builtins-loong64.cc 145 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73
// Copyright 2021 the V8 project authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.

#if V8_TARGET_ARCH_LOONG64

#include "src/api/api-arguments.h"
#include "src/codegen/code-factory.h"
#include "src/codegen/interface-descriptors-inl.h"
#include "src/debug/debug.h"
#include "src/deoptimizer/deoptimizer.h"
#include "src/execution/frame-constants.h"
#include "src/execution/frames.h"
#include "src/logging/counters.h"
// For interpreter_entry_return_pc_offset. TODO(jkummerow): Drop.
#include "src/codegen/loong64/constants-loong64.h"
#include "src/codegen/macro-assembler-inl.h"
#include "src/codegen/register-configuration.h"
#include "src/heap/heap-inl.h"
#include "src/objects/cell.h"
#include "src/objects/foreign.h"
#include "src/objects/heap-number.h"
#include "src/objects/js-generator.h"
#include "src/objects/objects-inl.h"
#include "src/objects/smi.h"
#include "src/runtime/runtime.h"

#if V8_ENABLE_WEBASSEMBLY
#include "src/wasm/wasm-linkage.h"
#include "src/wasm/wasm-objects.h"
#endif  // V8_ENABLE_WEBASSEMBLY

namespace v8 {
namespace internal {

#define __ ACCESS_MASM(masm)

void Builtins::Generate_Adaptor(MacroAssembler* masm, Address address) {
  __ li(kJavaScriptCallExtraArg1Register, ExternalReference::Create(address));
  __ Jump(BUILTIN_CODE(masm->isolate(), AdaptorWithBuiltinExitFrame),
          RelocInfo::CODE_TARGET);
}

static void GenerateTailCallToReturnedCode(MacroAssembler* masm,
                                           Runtime::FunctionId function_id) {
  // ----------- S t a t e -------------
  //  -- a0 : actual argument count
  //  -- a1 : target function (preserved for callee)
  //  -- a3 : new target (preserved for callee)
  // -----------------------------------
  {
    FrameScope scope(masm, StackFrame::INTERNAL);
    // Push a copy of the target function, the new target and the actual
    // argument count.
    // Push function as parameter to the runtime call.
    __ SmiTag(kJavaScriptCallArgCountRegister);
    __ Push(kJavaScriptCallTargetRegister, kJavaScriptCallNewTargetRegister,
            kJavaScriptCallArgCountRegister, kJavaScriptCallTargetRegister);

    __ CallRuntime(function_id, 1);
    __ LoadCodeObjectEntry(a2, a0);
    // Restore target function, new target and actual argument count.
    __ Pop(kJavaScriptCallTargetRegister, kJavaScriptCallNewTargetRegister,
           kJavaScriptCallArgCountRegister);
    __ SmiUntag(kJavaScriptCallArgCountRegister);
  }

  static_assert(kJavaScriptCallCodeStartRegister == a2, "ABI mismatch");
  __ Jump(a2);
}

namespace {

74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101
enum class ArgumentsElementType {
  kRaw,    // Push arguments as they are.
  kHandle  // Dereference arguments before pushing.
};

void Generate_PushArguments(MacroAssembler* masm, Register array, Register argc,
                            Register scratch, Register scratch2,
                            ArgumentsElementType element_type) {
  DCHECK(!AreAliased(array, argc, scratch));
  Label loop, entry;
  if (kJSArgcIncludesReceiver) {
    __ Sub_d(scratch, argc, Operand(kJSArgcReceiverSlots));
  } else {
    __ mov(scratch, argc);
  }
  __ Branch(&entry);
  __ bind(&loop);
  __ Alsl_d(scratch2, scratch, array, kPointerSizeLog2, t7);
  __ Ld_d(scratch2, MemOperand(scratch2, 0));
  if (element_type == ArgumentsElementType::kHandle) {
    __ Ld_d(scratch2, MemOperand(scratch2, 0));
  }
  __ Push(scratch2);
  __ bind(&entry);
  __ Add_d(scratch, scratch, Operand(-1));
  __ Branch(&loop, greater_equal, scratch, Operand(zero_reg));
}

102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120
void Generate_JSBuiltinsConstructStubHelper(MacroAssembler* masm) {
  // ----------- S t a t e -------------
  //  -- a0     : number of arguments
  //  -- a1     : constructor function
  //  -- a3     : new target
  //  -- cp     : context
  //  -- ra     : return address
  //  -- sp[...]: constructor arguments
  // -----------------------------------

  // Enter a construct frame.
  {
    FrameScope scope(masm, StackFrame::CONSTRUCT);

    // Preserve the incoming parameters on the stack.
    __ SmiTag(a0);
    __ Push(cp, a0);
    __ SmiUntag(a0);

121
    // Set up pointer to first argument (skip receiver).
122 123 124 125
    __ Add_d(
        t2, fp,
        Operand(StandardFrameConstants::kCallerSPOffset + kSystemPointerSize));
    // Copy arguments and receiver to the expression stack.
126 127 128
    // t2: Pointer to start of arguments.
    // a0: Number of arguments.
    Generate_PushArguments(masm, t2, a0, t3, t0, ArgumentsElementType::kRaw);
129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145
    // The receiver for the builtin/api call.
    __ PushRoot(RootIndex::kTheHoleValue);

    // Call the function.
    // a0: number of arguments (untagged)
    // a1: constructor function
    // a3: new target
    __ InvokeFunctionWithNewTarget(a1, a3, a0, InvokeType::kCall);

    // Restore context from the frame.
    __ Ld_d(cp, MemOperand(fp, ConstructFrameConstants::kContextOffset));
    // Restore smi-tagged arguments count from the frame.
    __ Ld_d(t3, MemOperand(fp, ConstructFrameConstants::kLengthOffset));
    // Leave construct frame.
  }

  // Remove caller arguments from the stack and return.
146 147 148 149 150
  __ DropArguments(t3, TurboAssembler::kCountIsSmi,
                   kJSArgcIncludesReceiver
                       ? TurboAssembler::kCountIncludesReceiver
                       : TurboAssembler::kCountExcludesReceiver,
                   t3);
151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255
  __ Ret();
}

}  // namespace

// The construct stub for ES5 constructor functions and ES6 class constructors.
void Builtins::Generate_JSConstructStubGeneric(MacroAssembler* masm) {
  // ----------- S t a t e -------------
  //  --      a0: number of arguments (untagged)
  //  --      a1: constructor function
  //  --      a3: new target
  //  --      cp: context
  //  --      ra: return address
  //  -- sp[...]: constructor arguments
  // -----------------------------------

  // Enter a construct frame.
  FrameScope scope(masm, StackFrame::MANUAL);
  Label post_instantiation_deopt_entry, not_create_implicit_receiver;
  __ EnterFrame(StackFrame::CONSTRUCT);

  // Preserve the incoming parameters on the stack.
  __ SmiTag(a0);
  __ Push(cp, a0, a1);
  __ PushRoot(RootIndex::kUndefinedValue);
  __ Push(a3);

  // ----------- S t a t e -------------
  //  --        sp[0*kPointerSize]: new target
  //  --        sp[1*kPointerSize]: padding
  //  -- a1 and sp[2*kPointerSize]: constructor function
  //  --        sp[3*kPointerSize]: number of arguments (tagged)
  //  --        sp[4*kPointerSize]: context
  // -----------------------------------

  __ Ld_d(t2, FieldMemOperand(a1, JSFunction::kSharedFunctionInfoOffset));
  __ Ld_wu(t2, FieldMemOperand(t2, SharedFunctionInfo::kFlagsOffset));
  __ DecodeField<SharedFunctionInfo::FunctionKindBits>(t2);
  __ JumpIfIsInRange(t2, kDefaultDerivedConstructor, kDerivedConstructor,
                     &not_create_implicit_receiver);

  // If not derived class constructor: Allocate the new receiver object.
  __ IncrementCounter(masm->isolate()->counters()->constructed_objects(), 1, t2,
                      t3);
  __ Call(BUILTIN_CODE(masm->isolate(), FastNewObject), RelocInfo::CODE_TARGET);
  __ Branch(&post_instantiation_deopt_entry);

  // Else: use TheHoleValue as receiver for constructor call
  __ bind(&not_create_implicit_receiver);
  __ LoadRoot(a0, RootIndex::kTheHoleValue);

  // ----------- S t a t e -------------
  //  --                          a0: receiver
  //  -- Slot 4 / sp[0*kPointerSize]: new target
  //  -- Slot 3 / sp[1*kPointerSize]: padding
  //  -- Slot 2 / sp[2*kPointerSize]: constructor function
  //  -- Slot 1 / sp[3*kPointerSize]: number of arguments (tagged)
  //  -- Slot 0 / sp[4*kPointerSize]: context
  // -----------------------------------
  // Deoptimizer enters here.
  masm->isolate()->heap()->SetConstructStubCreateDeoptPCOffset(
      masm->pc_offset());
  __ bind(&post_instantiation_deopt_entry);

  // Restore new target.
  __ Pop(a3);

  // Push the allocated receiver to the stack.
  __ Push(a0);

  // We need two copies because we may have to return the original one
  // and the calling conventions dictate that the called function pops the
  // receiver. The second copy is pushed after the arguments, we saved in a6
  // since a0 will store the return value of callRuntime.
  __ mov(a6, a0);

  // Set up pointer to last argument.
  __ Add_d(
      t2, fp,
      Operand(StandardFrameConstants::kCallerSPOffset + kSystemPointerSize));

  // ----------- S t a t e -------------
  //  --                 r3: new target
  //  -- sp[0*kPointerSize]: implicit receiver
  //  -- sp[1*kPointerSize]: implicit receiver
  //  -- sp[2*kPointerSize]: padding
  //  -- sp[3*kPointerSize]: constructor function
  //  -- sp[4*kPointerSize]: number of arguments (tagged)
  //  -- sp[5*kPointerSize]: context
  // -----------------------------------

  // Restore constructor function and argument count.
  __ Ld_d(a1, MemOperand(fp, ConstructFrameConstants::kConstructorOffset));
  __ Ld_d(a0, MemOperand(fp, ConstructFrameConstants::kLengthOffset));
  __ SmiUntag(a0);

  Label stack_overflow;
  __ StackOverflowCheck(a0, t0, t1, &stack_overflow);

  // TODO(victorgomes): When the arguments adaptor is completely removed, we
  // should get the formal parameter count and copy the arguments in its
  // correct position (including any undefined), instead of delaying this to
  // InvokeFunction.

  // Copy arguments and receiver to the expression stack.
256 257 258
  // t2: Pointer to start of argument.
  // a0: Number of arguments.
  Generate_PushArguments(masm, t2, a0, t0, t1, ArgumentsElementType::kRaw);
259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303
  // We need two copies because we may have to return the original one
  // and the calling conventions dictate that the called function pops the
  // receiver. The second copy is pushed after the arguments,
  __ Push(a6);

  // Call the function.
  __ InvokeFunctionWithNewTarget(a1, a3, a0, InvokeType::kCall);

  // ----------- S t a t e -------------
  //  --                 s0: constructor result
  //  -- sp[0*kPointerSize]: implicit receiver
  //  -- sp[1*kPointerSize]: padding
  //  -- sp[2*kPointerSize]: constructor function
  //  -- sp[3*kPointerSize]: number of arguments
  //  -- sp[4*kPointerSize]: context
  // -----------------------------------

  // Store offset of return address for deoptimizer.
  masm->isolate()->heap()->SetConstructStubInvokeDeoptPCOffset(
      masm->pc_offset());

  // If the result is an object (in the ECMA sense), we should get rid
  // of the receiver and use the result; see ECMA-262 section 13.2.2-7
  // on page 74.
  Label use_receiver, do_throw, leave_and_return, check_receiver;

  // If the result is undefined, we jump out to using the implicit receiver.
  __ JumpIfNotRoot(a0, RootIndex::kUndefinedValue, &check_receiver);

  // Otherwise we do a smi check and fall through to check if the return value
  // is a valid receiver.

  // Throw away the result of the constructor invocation and use the
  // on-stack receiver as the result.
  __ bind(&use_receiver);
  __ Ld_d(a0, MemOperand(sp, 0 * kPointerSize));
  __ JumpIfRoot(a0, RootIndex::kTheHoleValue, &do_throw);

  __ bind(&leave_and_return);
  // Restore smi-tagged arguments count from the frame.
  __ Ld_d(a1, MemOperand(fp, ConstructFrameConstants::kLengthOffset));
  // Leave construct frame.
  __ LeaveFrame(StackFrame::CONSTRUCT);

  // Remove caller arguments from the stack and return.
304 305 306 307 308
  __ DropArguments(a1, TurboAssembler::kCountIsSmi,
                   kJSArgcIncludesReceiver
                       ? TurboAssembler::kCountIncludesReceiver
                       : TurboAssembler::kCountExcludesReceiver,
                   a4);
309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338
  __ Ret();

  __ bind(&check_receiver);
  __ JumpIfSmi(a0, &use_receiver);

  // If the type of the result (stored in its map) is less than
  // FIRST_JS_RECEIVER_TYPE, it is not an object in the ECMA sense.
  __ GetObjectType(a0, t2, t2);
  STATIC_ASSERT(LAST_JS_RECEIVER_TYPE == LAST_TYPE);
  __ Branch(&leave_and_return, greater_equal, t2,
            Operand(FIRST_JS_RECEIVER_TYPE));
  __ Branch(&use_receiver);

  __ bind(&do_throw);
  // Restore the context from the frame.
  __ Ld_d(cp, MemOperand(fp, ConstructFrameConstants::kContextOffset));
  __ CallRuntime(Runtime::kThrowConstructorReturnedNonObject);
  __ break_(0xCC);

  __ bind(&stack_overflow);
  // Restore the context from the frame.
  __ Ld_d(cp, MemOperand(fp, ConstructFrameConstants::kContextOffset));
  __ CallRuntime(Runtime::kThrowStackOverflow);
  __ break_(0xCC);
}

void Builtins::Generate_JSBuiltinsConstructStub(MacroAssembler* masm) {
  Generate_JSBuiltinsConstructStubHelper(masm);
}

339 340 341 342 343 344 345 346 347 348
static void AssertCodeIsBaseline(MacroAssembler* masm, Register code,
                                 Register scratch) {
  DCHECK(!AreAliased(code, scratch));
  // Verify that the code kind is baseline code via the CodeKind.
  __ Ld_d(scratch, FieldMemOperand(code, Code::kFlagsOffset));
  __ DecodeField<Code::KindField>(scratch);
  __ Assert(eq, AbortReason::kExpectedBaselineData, scratch,
            Operand(static_cast<int>(CodeKind::BASELINE)));
}

349 350 351 352 353 354 355 356 357
// TODO(v8:11429): Add a path for "not_compiled" and unify the two uses under
// the more general dispatch.
static void GetSharedFunctionInfoBytecodeOrBaseline(MacroAssembler* masm,
                                                    Register sfi_data,
                                                    Register scratch1,
                                                    Label* is_baseline) {
  Label done;

  __ GetObjectType(sfi_data, scratch1, scratch1);
358 359 360 361 362 363 364 365 366
  if (FLAG_debug_code) {
    Label not_baseline;
    __ Branch(&not_baseline, ne, scratch1, Operand(CODET_TYPE));
    AssertCodeIsBaseline(masm, sfi_data, scratch1);
    __ Branch(is_baseline);
    __ bind(&not_baseline);
  } else {
    __ Branch(is_baseline, eq, scratch1, Operand(CODET_TYPE));
  }
367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429
  __ Branch(&done, ne, scratch1, Operand(INTERPRETER_DATA_TYPE));
  __ Ld_d(sfi_data,
          FieldMemOperand(sfi_data, InterpreterData::kBytecodeArrayOffset));

  __ bind(&done);
}

// static
void Builtins::Generate_ResumeGeneratorTrampoline(MacroAssembler* masm) {
  // ----------- S t a t e -------------
  //  -- a0 : the value to pass to the generator
  //  -- a1 : the JSGeneratorObject to resume
  //  -- ra : return address
  // -----------------------------------
  // Store input value into generator object.
  __ St_d(a0, FieldMemOperand(a1, JSGeneratorObject::kInputOrDebugPosOffset));
  __ RecordWriteField(a1, JSGeneratorObject::kInputOrDebugPosOffset, a0,
                      kRAHasNotBeenSaved, SaveFPRegsMode::kIgnore);
  // Check that a1 is still valid, RecordWrite might have clobbered it.
  __ AssertGeneratorObject(a1);

  // Load suspended function and context.
  __ Ld_d(a4, FieldMemOperand(a1, JSGeneratorObject::kFunctionOffset));
  __ Ld_d(cp, FieldMemOperand(a4, JSFunction::kContextOffset));

  // Flood function if we are stepping.
  Label prepare_step_in_if_stepping, prepare_step_in_suspended_generator;
  Label stepping_prepared;
  ExternalReference debug_hook =
      ExternalReference::debug_hook_on_function_call_address(masm->isolate());
  __ li(a5, debug_hook);
  __ Ld_b(a5, MemOperand(a5, 0));
  __ Branch(&prepare_step_in_if_stepping, ne, a5, Operand(zero_reg));

  // Flood function if we need to continue stepping in the suspended generator.
  ExternalReference debug_suspended_generator =
      ExternalReference::debug_suspended_generator_address(masm->isolate());
  __ li(a5, debug_suspended_generator);
  __ Ld_d(a5, MemOperand(a5, 0));
  __ Branch(&prepare_step_in_suspended_generator, eq, a1, Operand(a5));
  __ bind(&stepping_prepared);

  // Check the stack for overflow. We are not trying to catch interruptions
  // (i.e. debug break and preemption) here, so check the "real stack limit".
  Label stack_overflow;
  __ LoadStackLimit(kScratchReg,
                    MacroAssembler::StackLimitKind::kRealStackLimit);
  __ Branch(&stack_overflow, lo, sp, Operand(kScratchReg));

  // ----------- S t a t e -------------
  //  -- a1    : the JSGeneratorObject to resume
  //  -- a4    : generator function
  //  -- cp    : generator context
  //  -- ra    : return address
  // -----------------------------------

  // Push holes for arguments to generator function. Since the parser forced
  // context allocation for any variables in generators, the actual argument
  // values have already been copied into the context and these dummy values
  // will never be used.
  __ Ld_d(a3, FieldMemOperand(a4, JSFunction::kSharedFunctionInfoOffset));
  __ Ld_hu(
      a3, FieldMemOperand(a3, SharedFunctionInfo::kFormalParameterCountOffset));
430 431 432
  if (kJSArgcIncludesReceiver) {
    __ Sub_d(a3, a3, Operand(kJSArgcReceiverSlots));
  }
433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764
  __ Ld_d(t1, FieldMemOperand(
                  a1, JSGeneratorObject::kParametersAndRegistersOffset));
  {
    Label done_loop, loop;
    __ bind(&loop);
    __ Sub_d(a3, a3, Operand(1));
    __ Branch(&done_loop, lt, a3, Operand(zero_reg));
    __ Alsl_d(kScratchReg, a3, t1, kPointerSizeLog2, t7);
    __ Ld_d(kScratchReg, FieldMemOperand(kScratchReg, FixedArray::kHeaderSize));
    __ Push(kScratchReg);
    __ Branch(&loop);
    __ bind(&done_loop);
    // Push receiver.
    __ Ld_d(kScratchReg,
            FieldMemOperand(a1, JSGeneratorObject::kReceiverOffset));
    __ Push(kScratchReg);
  }

  // Underlying function needs to have bytecode available.
  if (FLAG_debug_code) {
    Label is_baseline;
    __ Ld_d(a3, FieldMemOperand(a4, JSFunction::kSharedFunctionInfoOffset));
    __ Ld_d(a3, FieldMemOperand(a3, SharedFunctionInfo::kFunctionDataOffset));
    GetSharedFunctionInfoBytecodeOrBaseline(masm, a3, t5, &is_baseline);
    __ GetObjectType(a3, a3, a3);
    __ Assert(eq, AbortReason::kMissingBytecodeArray, a3,
              Operand(BYTECODE_ARRAY_TYPE));
    __ bind(&is_baseline);
  }

  // Resume (Ignition/TurboFan) generator object.
  {
    __ Ld_d(a0, FieldMemOperand(a4, JSFunction::kSharedFunctionInfoOffset));
    __ Ld_hu(a0, FieldMemOperand(
                     a0, SharedFunctionInfo::kFormalParameterCountOffset));
    // We abuse new.target both to indicate that this is a resume call and to
    // pass in the generator object.  In ordinary calls, new.target is always
    // undefined because generator functions are non-constructable.
    __ Move(a3, a1);
    __ Move(a1, a4);
    static_assert(kJavaScriptCallCodeStartRegister == a2, "ABI mismatch");
    __ Ld_d(a2, FieldMemOperand(a1, JSFunction::kCodeOffset));
    __ JumpCodeObject(a2);
  }

  __ bind(&prepare_step_in_if_stepping);
  {
    FrameScope scope(masm, StackFrame::INTERNAL);
    __ Push(a1, a4);
    // Push hole as receiver since we do not use it for stepping.
    __ PushRoot(RootIndex::kTheHoleValue);
    __ CallRuntime(Runtime::kDebugOnFunctionCall);
    __ Pop(a1);
  }
  __ Ld_d(a4, FieldMemOperand(a1, JSGeneratorObject::kFunctionOffset));
  __ Branch(&stepping_prepared);

  __ bind(&prepare_step_in_suspended_generator);
  {
    FrameScope scope(masm, StackFrame::INTERNAL);
    __ Push(a1);
    __ CallRuntime(Runtime::kDebugPrepareStepInSuspendedGenerator);
    __ Pop(a1);
  }
  __ Ld_d(a4, FieldMemOperand(a1, JSGeneratorObject::kFunctionOffset));
  __ Branch(&stepping_prepared);

  __ bind(&stack_overflow);
  {
    FrameScope scope(masm, StackFrame::INTERNAL);
    __ CallRuntime(Runtime::kThrowStackOverflow);
    __ break_(0xCC);  // This should be unreachable.
  }
}

void Builtins::Generate_ConstructedNonConstructable(MacroAssembler* masm) {
  FrameScope scope(masm, StackFrame::INTERNAL);
  __ Push(a1);
  __ CallRuntime(Runtime::kThrowConstructedNonConstructable);
}

// Clobbers scratch1 and scratch2; preserves all other registers.
static void Generate_CheckStackOverflow(MacroAssembler* masm, Register argc,
                                        Register scratch1, Register scratch2) {
  // Check the stack for overflow. We are not trying to catch
  // interruptions (e.g. debug break and preemption) here, so the "real stack
  // limit" is checked.
  Label okay;
  __ LoadStackLimit(scratch1, MacroAssembler::StackLimitKind::kRealStackLimit);
  // Make a2 the space we have left. The stack might already be overflowed
  // here which will cause r2 to become negative.
  __ sub_d(scratch1, sp, scratch1);
  // Check if the arguments will overflow the stack.
  __ slli_d(scratch2, argc, kPointerSizeLog2);
  __ Branch(&okay, gt, scratch1, Operand(scratch2));  // Signed comparison.

