test-macro-assembler-mips.cc 45.4 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
// Copyright 2013 the V8 project authors. All rights reserved.
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
//     * Redistributions of source code must retain the above copyright
//       notice, this list of conditions and the following disclaimer.
//     * Redistributions in binary form must reproduce the above
//       copyright notice, this list of conditions and the following
//       disclaimer in the documentation and/or other materials provided
//       with the distribution.
//     * Neither the name of Google Inc. nor the names of its
//       contributors may be used to endorse or promote products derived
//       from this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

#include <stdlib.h>
29
#include <iostream>  // NOLINT(readability/streams)
30

31
#include "src/api-inl.h"
32
#include "src/base/utils/random-number-generator.h"
33 34
#include "src/macro-assembler.h"
#include "src/mips/macro-assembler-mips.h"
35
#include "src/objects-inl.h"
36
#include "src/simulator.h"
37 38
#include "src/v8.h"
#include "test/cctest/cctest.h"
39

40 41
namespace v8 {
namespace internal {
42

43 44 45 46
// TODO(mips): Refine these signatures per test case.
using F1 = Object*(int x, int p1, int p2, int p3, int p4);
using F3 = Object*(void* p, int p1, int p2, int p3, int p4);
using F4 = Object*(void* p0, void* p1, int p2, int p3, int p4);
47 48 49

#define __ masm->

50 51 52
TEST(BYTESWAP) {
  CcTest::InitializeVM();
  Isolate* isolate = CcTest::i_isolate();
53
  HandleScope scope(isolate);
54 55

  struct T {
56 57 58
    uint32_t s4;
    uint32_t s2;
    uint32_t u2;
59
  };
60

61
  T t;
62 63
  uint32_t test_values[] = {0x5612FFCD, 0x9D327ACC, 0x781A15C3, 0xFCDE,    0x9F,
                            0xC81A15C3, 0x80000000, 0xFFFFFFFF, 0x00008000};
64

65
  MacroAssembler assembler(isolate, nullptr, 0,
66
                           v8::internal::CodeObjectRequired::kYes);
67

68 69
  MacroAssembler* masm = &assembler;

70 71 72 73 74 75 76 77 78 79 80 81 82 83
  __ lw(a1, MemOperand(a0, offsetof(T, s4)));
  __ nop();
  __ ByteSwapSigned(a1, a1, 4);
  __ sw(a1, MemOperand(a0, offsetof(T, s4)));

  __ lw(a1, MemOperand(a0, offsetof(T, s2)));
  __ nop();
  __ ByteSwapSigned(a1, a1, 2);
  __ sw(a1, MemOperand(a0, offsetof(T, s2)));

  __ lw(a1, MemOperand(a0, offsetof(T, u2)));
  __ nop();
  __ ByteSwapUnsigned(a1, a1, 2);
  __ sw(a1, MemOperand(a0, offsetof(T, u2)));
84 85 86 87 88

  __ jr(ra);
  __ nop();

  CodeDesc desc;
89
  masm->GetCode(isolate, &desc);
90 91
  Handle<Code> code =
      isolate->factory()->NewCode(desc, Code::STUB, Handle<Code>());
92
  auto f = GeneratedCode<F3>::FromCode(*code);
93 94 95 96 97 98 99 100 101 102 103 104 105 106 107

  for (size_t i = 0; i < arraysize(test_values); i++) {
    int16_t in_s2 = static_cast<int16_t>(test_values[i]);
    uint16_t in_u2 = static_cast<uint16_t>(test_values[i]);

    t.s4 = test_values[i];
    t.s2 = static_cast<uint64_t>(in_s2);
    t.u2 = static_cast<uint64_t>(in_u2);

    f.Call(&t, 0, 0, 0, 0);

    CHECK_EQ(ByteReverse(test_values[i]), t.s4);
    CHECK_EQ(ByteReverse<int16_t>(in_s2), static_cast<int16_t>(t.s2));
    CHECK_EQ(ByteReverse<uint16_t>(in_u2), static_cast<uint16_t>(t.u2));
  }
108
}
109

