TMP0, TISNIL, AT // Clear missing parameters. 4051 | movn CARG1, TISNIL, AT // Clear old fixarg slot (help the GC). 4052 | sw TMP3, 8+LO(TMP1) 4053 | addiu TMP2, TMP2, -1 4054 | sw TMP0, 8+HI(TMP1) 4055 | addiu TMP1, TMP1, 8 4056 | sw CARG1, HI(RA) 4057 | bnez TMP2, <1 4058 |. addiu RA, RA, 8 4059 |3: 4060 | ins_next2 4061 break; 4062 4063 case BC_FUNCC: 4064 case BC_FUNCCW: 4065 | // BASE = new base, RA = BASE+framesize*8, RB = CFUNC, RC = nargs*8 4066 if (op == BC_FUNCC) { 4067 | lw CFUNCADDR, CFUNC:RB->f 4068 } else { 4069 | lw CFUNCADDR, DISPATCH_GL(wrapf)(DISPATCH) 4070 } 4071 | addu TMP1, RA, NARGS8:RC 4072 | lw TMP2, L->maxstack 4073 | addu RC, BASE, NARGS8:RC 4074 | sw BASE, L->base 4075 | sltu AT, TMP2, TMP1 4076 | sw RC, L->top 4077 | li_vmstate C 4078 if (op == BC_FUNCCW) { 4079 | lw CARG2, CFUNC:RB->f 4080 } 4081 | bnez AT, ->vm_growstack_c // Need to grow stack. 4082 |. move CARG1, L 4083 | jalr CFUNCADDR // (lua_State *L [, lua_CFunction f]) 4084 |. st_vmstate 4085 | // Returns nresults. 4086 | lw BASE, L->base 4087 | sll RD, CRET1, 3 4088 | lw TMP1, L->top 4089 | li_vmstate INTERP 4090 | lw PC, FRAME_PC(BASE) // Fetch PC of caller. 4091 | subu RA, TMP1, RD // RA = L->top - nresults*8 4092 | b ->vm_returnc 4093 |. st_vmstate 4094 break; 4095 4096 /* ---------------------------------------------------------------------- */ 4097 4098 default: 4099 fprintf(stderr, "Error: undefined opcode BC_%s\n", bc_names[op]); 4100 exit(2); 4101 break; 4102 } 4103} 4104 4105static int build_backend(BuildCtx *ctx) 4106{ 4107 int op; 4108 4109 dasm_growpc(Dst, BC__MAX); 4110 4111 build_subroutines(ctx); 4112 4113 |.code_op 4114 for (op = 0; op < BC__MAX; op++) 4115 build_ins(ctx, (BCOp)op, op); 4116 4117 return BC__MAX; 4118} 4119 4120/* Emit pseudo frame-info for all assembler functions. */ 4121static void emit_asm_debug(BuildCtx *ctx) 4122{ 4123 int fcofs = (int)((uint8_t *)ctx->glob[GLOB_vm_ffi_call] - ctx->code); 4124 int i; 4125 switch (ctx->mode) { 4126 case BUILD_elfasm: 4127 fprintf(ctx->fp, "\t.section .debug_frame,\"\",@progbits\n"); 4128 fprintf(ctx->fp, 4129 ".Lframe0:\n" 4130 "\t.4byte .LECIE0-.LSCIE0\n" 4131 ".LSCIE0:\n" 4132 "\t.4byte 0xffffffff\n" 4133 "\t.byte 0x1\n" 4134 "\t.string \"\"\n" 4135 "\t.uleb128 0x1\n" 4136 "\t.sleb128 -4\n" 4137 "\t.byte 31\n" 4138 "\t.byte 0xc\n\t.uleb128 29\n\t.uleb128 0\n" 4139 "\t.align 2\n" 4140 ".LECIE0:\n\n"); 4141 fprintf(ctx->fp, 4142 ".LSFDE0:\n" 4143 "\t.4byte .LEFDE0-.LASFDE0\n" 4144 ".LASFDE0:\n" 4145 "\t.4byte .Lframe0\n" 4146 "\t.4byte .Lbegin\n" 4147 "\t.4byte %d\n" 4148 "\t.byte 0xe\n\t.uleb128 %d\n" 4149 "\t.byte 0x9f\n\t.sleb128 1\n" 4150 "\t.byte 0x9e\n\t.sleb128 2\n", 4151 fcofs, CFRAME_SIZE); 4152 for (i = 23; i >= 