  // Out of stack space.
  __ CallRuntime(Runtime::kThrowStackOverflow);

  __ bind(&okay);
}

namespace {

// Called with the native C calling convention. The corresponding function
// signature is either:
//
//   using JSEntryFunction = GeneratedCode<Address(
//       Address root_register_value, Address new_target, Address target,
//       Address receiver, intptr_t argc, Address** args)>;
// or
//   using JSEntryFunction = GeneratedCode<Address(
//       Address root_register_value, MicrotaskQueue* microtask_queue)>;
void Generate_JSEntryVariant(MacroAssembler* masm, StackFrame::Type type,
                             Builtin entry_trampoline) {
  Label invoke, handler_entry, exit;

  {
    NoRootArrayScope no_root_array(masm);

    // Registers:
    //  either
    //   a0: root register value
    //   a1: entry address
    //   a2: function
    //   a3: receiver
    //   a4: argc
    //   a5: argv
    //  or
    //   a0: root register value
    //   a1: microtask_queue

    // Save callee saved registers on the stack.
    __ MultiPush(kCalleeSaved | ra.bit());

    // Save callee-saved FPU registers.
    __ MultiPushFPU(kCalleeSavedFPU);
    // Set up the reserved register for 0.0.
    __ Move(kDoubleRegZero, 0.0);

    // Initialize the root register.
    // C calling convention. The first argument is passed in a0.
    __ mov(kRootRegister, a0);
  }

  // a1: entry address
  // a2: function
  // a3: receiver
  // a4: argc
  // a5: argv

  // We build an EntryFrame.
  __ li(s1, Operand(-1));  // Push a bad frame pointer to fail if it is used.
  __ li(s2, Operand(StackFrame::TypeToMarker(type)));
  __ li(s3, Operand(StackFrame::TypeToMarker(type)));
  ExternalReference c_entry_fp = ExternalReference::Create(
      IsolateAddressId::kCEntryFPAddress, masm->isolate());
  __ li(s5, c_entry_fp);
  __ Ld_d(s4, MemOperand(s5, 0));
  __ Push(s1, s2, s3, s4);

  // Clear c_entry_fp, now we've pushed its previous value to the stack.
  // If the c_entry_fp is not already zero and we don't clear it, the
  // SafeStackFrameIterator will assume we are executing C++ and miss the JS
  // frames on top.
  __ St_d(zero_reg, MemOperand(s5, 0));

  // Set up frame pointer for the frame to be pushed.
  __ addi_d(fp, sp, -EntryFrameConstants::kCallerFPOffset);

  // Registers:
  //  either
  //   a1: entry address
  //   a2: function
  //   a3: receiver
  //   a4: argc
  //   a5: argv
  //  or
  //   a1: microtask_queue
  //
  // Stack:
  // caller fp          |
  // function slot      | entry frame
  // context slot       |
  // bad fp (0xFF...F)  |
  // callee saved registers + ra
  // [ O32: 4 args slots]
  // args

  // If this is the outermost JS call, set js_entry_sp value.
  Label non_outermost_js;
  ExternalReference js_entry_sp = ExternalReference::Create(
      IsolateAddressId::kJSEntrySPAddress, masm->isolate());
  __ li(s1, js_entry_sp);
  __ Ld_d(s2, MemOperand(s1, 0));
  __ Branch(&non_outermost_js, ne, s2, Operand(zero_reg));
  __ St_d(fp, MemOperand(s1, 0));
  __ li(s3, Operand(StackFrame::OUTERMOST_JSENTRY_FRAME));
  Label cont;
  __ b(&cont);
  __ nop();  // Branch delay slot nop.
  __ bind(&non_outermost_js);
  __ li(s3, Operand(StackFrame::INNER_JSENTRY_FRAME));
  __ bind(&cont);
  __ Push(s3);

  // Jump to a faked try block that does the invoke, with a faked catch
  // block that sets the pending exception.
  __ jmp(&invoke);
  __ bind(&handler_entry);

  // Store the current pc as the handler offset. It's used later to create the
  // handler table.
  masm->isolate()->builtins()->SetJSEntryHandlerOffset(handler_entry.pos());

  // Caught exception: Store result (exception) in the pending exception
  // field in the JSEnv and return a failure sentinel.  Coming in here the
  // fp will be invalid because the PushStackHandler below sets it to 0 to
  // signal the existence of the JSEntry frame.
  __ li(s1, ExternalReference::Create(
                IsolateAddressId::kPendingExceptionAddress, masm->isolate()));
  __ St_d(a0,
          MemOperand(s1, 0));  // We come back from 'invoke'. result is in a0.
  __ LoadRoot(a0, RootIndex::kException);
  __ b(&exit);  // b exposes branch delay slot.
  __ nop();     // Branch delay slot nop.

  // Invoke: Link this frame into the handler chain.
  __ bind(&invoke);
  __ PushStackHandler();
  // If an exception not caught by another handler occurs, this handler
  // returns control to the code after the bal(&invoke) above, which
  // restores all kCalleeSaved registers (including cp and fp) to their
  // saved values before returning a failure to C.
  //
  // Registers:
  //  either
  //   a0: root register value
  //   a1: entry address
  //   a2: function
  //   a3: receiver
  //   a4: argc
  //   a5: argv
  //  or
  //   a0: root register value
  //   a1: microtask_queue
  //
  // Stack:
  // handler frame
  // entry frame
  // callee saved registers + ra
  // [ O32: 4 args slots]
  // args
  //
  // Invoke the function by calling through JS entry trampoline builtin and
  // pop the faked function when we return.

  Handle<Code> trampoline_code =
      masm->isolate()->builtins()->code_handle(entry_trampoline);
  __ Call(trampoline_code, RelocInfo::CODE_TARGET);

  // Unlink this frame from the handler chain.
  __ PopStackHandler();

  __ bind(&exit);  // a0 holds result
  // Check if the current stack frame is marked as the outermost JS frame.
  Label non_outermost_js_2;
  __ Pop(a5);
  __ Branch(&non_outermost_js_2, ne, a5,
            Operand(StackFrame::OUTERMOST_JSENTRY_FRAME));
  __ li(a5, js_entry_sp);
  __ St_d(zero_reg, MemOperand(a5, 0));
  __ bind(&non_outermost_js_2);

  // Restore the top frame descriptors from the stack.
  __ Pop(a5);
  __ li(a4, ExternalReference::Create(IsolateAddressId::kCEntryFPAddress,
                                      masm->isolate()));
  __ St_d(a5, MemOperand(a4, 0));

  // Reset the stack to the callee saved registers.
  __ addi_d(sp, sp, -EntryFrameConstants::kCallerFPOffset);

  // Restore callee-saved fpu registers.
  __ MultiPopFPU(kCalleeSavedFPU);

  // Restore callee saved registers from the stack.
  __ MultiPop(kCalleeSaved | ra.bit());
  // Return.
  __ Jump(ra);
}

}  // namespace

void Builtins::Generate_JSEntry(MacroAssembler* masm) {
  Generate_JSEntryVariant(masm, StackFrame::ENTRY, Builtin::kJSEntryTrampoline);
}

void Builtins::Generate_JSConstructEntry(MacroAssembler* masm) {
  Generate_JSEntryVariant(masm, StackFrame::CONSTRUCT_ENTRY,
                          Builtin::kJSConstructEntryTrampoline);
}

void Builtins::Generate_JSRunMicrotasksEntry(MacroAssembler* masm) {
  Generate_JSEntryVariant(masm, StackFrame::ENTRY,
                          Builtin::kRunMicrotasksTrampoline);
}

static void Generate_JSEntryTrampolineHelper(MacroAssembler* masm,
                                             bool is_construct) {
  // ----------- S t a t e -------------
  //  -- a1: new.target
  //  -- a2: function
  //  -- a3: receiver_pointer
  //  -- a4: argc
  //  -- a5: argv
  // -----------------------------------

  // Enter an internal frame.
  {
    FrameScope scope(masm, StackFrame::INTERNAL);

    // Setup the context (we need to use the caller context from the isolate).
    ExternalReference context_address = ExternalReference::Create(
        IsolateAddressId::kContextAddress, masm->isolate());
    __ li(cp, context_address);
    __ Ld_d(cp, MemOperand(cp, 0));

    // Push the function and the receiver onto the stack.
    __ Push(a2);

    // Check if we have enough stack space to push all arguments.
765 766 767 768 769
    if (kJSArgcIncludesReceiver) {
      __ mov(a6, a4);
    } else {
      __ addi_d(a6, a4, 1);
    }
770 771
    Generate_CheckStackOverflow(masm, a6, a0, s2);

772
    // Copy arguments to the stack.
773 774
    // a4: argc
    // a5: argv, i.e. points to first arg
775
    Generate_PushArguments(masm, a5, a4, s1, s2, ArgumentsElementType::kHandle);
776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849

    // Push the receive.
    __ Push(a3);

    // a0: argc
    // a1: function
    // a3: new.target
    __ mov(a3, a1);
    __ mov(a1, a2);
    __ mov(a0, a4);

    // Initialize all JavaScript callee-saved registers, since they will be seen
    // by the garbage collector as part of handlers.
    __ LoadRoot(a4, RootIndex::kUndefinedValue);
    __ mov(a5, a4);
    __ mov(s1, a4);
    __ mov(s2, a4);
    __ mov(s3, a4);
    __ mov(s4, a4);
    __ mov(s5, a4);
    // s6 holds the root address. Do not clobber.
    // s7 is cp. Do not init.

    // Invoke the code.
    Handle<Code> builtin = is_construct
                               ? BUILTIN_CODE(masm->isolate(), Construct)
                               : masm->isolate()->builtins()->Call();
    __ Call(builtin, RelocInfo::CODE_TARGET);

    // Leave internal frame.
  }
  __ Jump(ra);
}

void Builtins::Generate_JSEntryTrampoline(MacroAssembler* masm) {
  Generate_JSEntryTrampolineHelper(masm, false);
}

void Builtins::Generate_JSConstructEntryTrampoline(MacroAssembler* masm) {
  Generate_JSEntryTrampolineHelper(masm, true);
}

void Builtins::Generate_RunMicrotasksTrampoline(MacroAssembler* masm) {
  // a1: microtask_queue
  __ mov(RunMicrotasksDescriptor::MicrotaskQueueRegister(), a1);
  __ Jump(BUILTIN_CODE(masm->isolate(), RunMicrotasks), RelocInfo::CODE_TARGET);
}

static void ReplaceClosureCodeWithOptimizedCode(MacroAssembler* masm,
                                                Register optimized_code,
                                                Register closure) {
  DCHECK(!AreAliased(optimized_code, closure));
  // Store code entry in the closure.
  __ St_d(optimized_code, FieldMemOperand(closure, JSFunction::kCodeOffset));
  __ RecordWriteField(closure, JSFunction::kCodeOffset, optimized_code,
                      kRAHasNotBeenSaved, SaveFPRegsMode::kIgnore,
                      RememberedSetAction::kOmit, SmiCheck::kOmit);
}

static void LeaveInterpreterFrame(MacroAssembler* masm, Register scratch1,
                                  Register scratch2) {
  Register params_size = scratch1;

  // Get the size of the formal parameters + receiver (in bytes).
  __ Ld_d(params_size,
          MemOperand(fp, InterpreterFrameConstants::kBytecodeArrayFromFp));
  __ Ld_w(params_size,
          FieldMemOperand(params_size, BytecodeArray::kParameterSizeOffset));

  Register actual_params_size = scratch2;
  // Compute the size of the actual parameters + receiver (in bytes).
  __ Ld_d(actual_params_size,
          MemOperand(fp, StandardFrameConstants::kArgCOffset));
  __ slli_d(actual_params_size, actual_params_size, kPointerSizeLog2);
850 851 852 853
  if (!kJSArgcIncludesReceiver) {
    __ Add_d(actual_params_size, actual_params_size,
             Operand(kSystemPointerSize));
  }
854 855 856 857 858 859 860 861 862 863

  // If actual is bigger than formal, then we should use it to free up the stack
  // arguments.
  __ slt(t2, params_size, actual_params_size);
  __ Movn(params_size, actual_params_size, t2);

  // Leave the frame (also dropping the register file).
  __ LeaveFrame(StackFrame::INTERPRETED);

  // Drop receiver + arguments.
864 865
  __ DropArguments(params_size, TurboAssembler::kCountIsBytes,
                   TurboAssembler::kCountIncludesReceiver);
866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231
}

// Tail-call |function_id| if |actual_marker| == |expected_marker|
static void TailCallRuntimeIfMarkerEquals(MacroAssembler* masm,
                                          Register actual_marker,
                                          OptimizationMarker expected_marker,
                                          Runtime::FunctionId function_id) {
  Label no_match;
  __ Branch(&no_match, ne, actual_marker, Operand(expected_marker));
  GenerateTailCallToReturnedCode(masm, function_id);
  __ bind(&no_match);
}

static void TailCallOptimizedCodeSlot(MacroAssembler* masm,
                                      Register optimized_code_entry) {
  // ----------- S t a t e -------------
  //  -- a0 : actual argument count
  //  -- a3 : new target (preserved for callee if needed, and caller)
  //  -- a1 : target function (preserved for callee if needed, and caller)
  // -----------------------------------
  DCHECK(!AreAliased(optimized_code_entry, a1, a3));

  Register closure = a1;
  Label heal_optimized_code_slot;

  // If the optimized code is cleared, go to runtime to update the optimization
  // marker field.
  __ LoadWeakValue(optimized_code_entry, optimized_code_entry,
                   &heal_optimized_code_slot);

  // Check if the optimized code is marked for deopt. If it is, call the
  // runtime to clear it.
  __ Ld_d(a6, FieldMemOperand(optimized_code_entry,
                              Code::kCodeDataContainerOffset));
  __ Ld_w(a6, FieldMemOperand(a6, CodeDataContainer::kKindSpecificFlagsOffset));
  __ And(a6, a6, Operand(1 << Code::kMarkedForDeoptimizationBit));
  __ Branch(&heal_optimized_code_slot, ne, a6, Operand(zero_reg));

  // Optimized code is good, get it into the closure and link the closure into
  // the optimized functions list, then tail call the optimized code.
  // The feedback vector is no longer used, so re-use it as a scratch
  // register.
  ReplaceClosureCodeWithOptimizedCode(masm, optimized_code_entry, closure);

  static_assert(kJavaScriptCallCodeStartRegister == a2, "ABI mismatch");
  __ LoadCodeObjectEntry(a2, optimized_code_entry);
  __ Jump(a2);

  // Optimized code slot contains deoptimized code or code is cleared and
  // optimized code marker isn't updated. Evict the code, update the marker
  // and re-enter the closure's code.
  __ bind(&heal_optimized_code_slot);
  GenerateTailCallToReturnedCode(masm, Runtime::kHealOptimizedCodeSlot);
}

static void MaybeOptimizeCode(MacroAssembler* masm, Register feedback_vector,
                              Register optimization_marker) {
  // ----------- S t a t e -------------
  //  -- a0 : actual argument count
  //  -- a3 : new target (preserved for callee if needed, and caller)
  //  -- a1 : target function (preserved for callee if needed, and caller)
  //  -- feedback vector (preserved for caller if needed)
  //  -- optimization_marker : a Smi containing a non-zero optimization marker.
  // -----------------------------------
  DCHECK(!AreAliased(feedback_vector, a1, a3, optimization_marker));

  // TODO(v8:8394): The logging of first execution will break if
  // feedback vectors are not allocated. We need to find a different way of
  // logging these events if required.
  TailCallRuntimeIfMarkerEquals(masm, optimization_marker,
                                OptimizationMarker::kLogFirstExecution,
                                Runtime::kFunctionFirstExecution);
  TailCallRuntimeIfMarkerEquals(masm, optimization_marker,
                                OptimizationMarker::kCompileOptimized,
                                Runtime::kCompileOptimized_NotConcurrent);
  TailCallRuntimeIfMarkerEquals(masm, optimization_marker,
                                OptimizationMarker::kCompileOptimizedConcurrent,
                                Runtime::kCompileOptimized_Concurrent);

  // Marker should be one of LogFirstExecution / CompileOptimized /
  // CompileOptimizedConcurrent. InOptimizationQueue and None shouldn't reach
  // here.
  if (FLAG_debug_code) {
    __ stop();
  }
}

// Advance the current bytecode offset. This simulates what all bytecode
// handlers do upon completion of the underlying operation. Will bail out to a
// label if the bytecode (without prefix) is a return bytecode. Will not advance
// the bytecode offset if the current bytecode is a JumpLoop, instead just
// re-executing the JumpLoop to jump to the correct bytecode.
static void AdvanceBytecodeOffsetOrReturn(MacroAssembler* masm,
                                          Register bytecode_array,
                                          Register bytecode_offset,
                                          Register bytecode, Register scratch1,
                                          Register scratch2, Register scratch3,
                                          Label* if_return) {
  Register bytecode_size_table = scratch1;

  // The bytecode offset value will be increased by one in wide and extra wide
  // cases. In the case of having a wide or extra wide JumpLoop bytecode, we
  // will restore the original bytecode. In order to simplify the code, we have
  // a backup of it.
  Register original_bytecode_offset = scratch3;
  DCHECK(!AreAliased(bytecode_array, bytecode_offset, bytecode,
                     bytecode_size_table, original_bytecode_offset));
  __ Move(original_bytecode_offset, bytecode_offset);
  __ li(bytecode_size_table, ExternalReference::bytecode_size_table_address());

  // Check if the bytecode is a Wide or ExtraWide prefix bytecode.
  Label process_bytecode, extra_wide;
  STATIC_ASSERT(0 == static_cast<int>(interpreter::Bytecode::kWide));
  STATIC_ASSERT(1 == static_cast<int>(interpreter::Bytecode::kExtraWide));
  STATIC_ASSERT(2 == static_cast<int>(interpreter::Bytecode::kDebugBreakWide));
  STATIC_ASSERT(3 ==
                static_cast<int>(interpreter::Bytecode::kDebugBreakExtraWide));
  __ Branch(&process_bytecode, hi, bytecode, Operand(3));
  __ And(scratch2, bytecode, Operand(1));
  __ Branch(&extra_wide, ne, scratch2, Operand(zero_reg));

  // Load the next bytecode and update table to the wide scaled table.
  __ Add_d(bytecode_offset, bytecode_offset, Operand(1));
  __ Add_d(scratch2, bytecode_array, bytecode_offset);
  __ Ld_bu(bytecode, MemOperand(scratch2, 0));
  __ Add_d(bytecode_size_table, bytecode_size_table,
           Operand(kByteSize * interpreter::Bytecodes::kBytecodeCount));
  __ jmp(&process_bytecode);

  __ bind(&extra_wide);
  // Load the next bytecode and update table to the extra wide scaled table.
  __ Add_d(bytecode_offset, bytecode_offset, Operand(1));
  __ Add_d(scratch2, bytecode_array, bytecode_offset);
  __ Ld_bu(bytecode, MemOperand(scratch2, 0));
  __ Add_d(bytecode_size_table, bytecode_size_table,
           Operand(2 * kByteSize * interpreter::Bytecodes::kBytecodeCount));