110 111 112 113 114 115 116 117
static void TestNaN(const char *code) {
  // NaN value is different on MIPS and x86 architectures, and TEST(NaNx)
  // tests checks the case where a x86 NaN value is serialized into the
  // snapshot on the simulator during cross compilation.
  v8::HandleScope scope(CcTest::isolate());
  v8::Local<v8::Context> context = CcTest::NewContext(PRINT_EXTENSION);
  v8::Context::Scope context_scope(context);

118 119 120 121
  v8::Local<v8::Script> script =
      v8::Script::Compile(context, v8_str(code)).ToLocalChecked();
  v8::Local<v8::Object> result =
      v8::Local<v8::Object>::Cast(script->Run(context).ToLocalChecked());
122
  i::Handle<i::JSReceiver> o = v8::Utils::OpenHandle(*result);
123 124
  i::Handle<i::JSArray> array1(reinterpret_cast<i::JSArray*>(*o),
                               o->GetIsolate());
125 126 127
  i::FixedDoubleArray* a = i::FixedDoubleArray::cast(array1->elements());
  double value = a->get_scalar(0);
  CHECK(std::isnan(value) &&
128
        bit_cast<uint64_t>(value) ==
129
            bit_cast<uint64_t>(std::numeric_limits<double>::quiet_NaN()));
130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151
}


TEST(NaN0) {
  TestNaN(
          "var result;"
          "for (var i = 0; i < 2; i++) {"
          "  result = new Array(Number.NaN, Number.POSITIVE_INFINITY);"
          "}"
          "result;");
}


TEST(NaN1) {
  TestNaN(
          "var result;"
          "for (var i = 0; i < 2; i++) {"
          "  result = [NaN];"
          "}"
          "result;");
}

152

153 154 155 156 157 158 159 160
TEST(jump_tables4) {
  // Similar to test-assembler-mips jump_tables1, with extra test for branch
  // trampoline required before emission of the dd table (where trampolines are
  // blocked), and proper transition to long-branch mode.
  // Regression test for v8:4294.
  CcTest::InitializeVM();
  Isolate* isolate = CcTest::i_isolate();
  HandleScope scope(isolate);
161
  MacroAssembler assembler(isolate, nullptr, 0,
162
                           v8::internal::CodeObjectRequired::kYes);
163 164 165 166 167 168
  MacroAssembler* masm = &assembler;

  const int kNumCases = 512;
  int values[kNumCases];
  isolate->random_number_generator()->NextBytes(values, sizeof(values));
  Label labels[kNumCases];
169
  Label near_start, end, done;
170

171
  __ Push(ra);
172 173 174 175 176 177 178 179 180 181 182
  __ mov(v0, zero_reg);

  __ Branch(&end);
  __ bind(&near_start);

  // Generate slightly less than 32K instructions, which will soon require
  // trampoline for branch distance fixup.
  for (int i = 0; i < 32768 - 256; ++i) {
    __ addiu(v0, v0, 1);
  }

183 184
  __ GenerateSwitchTable(a0, kNumCases,
                         [&labels](size_t i) { return labels + i; });
185 186 187

  for (int i = 0; i < kNumCases; ++i) {
    __ bind(&labels[i]);
188
    __ li(v0, values[i]);
189 190 191 192
    __ Branch(&done);
  }

  __ bind(&done);
193
  __ Pop(ra);
194 195 196 197 198 199 200
  __ jr(ra);
  __ nop();

  __ bind(&end);
  __ Branch(&near_start);

  CodeDesc desc;
201
  masm->GetCode(isolate, &desc);
202 203
  Handle<Code> code =
      isolate->factory()->NewCode(desc, Code::STUB, Handle<Code>());
204
#ifdef OBJECT_PRINT
205
  code->Print(std::cout);
206
#endif
207
  auto f = GeneratedCode<F1>::FromCode(*code);
208
  for (int i = 0; i < kNumCases; ++i) {
209
    int res = reinterpret_cast<int>(f.Call(i, 0, 0, 0, 0));
210 211 212 213 214 215
    ::printf("f(%d) = %d\n", i, res);
    CHECK_EQ(values[i], res);
  }
}


216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233
TEST(jump_tables5) {
  if (!IsMipsArchVariant(kMips32r6)) return;