16; i--) 4153 fprintf(ctx->fp, "\t.byte %d\n\t.uleb128 %d\n", 0x80+i, 26-i); 4154 for (i = 30; i >= 20; i -= 2) 4155 fprintf(ctx->fp, "\t.byte %d\n\t.uleb128 %d\n", 0x80+32+i, 42-i); 4156 fprintf(ctx->fp, 4157 "\t.align 2\n" 4158 ".LEFDE0:\n\n"); 4159#if LJ_HASFFI 4160 fprintf(ctx->fp, 4161 ".LSFDE1:\n" 4162 "\t.4byte .LEFDE1-.LASFDE1\n" 4163 ".LASFDE1:\n" 4164 "\t.4byte .Lframe0\n" 4165 "\t.4byte lj_vm_ffi_call\n" 4166 "\t.4byte %d\n" 4167 "\t.byte 0x9f\n\t.uleb128 1\n" 4168 "\t.byte 0x90\n\t.uleb128 2\n" 4169 "\t.byte 0xd\n\t.uleb128 0x10\n" 4170 "\t.align 2\n" 4171 ".LEFDE1:\n\n", (int)ctx->codesz - fcofs); 4172#endif 4173#if !LJ_NO_UNWIND 4174 fprintf(ctx->fp, "\t.section .eh_frame,\"aw\",@progbits\n"); 4175 fprintf(ctx->fp, 4176 "\t.globl lj_err_unwind_dwarf\n" 4177 ".Lframe1:\n" 4178 "\t.4byte .LECIE1-.LSCIE1\n" 4179 ".LSCIE1:\n" 4180 "\t.4byte 0\n" 4181 "\t.byte 0x1\n" 4182 "\t.string \"zPR\"\n" 4183 "\t.uleb128 0x1\n" 4184 "\t.sleb128 -4\n" 4185 "\t.byte 31\n" 4186 "\t.uleb128 6\n" /* augmentation length */ 4187 "\t.byte 0\n" 4188 "\t.4byte lj_err_unwind_dwarf\n" 4189 "\t.byte 0\n" 4190 "\t.byte 0xc\n\t.uleb128 29\n\t.uleb128 0\n" 4191 "\t.align 2\n" 4192 ".LECIE1:\n\n"); 4193 fprintf(ctx->fp, 4194 ".LSFDE2:\n" 4195 "\t.4byte .LEFDE2-.LASFDE2\n" 4196 ".LASFDE2:\n" 4197 "\t.4byte .LASFDE2-.Lframe1\n" 4198 "\t.4byte .Lbegin\n" 4199 "\t.4byte %d\n" 4200 "\t.uleb128 0\n" /* augmentation length */ 4201 "\t.byte 0xe\n\t.uleb128 %d\n" 4202 "\t.byte 0x9f\n\t.sleb128 1\n" 4203 "\t.byte 0x9e\n\t.sleb128 2\n", 4204 fcofs, CFRAME_SIZE); 4205 for (i = 23; i >= 16; i--) 4206 fprintf(ctx->fp, "\t.byte %d\n\t.uleb128 %d\n", 0x80+i, 26-i); 4207 for (i = 30; i >= 20; i -= 2) 4208 fprintf(ctx->fp, "\t.byte %d\n\t.uleb128 %d\n", 0x80+32+i, 42-i); 4209 fprintf(ctx->fp, 4210 "\t.align 2\n" 4211 ".LEFDE2:\n\n"); 4212#if LJ_HASFFI 4213 fprintf(ctx->fp, 4214 ".Lframe2:\n" 4215 "\t.4byte .LECIE2-.LSCIE2\n" 4216 ".LSCIE2:\n" 4217 "\t.4byte 0\n" 4218 "\t.byte 0x1\n" 4219 "\t.string \"zR\"\n" 4220 "\t.uleb128 0x1\n" 4221 "\t.sleb128 -4\n" 4222 "\t.byte 31\n" 4223 "\t.uleb128 1\n" /* augmentation length */ 4224 "\t.byte 0\n" 4225 "\t.byte 0xc\n\t.uleb128 29\n\t.uleb128 0\n" 4226 "\t.align 2\n" 4227 ".LECIE2:\n\n"); 4228 fprintf(ctx->fp, 4229 ".LSFDE3:\n" 4230 "\t.4byte .LEFDE3-.LASFDE3\n" 4231 ".LASFDE3:\n" 4232 "\t.4byte .LASFDE3-.Lframe2\n" 4233 "\t.4byte lj_vm_ffi_call\n" 4234 "\t.4byte %d\n" 4235 "\t.uleb128 0\n" /* augmentation length */ 4236 "\t.byte 0x9f\n\t.uleb128 1\n" 4237 "\t.byte 0x90\n\t.uleb128 2\n" 4238 "\t.byte 0xd\n\t.uleb128 0x10\n" 4239 "\t.align 2\n" 4240 ".LEFDE3:\n\n", (int)ctx->codesz - fcofs); 4241#endif 4242#endif 4243 break; 4244 default: 4245 break; 4246 } 4247} 4248 4249