  __ bind(&process_bytecode);

// Bailout to the return label if this is a return bytecode.
#define JUMP_IF_EQUAL(NAME)          \
  __ Branch(if_return, eq, bytecode, \
            Operand(static_cast<int>(interpreter::Bytecode::k##NAME)));
  RETURN_BYTECODE_LIST(JUMP_IF_EQUAL)
#undef JUMP_IF_EQUAL

  // If this is a JumpLoop, re-execute it to perform the jump to the beginning
  // of the loop.
  Label end, not_jump_loop;
  __ Branch(&not_jump_loop, ne, bytecode,
            Operand(static_cast<int>(interpreter::Bytecode::kJumpLoop)));
  // We need to restore the original bytecode_offset since we might have
  // increased it to skip the wide / extra-wide prefix bytecode.
  __ Move(bytecode_offset, original_bytecode_offset);
  __ jmp(&end);

  __ bind(&not_jump_loop);
  // Otherwise, load the size of the current bytecode and advance the offset.
  __ Add_d(scratch2, bytecode_size_table, bytecode);
  __ Ld_b(scratch2, MemOperand(scratch2, 0));
  __ Add_d(bytecode_offset, bytecode_offset, scratch2);

  __ bind(&end);
}

// Read off the optimization state in the feedback vector and check if there
// is optimized code or a optimization marker that needs to be processed.
static void LoadOptimizationStateAndJumpIfNeedsProcessing(
    MacroAssembler* masm, Register optimization_state, Register feedback_vector,
    Label* has_optimized_code_or_marker) {
  ASM_CODE_COMMENT(masm);
  Register scratch = t2;
  // TODO(liuyu): Remove CHECK
  CHECK_NE(t2, optimization_state);
  CHECK_NE(t2, feedback_vector);
  __ Ld_w(optimization_state,
          FieldMemOperand(feedback_vector, FeedbackVector::kFlagsOffset));
  __ And(
      scratch, optimization_state,
      Operand(FeedbackVector::kHasOptimizedCodeOrCompileOptimizedMarkerMask));
  __ Branch(has_optimized_code_or_marker, ne, scratch, Operand(zero_reg));
}

static void MaybeOptimizeCodeOrTailCallOptimizedCodeSlot(
    MacroAssembler* masm, Register optimization_state,
    Register feedback_vector) {
  ASM_CODE_COMMENT(masm);
  Label maybe_has_optimized_code;
  // Check if optimized code marker is available
  {
    UseScratchRegisterScope temps(masm);
    Register scratch = temps.Acquire();
    __ And(
        scratch, optimization_state,
        Operand(FeedbackVector::kHasCompileOptimizedOrLogFirstExecutionMarker));
    __ Branch(&maybe_has_optimized_code, eq, scratch, Operand(zero_reg));
  }

  Register optimization_marker = optimization_state;
  __ DecodeField<FeedbackVector::OptimizationMarkerBits>(optimization_marker);
  MaybeOptimizeCode(masm, feedback_vector, optimization_marker);

  __ bind(&maybe_has_optimized_code);
  Register optimized_code_entry = optimization_state;
  __ Ld_d(optimization_marker,
          FieldMemOperand(feedback_vector,
                          FeedbackVector::kMaybeOptimizedCodeOffset));

  TailCallOptimizedCodeSlot(masm, optimized_code_entry);
}

// static
void Builtins::Generate_BaselineOutOfLinePrologue(MacroAssembler* masm) {
  UseScratchRegisterScope temps(masm);
  temps.Include(s1.bit() | s2.bit());
  temps.Exclude(t7.bit());
  auto descriptor =
      Builtins::CallInterfaceDescriptorFor(Builtin::kBaselineOutOfLinePrologue);
  Register closure = descriptor.GetRegisterParameter(
      BaselineOutOfLinePrologueDescriptor::kClosure);
  // Load the feedback vector from the closure.
  Register feedback_vector = temps.Acquire();
  __ Ld_d(feedback_vector,
          FieldMemOperand(closure, JSFunction::kFeedbackCellOffset));
  __ Ld_d(feedback_vector,
          FieldMemOperand(feedback_vector, Cell::kValueOffset));
  if (FLAG_debug_code) {
    UseScratchRegisterScope temps(masm);
    Register scratch = temps.Acquire();
    __ GetObjectType(feedback_vector, scratch, scratch);
    __ Assert(eq, AbortReason::kExpectedFeedbackVector, scratch,
              Operand(FEEDBACK_VECTOR_TYPE));
  }
  // Check for an optimization marker.
  Label has_optimized_code_or_marker;
  Register optimization_state = no_reg;
  {
    UseScratchRegisterScope temps(masm);
    optimization_state = temps.Acquire();
    // optimization_state will be used only in |has_optimized_code_or_marker|
    // and outside it can be reused.
    LoadOptimizationStateAndJumpIfNeedsProcessing(
        masm, optimization_state, feedback_vector,
        &has_optimized_code_or_marker);
  }
  // Increment invocation count for the function.
  {
    UseScratchRegisterScope temps(masm);
    Register invocation_count = temps.Acquire();
    __ Ld_w(invocation_count,
            FieldMemOperand(feedback_vector,
                            FeedbackVector::kInvocationCountOffset));
    __ Add_w(invocation_count, invocation_count, Operand(1));
    __ St_w(invocation_count,
            FieldMemOperand(feedback_vector,
                            FeedbackVector::kInvocationCountOffset));
  }

  FrameScope frame_scope(masm, StackFrame::MANUAL);
  {
    ASM_CODE_COMMENT_STRING(masm, "Frame Setup");
    // Normally the first thing we'd do here is Push(ra, fp), but we already
    // entered the frame in BaselineCompiler::Prologue, as we had to use the
    // value ra before the call to this BaselineOutOfLinePrologue builtin.
    Register callee_context = descriptor.GetRegisterParameter(
        BaselineOutOfLinePrologueDescriptor::kCalleeContext);
    Register callee_js_function = descriptor.GetRegisterParameter(
        BaselineOutOfLinePrologueDescriptor::kClosure);
    __ Push(callee_context, callee_js_function);
    DCHECK_EQ(callee_js_function, kJavaScriptCallTargetRegister);
    DCHECK_EQ(callee_js_function, kJSFunctionRegister);

    Register argc = descriptor.GetRegisterParameter(
        BaselineOutOfLinePrologueDescriptor::kJavaScriptCallArgCount);
    // We'll use the bytecode for both code age/OSR resetting, and pushing onto
    // the frame, so load it into a register.
    Register bytecodeArray = descriptor.GetRegisterParameter(
        BaselineOutOfLinePrologueDescriptor::kInterpreterBytecodeArray);

    // Reset code age and the OSR arming. The OSR field and BytecodeAgeOffset
    // are 8-bit fields next to each other, so we could just optimize by writing
    // a 16-bit. These static asserts guard our assumption is valid.
    STATIC_ASSERT(BytecodeArray::kBytecodeAgeOffset ==
                  BytecodeArray::kOsrLoopNestingLevelOffset + kCharSize);
    STATIC_ASSERT(BytecodeArray::kNoAgeBytecodeAge == 0);
    __ St_h(zero_reg,
            FieldMemOperand(bytecodeArray,
                            BytecodeArray::kOsrLoopNestingLevelOffset));

    __ Push(argc, bytecodeArray);

    // Baseline code frames store the feedback vector where interpreter would
    // store the bytecode offset.
    if (FLAG_debug_code) {
      UseScratchRegisterScope temps(masm);
      Register invocation_count = temps.Acquire();
      __ GetObjectType(feedback_vector, invocation_count, invocation_count);
      __ Assert(eq, AbortReason::kExpectedFeedbackVector, invocation_count,
                Operand(FEEDBACK_VECTOR_TYPE));
    }
    // Our stack is currently aligned. We have have to push something along with
    // the feedback vector to keep it that way -- we may as well start
    // initialising the register frame.
    // TODO(v8:11429,leszeks): Consider guaranteeing that this call leaves
    // `undefined` in the accumulator register, to skip the load in the baseline
    // code.
    __ Push(feedback_vector);
  }

  Label call_stack_guard;
  Register frame_size = descriptor.GetRegisterParameter(
      BaselineOutOfLinePrologueDescriptor::kStackFrameSize);
  {
    ASM_CODE_COMMENT_STRING(masm, "Stack/interrupt check");
    // Stack check. This folds the checks for both the interrupt stack limit
    // check and the real stack limit into one by just checking for the
    // interrupt limit. The interrupt limit is either equal to the real stack
    // limit or tighter. By ensuring we have space until that limit after
    // building the frame we can quickly precheck both at once.
    UseScratchRegisterScope temps(masm);
    Register sp_minus_frame_size = temps.Acquire();
    __ Sub_d(sp_minus_frame_size, sp, frame_size);
    Register interrupt_limit = temps.Acquire();
    __ LoadStackLimit(interrupt_limit,
                      MacroAssembler::StackLimitKind::kInterruptStackLimit);
    __ Branch(&call_stack_guard, Uless, sp_minus_frame_size,
              Operand(interrupt_limit));
  }

  // Do "fast" return to the caller pc in ra.
  // TODO(v8:11429): Document this frame setup better.
  __ Ret();

  __ bind(&has_optimized_code_or_marker);
  {
    ASM_CODE_COMMENT_STRING(masm, "Optimized marker check");
    UseScratchRegisterScope temps(masm);
    temps.Exclude(optimization_state);
    // Ensure the optimization_state is not allocated again.
    // Drop the frame created by the baseline call.
    __ Pop(ra, fp);
    MaybeOptimizeCodeOrTailCallOptimizedCodeSlot(masm, optimization_state,
                                                 feedback_vector);
    __ Trap();
  }

  __ bind(&call_stack_guard);
  {
    ASM_CODE_COMMENT_STRING(masm, "Stack/interrupt call");
    FrameScope frame_scope(masm, StackFrame::INTERNAL);
    // Save incoming new target or generator
    __ Push(kJavaScriptCallNewTargetRegister);
    __ SmiTag(frame_size);
    __ Push(frame_size);
    __ CallRuntime(Runtime::kStackGuardWithGap);
    __ Pop(kJavaScriptCallNewTargetRegister);
  }
  __ Ret();
  temps.Exclude(s1.bit() | s2.bit());
}

// Generate code for entering a JS function with the interpreter.
// On entry to the function the receiver and arguments have been pushed on the
// stack left to right.
//
// The live registers are:
1232
//   o a0 : actual argument count
1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458
//   o a1: the JS function object being called.
//   o a3: the incoming new target or generator object
//   o cp: our context
//   o fp: the caller's frame pointer
//   o sp: stack pointer
//   o ra: return address
//
// The function builds an interpreter frame.  See InterpreterFrameConstants in
// frame-constants.h for its layout.
void Builtins::Generate_InterpreterEntryTrampoline(MacroAssembler* masm) {
  Register closure = a1;
  Register feedback_vector = a2;

  // Get the bytecode array from the function object and load it into
  // kInterpreterBytecodeArrayRegister.
  __ Ld_d(kScratchReg,
          FieldMemOperand(closure, JSFunction::kSharedFunctionInfoOffset));
  __ Ld_d(
      kInterpreterBytecodeArrayRegister,
      FieldMemOperand(kScratchReg, SharedFunctionInfo::kFunctionDataOffset));
  Label is_baseline;
  GetSharedFunctionInfoBytecodeOrBaseline(
      masm, kInterpreterBytecodeArrayRegister, kScratchReg, &is_baseline);

  // The bytecode array could have been flushed from the shared function info,
  // if so, call into CompileLazy.
  Label compile_lazy;
  __ GetObjectType(kInterpreterBytecodeArrayRegister, kScratchReg, kScratchReg);
  __ Branch(&compile_lazy, ne, kScratchReg, Operand(BYTECODE_ARRAY_TYPE));

  // Load the feedback vector from the closure.
  __ Ld_d(feedback_vector,
          FieldMemOperand(closure, JSFunction::kFeedbackCellOffset));
  __ Ld_d(feedback_vector,
          FieldMemOperand(feedback_vector, Cell::kValueOffset));

  Label push_stack_frame;
  // Check if feedback vector is valid. If valid, check for optimized code
  // and update invocation count. Otherwise, setup the stack frame.
  __ Ld_d(a4, FieldMemOperand(feedback_vector, HeapObject::kMapOffset));
  __ Ld_hu(a4, FieldMemOperand(a4, Map::kInstanceTypeOffset));
  __ Branch(&push_stack_frame, ne, a4, Operand(FEEDBACK_VECTOR_TYPE));

  // Read off the optimization state in the feedback vector, and if there
  // is optimized code or an optimization marker, call that instead.
  Register optimization_state = a4;
  __ Ld_w(optimization_state,
          FieldMemOperand(feedback_vector, FeedbackVector::kFlagsOffset));

  // Check if the optimized code slot is not empty or has a optimization marker.
  Label has_optimized_code_or_marker;

  __ andi(t0, optimization_state,
          FeedbackVector::kHasOptimizedCodeOrCompileOptimizedMarkerMask);
  __ Branch(&has_optimized_code_or_marker, ne, t0, Operand(zero_reg));

  Label not_optimized;
  __ bind(&not_optimized);

  // Increment invocation count for the function.
  __ Ld_w(a4, FieldMemOperand(feedback_vector,
                              FeedbackVector::kInvocationCountOffset));
  __ Add_w(a4, a4, Operand(1));
  __ St_w(a4, FieldMemOperand(feedback_vector,
                              FeedbackVector::kInvocationCountOffset));

  // Open a frame scope to indicate that there is a frame on the stack.  The
  // MANUAL indicates that the scope shouldn't actually generate code to set up
  // the frame (that is done below).
  __ bind(&push_stack_frame);
  FrameScope frame_scope(masm, StackFrame::MANUAL);
  __ PushStandardFrame(closure);

  // Reset code age and the OSR arming. The OSR field and BytecodeAgeOffset are
  // 8-bit fields next to each other, so we could just optimize by writing a
  // 16-bit. These static asserts guard our assumption is valid.
  STATIC_ASSERT(BytecodeArray::kBytecodeAgeOffset ==
                BytecodeArray::kOsrLoopNestingLevelOffset + kCharSize);
  STATIC_ASSERT(BytecodeArray::kNoAgeBytecodeAge == 0);
  __ St_h(zero_reg, FieldMemOperand(kInterpreterBytecodeArrayRegister,
                                    BytecodeArray::kOsrLoopNestingLevelOffset));

  // Load initial bytecode offset.
  __ li(kInterpreterBytecodeOffsetRegister,
        Operand(BytecodeArray::kHeaderSize - kHeapObjectTag));

  // Push bytecode array and Smi tagged bytecode array offset.
  __ SmiTag(a4, kInterpreterBytecodeOffsetRegister);
  __ Push(kInterpreterBytecodeArrayRegister, a4);

  // Allocate the local and temporary register file on the stack.
  Label stack_overflow;
  {
    // Load frame size (word) from the BytecodeArray object.
    __ Ld_w(a4, FieldMemOperand(kInterpreterBytecodeArrayRegister,
                                BytecodeArray::kFrameSizeOffset));

    // Do a stack check to ensure we don't go over the limit.
    __ Sub_d(a5, sp, Operand(a4));
    __ LoadStackLimit(a2, MacroAssembler::StackLimitKind::kRealStackLimit);
    __ Branch(&stack_overflow, lo, a5, Operand(a2));

    // If ok, push undefined as the initial value for all register file entries.
    Label loop_header;
    Label loop_check;
    __ LoadRoot(kInterpreterAccumulatorRegister, RootIndex::kUndefinedValue);
    __ Branch(&loop_check);
    __ bind(&loop_header);
    // TODO(rmcilroy): Consider doing more than one push per loop iteration.
    __ Push(kInterpreterAccumulatorRegister);
    // Continue loop if not done.
    __ bind(&loop_check);
    __ Sub_d(a4, a4, Operand(kPointerSize));
    __ Branch(&loop_header, ge, a4, Operand(zero_reg));
  }

  // If the bytecode array has a valid incoming new target or generator object
  // register, initialize it with incoming value which was passed in r3.
  Label no_incoming_new_target_or_generator_register;
  __ Ld_w(a5, FieldMemOperand(
                  kInterpreterBytecodeArrayRegister,
                  BytecodeArray::kIncomingNewTargetOrGeneratorRegisterOffset));
  __ Branch(&no_incoming_new_target_or_generator_register, eq, a5,
            Operand(zero_reg));
  __ Alsl_d(a5, a5, fp, kPointerSizeLog2, t7);
  __ St_d(a3, MemOperand(a5, 0));
  __ bind(&no_incoming_new_target_or_generator_register);

  // Perform interrupt stack check.
  // TODO(solanes): Merge with the real stack limit check above.
  Label stack_check_interrupt, after_stack_check_interrupt;
  __ LoadStackLimit(a5, MacroAssembler::StackLimitKind::kInterruptStackLimit);
  __ Branch(&stack_check_interrupt, lo, sp, Operand(a5));
  __ bind(&after_stack_check_interrupt);

  // The accumulator is already loaded with undefined.

  // Load the dispatch table into a register and dispatch to the bytecode
  // handler at the current bytecode offset.
  Label do_dispatch;
  __ bind(&do_dispatch);
  __ li(kInterpreterDispatchTableRegister,
        ExternalReference::interpreter_dispatch_table_address(masm->isolate()));
  __ Add_d(t5, kInterpreterBytecodeArrayRegister,
           kInterpreterBytecodeOffsetRegister);
  __ Ld_bu(a7, MemOperand(t5, 0));
  __ Alsl_d(kScratchReg, a7, kInterpreterDispatchTableRegister,
            kPointerSizeLog2, t7);
  __ Ld_d(kJavaScriptCallCodeStartRegister, MemOperand(kScratchReg, 0));
  __ Call(kJavaScriptCallCodeStartRegister);
  masm->isolate()->heap()->SetInterpreterEntryReturnPCOffset(masm->pc_offset());

  // Any returns to the entry trampoline are either due to the return bytecode
  // or the interpreter tail calling a builtin and then a dispatch.

  // Get bytecode array and bytecode offset from the stack frame.
  __ Ld_d(kInterpreterBytecodeArrayRegister,
          MemOperand(fp, InterpreterFrameConstants::kBytecodeArrayFromFp));
  __ Ld_d(kInterpreterBytecodeOffsetRegister,
          MemOperand(fp, InterpreterFrameConstants::kBytecodeOffsetFromFp));
  __ SmiUntag(kInterpreterBytecodeOffsetRegister);

  // Either return, or advance to the next bytecode and dispatch.
  Label do_return;
  __ Add_d(a1, kInterpreterBytecodeArrayRegister,
           kInterpreterBytecodeOffsetRegister);
  __ Ld_bu(a1, MemOperand(a1, 0));
  AdvanceBytecodeOffsetOrReturn(masm, kInterpreterBytecodeArrayRegister,
                                kInterpreterBytecodeOffsetRegister, a1, a2, a3,
                                a4, &do_return);
  __ jmp(&do_dispatch);

  __ bind(&do_return);
  // The return value is in a0.
  LeaveInterpreterFrame(masm, t0, t1);
  __ Jump(ra);

  __ bind(&stack_check_interrupt);
  // Modify the bytecode offset in the stack to be kFunctionEntryBytecodeOffset
  // for the call to the StackGuard.
  __ li(kInterpreterBytecodeOffsetRegister,
        Operand(Smi::FromInt(BytecodeArray::kHeaderSize - kHeapObjectTag +
                             kFunctionEntryBytecodeOffset)));
  __ St_d(kInterpreterBytecodeOffsetRegister,
          MemOperand(fp, InterpreterFrameConstants::kBytecodeOffsetFromFp));
  __ CallRuntime(Runtime::kStackGuard);

  // After the call, restore the bytecode array, bytecode offset and accumulator
  // registers again. Also, restore the bytecode offset in the stack to its
  // previous value.
  __ Ld_d(kInterpreterBytecodeArrayRegister,
          MemOperand(fp, InterpreterFrameConstants::kBytecodeArrayFromFp));
  __ li(kInterpreterBytecodeOffsetRegister,
        Operand(BytecodeArray::kHeaderSize - kHeapObjectTag));
  __ LoadRoot(kInterpreterAccumulatorRegister, RootIndex::kUndefinedValue);

  __ SmiTag(a5, kInterpreterBytecodeOffsetRegister);
  __ St_d(a5, MemOperand(fp, InterpreterFrameConstants::kBytecodeOffsetFromFp));

  __ jmp(&after_stack_check_interrupt);

  __ bind(&has_optimized_code_or_marker);
  MaybeOptimizeCodeOrTailCallOptimizedCodeSlot(masm, optimization_state,
                                               feedback_vector);

  __ bind(&is_baseline);
  {
    // Load the feedback vector from the closure.
    __ Ld_d(feedback_vector,
            FieldMemOperand(closure, JSFunction::kFeedbackCellOffset));
    __ Ld_d(feedback_vector,
            FieldMemOperand(feedback_vector, Cell::kValueOffset));

    Label install_baseline_code;
    // Check if feedback vector is valid. If not, call prepare for baseline to
    // allocate it.
    __ Ld_d(t0, FieldMemOperand(feedback_vector, HeapObject::kMapOffset));
    __ Ld_hu(t0, FieldMemOperand(t0, Map::kInstanceTypeOffset));
    __ Branch(&install_baseline_code, ne, t0, Operand(FEEDBACK_VECTOR_TYPE));