  // Similar to test-assembler-mips jump_tables1, with extra test for emitting a
  // compact branch instruction before emission of the dd table.
  CcTest::InitializeVM();
  Isolate* isolate = CcTest::i_isolate();
  HandleScope scope(isolate);
  MacroAssembler assembler(isolate, nullptr, 0,
                           v8::internal::CodeObjectRequired::kYes);
  MacroAssembler* masm = &assembler;

  const int kNumCases = 512;
  int values[kNumCases];
  isolate->random_number_generator()->NextBytes(values, sizeof(values));
  Label labels[kNumCases];
  Label done;

234
  __ Push(ra);
235 236

  {
237
    __ BlockTrampolinePoolFor(kNumCases + 6 + 1);
238
    PredictableCodeSizeScope predictable(
239
        masm, kNumCases * kPointerSize + ((6 + 1) * kInstrSize));
240 241

    __ addiupc(at, 6 + 1);
242
    __ Lsa(at, at, a0, 2);
243
    __ lw(at, MemOperand(at));
244 245 246
    __ jalr(at);
    __ nop();  // Branch delay slot nop.
    __ bc(&done);
247 248
    // A nop instruction must be generated by the forbidden slot guard
    // (Assembler::dd(Label*)).
249 250 251 252 253 254 255
    for (int i = 0; i < kNumCases; ++i) {
      __ dd(&labels[i]);
    }
  }

  for (int i = 0; i < kNumCases; ++i) {
    __ bind(&labels[i]);
256
    __ li(v0, values[i]);
257 258 259 260 261
    __ jr(ra);
    __ nop();
  }

  __ bind(&done);
262
  __ Pop(ra);
263 264 265 266
  __ jr(ra);
  __ nop();

  CodeDesc desc;
267
  masm->GetCode(isolate, &desc);
268 269
  Handle<Code> code =
      isolate->factory()->NewCode(desc, Code::STUB, Handle<Code>());
270
#ifdef OBJECT_PRINT
271
  code->Print(std::cout);
272
#endif
273
  auto f = GeneratedCode<F1>::FromCode(*code);
274
  for (int i = 0; i < kNumCases; ++i) {
275
    int32_t res = reinterpret_cast<int32_t>(f.Call(i, 0, 0, 0, 0));
276
    ::printf("f(%d) = %d\n", i, res);
277 278 279 280
    CHECK_EQ(values[i], res);
  }
}

281 282 283 284 285 286 287 288 289 290 291 292 293 294
TEST(jump_tables6) {
  // Similar to test-assembler-mips jump_tables1, with extra test for branch
  // trampoline required after emission of the dd table (where trampolines are
  // blocked). This test checks if number of really generated instructions is
  // greater than number of counted instructions from code, as we are expecting
  // generation of trampoline in this case (when number of kFillInstr
  // instructions is close to 32K)
  CcTest::InitializeVM();
  Isolate* isolate = CcTest::i_isolate();
  HandleScope scope(isolate);
  MacroAssembler assembler(isolate, nullptr, 0,
                           v8::internal::CodeObjectRequired::kYes);
  MacroAssembler* masm = &assembler;

295 296 297 298 299 300
  const int kSwitchTableCases = 40;

  const int kMaxBranchOffset = Assembler::kMaxBranchOffset;
  const int kTrampolineSlotsSize = Assembler::kTrampolineSlotsSize;
  const int kSwitchTablePrologueSize = MacroAssembler::kSwitchTablePrologueSize;

301 302
  const int kMaxOffsetForTrampolineStart =
      kMaxBranchOffset - 16 * kTrampolineSlotsSize;
303 304
  const int kFillInstr = (kMaxOffsetForTrampolineStart / kInstrSize) -
                         (kSwitchTablePrologueSize + kSwitchTableCases) - 20;
305

306
  int values[kSwitchTableCases];
307
  isolate->random_number_generator()->NextBytes(values, sizeof(values));
308
  Label labels[kSwitchTableCases];
309 310 311 312 313 314 315 316 317
  Label near_start, end, done;

  __ Push(ra);
  __ mov(v0, zero_reg);

  int offs1 = masm->pc_offset();
  int gen_insn = 0;

  __ Branch(&end);
318
  gen_insn += Assembler::IsCompactBranchSupported() ? 1 : 2;
319 320 321 322 323 324 325 326 327
  __ bind(&near_start);

  // Generate slightly less than 32K instructions, which will soon require
  // trampoline for branch distance fixup.
  for (int i = 0; i < kFillInstr; ++i) {
    __ addiu(v0, v0, 1);
  }
  gen_insn += kFillInstr;

328
  __ GenerateSwitchTable(a0, kSwitchTableCases,
329
                         [&labels](size_t i) { return labels + i; });
330
  gen_insn += (kSwitchTablePrologueSize + kSwitchTableCases);
331

332
  for (int i = 0; i < kSwitchTableCases; ++i) {
333 334 335 336
    __ bind(&labels[i]);
    __ li(v0, values[i]);
    __ Branch(&done);
  }
337 338
  gen_insn +=
      ((Assembler::IsCompactBranchSupported() ? 3 : 4) * kSwitchTableCases);
339 340 341 342 343 344

  // If offset from here to first branch instr is greater than max allowed
  // offset for trampoline ...
  CHECK_LT(kMaxOffsetForTrampolineStart, masm->pc_offset() - offs1);
  // ... number of generated instructions must be greater then "gen_insn",
  // as we are expecting trampoline generation
345
  CHECK_LT(gen_insn, (masm->pc_offset() - offs1) / kInstrSize);
346 347 348 349 350 351 352 353 354 355

  __ bind(&done);
  __ Pop(ra);
  __ jr(ra);
  __ nop();

  __ bind(&end);
  __ Branch(&near_start);

  CodeDesc desc;
356
  masm->GetCode(isolate, &desc);
357 358
  Handle<Code> code =
      isolate->factory()->NewCode(desc, Code::STUB, Handle<Code>());
359
#ifdef OBJECT_PRINT
360
  code->Print(std::cout);
361
#endif
362
  auto f = GeneratedCode<F1>::FromCode(*code);
363
  for (int i = 0; i < kSwitchTableCases; ++i) {
364
    int res = reinterpret_cast<int>(f.Call(i, 0, 0, 0, 0));
365 366 367 368
    ::printf("f(%d) = %d\n", i, res);
    CHECK_EQ(values[i], res);
  }
}
369

370 371 372 373 374 375 376 377 378 379 380 381
static uint32_t run_lsa(uint32_t rt, uint32_t rs, int8_t sa) {
  Isolate* isolate = CcTest::i_isolate();
  HandleScope scope(isolate);
  MacroAssembler assembler(isolate, nullptr, 0,
                           v8::internal::CodeObjectRequired::kYes);
  MacroAssembler* masm = &assembler;

  __ Lsa(v0, a0, a1, sa);
  __ jr(ra);
  __ nop();

  CodeDesc desc;
382
  assembler.GetCode(isolate, &desc);
383 384
  Handle<Code> code =
      isolate->factory()->NewCode(desc, Code::STUB, Handle<Code>());
385

386
  auto f = GeneratedCode<F1>::FromCode(*code);
387

388
  uint32_t res = reinterpret_cast<uint32_t>(f.Call(rt, rs, 0, 0, 0));
389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405

  return res;
}


TEST(Lsa) {
  CcTest::InitializeVM();
  struct TestCaseLsa {
    int32_t rt;
    int32_t rs;
    uint8_t sa;
    uint32_t expected_res;
  };

  struct TestCaseLsa tc[] = {// rt, rs, sa, expected_res
                             {0x4, 0x1, 1, 0x6},
                             {0x4, 0x1, 2, 0x8},
406
                             {0x4, 0x1, 3, 0xC},
407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422
                             {0x4, 0x1, 4, 0x14},
                             {0x4, 0x1, 5, 0x24},
                             {0x0, 0x1, 1, 0x2},
                             {0x0, 0x1, 2, 0x4},
                             {0x0, 0x1, 3, 0x8},
                             {0x0, 0x1, 4, 0x10},
                             {0x0, 0x1, 5, 0x20},
                             {0x4, 0x0, 1, 0x4},
                             {0x4, 0x0, 2, 0x4},
                             {0x4, 0x0, 3, 0x4},
                             {0x4, 0x0, 4, 0x4},
                             {0x4, 0x0, 5, 0x4},