    // Check for an optimization marker.
    LoadOptimizationStateAndJumpIfNeedsProcessing(
        masm, optimization_state, feedback_vector,
        &has_optimized_code_or_marker);

    // Load the baseline code into the closure.
1459
    __ Move(a2, kInterpreterBytecodeArrayRegister);
1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497
    static_assert(kJavaScriptCallCodeStartRegister == a2, "ABI mismatch");
    ReplaceClosureCodeWithOptimizedCode(masm, a2, closure);
    __ JumpCodeObject(a2);

    __ bind(&install_baseline_code);
    GenerateTailCallToReturnedCode(masm, Runtime::kInstallBaselineCode);
  }

  __ bind(&compile_lazy);
  GenerateTailCallToReturnedCode(masm, Runtime::kCompileLazy);
  // Unreachable code.
  __ break_(0xCC);

  __ bind(&stack_overflow);
  __ CallRuntime(Runtime::kThrowStackOverflow);
  // Unreachable code.
  __ break_(0xCC);
}

static void GenerateInterpreterPushArgs(MacroAssembler* masm, Register num_args,
                                        Register start_address,
                                        Register scratch, Register scratch2) {
  // Find the address of the last argument.
  __ Sub_d(scratch, num_args, Operand(1));
  __ slli_d(scratch, scratch, kPointerSizeLog2);
  __ Sub_d(start_address, start_address, scratch);

  // Push the arguments.
  __ PushArray(start_address, num_args, scratch, scratch2,
               TurboAssembler::PushArrayOrder::kReverse);
}

// static
void Builtins::Generate_InterpreterPushArgsThenCallImpl(
    MacroAssembler* masm, ConvertReceiverMode receiver_mode,
    InterpreterPushArgsMode mode) {
  DCHECK(mode != InterpreterPushArgsMode::kArrayFunction);
  // ----------- S t a t e -------------
1498
  //  -- a0 : the number of arguments
1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509
  //  -- a2 : the address of the first argument to be pushed. Subsequent
  //          arguments should be consecutive above this, in the same order as
  //          they are to be pushed onto the stack.
  //  -- a1 : the target to call (can be any Object).
  // -----------------------------------
  Label stack_overflow;
  if (mode == InterpreterPushArgsMode::kWithFinalSpread) {
    // The spread argument should not be pushed.
    __ Sub_d(a0, a0, Operand(1));
  }

1510 1511 1512 1513 1514 1515 1516
  const bool skip_receiver =
      receiver_mode == ConvertReceiverMode::kNullOrUndefined;
  if (kJSArgcIncludesReceiver && skip_receiver) {
    __ Sub_d(a3, a0, Operand(kJSArgcReceiverSlots));
  } else if (!kJSArgcIncludesReceiver && !skip_receiver) {
    __ Add_d(a3, a0, Operand(1));
  } else {
1517 1518 1519
    __ mov(a3, a0);
  }

1520 1521
  __ StackOverflowCheck(a3, a4, t0, &stack_overflow);

1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556
  // This function modifies a2, t0 and a4.
  GenerateInterpreterPushArgs(masm, a3, a2, a4, t0);

  if (receiver_mode == ConvertReceiverMode::kNullOrUndefined) {
    __ PushRoot(RootIndex::kUndefinedValue);
  }

  if (mode == InterpreterPushArgsMode::kWithFinalSpread) {
    // Pass the spread in the register a2.
    // a2 already points to the penultime argument, the spread
    // is below that.
    __ Ld_d(a2, MemOperand(a2, -kSystemPointerSize));
  }

  // Call the target.
  if (mode == InterpreterPushArgsMode::kWithFinalSpread) {
    __ Jump(BUILTIN_CODE(masm->isolate(), CallWithSpread),
            RelocInfo::CODE_TARGET);
  } else {
    __ Jump(masm->isolate()->builtins()->Call(ConvertReceiverMode::kAny),
            RelocInfo::CODE_TARGET);
  }

  __ bind(&stack_overflow);
  {
    __ TailCallRuntime(Runtime::kThrowStackOverflow);
    // Unreachable code.
    __ break_(0xCC);
  }
}

// static
void Builtins::Generate_InterpreterPushArgsThenConstructImpl(
    MacroAssembler* masm, InterpreterPushArgsMode mode) {
  // ----------- S t a t e -------------
1557
  // -- a0 : argument count
1558 1559 1560 1561 1562 1563
  // -- a3 : new target
  // -- a1 : constructor to call
  // -- a2 : allocation site feedback if available, undefined otherwise.
  // -- a4 : address of the first argument
  // -----------------------------------
  Label stack_overflow;
1564
  __ StackOverflowCheck(a0, a5, t0, &stack_overflow);
1565 1566 1567 1568 1569 1570

  if (mode == InterpreterPushArgsMode::kWithFinalSpread) {
    // The spread argument should not be pushed.
    __ Sub_d(a0, a0, Operand(1));
  }

1571 1572 1573 1574 1575 1576
  Register argc_without_receiver = a0;
  if (kJSArgcIncludesReceiver) {
    argc_without_receiver = a6;
    __ Sub_d(argc_without_receiver, a0, Operand(kJSArgcReceiverSlots));
  }

1577
  // Push the arguments, This function modifies t0, a4 and a5.
1578
  GenerateInterpreterPushArgs(masm, argc_without_receiver, a4, a5, t0);
1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776

  // Push a slot for the receiver.
  __ Push(zero_reg);

  if (mode == InterpreterPushArgsMode::kWithFinalSpread) {
    // Pass the spread in the register a2.
    // a4 already points to the penultimate argument, the spread
    // lies in the next interpreter register.
    __ Ld_d(a2, MemOperand(a4, -kSystemPointerSize));
  } else {
    __ AssertUndefinedOrAllocationSite(a2, t0);
  }

  if (mode == InterpreterPushArgsMode::kArrayFunction) {
    __ AssertFunction(a1);

    // Tail call to the function-specific construct stub (still in the caller
    // context at this point).
    __ Jump(BUILTIN_CODE(masm->isolate(), ArrayConstructorImpl),
            RelocInfo::CODE_TARGET);
  } else if (mode == InterpreterPushArgsMode::kWithFinalSpread) {
    // Call the constructor with a0, a1, and a3 unmodified.
    __ Jump(BUILTIN_CODE(masm->isolate(), ConstructWithSpread),
            RelocInfo::CODE_TARGET);
  } else {
    DCHECK_EQ(InterpreterPushArgsMode::kOther, mode);
    // Call the constructor with a0, a1, and a3 unmodified.
    __ Jump(BUILTIN_CODE(masm->isolate(), Construct), RelocInfo::CODE_TARGET);
  }

  __ bind(&stack_overflow);
  {
    __ TailCallRuntime(Runtime::kThrowStackOverflow);
    // Unreachable code.
    __ break_(0xCC);
  }
}

static void Generate_InterpreterEnterBytecode(MacroAssembler* masm) {
  // Set the return address to the correct point in the interpreter entry
  // trampoline.
  Label builtin_trampoline, trampoline_loaded;
  Smi interpreter_entry_return_pc_offset(
      masm->isolate()->heap()->interpreter_entry_return_pc_offset());
  DCHECK_NE(interpreter_entry_return_pc_offset, Smi::zero());

  // If the SFI function_data is an InterpreterData, the function will have a
  // custom copy of the interpreter entry trampoline for profiling. If so,
  // get the custom trampoline, otherwise grab the entry address of the global
  // trampoline.
  __ Ld_d(t0, MemOperand(fp, StandardFrameConstants::kFunctionOffset));
  __ Ld_d(t0, FieldMemOperand(t0, JSFunction::kSharedFunctionInfoOffset));
  __ Ld_d(t0, FieldMemOperand(t0, SharedFunctionInfo::kFunctionDataOffset));
  __ GetObjectType(t0, kInterpreterDispatchTableRegister,
                   kInterpreterDispatchTableRegister);
  __ Branch(&builtin_trampoline, ne, kInterpreterDispatchTableRegister,
            Operand(INTERPRETER_DATA_TYPE));

  __ Ld_d(t0,
          FieldMemOperand(t0, InterpreterData::kInterpreterTrampolineOffset));
  __ Add_d(t0, t0, Operand(Code::kHeaderSize - kHeapObjectTag));
  __ Branch(&trampoline_loaded);

  __ bind(&builtin_trampoline);
  __ li(t0, ExternalReference::
                address_of_interpreter_entry_trampoline_instruction_start(
                    masm->isolate()));
  __ Ld_d(t0, MemOperand(t0, 0));

  __ bind(&trampoline_loaded);
  __ Add_d(ra, t0, Operand(interpreter_entry_return_pc_offset.value()));

  // Initialize the dispatch table register.
  __ li(kInterpreterDispatchTableRegister,
        ExternalReference::interpreter_dispatch_table_address(masm->isolate()));

  // Get the bytecode array pointer from the frame.
  __ Ld_d(kInterpreterBytecodeArrayRegister,
          MemOperand(fp, InterpreterFrameConstants::kBytecodeArrayFromFp));

  if (FLAG_debug_code) {
    // Check function data field is actually a BytecodeArray object.
    __ SmiTst(kInterpreterBytecodeArrayRegister, kScratchReg);
    __ Assert(ne,
              AbortReason::kFunctionDataShouldBeBytecodeArrayOnInterpreterEntry,
              kScratchReg, Operand(zero_reg));
    __ GetObjectType(kInterpreterBytecodeArrayRegister, a1, a1);
    __ Assert(eq,
              AbortReason::kFunctionDataShouldBeBytecodeArrayOnInterpreterEntry,
              a1, Operand(BYTECODE_ARRAY_TYPE));
  }

  // Get the target bytecode offset from the frame.
  __ SmiUntag(kInterpreterBytecodeOffsetRegister,
              MemOperand(fp, InterpreterFrameConstants::kBytecodeOffsetFromFp));

  if (FLAG_debug_code) {
    Label okay;
    __ Branch(&okay, ge, kInterpreterBytecodeOffsetRegister,
              Operand(BytecodeArray::kHeaderSize - kHeapObjectTag));
    // Unreachable code.
    __ break_(0xCC);
    __ bind(&okay);
  }

  // Dispatch to the target bytecode.
  __ Add_d(a1, kInterpreterBytecodeArrayRegister,
           kInterpreterBytecodeOffsetRegister);
  __ Ld_bu(a7, MemOperand(a1, 0));
  __ Alsl_d(a1, a7, kInterpreterDispatchTableRegister, kPointerSizeLog2, t7);
  __ Ld_d(kJavaScriptCallCodeStartRegister, MemOperand(a1, 0));
  __ Jump(kJavaScriptCallCodeStartRegister);
}

void Builtins::Generate_InterpreterEnterAtNextBytecode(MacroAssembler* masm) {
  // Advance the current bytecode offset stored within the given interpreter
  // stack frame. This simulates what all bytecode handlers do upon completion
  // of the underlying operation.
  __ Ld_d(kInterpreterBytecodeArrayRegister,
          MemOperand(fp, InterpreterFrameConstants::kBytecodeArrayFromFp));
  __ Ld_d(kInterpreterBytecodeOffsetRegister,
          MemOperand(fp, InterpreterFrameConstants::kBytecodeOffsetFromFp));
  __ SmiUntag(kInterpreterBytecodeOffsetRegister);

  Label enter_bytecode, function_entry_bytecode;
  __ Branch(&function_entry_bytecode, eq, kInterpreterBytecodeOffsetRegister,
            Operand(BytecodeArray::kHeaderSize - kHeapObjectTag +
                    kFunctionEntryBytecodeOffset));

  // Load the current bytecode.
  __ Add_d(a1, kInterpreterBytecodeArrayRegister,
           kInterpreterBytecodeOffsetRegister);
  __ Ld_bu(a1, MemOperand(a1, 0));

  // Advance to the next bytecode.
  Label if_return;
  AdvanceBytecodeOffsetOrReturn(masm, kInterpreterBytecodeArrayRegister,
                                kInterpreterBytecodeOffsetRegister, a1, a2, a3,
                                a4, &if_return);

  __ bind(&enter_bytecode);
  // Convert new bytecode offset to a Smi and save in the stackframe.
  __ SmiTag(a2, kInterpreterBytecodeOffsetRegister);
  __ St_d(a2, MemOperand(fp, InterpreterFrameConstants::kBytecodeOffsetFromFp));

  Generate_InterpreterEnterBytecode(masm);

  __ bind(&function_entry_bytecode);
  // If the code deoptimizes during the implicit function entry stack interrupt
  // check, it will have a bailout ID of kFunctionEntryBytecodeOffset, which is
  // not a valid bytecode offset. Detect this case and advance to the first
  // actual bytecode.
  __ li(kInterpreterBytecodeOffsetRegister,
        Operand(BytecodeArray::kHeaderSize - kHeapObjectTag));
  __ Branch(&enter_bytecode);

  // We should never take the if_return path.
  __ bind(&if_return);
  __ Abort(AbortReason::kInvalidBytecodeAdvance);
}

void Builtins::Generate_InterpreterEnterAtBytecode(MacroAssembler* masm) {
  Generate_InterpreterEnterBytecode(masm);
}

namespace {
void Generate_ContinueToBuiltinHelper(MacroAssembler* masm,
                                      bool java_script_builtin,
                                      bool with_result) {
  const RegisterConfiguration* config(RegisterConfiguration::Default());
  int allocatable_register_count = config->num_allocatable_general_registers();
  UseScratchRegisterScope temps(masm);
  Register scratch = temps.Acquire();
  if (with_result) {
    if (java_script_builtin) {
      __ mov(scratch, a0);
    } else {
      // Overwrite the hole inserted by the deoptimizer with the return value
      // from the LAZY deopt point.
      __ St_d(
          a0,
          MemOperand(
              sp, config->num_allocatable_general_registers() * kPointerSize +
                      BuiltinContinuationFrameConstants::kFixedFrameSize));
    }
  }
  for (int i = allocatable_register_count - 1; i >= 0; --i) {
    int code = config->GetAllocatableGeneralCode(i);
    __ Pop(Register::from_code(code));
    if (java_script_builtin && code == kJavaScriptCallArgCountRegister.code()) {
      __ SmiUntag(Register::from_code(code));
    }
  }

  if (with_result && java_script_builtin) {
    // Overwrite the hole inserted by the deoptimizer with the return value from
    // the LAZY deopt point. t0 contains the arguments count, the return value
    // from LAZY is always the last argument.
1777 1778 1779 1780
    constexpr int return_value_offset =
        BuiltinContinuationFrameConstants::kFixedSlotCount -
        kJSArgcReceiverSlots;
    __ Add_d(a0, a0, Operand(return_value_offset));
1781 1782 1783
    __ Alsl_d(t0, a0, sp, kSystemPointerSizeLog2, t7);
    __ St_d(scratch, MemOperand(t0, 0));
    // Recover arguments count.
1784
    __ Sub_d(a0, a0, Operand(return_value_offset));
1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856
  }

  __ Ld_d(
      fp,
      MemOperand(sp, BuiltinContinuationFrameConstants::kFixedFrameSizeFromFp));
  // Load builtin index (stored as a Smi) and use it to get the builtin start
  // address from the builtins table.
  __ Pop(t0);
  __ Add_d(sp, sp,
           Operand(BuiltinContinuationFrameConstants::kFixedFrameSizeFromFp));
  __ Pop(ra);
  __ LoadEntryFromBuiltinIndex(t0);
  __ Jump(t0);
}
}  // namespace

void Builtins::Generate_ContinueToCodeStubBuiltin(MacroAssembler* masm) {
  Generate_ContinueToBuiltinHelper(masm, false, false);
}

void Builtins::Generate_ContinueToCodeStubBuiltinWithResult(
    MacroAssembler* masm) {
  Generate_ContinueToBuiltinHelper(masm, false, true);
}

void Builtins::Generate_ContinueToJavaScriptBuiltin(MacroAssembler* masm) {
  Generate_ContinueToBuiltinHelper(masm, true, false);
}

void Builtins::Generate_ContinueToJavaScriptBuiltinWithResult(
    MacroAssembler* masm) {
  Generate_ContinueToBuiltinHelper(masm, true, true);
}

void Builtins::Generate_NotifyDeoptimized(MacroAssembler* masm) {
  {
    FrameScope scope(masm, StackFrame::INTERNAL);
    __ CallRuntime(Runtime::kNotifyDeoptimized);
  }

  DCHECK_EQ(kInterpreterAccumulatorRegister.code(), a0.code());
  __ Ld_d(a0, MemOperand(sp, 0 * kPointerSize));
  __ Add_d(sp, sp, Operand(1 * kPointerSize));  // Remove state.
  __ Ret();
}

namespace {

void Generate_OSREntry(MacroAssembler* masm, Register entry_address,
                       Operand offset = Operand(zero_reg)) {
  __ Add_d(ra, entry_address, offset);
  // And "return" to the OSR entry point of the function.
  __ Ret();
}

void OnStackReplacement(MacroAssembler* masm, bool is_interpreter) {
  {
    FrameScope scope(masm, StackFrame::INTERNAL);
    __ CallRuntime(Runtime::kCompileForOnStackReplacement);
  }

  // If the code object is null, just return to the caller.
  __ Ret(eq, a0, Operand(Smi::zero()));

  if (is_interpreter) {
    // Drop the handler frame that is be sitting on top of the actual
    // JavaScript frame. This is the case then OSR is triggered from bytecode.
    __ LeaveFrame(StackFrame::STUB);
  }

  // Load deoptimization data from the code object.
  // <deopt_data> = <code>[#deoptimization_data_offset]
1857 1858
  __ Ld_d(a1, MemOperand(a0, Code::kDeoptimizationDataOrInterpreterDataOffset -
                                 kHeapObjectTag));
1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904

  // Load the OSR entrypoint offset from the deoptimization data.
  // <osr_offset> = <deopt_data>[#header_size + #osr_pc_offset]
  __ SmiUntag(a1, MemOperand(a1, FixedArray::OffsetOfElementAt(
                                     DeoptimizationData::kOsrPcOffsetIndex) -
                                     kHeapObjectTag));

  // Compute the target address = code_obj + header_size + osr_offset
  // <entry_addr> = <code_obj> + #header_size + <osr_offset>
  __ Add_d(a0, a0, a1);
  Generate_OSREntry(masm, a0, Operand(Code::kHeaderSize - kHeapObjectTag));
}
}  // namespace

void Builtins::Generate_InterpreterOnStackReplacement(MacroAssembler* masm) {
  return OnStackReplacement(masm, true);
}

void Builtins::Generate_BaselineOnStackReplacement(MacroAssembler* masm) {
  __ Ld_d(kContextRegister,
          MemOperand(fp, StandardFrameConstants::kContextOffset));
  return OnStackReplacement(masm, false);
}

// static
void Builtins::Generate_FunctionPrototypeApply(MacroAssembler* masm) {
  // ----------- S t a t e -------------
  //  -- a0    : argc
  //  -- sp[0] : receiver
  //  -- sp[4] : thisArg
  //  -- sp[8] : argArray
  // -----------------------------------

  Register argc = a0;
  Register arg_array = a2;
  Register receiver = a1;
  Register this_arg = a5;
  Register undefined_value = a3;
  Register scratch = a4;

  __ LoadRoot(undefined_value, RootIndex::kUndefinedValue);

  // 1. Load receiver into a1, argArray into a2 (if present), remove all
  // arguments from the stack (including the receiver), and push thisArg (if
  // present) instead.
  {
1905
    __ Sub_d(scratch, argc, JSParameterCount(0));
1906 1907 1908 1909 1910 1911 1912
    __ Ld_d(this_arg, MemOperand(sp, kPointerSize));
    __ Ld_d(arg_array, MemOperand(sp, 2 * kPointerSize));
    __ Movz(arg_array, undefined_value, scratch);  // if argc == 0
    __ Movz(this_arg, undefined_value, scratch);   // if argc == 0
    __ Sub_d(scratch, scratch, Operand(1));
    __ Movz(arg_array, undefined_value, scratch);  // if argc == 1
    __ Ld_d(receiver, MemOperand(sp, 0));
1913 1914 1915 1916
    __ DropArgumentsAndPushNewReceiver(
        argc, this_arg, TurboAssembler::kCountIsInteger,
        kJSArgcIncludesReceiver ? TurboAssembler::kCountIncludesReceiver
                                : TurboAssembler::kCountExcludesReceiver);
1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942
  }

  // ----------- S t a t e -------------
  //  -- a2    : argArray
  //  -- a1    : receiver
  //  -- a3    : undefined root value
  //  -- sp[0] : thisArg
  // -----------------------------------

  // 2. We don't need to check explicitly for callable receiver here,
  // since that's the first thing the Call/CallWithArrayLike builtins
  // will do.

  // 3. Tail call with no arguments if argArray is null or undefined.
  Label no_arguments;
  __ JumpIfRoot(arg_array, RootIndex::kNullValue, &no_arguments);
  __ Branch(&no_arguments, eq, arg_array, Operand(undefined_value));

  // 4a. Apply the receiver to the given argArray.
  __ Jump(BUILTIN_CODE(masm->isolate(), CallWithArrayLike),
          RelocInfo::CODE_TARGET);

  // 4b. The argArray is either null or undefined, so we tail call without any
  // arguments to the receiver.
  __ bind(&no_arguments);
  {
1943
    __ li(a0, JSParameterCount(0));
1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957
    DCHECK(receiver == a1);
    __ Jump(masm->isolate()->builtins()->Call(), RelocInfo::CODE_TARGET);
  }
}

// static
void Builtins::Generate_FunctionPrototypeCall(MacroAssembler* masm) {
  // 1. Get the callable to call (passed as receiver) from the stack.
  { __ Pop(a1); }

  // 2. Make sure we have at least one argument.
  // a0: actual number of arguments
  {
    Label done;
1958
    __ Branch(&done, ne, a0, Operand(JSParameterCount(0)));
1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995
    __ PushRoot(RootIndex::kUndefinedValue);
    __ Add_d(a0, a0, Operand(1));
    __ bind(&done);
  }

  // 3. Adjust the actual number of arguments.
  __ addi_d(a0, a0, -1);

  // 4. Call the callable.
  __ Jump(masm->isolate()->builtins()->Call(), RelocInfo::CODE_TARGET);
}

void Builtins::Generate_ReflectApply(MacroAssembler* masm) {
  // ----------- S t a t e -------------
  //  -- a0     : argc
  //  -- sp[0]  : receiver
  //  -- sp[8]  : target         (if argc >= 1)
  //  -- sp[16] : thisArgument   (if argc >= 2)
  //  -- sp[24] : argumentsList  (if argc == 3)
  // -----------------------------------

  Register argc = a0;
  Register arguments_list = a2;
  Register target = a1;
  Register this_argument = a5;
  Register undefined_value = a3;
  Register scratch = a4;

  __ LoadRoot(undefined_value, RootIndex::kUndefinedValue);

  // 1. Load target into a1 (if present), argumentsList into a2 (if present),
  // remove all arguments from the stack (including the receiver), and push
  // thisArgument (if present) instead.
  {
    // Claim (3 - argc) dummy arguments form the stack, to put the stack in a
    // consistent state for a simple pop operation.

1996
    __ Sub_d(scratch, argc, Operand(JSParameterCount(0)));
1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
    __ Ld_d(target, MemOperand(sp, kPointerSize));
    __ Ld_d(this_argument, MemOperand(sp, 2 * kPointerSize));
    __ Ld_d(arguments_list, MemOperand(sp, 3 * kPointerSize));
    __ Movz(arguments_list, undefined_value, scratch);  // if argc == 0
    __ Movz(this_argument, undefined_value, scratch);   // if argc == 0
    __ Movz(target, undefined_value, scratch);          // if argc == 0
    __ Sub_d(scratch, scratch, Operand(1));
    __ Movz(arguments_list, undefined_value, scratch);  // if argc == 1
    __ Movz(this_argument, undefined_value, scratch);   // if argc == 1
    __ Sub_d(scratch, scratch, Operand(1));
    __ Movz(arguments_list, undefined_value, scratch);  // if argc == 2

2009 2010 2011 2012
    __ DropArgumentsAndPushNewReceiver(
        argc, this_argument, TurboAssembler::kCountIsInteger,
        kJSArgcIncludesReceiver ? TurboAssembler::kCountIncludesReceiver
                                : TurboAssembler::kCountExcludesReceiver);
2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056
  }

  // ----------- S t a t e -------------
  //  -- a2    : argumentsList
  //  -- a1    : target
  //  -- a3    : undefined root value
  //  -- sp[0] : thisArgument
  // -----------------------------------

  // 2. We don't need to check explicitly for callable target here,
  // since that's the first thing the Call/CallWithArrayLike builtins
  // will do.

  // 3. Apply the target to the given argumentsList.
  __ Jump(BUILTIN_CODE(masm->isolate(), CallWithArrayLike),
          RelocInfo::CODE_TARGET);
}

void Builtins::Generate_ReflectConstruct(MacroAssembler* masm) {
  // ----------- S t a t e -------------
  //  -- a0     : argc
  //  -- sp[0]   : receiver
  //  -- sp[8]   : target
  //  -- sp[16]  : argumentsList
  //  -- sp[24]  : new.target (optional)
  // -----------------------------------

  Register argc = a0;
  Register arguments_list = a2;
  Register target = a1;
  Register new_target = a3;
  Register undefined_value = a4;
  Register scratch = a5;

  __ LoadRoot(undefined_value, RootIndex::kUndefinedValue);

  // 1. Load target into a1 (if present), argumentsList into a2 (if present),
  // new.target into a3 (if present, otherwise use target), remove all
  // arguments from the stack (including the receiver), and push thisArgument
  // (if present) instead.
  {
    // Claim (3 - argc) dummy arguments form the stack, to put the stack in a
    // consistent state for a simple pop operation.

2057
    __ Sub_d(scratch, argc, Operand(JSParameterCount(0)));
2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069
    __ Ld_d(target, MemOperand(sp, kPointerSize));
    __ Ld_d(arguments_list, MemOperand(sp, 2 * kPointerSize));
    __ Ld_d(new_target, MemOperand(sp, 3 * kPointerSize));
    __ Movz(arguments_list, undefined_value, scratch);  // if argc == 0
    __ Movz(new_target, undefined_value, scratch);      // if argc == 0
    __ Movz(target, undefined_value, scratch);          // if argc == 0
    __ Sub_d(scratch, scratch, Operand(1));
    __ Movz(arguments_list, undefined_value, scratch);  // if argc == 1
    __ Movz(new_target, target, scratch);               // if argc == 1
    __ Sub_d(scratch, scratch, Operand(1));
    __ Movz(new_target, target, scratch);  // if argc == 2

2070 2071 2072 2073
    __ DropArgumentsAndPushNewReceiver(
        argc, undefined_value, TurboAssembler::kCountIsInteger,
        kJSArgcIncludesReceiver ? TurboAssembler::kCountIncludesReceiver
                                : TurboAssembler::kCountExcludesReceiver);
2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095
  }

  // ----------- S t a t e -------------
  //  -- a2    : argumentsList
  //  -- a1    : target
  //  -- a3    : new.target
  //  -- sp[0] : receiver (undefined)
  // -----------------------------------

  // 2. We don't need to check explicitly for constructor target here,
  // since that's the first thing the Construct/ConstructWithArrayLike
  // builtins will do.

  // 3. We don't need to check explicitly for constructor new.target here,
  // since that's the second thing the Construct/ConstructWithArrayLike
  // builtins will do.

  // 4. Construct the target with the given new.target and argumentsList.
  __ Jump(BUILTIN_CODE(masm->isolate(), ConstructWithArrayLike),
          RelocInfo::CODE_TARGET);
}

2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142
namespace {

// Allocate new stack space for |count| arguments and shift all existing
// arguments already on the stack. |pointer_to_new_space_out| points to the
// first free slot on the stack to copy additional arguments to and
// |argc_in_out| is updated to include |count|.
void Generate_AllocateSpaceAndShiftExistingArguments(
    MacroAssembler* masm, Register count, Register argc_in_out,
    Register pointer_to_new_space_out, Register scratch1, Register scratch2,
    Register scratch3) {
  DCHECK(!AreAliased(count, argc_in_out, pointer_to_new_space_out, scratch1,
                     scratch2));
  Register old_sp = scratch1;
  Register new_space = scratch2;
  __ mov(old_sp, sp);
  __ slli_d(new_space, count, kPointerSizeLog2);
  __ Sub_d(sp, sp, Operand(new_space));

  Register end = scratch2;
  Register value = scratch3;
  Register dest = pointer_to_new_space_out;
  __ mov(dest, sp);
  __ Alsl_d(end, argc_in_out, old_sp, kSystemPointerSizeLog2);
  Label loop, done;
  if (kJSArgcIncludesReceiver) {
    __ Branch(&done, ge, old_sp, Operand(end));
  } else {
    __ Branch(&done, gt, old_sp, Operand(end));
  }
  __ bind(&loop);
  __ Ld_d(value, MemOperand(old_sp, 0));
  __ St_d(value, MemOperand(dest, 0));
  __ Add_d(old_sp, old_sp, Operand(kSystemPointerSize));
  __ Add_d(dest, dest, Operand(kSystemPointerSize));
  if (kJSArgcIncludesReceiver) {
    __ Branch(&loop, lt, old_sp, Operand(end));
  } else {
    __ Branch(&loop, le, old_sp, Operand(end));
  }
  __ bind(&done);

  // Update total number of arguments.
  __ Add_d(argc_in_out, argc_in_out, count);
}

}  // namespace

2143 2144 2145 2146 2147
// static
void Builtins::Generate_CallOrConstructVarargs(MacroAssembler* masm,
                                               Handle<Code> code) {
  // ----------- S t a t e -------------
  //  -- a1 : target
2148
  //  -- a0 : number of parameters on the stack
2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176
  //  -- a2 : arguments list (a FixedArray)
  //  -- a4 : len (number of elements to push from args)
  //  -- a3 : new.target (for [[Construct]])
  // -----------------------------------
  if (FLAG_debug_code) {
    // Allow a2 to be a FixedArray, or a FixedDoubleArray if a4 == 0.
    Label ok, fail;
    __ AssertNotSmi(a2);
    __ GetObjectType(a2, t8, t8);
    __ Branch(&ok, eq, t8, Operand(FIXED_ARRAY_TYPE));
    __ Branch(&fail, ne, t8, Operand(FIXED_DOUBLE_ARRAY_TYPE));
    __ Branch(&ok, eq, a4, Operand(zero_reg));
    // Fall through.
    __ bind(&fail);
    __ Abort(AbortReason::kOperandIsNotAFixedArray);

    __ bind(&ok);
  }

  Register args = a2;
  Register len = a4;

  // Check for stack overflow.
  Label stack_overflow;
  __ StackOverflowCheck(len, kScratchReg, a5, &stack_overflow);

  // Move the arguments already in the stack,
  // including the receiver and the return address.
2177 2178 2179 2180
  // a4: Number of arguments to make room for.
  // a0: Number of arguments already on the stack.
  // a7: Points to first free slot on the stack after arguments were shifted.
  Generate_AllocateSpaceAndShiftExistingArguments(masm, a4, a0, a7, a6, t0, t1);
2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217

  // Push arguments onto the stack (thisArgument is already on the stack).
  {
    Label done, push, loop;
    Register src = a6;
    Register scratch = len;

    __ addi_d(src, args, FixedArray::kHeaderSize - kHeapObjectTag);
    __ Branch(&done, eq, len, Operand(zero_reg));
    __ slli_d(scratch, len, kPointerSizeLog2);
    __ Sub_d(scratch, sp, Operand(scratch));
    __ LoadRoot(t1, RootIndex::kTheHoleValue);
    __ bind(&loop);
    __ Ld_d(a5, MemOperand(src, 0));
    __ addi_d(src, src, kPointerSize);
    __ Branch(&push, ne, a5, Operand(t1));
    __ LoadRoot(a5, RootIndex::kUndefinedValue);
    __ bind(&push);
    __ St_d(a5, MemOperand(a7, 0));
    __ Add_d(a7, a7, Operand(kSystemPointerSize));
    __ Add_d(scratch, scratch, Operand(kSystemPointerSize));
    __ Branch(&loop, ne, scratch, Operand(sp));
    __ bind(&done);
  }

  // Tail-call to the actual Call or Construct builtin.
  __ Jump(code, RelocInfo::CODE_TARGET);

  __ bind(&stack_overflow);
  __ TailCallRuntime(Runtime::kThrowStackOverflow);
}

// static
void Builtins::Generate_CallOrConstructForwardVarargs(MacroAssembler* masm,
                                                      CallOrConstructMode mode,
                                                      Handle<Code> code) {
  // ----------- S t a t e -------------
2218
  //  -- a0 : the number of arguments
2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243
  //  -- a3 : the new.target (for [[Construct]] calls)
  //  -- a1 : the target to call (can be any Object)
  //  -- a2 : start index (to support rest parameters)
  // -----------------------------------

  // Check if new.target has a [[Construct]] internal method.
  if (mode == CallOrConstructMode::kConstruct) {
    Label new_target_constructor, new_target_not_constructor;
    __ JumpIfSmi(a3, &new_target_not_constructor);
    __ Ld_d(t1, FieldMemOperand(a3, HeapObject::kMapOffset));
    __ Ld_bu(t1, FieldMemOperand(t1, Map::kBitFieldOffset));
    __ And(t1, t1, Operand(Map::Bits1::IsConstructorBit::kMask));
    __ Branch(&new_target_constructor, ne, t1, Operand(zero_reg));
    __ bind(&new_target_not_constructor);
    {
      FrameScope scope(masm, StackFrame::MANUAL);
      __ EnterFrame(StackFrame::INTERNAL);
      __ Push(a3);
      __ CallRuntime(Runtime::kThrowNotConstructor);
    }
    __ bind(&new_target_constructor);
  }

  Label stack_done, stack_overflow;
  __ Ld_d(a7, MemOperand(fp, StandardFrameConstants::kArgCOffset));
2244 2245 2246 2247
  if (kJSArgcIncludesReceiver) {
    __ Sub_d(a7, a7, Operand(kJSArgcReceiverSlots));
  }
  __ Sub_d(a7, a7, a2);
2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262
  __ Branch(&stack_done, le, a7, Operand(zero_reg));
  {
    // Check for stack overflow.
    __ StackOverflowCheck(a7, a4, a5, &stack_overflow);

    // Forward the arguments from the caller frame.

    // Point to the first argument to copy (skipping the receiver).
    __ Add_d(a6, fp,
             Operand(CommonFrameConstants::kFixedFrameSizeAboveFp +
                     kSystemPointerSize));
    __ Alsl_d(a6, a2, a6, kSystemPointerSizeLog2, t7);

    // Move the arguments already in the stack,
    // including the receiver and the return address.
2263 2264 2265 2266 2267
    // a7: Number of arguments to make room for.
    // a0: Number of arguments already on the stack.
    // a2: Points to first free slot on the stack after arguments were shifted.
    Generate_AllocateSpaceAndShiftExistingArguments(masm, a7, a0, a2, t0, t1,
                                                    t2);
2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297

    // Copy arguments from the caller frame.
    // TODO(victorgomes): Consider using forward order as potentially more cache
    // friendly.
    {
      Label loop;
      __ bind(&loop);
      {
        __ Sub_w(a7, a7, Operand(1));
        __ Alsl_d(t0, a7, a6, kPointerSizeLog2, t7);
        __ Ld_d(kScratchReg, MemOperand(t0, 0));
        __ Alsl_d(t0, a7, a2, kPointerSizeLog2, t7);
        __ St_d(kScratchReg, MemOperand(t0, 0));
        __ Branch(&loop, ne, a7, Operand(zero_reg));
      }
    }
  }
  __ Branch(&stack_done);
  __ bind(&stack_overflow);
  __ TailCallRuntime(Runtime::kThrowStackOverflow);
  __ bind(&stack_done);

  // Tail-call to the {code} handler.
  __ Jump(code, RelocInfo::CODE_TARGET);
}

// static
void Builtins::Generate_CallFunction(MacroAssembler* masm,
                                     ConvertReceiverMode mode) {
  // ----------- S t a t e -------------
2298
  //  -- a0 : the number of arguments
2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317
  //  -- a1 : the function to call (checked to be a JSFunction)
  // -----------------------------------
  __ AssertFunction(a1);

  __ Ld_d(a2, FieldMemOperand(a1, JSFunction::kSharedFunctionInfoOffset));

  // Enter the context of the function; ToObject has to run in the function
  // context, and we also need to take the global proxy from the function
  // context in case of conversion.
  __ Ld_d(cp, FieldMemOperand(a1, JSFunction::kContextOffset));
  // We need to convert the receiver for non-native sloppy mode functions.
  Label done_convert;
  __ Ld_wu(a3, FieldMemOperand(a2, SharedFunctionInfo::kFlagsOffset));
  __ And(kScratchReg, a3,
         Operand(SharedFunctionInfo::IsNativeBit::kMask |
                 SharedFunctionInfo::IsStrictBit::kMask));
  __ Branch(&done_convert, ne, kScratchReg, Operand(zero_reg));
  {
    // ----------- S t a t e -------------
2318
    //  -- a0 : the number of arguments
2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369
    //  -- a1 : the function to call (checked to be a JSFunction)
    //  -- a2 : the shared function info.
    //  -- cp : the function context.
    // -----------------------------------

    if (mode == ConvertReceiverMode::kNullOrUndefined) {
      // Patch receiver to global proxy.
      __ LoadGlobalProxy(a3);
    } else {
      Label convert_to_object, convert_receiver;
      __ LoadReceiver(a3, a0);
      __ JumpIfSmi(a3, &convert_to_object);
      STATIC_ASSERT(LAST_JS_RECEIVER_TYPE == LAST_TYPE);
      __ GetObjectType(a3, a4, a4);
      __ Branch(&done_convert, hs, a4, Operand(FIRST_JS_RECEIVER_TYPE));
      if (mode != ConvertReceiverMode::kNotNullOrUndefined) {
        Label convert_global_proxy;
        __ JumpIfRoot(a3, RootIndex::kUndefinedValue, &convert_global_proxy);
        __ JumpIfNotRoot(a3, RootIndex::kNullValue, &convert_to_object);
        __ bind(&convert_global_proxy);
        {
          // Patch receiver to global proxy.
          __ LoadGlobalProxy(a3);
        }
        __ Branch(&convert_receiver);
      }
      __ bind(&convert_to_object);
      {
        // Convert receiver using ToObject.
        // TODO(bmeurer): Inline the allocation here to avoid building the frame
        // in the fast case? (fall back to AllocateInNewSpace?)
        FrameScope scope(masm, StackFrame::INTERNAL);
        __ SmiTag(a0);
        __ Push(a0, a1);
        __ mov(a0, a3);
        __ Push(cp);
        __ Call(BUILTIN_CODE(masm->isolate(), ToObject),
                RelocInfo::CODE_TARGET);
        __ Pop(cp);
        __ mov(a3, a0);
        __ Pop(a0, a1);
        __ SmiUntag(a0);
      }
      __ Ld_d(a2, FieldMemOperand(a1, JSFunction::kSharedFunctionInfoOffset));
      __ bind(&convert_receiver);
    }
    __ StoreReceiver(a3, a0, kScratchReg);
  }
  __ bind(&done_convert);

  // ----------- S t a t e -------------
2370
  //  -- a0 : the number of arguments
2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383
  //  -- a1 : the function to call (checked to be a JSFunction)
  //  -- a2 : the shared function info.
  //  -- cp : the function context.
  // -----------------------------------

  __ Ld_hu(
      a2, FieldMemOperand(a2, SharedFunctionInfo::kFormalParameterCountOffset));
  __ InvokeFunctionCode(a1, no_reg, a2, a0, InvokeType::kJump);
}

// static
void Builtins::Generate_CallBoundFunctionImpl(MacroAssembler* masm) {
  // ----------- S t a t e -------------
2384
  //  -- a0 : the number of arguments
2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399
  //  -- a1 : the function to call (checked to be a JSBoundFunction)
  // -----------------------------------
  __ AssertBoundFunction(a1);

  // Patch the receiver to [[BoundThis]].
  {
    __ Ld_d(t0, FieldMemOperand(a1, JSBoundFunction::kBoundThisOffset));
    __ StoreReceiver(t0, a0, kScratchReg);
  }

  // Load [[BoundArguments]] into a2 and length of that into a4.
  __ Ld_d(a2, FieldMemOperand(a1, JSBoundFunction::kBoundArgumentsOffset));
  __ SmiUntag(a4, FieldMemOperand(a2, FixedArray::kLengthOffset));

  // ----------- S t a t e -------------
2400
  //  -- a0 : the number of arguments
2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454
  //  -- a1 : the function to call (checked to be a JSBoundFunction)
  //  -- a2 : the [[BoundArguments]] (implemented as FixedArray)
  //  -- a4 : the number of [[BoundArguments]]
  // -----------------------------------

  // Reserve stack space for the [[BoundArguments]].
  {
    Label done;
    __ slli_d(a5, a4, kPointerSizeLog2);
    __ Sub_d(t0, sp, Operand(a5));
    // Check the stack for overflow. We are not trying to catch interruptions
    // (i.e. debug break and preemption) here, so check the "real stack limit".
    __ LoadStackLimit(kScratchReg,
                      MacroAssembler::StackLimitKind::kRealStackLimit);
    __ Branch(&done, hs, t0, Operand(kScratchReg));
    {
      FrameScope scope(masm, StackFrame::MANUAL);
      __ EnterFrame(StackFrame::INTERNAL);
      __ CallRuntime(Runtime::kThrowStackOverflow);
    }
    __ bind(&done);
  }

  // Pop receiver.
  __ Pop(t0);

  // Push [[BoundArguments]].
  {
    Label loop, done_loop;
    __ SmiUntag(a4, FieldMemOperand(a2, FixedArray::kLengthOffset));
    __ Add_d(a0, a0, Operand(a4));
    __ Add_d(a2, a2, Operand(FixedArray::kHeaderSize - kHeapObjectTag));
    __ bind(&loop);
    __ Sub_d(a4, a4, Operand(1));
    __ Branch(&done_loop, lt, a4, Operand(zero_reg));
    __ Alsl_d(a5, a4, a2, kPointerSizeLog2, t7);
    __ Ld_d(kScratchReg, MemOperand(a5, 0));
    __ Push(kScratchReg);
    __ Branch(&loop);
    __ bind(&done_loop);
  }