                             // Shift overflow.
                             {0x4, INT32_MAX, 1, 0x2},
                             {0x4, INT32_MAX >> 1, 2, 0x0},
423 424 425
                             {0x4, INT32_MAX >> 2, 3, 0xFFFFFFFC},
                             {0x4, INT32_MAX >> 3, 4, 0xFFFFFFF4},
                             {0x4, INT32_MAX >> 4, 5, 0xFFFFFFE4},
426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449

                             // Signed addition overflow.
                             {INT32_MAX - 1, 0x1, 1, 0x80000000},
                             {INT32_MAX - 3, 0x1, 2, 0x80000000},
                             {INT32_MAX - 7, 0x1, 3, 0x80000000},
                             {INT32_MAX - 15, 0x1, 4, 0x80000000},
                             {INT32_MAX - 31, 0x1, 5, 0x80000000},

                             // Addition overflow.
                             {-2, 0x1, 1, 0x0},
                             {-4, 0x1, 2, 0x0},
                             {-8, 0x1, 3, 0x0},
                             {-16, 0x1, 4, 0x0},
                             {-32, 0x1, 5, 0x0}};

  size_t nr_test_cases = sizeof(tc) / sizeof(TestCaseLsa);
  for (size_t i = 0; i < nr_test_cases; ++i) {
    uint32_t res = run_lsa(tc[i].rt, tc[i].rs, tc[i].sa);
    PrintF("0x%x =? 0x%x == lsa(v0, %x, %x, %hhu)\n", tc[i].expected_res, res,
           tc[i].rt, tc[i].rs, tc[i].sa);
    CHECK_EQ(tc[i].expected_res, res);
  }
}

450
static const std::vector<uint32_t> cvt_trunc_uint32_test_values() {
451 452 453
  static const uint32_t kValues[] = {0x00000000, 0x00000001, 0x00FFFF00,
                                     0x7FFFFFFF, 0x80000000, 0x80000001,
                                     0x80FFFF00, 0x8FFFFFFF, 0xFFFFFFFF};
454 455 456
  return std::vector<uint32_t>(&kValues[0], &kValues[arraysize(kValues)]);
}

457
static const std::vector<int32_t> cvt_trunc_int32_test_values() {
458 459
  static const int32_t kValues[] = {
      static_cast<int32_t>(0x00000000), static_cast<int32_t>(0x00000001),
460
      static_cast<int32_t>(0x00FFFF00), static_cast<int32_t>(0x7FFFFFFF),
461
      static_cast<int32_t>(0x80000000), static_cast<int32_t>(0x80000001),
462 463
      static_cast<int32_t>(0x80FFFF00), static_cast<int32_t>(0x8FFFFFFF),
      static_cast<int32_t>(0xFFFFFFFF)};
464 465 466 467
  return std::vector<int32_t>(&kValues[0], &kValues[arraysize(kValues)]);
}

// Helper macros that can be used in FOR_INT32_INPUTS(i) { ... *i ... }
468 469
#define FOR_INPUTS(ctype, itype, var, test_vector)           \
  std::vector<ctype> var##_vec = test_vector();              \
470 471 472
  for (std::vector<ctype>::iterator var = var##_vec.begin(); \
       var != var##_vec.end(); ++var)

473 474 475 476 477 478 479
#define FOR_INPUTS2(ctype, itype, var, var2, test_vector)  \
  std::vector<ctype> var##_vec = test_vector();            \
  std::vector<ctype>::iterator var;                        \
  std::vector<ctype>::reverse_iterator var2;               \
  for (var = var##_vec.begin(), var2 = var##_vec.rbegin(); \
       var != var##_vec.end(); ++var, ++var2)