  // Push receiver.
  __ Push(t0);

  // Call the [[BoundTargetFunction]] via the Call builtin.
  __ Ld_d(a1, FieldMemOperand(a1, JSBoundFunction::kBoundTargetFunctionOffset));
  __ Jump(BUILTIN_CODE(masm->isolate(), Call_ReceiverIsAny),
          RelocInfo::CODE_TARGET);
}

// static
void Builtins::Generate_Call(MacroAssembler* masm, ConvertReceiverMode mode) {
  // ----------- S t a t e -------------
2455
  //  -- a0 : the number of arguments
2456 2457 2458
  //  -- a1 : the target to call (can be any Object).
  // -----------------------------------

2459 2460 2461 2462 2463 2464 2465
  Register argc = a0;
  Register target = a1;
  Register map = t1;
  Register instance_type = t2;
  Register scratch = t8;
  DCHECK(!AreAliased(argc, target, map, instance_type, scratch));

2466
  Label non_callable, class_constructor;
2467 2468 2469 2470
  __ JumpIfSmi(target, &non_callable);
  __ LoadMap(map, target);
  __ GetInstanceTypeRange(map, instance_type, FIRST_CALLABLE_JS_FUNCTION_TYPE,
                          scratch);
2471
  __ Jump(masm->isolate()->builtins()->CallFunction(mode),
2472
          RelocInfo::CODE_TARGET, ls, scratch,
2473 2474
          Operand(LAST_CALLABLE_JS_FUNCTION_TYPE -
                  FIRST_CALLABLE_JS_FUNCTION_TYPE));
2475
  __ Jump(BUILTIN_CODE(masm->isolate(), CallBoundFunction),
2476 2477
          RelocInfo::CODE_TARGET, eq, instance_type,
          Operand(JS_BOUND_FUNCTION_TYPE));
2478 2479

  // Check if target has a [[Call]] internal method.
2480 2481 2482 2483 2484 2485 2486
  {
    Register flags = t1;
    __ Ld_bu(flags, FieldMemOperand(map, Map::kBitFieldOffset));
    map = no_reg;
    __ And(flags, flags, Operand(Map::Bits1::IsCallableBit::kMask));
    __ Branch(&non_callable, eq, flags, Operand(zero_reg));
  }
2487 2488

  __ Jump(BUILTIN_CODE(masm->isolate(), CallProxy), RelocInfo::CODE_TARGET, eq,
2489
          instance_type, Operand(JS_PROXY_TYPE));
2490

2491 2492
  // ES6 section 9.2.1 [[Call]] ( thisArgument, argumentsList)
  // Check that the function is not a "classConstructor".
2493 2494
  __ Branch(&class_constructor, eq, instance_type,
            Operand(JS_CLASS_CONSTRUCTOR_TYPE));
2495

2496 2497 2498
  // 2. Call to something else, which might have a [[Call]] internal method (if
  // not we raise an exception).
  // Overwrite the original receiver with the (original) target.
2499
  __ StoreReceiver(target, argc, kScratchReg);
2500
  // Let the "call_as_function_delegate" take care of the rest.
2501
  __ LoadNativeContextSlot(target, Context::CALL_AS_FUNCTION_DELEGATE_INDEX);
2502 2503 2504 2505 2506 2507 2508 2509
  __ Jump(masm->isolate()->builtins()->CallFunction(
              ConvertReceiverMode::kNotNullOrUndefined),
          RelocInfo::CODE_TARGET);

  // 3. Call to something that is not callable.
  __ bind(&non_callable);
  {
    FrameScope scope(masm, StackFrame::INTERNAL);
2510
    __ Push(target);
2511 2512
    __ CallRuntime(Runtime::kThrowCalledNonCallable);
  }
2513 2514 2515 2516 2517

  // 4. The function is a "classConstructor", need to raise an exception.
  __ bind(&class_constructor);
  {
    FrameScope frame(masm, StackFrame::INTERNAL);
2518
    __ Push(target);
2519 2520
    __ CallRuntime(Runtime::kThrowConstructorNonCallableError);
  }
2521 2522 2523 2524
}

void Builtins::Generate_ConstructFunction(MacroAssembler* masm) {
  // ----------- S t a t e -------------
2525
  //  -- a0 : the number of arguments
2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554
  //  -- a1 : the constructor to call (checked to be a JSFunction)
  //  -- a3 : the new target (checked to be a constructor)
  // -----------------------------------
  __ AssertConstructor(a1);
  __ AssertFunction(a1);

  // Calling convention for function specific ConstructStubs require
  // a2 to contain either an AllocationSite or undefined.
  __ LoadRoot(a2, RootIndex::kUndefinedValue);

  Label call_generic_stub;

  // Jump to JSBuiltinsConstructStub or JSConstructStubGeneric.
  __ Ld_d(a4, FieldMemOperand(a1, JSFunction::kSharedFunctionInfoOffset));
  __ Ld_wu(a4, FieldMemOperand(a4, SharedFunctionInfo::kFlagsOffset));
  __ And(a4, a4, Operand(SharedFunctionInfo::ConstructAsBuiltinBit::kMask));
  __ Branch(&call_generic_stub, eq, a4, Operand(zero_reg));

  __ Jump(BUILTIN_CODE(masm->isolate(), JSBuiltinsConstructStub),
          RelocInfo::CODE_TARGET);

  __ bind(&call_generic_stub);
  __ Jump(BUILTIN_CODE(masm->isolate(), JSConstructStubGeneric),
          RelocInfo::CODE_TARGET);
}

// static
void Builtins::Generate_ConstructBoundFunction(MacroAssembler* masm) {
  // ----------- S t a t e -------------
2555
  //  -- a0 : the number of arguments
2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566
  //  -- a1 : the function to call (checked to be a JSBoundFunction)
  //  -- a3 : the new target (checked to be a constructor)
  // -----------------------------------
  __ AssertConstructor(a1);
  __ AssertBoundFunction(a1);

  // Load [[BoundArguments]] into a2 and length of that into a4.
  __ Ld_d(a2, FieldMemOperand(a1, JSBoundFunction::kBoundArgumentsOffset));
  __ SmiUntag(a4, FieldMemOperand(a2, FixedArray::kLengthOffset));

  // ----------- S t a t e -------------
2567
  //  -- a0 : the number of arguments
2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630
  //  -- a1 : the function to call (checked to be a JSBoundFunction)
  //  -- a2 : the [[BoundArguments]] (implemented as FixedArray)
  //  -- a3 : the new target (checked to be a constructor)
  //  -- a4 : the number of [[BoundArguments]]
  // -----------------------------------

  // Reserve stack space for the [[BoundArguments]].
  {
    Label done;
    __ slli_d(a5, a4, kPointerSizeLog2);
    __ Sub_d(t0, sp, Operand(a5));
    // Check the stack for overflow. We are not trying to catch interruptions
    // (i.e. debug break and preemption) here, so check the "real stack limit".
    __ LoadStackLimit(kScratchReg,
                      MacroAssembler::StackLimitKind::kRealStackLimit);
    __ Branch(&done, hs, t0, Operand(kScratchReg));
    {
      FrameScope scope(masm, StackFrame::MANUAL);
      __ EnterFrame(StackFrame::INTERNAL);
      __ CallRuntime(Runtime::kThrowStackOverflow);
    }
    __ bind(&done);
  }

  // Pop receiver.
  __ Pop(t0);

  // Push [[BoundArguments]].
  {
    Label loop, done_loop;
    __ SmiUntag(a4, FieldMemOperand(a2, FixedArray::kLengthOffset));
    __ Add_d(a0, a0, Operand(a4));
    __ Add_d(a2, a2, Operand(FixedArray::kHeaderSize - kHeapObjectTag));
    __ bind(&loop);
    __ Sub_d(a4, a4, Operand(1));
    __ Branch(&done_loop, lt, a4, Operand(zero_reg));
    __ Alsl_d(a5, a4, a2, kPointerSizeLog2, t7);
    __ Ld_d(kScratchReg, MemOperand(a5, 0));
    __ Push(kScratchReg);
    __ Branch(&loop);
    __ bind(&done_loop);
  }

  // Push receiver.
  __ Push(t0);

  // Patch new.target to [[BoundTargetFunction]] if new.target equals target.
  {
    Label skip_load;
    __ Branch(&skip_load, ne, a1, Operand(a3));
    __ Ld_d(a3,
            FieldMemOperand(a1, JSBoundFunction::kBoundTargetFunctionOffset));
    __ bind(&skip_load);
  }

  // Construct the [[BoundTargetFunction]] via the Construct builtin.
  __ Ld_d(a1, FieldMemOperand(a1, JSBoundFunction::kBoundTargetFunctionOffset));
  __ Jump(BUILTIN_CODE(masm->isolate(), Construct), RelocInfo::CODE_TARGET);
}

// static
void Builtins::Generate_Construct(MacroAssembler* masm) {
  // ----------- S t a t e -------------
2631
  //  -- a0 : the number of arguments
2632 2633 2634 2635 2636
  //  -- a1 : the constructor to call (can be any Object)
  //  -- a3 : the new target (either the same as the constructor or
  //          the JSFunction on which new was invoked initially)
  // -----------------------------------

2637 2638 2639 2640 2641 2642 2643
  Register argc = a0;
  Register target = a1;
  Register map = t1;
  Register instance_type = t2;
  Register scratch = t8;
  DCHECK(!AreAliased(argc, target, map, instance_type, scratch));

2644 2645
  // Check if target is a Smi.
  Label non_constructor, non_proxy;
2646
  __ JumpIfSmi(target, &non_constructor);
2647 2648

  // Check if target has a [[Construct]] internal method.
2649 2650 2651 2652 2653 2654 2655
  __ Ld_d(map, FieldMemOperand(target, HeapObject::kMapOffset));
  {
    Register flags = t3;
    __ Ld_bu(flags, FieldMemOperand(map, Map::kBitFieldOffset));
    __ And(flags, flags, Operand(Map::Bits1::IsConstructorBit::kMask));
    __ Branch(&non_constructor, eq, flags, Operand(zero_reg));
  }
2656 2657

  // Dispatch based on instance type.
2658
  __ GetInstanceTypeRange(map, instance_type, FIRST_JS_FUNCTION_TYPE, scratch);
2659
  __ Jump(BUILTIN_CODE(masm->isolate(), ConstructFunction),
2660
          RelocInfo::CODE_TARGET, ls, scratch,
2661 2662 2663 2664 2665
          Operand(LAST_JS_FUNCTION_TYPE - FIRST_JS_FUNCTION_TYPE));

  // Only dispatch to bound functions after checking whether they are
  // constructors.
  __ Jump(BUILTIN_CODE(masm->isolate(), ConstructBoundFunction),
2666 2667
          RelocInfo::CODE_TARGET, eq, instance_type,
          Operand(JS_BOUND_FUNCTION_TYPE));
2668 2669

  // Only dispatch to proxies after checking whether they are constructors.
2670
  __ Branch(&non_proxy, ne, instance_type, Operand(JS_PROXY_TYPE));
2671 2672 2673 2674 2675 2676 2677
  __ Jump(BUILTIN_CODE(masm->isolate(), ConstructProxy),
          RelocInfo::CODE_TARGET);

  // Called Construct on an exotic Object with a [[Construct]] internal method.
  __ bind(&non_proxy);
  {
    // Overwrite the original receiver with the (original) target.
2678
    __ StoreReceiver(target, argc, kScratchReg);
2679
    // Let the "call_as_constructor_delegate" take care of the rest.
2680 2681
    __ LoadNativeContextSlot(target,
                             Context::CALL_AS_CONSTRUCTOR_DELEGATE_INDEX);
2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773
    __ Jump(masm->isolate()->builtins()->CallFunction(),
            RelocInfo::CODE_TARGET);
  }

  // Called Construct on an Object that doesn't have a [[Construct]] internal
  // method.
  __ bind(&non_constructor);
  __ Jump(BUILTIN_CODE(masm->isolate(), ConstructedNonConstructable),
          RelocInfo::CODE_TARGET);
}

#if V8_ENABLE_WEBASSEMBLY
void Builtins::Generate_WasmCompileLazy(MacroAssembler* masm) {
  // The function index was put in t0 by the jump table trampoline.
  // Convert to Smi for the runtime call
  __ SmiTag(kWasmCompileLazyFuncIndexRegister);
  {
    HardAbortScope hard_abort(masm);  // Avoid calls to Abort.
    FrameScope scope(masm, StackFrame::WASM_COMPILE_LAZY);

    // Save all parameter registers (see wasm-linkage.h). They might be
    // overwritten in the runtime call below. We don't have any callee-saved
    // registers in wasm, so no need to store anything else.
    RegList gp_regs = 0;
    for (Register gp_param_reg : wasm::kGpParamRegisters) {
      gp_regs |= gp_param_reg.bit();
    }

    RegList fp_regs = 0;
    for (DoubleRegister fp_param_reg : wasm::kFpParamRegisters) {
      fp_regs |= fp_param_reg.bit();
    }

    CHECK_EQ(NumRegs(gp_regs), arraysize(wasm::kGpParamRegisters));
    CHECK_EQ(NumRegs(fp_regs), arraysize(wasm::kFpParamRegisters));
    CHECK_EQ(WasmCompileLazyFrameConstants::kNumberOfSavedGpParamRegs,
             NumRegs(gp_regs));
    CHECK_EQ(WasmCompileLazyFrameConstants::kNumberOfSavedFpParamRegs,
             NumRegs(fp_regs));

    __ MultiPush(gp_regs);
    __ MultiPushFPU(fp_regs);

    // kFixedFrameSizeFromFp is hard coded to include space for Simd
    // registers, so we still need to allocate extra (unused) space on the stack
    // as if they were saved.
    __ Sub_d(sp, sp, base::bits::CountPopulation(fp_regs) * kDoubleSize);

    // Pass instance and function index as an explicit arguments to the runtime
    // function.
    __ Push(kWasmInstanceRegister, kWasmCompileLazyFuncIndexRegister);
    // Initialize the JavaScript context with 0. CEntry will use it to
    // set the current context on the isolate.
    __ Move(kContextRegister, Smi::zero());
    __ CallRuntime(Runtime::kWasmCompileLazy, 2);
    __ mov(t8, a0);

    __ Add_d(sp, sp, base::bits::CountPopulation(fp_regs) * kDoubleSize);
    // Restore registers.
    __ MultiPopFPU(fp_regs);
    __ MultiPop(gp_regs);
  }
  // Finally, jump to the entrypoint.
  __ Jump(t8);
}

void Builtins::Generate_WasmDebugBreak(MacroAssembler* masm) {
  HardAbortScope hard_abort(masm);  // Avoid calls to Abort.
  {
    FrameScope scope(masm, StackFrame::WASM_DEBUG_BREAK);

    // Save all parameter registers. They might hold live values, we restore
    // them after the runtime call.
    __ MultiPush(WasmDebugBreakFrameConstants::kPushedGpRegs);
    __ MultiPushFPU(WasmDebugBreakFrameConstants::kPushedFpRegs);

    // Initialize the JavaScript context with 0. CEntry will use it to
    // set the current context on the isolate.
    __ Move(cp, Smi::zero());
    __ CallRuntime(Runtime::kWasmDebugBreak, 0);

    // Restore registers.
    __ MultiPopFPU(WasmDebugBreakFrameConstants::kPushedFpRegs);
    __ MultiPop(WasmDebugBreakFrameConstants::kPushedGpRegs);
  }
  __ Ret();
}

void Builtins::Generate_GenericJSToWasmWrapper(MacroAssembler* masm) {
  __ Trap();
}

2774 2775 2776 2777 2778
void Builtins::Generate_WasmReturnPromiseOnSuspend(MacroAssembler* masm) {
  // TODO(v8:12191): Implement for this platform.
  __ Trap();
}

2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172
void Builtins::Generate_WasmOnStackReplace(MacroAssembler* masm) {
  // Only needed on x64.
  __ Trap();
}

#endif  // V8_ENABLE_WEBASSEMBLY

void Builtins::Generate_CEntry(MacroAssembler* masm, int result_size,
                               SaveFPRegsMode save_doubles, ArgvMode argv_mode,
                               bool builtin_exit_frame) {
  // Called from JavaScript; parameters are on stack as if calling JS function
  // a0: number of arguments including receiver
  // a1: pointer to builtin function
  // fp: frame pointer    (restored after C call)
  // sp: stack pointer    (restored as callee's sp after C call)
  // cp: current context  (C callee-saved)
  //
  // If argv_mode == ArgvMode::kRegister:
  // a2: pointer to the first argument

  if (argv_mode == ArgvMode::kRegister) {
    // Move argv into the correct register.
    __ mov(s1, a2);
  } else {
    // Compute the argv pointer in a callee-saved register.
    __ Alsl_d(s1, a0, sp, kPointerSizeLog2, t7);
    __ Sub_d(s1, s1, kPointerSize);
  }

  // Enter the exit frame that transitions from JavaScript to C++.
  FrameScope scope(masm, StackFrame::MANUAL);
  __ EnterExitFrame(
      save_doubles == SaveFPRegsMode::kSave, 0,
      builtin_exit_frame ? StackFrame::BUILTIN_EXIT : StackFrame::EXIT);

  // s0: number of arguments  including receiver (C callee-saved)
  // s1: pointer to first argument (C callee-saved)
  // s2: pointer to builtin function (C callee-saved)

  // Prepare arguments for C routine.
  // a0 = argc
  __ mov(s0, a0);
  __ mov(s2, a1);

  // We are calling compiled C/C++ code. a0 and a1 hold our two arguments. We
  // also need to reserve the 4 argument slots on the stack.

  __ AssertStackIsAligned();

  // a0 = argc, a1 = argv, a2 = isolate
  __ li(a2, ExternalReference::isolate_address(masm->isolate()));
  __ mov(a1, s1);

  __ StoreReturnAddressAndCall(s2);

  // Result returned in a0 or a1:a0 - do not destroy these registers!

  // Check result for exception sentinel.
  Label exception_returned;
  __ LoadRoot(a4, RootIndex::kException);
  __ Branch(&exception_returned, eq, a4, Operand(a0));

  // Check that there is no pending exception, otherwise we
  // should have returned the exception sentinel.
  if (FLAG_debug_code) {
    Label okay;
    ExternalReference pending_exception_address = ExternalReference::Create(
        IsolateAddressId::kPendingExceptionAddress, masm->isolate());
    __ li(a2, pending_exception_address);
    __ Ld_d(a2, MemOperand(a2, 0));
    __ LoadRoot(a4, RootIndex::kTheHoleValue);
    // Cannot use check here as it attempts to generate call into runtime.
    __ Branch(&okay, eq, a4, Operand(a2));
    __ stop();
    __ bind(&okay);
  }

  // Exit C frame and return.
  // a0:a1: result
  // sp: stack pointer
  // fp: frame pointer
  Register argc = argv_mode == ArgvMode::kRegister
                      // We don't want to pop arguments so set argc to no_reg.
                      ? no_reg
                      // s0: still holds argc (callee-saved).
                      : s0;
  __ LeaveExitFrame(save_doubles == SaveFPRegsMode::kSave, argc, EMIT_RETURN);

  // Handling of exception.
  __ bind(&exception_returned);

  ExternalReference pending_handler_context_address = ExternalReference::Create(
      IsolateAddressId::kPendingHandlerContextAddress, masm->isolate());
  ExternalReference pending_handler_entrypoint_address =
      ExternalReference::Create(
          IsolateAddressId::kPendingHandlerEntrypointAddress, masm->isolate());
  ExternalReference pending_handler_fp_address = ExternalReference::Create(
      IsolateAddressId::kPendingHandlerFPAddress, masm->isolate());
  ExternalReference pending_handler_sp_address = ExternalReference::Create(
      IsolateAddressId::kPendingHandlerSPAddress, masm->isolate());

  // Ask the runtime for help to determine the handler. This will set a0 to
  // contain the current pending exception, don't clobber it.
  ExternalReference find_handler =
      ExternalReference::Create(Runtime::kUnwindAndFindExceptionHandler);
  {
    FrameScope scope(masm, StackFrame::MANUAL);
    __ PrepareCallCFunction(3, 0, a0);
    __ mov(a0, zero_reg);
    __ mov(a1, zero_reg);
    __ li(a2, ExternalReference::isolate_address(masm->isolate()));
    __ CallCFunction(find_handler, 3);
  }

  // Retrieve the handler context, SP and FP.
  __ li(cp, pending_handler_context_address);
  __ Ld_d(cp, MemOperand(cp, 0));
  __ li(sp, pending_handler_sp_address);
  __ Ld_d(sp, MemOperand(sp, 0));
  __ li(fp, pending_handler_fp_address);
  __ Ld_d(fp, MemOperand(fp, 0));

  // If the handler is a JS frame, restore the context to the frame. Note that
  // the context will be set to (cp == 0) for non-JS frames.
  Label zero;
  __ Branch(&zero, eq, cp, Operand(zero_reg));
  __ St_d(cp, MemOperand(fp, StandardFrameConstants::kContextOffset));
  __ bind(&zero);

  // Clear c_entry_fp, like we do in `LeaveExitFrame`.
  {
    UseScratchRegisterScope temps(masm);
    Register scratch = temps.Acquire();
    __ li(scratch, ExternalReference::Create(IsolateAddressId::kCEntryFPAddress,
                                             masm->isolate()));
    __ St_d(zero_reg, MemOperand(scratch, 0));
  }

  // Compute the handler entry address and jump to it.
  __ li(t7, pending_handler_entrypoint_address);
  __ Ld_d(t7, MemOperand(t7, 0));
  __ Jump(t7);
}

void Builtins::Generate_DoubleToI(MacroAssembler* masm) {
  Label done;
  Register result_reg = t0;

  Register scratch = GetRegisterThatIsNotOneOf(result_reg);
  Register scratch2 = GetRegisterThatIsNotOneOf(result_reg, scratch);
  Register scratch3 = GetRegisterThatIsNotOneOf(result_reg, scratch, scratch2);
  DoubleRegister double_scratch = kScratchDoubleReg;

  // Account for saved regs.
  const int kArgumentOffset = 4 * kPointerSize;

  __ Push(result_reg);
  __ Push(scratch, scratch2, scratch3);

  // Load double input.
  __ Fld_d(double_scratch, MemOperand(sp, kArgumentOffset));

  // Try a conversion to a signed integer.
  __ ftintrz_w_d(double_scratch, double_scratch);
  // Move the converted value into the result register.
  __ movfr2gr_s(scratch3, double_scratch);

  // Retrieve and restore the FCSR.
  __ movfcsr2gr(scratch);

  // Check for overflow and NaNs.
  __ And(scratch, scratch,
         kFCSRExceptionCauseMask ^ kFCSRDivideByZeroCauseMask);
  // If we had no exceptions then set result_reg and we are done.
  Label error;
  __ Branch(&error, ne, scratch, Operand(zero_reg));
  __ Move(result_reg, scratch3);
  __ Branch(&done);
  __ bind(&error);

  // Load the double value and perform a manual truncation.
  Register input_high = scratch2;
  Register input_low = scratch3;

  __ Ld_w(input_low,
          MemOperand(sp, kArgumentOffset + Register::kMantissaOffset));
  __ Ld_w(input_high,
          MemOperand(sp, kArgumentOffset + Register::kExponentOffset));

  Label normal_exponent;
  // Extract the biased exponent in result.
  __ bstrpick_w(result_reg, input_high,
                HeapNumber::kExponentShift + HeapNumber::kExponentBits - 1,
                HeapNumber::kExponentShift);

  // Check for Infinity and NaNs, which should return 0.
  __ Sub_w(scratch, result_reg, HeapNumber::kExponentMask);
  __ Movz(result_reg, zero_reg, scratch);
  __ Branch(&done, eq, scratch, Operand(zero_reg));

  // Express exponent as delta to (number of mantissa bits + 31).
  __ Sub_w(result_reg, result_reg,
           Operand(HeapNumber::kExponentBias + HeapNumber::kMantissaBits + 31));

  // If the delta is strictly positive, all bits would be shifted away,
  // which means that we can return 0.
  __ Branch(&normal_exponent, le, result_reg, Operand(zero_reg));
  __ mov(result_reg, zero_reg);
  __ Branch(&done);

  __ bind(&normal_exponent);
  const int kShiftBase = HeapNumber::kNonMantissaBitsInTopWord - 1;
  // Calculate shift.
  __ Add_w(scratch, result_reg,
           Operand(kShiftBase + HeapNumber::kMantissaBits));

  // Save the sign.
  Register sign = result_reg;
  result_reg = no_reg;
  __ And(sign, input_high, Operand(HeapNumber::kSignMask));

  // On ARM shifts > 31 bits are valid and will result in zero. On LOONG64 we
  // need to check for this specific case.
  Label high_shift_needed, high_shift_done;
  __ Branch(&high_shift_needed, lt, scratch, Operand(32));
  __ mov(input_high, zero_reg);
  __ Branch(&high_shift_done);
  __ bind(&high_shift_needed);

  // Set the implicit 1 before the mantissa part in input_high.
  __ Or(input_high, input_high,
        Operand(1 << HeapNumber::kMantissaBitsInTopWord));
  // Shift the mantissa bits to the correct position.
  // We don't need to clear non-mantissa bits as they will be shifted away.
  // If they weren't, it would mean that the answer is in the 32bit range.
  __ sll_w(input_high, input_high, scratch);

  __ bind(&high_shift_done);

  // Replace the shifted bits with bits from the lower mantissa word.
  Label pos_shift, shift_done;
  __ li(kScratchReg, 32);
  __ sub_w(scratch, kScratchReg, scratch);
  __ Branch(&pos_shift, ge, scratch, Operand(zero_reg));

  // Negate scratch.
  __ Sub_w(scratch, zero_reg, scratch);
  __ sll_w(input_low, input_low, scratch);
  __ Branch(&shift_done);

  __ bind(&pos_shift);
  __ srl_w(input_low, input_low, scratch);

  __ bind(&shift_done);
  __ Or(input_high, input_high, Operand(input_low));
  // Restore sign if necessary.
  __ mov(scratch, sign);
  result_reg = sign;
  sign = no_reg;
  __ Sub_w(result_reg, zero_reg, input_high);
  __ Movz(result_reg, input_high, scratch);

  __ bind(&done);

  __ St_d(result_reg, MemOperand(sp, kArgumentOffset));
  __ Pop(scratch, scratch2, scratch3);
  __ Pop(result_reg);
  __ Ret();
}

namespace {

int AddressOffset(ExternalReference ref0, ExternalReference ref1) {
  int64_t offset = (ref0.address() - ref1.address());
  DCHECK(static_cast<int>(offset) == offset);
  return static_cast<int>(offset);
}

// Calls an API function.  Allocates HandleScope, extracts returned value
// from handle and propagates exceptions.  Restores context.  stack_space
// - space to be unwound on exit (includes the call JS arguments space and
// the additional space allocated for the fast call).
void CallApiFunctionAndReturn(MacroAssembler* masm, Register function_address,
                              ExternalReference thunk_ref, int stack_space,
                              MemOperand* stack_space_operand,
                              MemOperand return_value_operand) {
  Isolate* isolate = masm->isolate();
  ExternalReference next_address =
      ExternalReference::handle_scope_next_address(isolate);
  const int kNextOffset = 0;
  const int kLimitOffset = AddressOffset(
      ExternalReference::handle_scope_limit_address(isolate), next_address);
  const int kLevelOffset = AddressOffset(
      ExternalReference::handle_scope_level_address(isolate), next_address);

  DCHECK(function_address == a1 || function_address == a2);

  Label profiler_enabled, end_profiler_check;
  __ li(t7, ExternalReference::is_profiling_address(isolate));
  __ Ld_b(t7, MemOperand(t7, 0));
  __ Branch(&profiler_enabled, ne, t7, Operand(zero_reg));
  __ li(t7, ExternalReference::address_of_runtime_stats_flag());
  __ Ld_w(t7, MemOperand(t7, 0));
  __ Branch(&profiler_enabled, ne, t7, Operand(zero_reg));
  {
    // Call the api function directly.
    __ mov(t7, function_address);
    __ Branch(&end_profiler_check);
  }

  __ bind(&profiler_enabled);
  {
    // Additional parameter is the address of the actual callback.
    __ li(t7, thunk_ref);
  }
  __ bind(&end_profiler_check);

  // Allocate HandleScope in callee-save registers.
  __ li(s5, next_address);
  __ Ld_d(s0, MemOperand(s5, kNextOffset));
  __ Ld_d(s1, MemOperand(s5, kLimitOffset));
  __ Ld_w(s2, MemOperand(s5, kLevelOffset));
  __ Add_w(s2, s2, Operand(1));
  __ St_w(s2, MemOperand(s5, kLevelOffset));

  __ StoreReturnAddressAndCall(t7);

  Label promote_scheduled_exception;
  Label delete_allocated_handles;
  Label leave_exit_frame;
  Label return_value_loaded;

  // Load value from ReturnValue.
  __ Ld_d(a0, return_value_operand);
  __ bind(&return_value_loaded);

  // No more valid handles (the result handle was the last one). Restore
  // previous handle scope.
  __ St_d(s0, MemOperand(s5, kNextOffset));
  if (FLAG_debug_code) {
    __ Ld_w(a1, MemOperand(s5, kLevelOffset));
    __ Check(eq, AbortReason::kUnexpectedLevelAfterReturnFromApiCall, a1,
             Operand(s2));
  }
  __ Sub_w(s2, s2, Operand(1));
  __ St_w(s2, MemOperand(s5, kLevelOffset));
  __ Ld_d(kScratchReg, MemOperand(s5, kLimitOffset));
  __ Branch(&delete_allocated_handles, ne, s1, Operand(kScratchReg));

  // Leave the API exit frame.
  __ bind(&leave_exit_frame);

  if (stack_space_operand == nullptr) {
    DCHECK_NE(stack_space, 0);
    __ li(s0, Operand(stack_space));
  } else {
    DCHECK_EQ(stack_space, 0);
    __ Ld_d(s0, *stack_space_operand);
  }

  static constexpr bool kDontSaveDoubles = false;
  static constexpr bool kRegisterContainsSlotCount = false;
  __ LeaveExitFrame(kDontSaveDoubles, s0, NO_EMIT_RETURN,
                    kRegisterContainsSlotCount);

  // Check if the function scheduled an exception.
  __ LoadRoot(a4, RootIndex::kTheHoleValue);
  __ li(kScratchReg, ExternalReference::scheduled_exception_address(isolate));
  __ Ld_d(a5, MemOperand(kScratchReg, 0));
  __ Branch(&promote_scheduled_exception, ne, a4, Operand(a5));

  __ Ret();

  // Re-throw by promoting a scheduled exception.
  __ bind(&promote_scheduled_exception);
  __ TailCallRuntime(Runtime::kPromoteScheduledException);

  // HandleScope limit has changed. Delete allocated extensions.
  __ bind(&delete_allocated_handles);
  __ St_d(s1, MemOperand(s5, kLimitOffset));
  __ mov(s0, a0);
  __ PrepareCallCFunction(1, s1);
  __ li(a0, ExternalReference::isolate_address(isolate));
  __ CallCFunction(ExternalReference::delete_handle_scope_extensions(), 1);
  __ mov(a0, s0);
  __ jmp(&leave_exit_frame);
}

}  // namespace

void Builtins::Generate_CallApiCallback(MacroAssembler* masm) {
  // ----------- S t a t e -------------
  //  -- cp                  : context
  //  -- a1                  : api function address
3173
  //  -- a2                  : arguments count
3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631
  //  -- a3                  : call data
  //  -- a0                  : holder
  //  -- sp[0]               : receiver
  //  -- sp[8]               : first argument
  //  -- ...
  //  -- sp[(argc) * 8]      : last argument
  // -----------------------------------

  Register api_function_address = a1;
  Register argc = a2;
  Register call_data = a3;
  Register holder = a0;
  Register scratch = t0;
  Register base = t1;  // For addressing MemOperands on the stack.

  DCHECK(!AreAliased(api_function_address, argc, call_data, holder, scratch,
                     base));

  using FCA = FunctionCallbackArguments;

  STATIC_ASSERT(FCA::kArgsLength == 6);
  STATIC_ASSERT(FCA::kNewTargetIndex == 5);
  STATIC_ASSERT(FCA::kDataIndex == 4);
  STATIC_ASSERT(FCA::kReturnValueOffset == 3);
  STATIC_ASSERT(FCA::kReturnValueDefaultValueIndex == 2);
  STATIC_ASSERT(FCA::kIsolateIndex == 1);
  STATIC_ASSERT(FCA::kHolderIndex == 0);

  // Set up FunctionCallbackInfo's implicit_args on the stack as follows:
  //
  // Target state:
  //   sp[0 * kPointerSize]: kHolder
  //   sp[1 * kPointerSize]: kIsolate
  //   sp[2 * kPointerSize]: undefined (kReturnValueDefaultValue)
  //   sp[3 * kPointerSize]: undefined (kReturnValue)
  //   sp[4 * kPointerSize]: kData
  //   sp[5 * kPointerSize]: undefined (kNewTarget)

  // Set up the base register for addressing through MemOperands. It will point
  // at the receiver (located at sp + argc * kPointerSize).
  __ Alsl_d(base, argc, sp, kPointerSizeLog2, t7);

  // Reserve space on the stack.
  __ Sub_d(sp, sp, Operand(FCA::kArgsLength * kPointerSize));

  // kHolder.
  __ St_d(holder, MemOperand(sp, 0 * kPointerSize));

  // kIsolate.
  __ li(scratch, ExternalReference::isolate_address(masm->isolate()));
  __ St_d(scratch, MemOperand(sp, 1 * kPointerSize));

  // kReturnValueDefaultValue and kReturnValue.
  __ LoadRoot(scratch, RootIndex::kUndefinedValue);
  __ St_d(scratch, MemOperand(sp, 2 * kPointerSize));
  __ St_d(scratch, MemOperand(sp, 3 * kPointerSize));

  // kData.
  __ St_d(call_data, MemOperand(sp, 4 * kPointerSize));

  // kNewTarget.
  __ St_d(scratch, MemOperand(sp, 5 * kPointerSize));

  // Keep a pointer to kHolder (= implicit_args) in a scratch register.
  // We use it below to set up the FunctionCallbackInfo object.
  __ mov(scratch, sp);

  // Allocate the v8::Arguments structure in the arguments' space since
  // it's not controlled by GC.
  static constexpr int kApiStackSpace = 4;
  static constexpr bool kDontSaveDoubles = false;
  FrameScope frame_scope(masm, StackFrame::MANUAL);
  __ EnterExitFrame(kDontSaveDoubles, kApiStackSpace);

  // EnterExitFrame may align the sp.

  // FunctionCallbackInfo::implicit_args_ (points at kHolder as set up above).
  // Arguments are after the return address (pushed by EnterExitFrame()).
  __ St_d(scratch, MemOperand(sp, 1 * kPointerSize));

  // FunctionCallbackInfo::values_ (points at the first varargs argument passed
  // on the stack).
  __ Add_d(scratch, scratch,
           Operand((FCA::kArgsLength + 1) * kSystemPointerSize));

  __ St_d(scratch, MemOperand(sp, 2 * kPointerSize));

  // FunctionCallbackInfo::length_.
  // Stored as int field, 32-bit integers within struct on stack always left
  // justified by n64 ABI.
  __ St_w(argc, MemOperand(sp, 3 * kPointerSize));

  // We also store the number of bytes to drop from the stack after returning
  // from the API function here.
  // Note: Unlike on other architectures, this stores the number of slots to
  // drop, not the number of bytes.
  __ Add_d(scratch, argc, Operand(FCA::kArgsLength + 1 /* receiver */));
  __ St_d(scratch, MemOperand(sp, 4 * kPointerSize));

  // v8::InvocationCallback's argument.
  DCHECK(!AreAliased(api_function_address, scratch, a0));
  __ Add_d(a0, sp, Operand(1 * kPointerSize));

  ExternalReference thunk_ref = ExternalReference::invoke_function_callback();

  // There are two stack slots above the arguments we constructed on the stack.
  // TODO(jgruber): Document what these arguments are.
  static constexpr int kStackSlotsAboveFCA = 2;
  MemOperand return_value_operand(
      fp, (kStackSlotsAboveFCA + FCA::kReturnValueOffset) * kPointerSize);

  static constexpr int kUseStackSpaceOperand = 0;
  MemOperand stack_space_operand(sp, 4 * kPointerSize);

  AllowExternalCallThatCantCauseGC scope(masm);
  CallApiFunctionAndReturn(masm, api_function_address, thunk_ref,
                           kUseStackSpaceOperand, &stack_space_operand,
                           return_value_operand);
}

void Builtins::Generate_CallApiGetter(MacroAssembler* masm) {
  // Build v8::PropertyCallbackInfo::args_ array on the stack and push property
  // name below the exit frame to make GC aware of them.
  STATIC_ASSERT(PropertyCallbackArguments::kShouldThrowOnErrorIndex == 0);
  STATIC_ASSERT(PropertyCallbackArguments::kHolderIndex == 1);
  STATIC_ASSERT(PropertyCallbackArguments::kIsolateIndex == 2);
  STATIC_ASSERT(PropertyCallbackArguments::kReturnValueDefaultValueIndex == 3);
  STATIC_ASSERT(PropertyCallbackArguments::kReturnValueOffset == 4);
  STATIC_ASSERT(PropertyCallbackArguments::kDataIndex == 5);
  STATIC_ASSERT(PropertyCallbackArguments::kThisIndex == 6);
  STATIC_ASSERT(PropertyCallbackArguments::kArgsLength == 7);

  Register receiver = ApiGetterDescriptor::ReceiverRegister();
  Register holder = ApiGetterDescriptor::HolderRegister();
  Register callback = ApiGetterDescriptor::CallbackRegister();
  Register scratch = a4;
  DCHECK(!AreAliased(receiver, holder, callback, scratch));

  Register api_function_address = a2;

  // Here and below +1 is for name() pushed after the args_ array.
  using PCA = PropertyCallbackArguments;
  __ Sub_d(sp, sp, (PCA::kArgsLength + 1) * kPointerSize);
  __ St_d(receiver, MemOperand(sp, (PCA::kThisIndex + 1) * kPointerSize));
  __ Ld_d(scratch, FieldMemOperand(callback, AccessorInfo::kDataOffset));
  __ St_d(scratch, MemOperand(sp, (PCA::kDataIndex + 1) * kPointerSize));
  __ LoadRoot(scratch, RootIndex::kUndefinedValue);
  __ St_d(scratch,
          MemOperand(sp, (PCA::kReturnValueOffset + 1) * kPointerSize));
  __ St_d(scratch, MemOperand(sp, (PCA::kReturnValueDefaultValueIndex + 1) *
                                      kPointerSize));
  __ li(scratch, ExternalReference::isolate_address(masm->isolate()));
  __ St_d(scratch, MemOperand(sp, (PCA::kIsolateIndex + 1) * kPointerSize));
  __ St_d(holder, MemOperand(sp, (PCA::kHolderIndex + 1) * kPointerSize));
  // should_throw_on_error -> false
  DCHECK_EQ(0, Smi::zero().ptr());
  __ St_d(zero_reg,
          MemOperand(sp, (PCA::kShouldThrowOnErrorIndex + 1) * kPointerSize));
  __ Ld_d(scratch, FieldMemOperand(callback, AccessorInfo::kNameOffset));
  __ St_d(scratch, MemOperand(sp, 0 * kPointerSize));

  // v8::PropertyCallbackInfo::args_ array and name handle.
  const int kStackUnwindSpace = PropertyCallbackArguments::kArgsLength + 1;

  // Load address of v8::PropertyAccessorInfo::args_ array and name handle.
  __ mov(a0, sp);                               // a0 = Handle<Name>
  __ Add_d(a1, a0, Operand(1 * kPointerSize));  // a1 = v8::PCI::args_

  const int kApiStackSpace = 1;
  FrameScope frame_scope(masm, StackFrame::MANUAL);
  __ EnterExitFrame(false, kApiStackSpace);

  // Create v8::PropertyCallbackInfo object on the stack and initialize
  // it's args_ field.
  __ St_d(a1, MemOperand(sp, 1 * kPointerSize));
  __ Add_d(a1, sp, Operand(1 * kPointerSize));
  // a1 = v8::PropertyCallbackInfo&

  ExternalReference thunk_ref =
      ExternalReference::invoke_accessor_getter_callback();

  __ Ld_d(scratch, FieldMemOperand(callback, AccessorInfo::kJsGetterOffset));
  __ Ld_d(api_function_address,
          FieldMemOperand(scratch, Foreign::kForeignAddressOffset));

  // +3 is to skip prolog, return address and name handle.
  MemOperand return_value_operand(
      fp, (PropertyCallbackArguments::kReturnValueOffset + 3) * kPointerSize);
  MemOperand* const kUseStackSpaceConstant = nullptr;
  CallApiFunctionAndReturn(masm, api_function_address, thunk_ref,
                           kStackUnwindSpace, kUseStackSpaceConstant,
                           return_value_operand);
}

void Builtins::Generate_DirectCEntry(MacroAssembler* masm) {
  // The sole purpose of DirectCEntry is for movable callers (e.g. any general
  // purpose Code object) to be able to call into C functions that may trigger
  // GC and thus move the caller.
  //
  // DirectCEntry places the return address on the stack (updated by the GC),
  // making the call GC safe. The irregexp backend relies on this.

  __ St_d(ra, MemOperand(sp, 0));  // Store the return address.
  __ Call(t7);                     // Call the C++ function.
  __ Ld_d(ra, MemOperand(sp, 0));  // Return to calling code.