480 481 482 483 484 485 486 487
#define FOR_ENUM_INPUTS(var, type, test_vector) \
  FOR_INPUTS(enum type, type, var, test_vector)
#define FOR_STRUCT_INPUTS(var, type, test_vector) \
  FOR_INPUTS(struct type, type, var, test_vector)
#define FOR_UINT32_INPUTS(var, test_vector) \
  FOR_INPUTS(uint32_t, uint32, var, test_vector)
#define FOR_INT32_INPUTS(var, test_vector) \
  FOR_INPUTS(int32_t, int32, var, test_vector)
488 489 490 491 492
#define FOR_INT32_INPUTS2(var, var2, test_vector) \
  FOR_INPUTS2(int32_t, int32, var, var2, test_vector)

#define FOR_UINT64_INPUTS(var, test_vector) \
  FOR_INPUTS(uint64_t, uint32, var, test_vector)
493 494 495

template <typename RET_TYPE, typename IN_TYPE, typename Func>
RET_TYPE run_Cvt(IN_TYPE x, Func GenerateConvertInstructionFunc) {
496
  typedef RET_TYPE(F_CVT)(IN_TYPE x0, int x1, int x2, int x3, int x4);
497 498 499 500 501 502 503 504 505 506 507 508 509 510

  Isolate* isolate = CcTest::i_isolate();
  HandleScope scope(isolate);
  MacroAssembler assm(isolate, nullptr, 0,
                      v8::internal::CodeObjectRequired::kYes);
  MacroAssembler* masm = &assm;

  __ mtc1(a0, f4);
  GenerateConvertInstructionFunc(masm);
  __ mfc1(v0, f2);
  __ jr(ra);
  __ nop();

  CodeDesc desc;
511
  assm.GetCode(isolate, &desc);
512 513
  Handle<Code> code =
      isolate->factory()->NewCode(desc, Code::STUB, Handle<Code>());
514

515
  auto f = GeneratedCode<F_CVT>::FromCode(*code);
516

517
  return reinterpret_cast<RET_TYPE>(f.Call(x, 0, 0, 0, 0));
518 519 520 521
}

TEST(cvt_s_w_Trunc_uw_s) {
  CcTest::InitializeVM();
522
  FOR_UINT32_INPUTS(i, cvt_trunc_uint32_test_values) {
523
    uint32_t input = *i;
524 525
    auto fn = [](MacroAssembler* masm) {
      __ cvt_s_w(f0, f4);
526
      __ Trunc_uw_s(f2, f0, f6);
527 528
    };
    CHECK_EQ(static_cast<float>(input), run_Cvt<uint32_t>(input, fn));
529 530 531 532 533
  }
}

TEST(cvt_d_w_Trunc_w_d) {
  CcTest::InitializeVM();
534
  FOR_INT32_INPUTS(i, cvt_trunc_int32_test_values) {
535
    int32_t input = *i;
536 537 538 539 540
    auto fn = [](MacroAssembler* masm) {
      __ cvt_d_w(f0, f4);
      __ Trunc_w_d(f2, f0);
    };
    CHECK_EQ(static_cast<double>(input), run_Cvt<int32_t>(input, fn));
541 542 543
  }
}

544 545
static const std::vector<int32_t> overflow_int32_test_values() {
  static const int32_t kValues[] = {
546 547 548 549 550
      static_cast<int32_t>(0xF0000000), static_cast<int32_t>(0x00000001),
      static_cast<int32_t>(0xFF000000), static_cast<int32_t>(0x0000F000),
      static_cast<int32_t>(0x0F000000), static_cast<int32_t>(0x991234AB),
      static_cast<int32_t>(0xB0FFFF01), static_cast<int32_t>(0x00006FFF),
      static_cast<int32_t>(0xFFFFFFFF)};
551 552 553
  return std::vector<int32_t>(&kValues[0], &kValues[arraysize(kValues)]);
}

554 555
TEST(OverflowInstructions) {
  CcTest::InitializeVM();
556
  Isolate* isolate = CcTest::i_isolate();
557
  HandleScope handles(isolate);
558

559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575
  struct T {
    int32_t lhs;
    int32_t rhs;
    int32_t output_add;
    int32_t output_add2;
    int32_t output_sub;
    int32_t output_sub2;
    int32_t output_mul;
    int32_t output_mul2;
    int32_t overflow_add;
    int32_t overflow_add2;
    int32_t overflow_sub;
    int32_t overflow_sub2;
    int32_t overflow_mul;
    int32_t overflow_mul2;
  };
  T t;
576 577 578