  // TODO(LOONG_dev): LOONG64 Check this assert.
  if (FLAG_debug_code && FLAG_enable_slow_asserts) {
    // In case of an error the return address may point to a memory area
    // filled with kZapValue by the GC. Dereference the address and check for
    // this.
    __ Ld_d(a4, MemOperand(ra, 0));
    __ Assert(ne, AbortReason::kReceivedInvalidReturnAddress, a4,
              Operand(reinterpret_cast<uint64_t>(kZapValue)));
  }

  __ Jump(ra);
}

namespace {

// This code tries to be close to ia32 code so that any changes can be
// easily ported.
void Generate_DeoptimizationEntry(MacroAssembler* masm,
                                  DeoptimizeKind deopt_kind) {
  Isolate* isolate = masm->isolate();

  // Unlike on ARM we don't save all the registers, just the useful ones.
  // For the rest, there are gaps on the stack, so the offsets remain the same.
  const int kNumberOfRegisters = Register::kNumRegisters;

  RegList restored_regs = kJSCallerSaved | kCalleeSaved;
  RegList saved_regs = restored_regs | sp.bit() | ra.bit();

  const int kDoubleRegsSize = kDoubleSize * DoubleRegister::kNumRegisters;

  // Save all double FPU registers before messing with them.
  __ Sub_d(sp, sp, Operand(kDoubleRegsSize));
  const RegisterConfiguration* config = RegisterConfiguration::Default();
  for (int i = 0; i < config->num_allocatable_double_registers(); ++i) {
    int code = config->GetAllocatableDoubleCode(i);
    const DoubleRegister fpu_reg = DoubleRegister::from_code(code);
    int offset = code * kDoubleSize;
    __ Fst_d(fpu_reg, MemOperand(sp, offset));
  }

  // Push saved_regs (needed to populate FrameDescription::registers_).
  // Leave gaps for other registers.
  __ Sub_d(sp, sp, kNumberOfRegisters * kPointerSize);
  for (int16_t i = kNumberOfRegisters - 1; i >= 0; i--) {
    if ((saved_regs & (1 << i)) != 0) {
      __ St_d(ToRegister(i), MemOperand(sp, kPointerSize * i));
    }
  }

  __ li(a2,
        ExternalReference::Create(IsolateAddressId::kCEntryFPAddress, isolate));
  __ St_d(fp, MemOperand(a2, 0));

  const int kSavedRegistersAreaSize =
      (kNumberOfRegisters * kPointerSize) + kDoubleRegsSize;

  __ li(a2, Operand(Deoptimizer::kFixedExitSizeMarker));
  // Get the address of the location in the code object (a3) (return
  // address for lazy deoptimization) and compute the fp-to-sp delta in
  // register a4.
  __ mov(a3, ra);
  __ Add_d(a4, sp, Operand(kSavedRegistersAreaSize));

  __ sub_d(a4, fp, a4);

  // Allocate a new deoptimizer object.
  __ PrepareCallCFunction(6, a5);
  // Pass six arguments, according to n64 ABI.
  __ mov(a0, zero_reg);
  Label context_check;
  __ Ld_d(a1, MemOperand(fp, CommonFrameConstants::kContextOrFrameTypeOffset));
  __ JumpIfSmi(a1, &context_check);
  __ Ld_d(a0, MemOperand(fp, StandardFrameConstants::kFunctionOffset));
  __ bind(&context_check);
  __ li(a1, Operand(static_cast<int>(deopt_kind)));
  // a2: bailout id already loaded.
  // a3: code address or 0 already loaded.
  // a4: already has fp-to-sp delta.
  __ li(a5, ExternalReference::isolate_address(isolate));

  // Call Deoptimizer::New().
  {
    AllowExternalCallThatCantCauseGC scope(masm);
    __ CallCFunction(ExternalReference::new_deoptimizer_function(), 6);
  }

  // Preserve "deoptimizer" object in register a0 and get the input
  // frame descriptor pointer to a1 (deoptimizer->input_);
  // Move deopt-obj to a0 for call to Deoptimizer::ComputeOutputFrames() below.
  __ Ld_d(a1, MemOperand(a0, Deoptimizer::input_offset()));

  // Copy core registers into FrameDescription::registers_[kNumRegisters].
  DCHECK_EQ(Register::kNumRegisters, kNumberOfRegisters);
  for (int i = 0; i < kNumberOfRegisters; i++) {
    int offset = (i * kPointerSize) + FrameDescription::registers_offset();
    if ((saved_regs & (1 << i)) != 0) {
      __ Ld_d(a2, MemOperand(sp, i * kPointerSize));
      __ St_d(a2, MemOperand(a1, offset));
    } else if (FLAG_debug_code) {
      __ li(a2, Operand(kDebugZapValue));
      __ St_d(a2, MemOperand(a1, offset));
    }
  }

  int double_regs_offset = FrameDescription::double_registers_offset();
  // Copy FPU registers to
  // double_registers_[DoubleRegister::kNumAllocatableRegisters]
  for (int i = 0; i < config->num_allocatable_double_registers(); ++i) {
    int code = config->GetAllocatableDoubleCode(i);
    int dst_offset = code * kDoubleSize + double_regs_offset;
    int src_offset = code * kDoubleSize + kNumberOfRegisters * kPointerSize;
    __ Fld_d(f0, MemOperand(sp, src_offset));
    __ Fst_d(f0, MemOperand(a1, dst_offset));
  }

  // Remove the saved registers from the stack.
  __ Add_d(sp, sp, Operand(kSavedRegistersAreaSize));

  // Compute a pointer to the unwinding limit in register a2; that is
  // the first stack slot not part of the input frame.
  __ Ld_d(a2, MemOperand(a1, FrameDescription::frame_size_offset()));
  __ add_d(a2, a2, sp);

  // Unwind the stack down to - but not including - the unwinding
  // limit and copy the contents of the activation frame to the input
  // frame description.
  __ Add_d(a3, a1, Operand(FrameDescription::frame_content_offset()));
  Label pop_loop;
  Label pop_loop_header;
  __ Branch(&pop_loop_header);
  __ bind(&pop_loop);
  __ Pop(a4);
  __ St_d(a4, MemOperand(a3, 0));
  __ addi_d(a3, a3, sizeof(uint64_t));
  __ bind(&pop_loop_header);
  __ BranchShort(&pop_loop, ne, a2, Operand(sp));
  // Compute the output frame in the deoptimizer.
  __ Push(a0);  // Preserve deoptimizer object across call.
  // a0: deoptimizer object; a1: scratch.
  __ PrepareCallCFunction(1, a1);
  // Call Deoptimizer::ComputeOutputFrames().
  {
    AllowExternalCallThatCantCauseGC scope(masm);
    __ CallCFunction(ExternalReference::compute_output_frames_function(), 1);
  }
  __ Pop(a0);  // Restore deoptimizer object (class Deoptimizer).

  __ Ld_d(sp, MemOperand(a0, Deoptimizer::caller_frame_top_offset()));

  // Replace the current (input) frame with the output frames.
  Label outer_push_loop, inner_push_loop, outer_loop_header, inner_loop_header;
  // Outer loop state: a4 = current "FrameDescription** output_",
  // a1 = one past the last FrameDescription**.
  __ Ld_w(a1, MemOperand(a0, Deoptimizer::output_count_offset()));
  __ Ld_d(a4, MemOperand(a0, Deoptimizer::output_offset()));  // a4 is output_.
  __ Alsl_d(a1, a1, a4, kPointerSizeLog2);
  __ Branch(&outer_loop_header);
  __ bind(&outer_push_loop);
  // Inner loop state: a2 = current FrameDescription*, a3 = loop index.
  __ Ld_d(a2, MemOperand(a4, 0));  // output_[ix]
  __ Ld_d(a3, MemOperand(a2, FrameDescription::frame_size_offset()));
  __ Branch(&inner_loop_header);
  __ bind(&inner_push_loop);
  __ Sub_d(a3, a3, Operand(sizeof(uint64_t)));
  __ Add_d(a6, a2, Operand(a3));
  __ Ld_d(a7, MemOperand(a6, FrameDescription::frame_content_offset()));
  __ Push(a7);
  __ bind(&inner_loop_header);
  __ BranchShort(&inner_push_loop, ne, a3, Operand(zero_reg));

  __ Add_d(a4, a4, Operand(kPointerSize));
  __ bind(&outer_loop_header);
  __ BranchShort(&outer_push_loop, lt, a4, Operand(a1));

  __ Ld_d(a1, MemOperand(a0, Deoptimizer::input_offset()));
  for (int i = 0; i < config->num_allocatable_double_registers(); ++i) {
    int code = config->GetAllocatableDoubleCode(i);
    const DoubleRegister fpu_reg = DoubleRegister::from_code(code);
    int src_offset = code * kDoubleSize + double_regs_offset;
    __ Fld_d(fpu_reg, MemOperand(a1, src_offset));
  }

  // Push pc and continuation from the last output frame.
  __ Ld_d(a6, MemOperand(a2, FrameDescription::pc_offset()));
  __ Push(a6);
  __ Ld_d(a6, MemOperand(a2, FrameDescription::continuation_offset()));
  __ Push(a6);

  // Technically restoring 'at' should work unless zero_reg is also restored
  // but it's safer to check for this.
  DCHECK(!(t7.bit() & restored_regs));
  // Restore the registers from the last output frame.
  __ mov(t7, a2);
  for (int i = kNumberOfRegisters - 1; i >= 0; i--) {
    int offset = (i * kPointerSize) + FrameDescription::registers_offset();
    if ((restored_regs & (1 << i)) != 0) {
      __ Ld_d(ToRegister(i), MemOperand(t7, offset));
    }
  }

  __ Pop(t7);  // Get continuation, leave pc on stack.
  __ Pop(ra);
  __ Jump(t7);
  __ stop();
}

}  // namespace

void Builtins::Generate_DeoptimizationEntry_Eager(MacroAssembler* masm) {
  Generate_DeoptimizationEntry(masm, DeoptimizeKind::kEager);
}

void Builtins::Generate_DeoptimizationEntry_Soft(MacroAssembler* masm) {
  Generate_DeoptimizationEntry(masm, DeoptimizeKind::kSoft);
}

void Builtins::Generate_DeoptimizationEntry_Bailout(MacroAssembler* masm) {
  Generate_DeoptimizationEntry(masm, DeoptimizeKind::kBailout);
}

void Builtins::Generate_DeoptimizationEntry_Lazy(MacroAssembler* masm) {
  Generate_DeoptimizationEntry(masm, DeoptimizeKind::kLazy);
}

namespace {

// Restarts execution either at the current or next (in execution order)
// bytecode. If there is baseline code on the shared function info, converts an
// interpreter frame into a baseline frame and continues execution in baseline
// code. Otherwise execution continues with bytecode.
void Generate_BaselineOrInterpreterEntry(MacroAssembler* masm,
                                         bool next_bytecode,
                                         bool is_osr = false) {
  Label start;
  __ bind(&start);

  // Get function from the frame.
  Register closure = a1;
  __ Ld_d(closure, MemOperand(fp, StandardFrameConstants::kFunctionOffset));

  // Get the Code object from the shared function info.
  Register code_obj = s1;
  __ Ld_d(code_obj,
          FieldMemOperand(closure, JSFunction::kSharedFunctionInfoOffset));
  __ Ld_d(code_obj,
          FieldMemOperand(code_obj, SharedFunctionInfo::kFunctionDataOffset));

  // Check if we have baseline code. For OSR entry it is safe to assume we
  // always have baseline code.
  if (!is_osr) {
    Label start_with_baseline;
    __ GetObjectType(code_obj, t2, t2);
3632
    __ Branch(&start_with_baseline, eq, t2, Operand(CODET_TYPE));
3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644

    // Start with bytecode as there is no baseline code.
    Builtin builtin_id = next_bytecode
                             ? Builtin::kInterpreterEnterAtNextBytecode
                             : Builtin::kInterpreterEnterAtBytecode;
    __ Jump(masm->isolate()->builtins()->code_handle(builtin_id),
            RelocInfo::CODE_TARGET);

    // Start with baseline code.
    __ bind(&start_with_baseline);
  } else if (FLAG_debug_code) {
    __ GetObjectType(code_obj, t2, t2);
3645
    __ Assert(eq, AbortReason::kExpectedBaselineData, t2, Operand(CODET_TYPE));
3646 3647
  }

3648 3649 3650
  if (FLAG_debug_code) {
    AssertCodeIsBaseline(masm, code_obj, t2);
  }
3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853

  // Replace BytecodeOffset with the feedback vector.
  Register feedback_vector = a2;
  __ Ld_d(feedback_vector,
          FieldMemOperand(closure, JSFunction::kFeedbackCellOffset));
  __ Ld_d(feedback_vector,
          FieldMemOperand(feedback_vector, Cell::kValueOffset));

  Label install_baseline_code;
  // Check if feedback vector is valid. If not, call prepare for baseline to
  // allocate it.
  __ GetObjectType(feedback_vector, t2, t2);
  __ Branch(&install_baseline_code, ne, t2, Operand(FEEDBACK_VECTOR_TYPE));

  // Save BytecodeOffset from the stack frame.
  __ SmiUntag(kInterpreterBytecodeOffsetRegister,
              MemOperand(fp, InterpreterFrameConstants::kBytecodeOffsetFromFp));
  // Replace BytecodeOffset with the feedback vector.
  __ St_d(feedback_vector,
          MemOperand(fp, InterpreterFrameConstants::kBytecodeOffsetFromFp));
  feedback_vector = no_reg;

  // Compute baseline pc for bytecode offset.
  ExternalReference get_baseline_pc_extref;
  if (next_bytecode || is_osr) {
    get_baseline_pc_extref =
        ExternalReference::baseline_pc_for_next_executed_bytecode();
  } else {
    get_baseline_pc_extref =
        ExternalReference::baseline_pc_for_bytecode_offset();
  }

  Register get_baseline_pc = a3;
  __ li(get_baseline_pc, get_baseline_pc_extref);

  // If the code deoptimizes during the implicit function entry stack interrupt
  // check, it will have a bailout ID of kFunctionEntryBytecodeOffset, which is
  // not a valid bytecode offset.
  // TODO(pthier): Investigate if it is feasible to handle this special case
  // in TurboFan instead of here.
  Label valid_bytecode_offset, function_entry_bytecode;
  if (!is_osr) {
    __ Branch(&function_entry_bytecode, eq, kInterpreterBytecodeOffsetRegister,
              Operand(BytecodeArray::kHeaderSize - kHeapObjectTag +
                      kFunctionEntryBytecodeOffset));
  }

  __ Sub_d(kInterpreterBytecodeOffsetRegister,
           kInterpreterBytecodeOffsetRegister,
           (BytecodeArray::kHeaderSize - kHeapObjectTag));

  __ bind(&valid_bytecode_offset);
  // Get bytecode array from the stack frame.
  __ Ld_d(kInterpreterBytecodeArrayRegister,
          MemOperand(fp, InterpreterFrameConstants::kBytecodeArrayFromFp));
  // Save the accumulator register, since it's clobbered by the below call.
  __ Push(kInterpreterAccumulatorRegister);
  {
    Register arg_reg_1 = a0;
    Register arg_reg_2 = a1;
    Register arg_reg_3 = a2;
    __ Move(arg_reg_1, code_obj);
    __ Move(arg_reg_2, kInterpreterBytecodeOffsetRegister);
    __ Move(arg_reg_3, kInterpreterBytecodeArrayRegister);
    FrameScope scope(masm, StackFrame::INTERNAL);
    __ CallCFunction(get_baseline_pc, 3, 0);
  }
  __ Add_d(code_obj, code_obj, kReturnRegister0);
  __ Pop(kInterpreterAccumulatorRegister);

  if (is_osr) {
    // Reset the OSR loop nesting depth to disarm back edges.
    // TODO(pthier): Separate baseline Sparkplug from TF arming and don't disarm
    // Sparkplug here.
    // TODO(liuyu): Remove Ld as arm64 after register reallocation.
    __ Ld_d(kInterpreterBytecodeArrayRegister,
            MemOperand(fp, InterpreterFrameConstants::kBytecodeArrayFromFp));
    __ St_h(zero_reg,
            FieldMemOperand(kInterpreterBytecodeArrayRegister,
                            BytecodeArray::kOsrLoopNestingLevelOffset));
    Generate_OSREntry(masm, code_obj,
                      Operand(Code::kHeaderSize - kHeapObjectTag));
  } else {
    __ Add_d(code_obj, code_obj, Code::kHeaderSize - kHeapObjectTag);
    __ Jump(code_obj);
  }
  __ Trap();  // Unreachable.

  if (!is_osr) {
    __ bind(&function_entry_bytecode);
    // If the bytecode offset is kFunctionEntryOffset, get the start address of
    // the first bytecode.
    __ mov(kInterpreterBytecodeOffsetRegister, zero_reg);
    if (next_bytecode) {
      __ li(get_baseline_pc,
            ExternalReference::baseline_pc_for_bytecode_offset());
    }
    __ Branch(&valid_bytecode_offset);
  }

  __ bind(&install_baseline_code);
  {
    FrameScope scope(masm, StackFrame::INTERNAL);
    __ Push(kInterpreterAccumulatorRegister);
    __ Push(closure);
    __ CallRuntime(Runtime::kInstallBaselineCode, 1);
    __ Pop(kInterpreterAccumulatorRegister);
  }
  // Retry from the start after installing baseline code.
  __ Branch(&start);
}

}  // namespace

void Builtins::Generate_BaselineOrInterpreterEnterAtBytecode(
    MacroAssembler* masm) {
  Generate_BaselineOrInterpreterEntry(masm, false);
}

void Builtins::Generate_BaselineOrInterpreterEnterAtNextBytecode(
    MacroAssembler* masm) {
  Generate_BaselineOrInterpreterEntry(masm, true);
}

void Builtins::Generate_InterpreterOnStackReplacement_ToBaseline(
    MacroAssembler* masm) {
  Generate_BaselineOrInterpreterEntry(masm, false, true);
}

void Builtins::Generate_DynamicCheckMapsTrampoline(MacroAssembler* masm) {
  Generate_DynamicCheckMapsTrampoline<DynamicCheckMapsDescriptor>(
      masm, BUILTIN_CODE(masm->isolate(), DynamicCheckMaps));
}

void Builtins::Generate_DynamicCheckMapsWithFeedbackVectorTrampoline(
    MacroAssembler* masm) {
  Generate_DynamicCheckMapsTrampoline<
      DynamicCheckMapsWithFeedbackVectorDescriptor>(
      masm, BUILTIN_CODE(masm->isolate(), DynamicCheckMapsWithFeedbackVector));
}

template <class Descriptor>
void Builtins::Generate_DynamicCheckMapsTrampoline(
    MacroAssembler* masm, Handle<Code> builtin_target) {
  FrameScope scope(masm, StackFrame::MANUAL);
  __ EnterFrame(StackFrame::INTERNAL);

  // Only save the registers that the DynamicCheckMaps builtin can clobber.
  Descriptor descriptor;
  RegList registers = descriptor.allocatable_registers();
  // FLAG_debug_code is enabled CSA checks will call C function and so we need
  // to save all CallerSaved registers too.
  if (FLAG_debug_code) registers |= kJSCallerSaved;
  __ MaybeSaveRegisters(registers);

  // Load the immediate arguments from the deopt exit to pass to the builtin.
  Register slot_arg = descriptor.GetRegisterParameter(Descriptor::kSlot);
  Register handler_arg = descriptor.GetRegisterParameter(Descriptor::kHandler);
  __ Ld_d(handler_arg, MemOperand(fp, CommonFrameConstants::kCallerPCOffset));
  __ Ld_d(
      slot_arg,
      MemOperand(handler_arg, Deoptimizer::kEagerWithResumeImmedArgs1PcOffset));
  __ Ld_d(
      handler_arg,
      MemOperand(handler_arg, Deoptimizer::kEagerWithResumeImmedArgs2PcOffset));
  __ Call(builtin_target, RelocInfo::CODE_TARGET);

  Label deopt, bailout;
  __ Branch(&deopt, ne, a0,
            Operand(static_cast<int64_t>(DynamicCheckMapsStatus::kSuccess)));

  __ MaybeRestoreRegisters(registers);
  __ LeaveFrame(StackFrame::INTERNAL);
  __ Ret();

  __ bind(&deopt);
  __ Branch(&bailout, eq, a0,
            Operand(static_cast<int64_t>(DynamicCheckMapsStatus::kBailout)));

  if (FLAG_debug_code) {
    __ Assert(eq, AbortReason::kUnexpectedDynamicCheckMapsStatus, a0,
              Operand(static_cast<int64_t>(DynamicCheckMapsStatus::kDeopt)));
  }
  __ MaybeRestoreRegisters(registers);
  __ LeaveFrame(StackFrame::INTERNAL);
  Handle<Code> deopt_eager = masm->isolate()->builtins()->code_handle(
      Deoptimizer::GetDeoptimizationEntry(DeoptimizeKind::kEager));
  __ Jump(deopt_eager, RelocInfo::CODE_TARGET);

  __ bind(&bailout);
  __ MaybeRestoreRegisters(registers);
  __ LeaveFrame(StackFrame::INTERNAL);
  Handle<Code> deopt_bailout = masm->isolate()->builtins()->code_handle(
      Deoptimizer::GetDeoptimizationEntry(DeoptimizeKind::kBailout));
  __ Jump(deopt_bailout, RelocInfo::CODE_TARGET);
}

#undef __

}  // namespace internal
}  // namespace v8

#endif  // V8_TARGET_ARCH_LOONG64