  FOR_INT32_INPUTS(i, overflow_int32_test_values) {
    FOR_INT32_INPUTS(j, overflow_int32_test_values) {
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
      int32_t ii = *i;
      int32_t jj = *j;
      int32_t expected_add, expected_sub, expected_mul;
      bool expected_add_ovf, expected_sub_ovf, expected_mul_ovf;
      MacroAssembler assembler(isolate, nullptr, 0,
                               v8::internal::CodeObjectRequired::kYes);
      MacroAssembler* masm = &assembler;

      __ lw(t0, MemOperand(a0, offsetof(T, lhs)));
      __ lw(t1, MemOperand(a0, offsetof(T, rhs)));

      __ AddOverflow(t2, t0, Operand(t1), t3);
      __ sw(t2, MemOperand(a0, offsetof(T, output_add)));
      __ sw(t3, MemOperand(a0, offsetof(T, overflow_add)));
      __ mov(t3, zero_reg);
      __ AddOverflow(t0, t0, Operand(t1), t3);
      __ sw(t0, MemOperand(a0, offsetof(T, output_add2)));
      __ sw(t3, MemOperand(a0, offsetof(T, overflow_add2)));

      __ lw(t0, MemOperand(a0, offsetof(T, lhs)));
      __ lw(t1, MemOperand(a0, offsetof(T, rhs)));

      __ SubOverflow(t2, t0, Operand(t1), t3);
      __ sw(t2, MemOperand(a0, offsetof(T, output_sub)));
      __ sw(t3, MemOperand(a0, offsetof(T, overflow_sub)));
      __ mov(t3, zero_reg);
      __ SubOverflow(t0, t0, Operand(t1), t3);
      __ sw(t0, MemOperand(a0, offsetof(T, output_sub2)));
      __ sw(t3, MemOperand(a0, offsetof(T, overflow_sub2)));

      __ lw(t0, MemOperand(a0, offsetof(T, lhs)));
      __ lw(t1, MemOperand(a0, offsetof(T, rhs)));

      __ MulOverflow(t2, t0, Operand(t1), t3);
      __ sw(t2, MemOperand(a0, offsetof(T, output_mul)));
      __ sw(t3, MemOperand(a0, offsetof(T, overflow_mul)));
      __ mov(t3, zero_reg);
      __ MulOverflow(t0, t0, Operand(t1), t3);
      __ sw(t0, MemOperand(a0, offsetof(T, output_mul2)));
      __ sw(t3, MemOperand(a0, offsetof(T, overflow_mul2)));

      __ jr(ra);
      __ nop();

      CodeDesc desc;
      masm->GetCode(isolate, &desc);
      Handle<Code> code =
          isolate->factory()->NewCode(desc, Code::STUB, Handle<Code>());
      auto f = GeneratedCode<F3>::FromCode(*code);
      t.lhs = ii;
      t.rhs = jj;
      f.Call(&t, 0, 0, 0, 0);

      expected_add_ovf = base::bits::SignedAddOverflow32(ii, jj, &expected_add);
      expected_sub_ovf = base::bits::SignedSubOverflow32(ii, jj, &expected_sub);
      expected_mul_ovf = base::bits::SignedMulOverflow32(ii, jj, &expected_mul);

      CHECK_EQ(expected_add_ovf, t.overflow_add < 0);
      CHECK_EQ(expected_sub_ovf, t.overflow_sub < 0);
      CHECK_EQ(expected_mul_ovf, t.overflow_mul != 0);

      CHECK_EQ(t.overflow_add, t.overflow_add2);
      CHECK_EQ(t.overflow_sub, t.overflow_sub2);
      CHECK_EQ(t.overflow_mul, t.overflow_mul2);

      CHECK_EQ(expected_add, t.output_add);
      CHECK_EQ(expected_add, t.output_add2);
      CHECK_EQ(expected_sub, t.output_sub);
      CHECK_EQ(expected_sub, t.output_sub2);
      if (!expected_mul_ovf) {
        CHECK_EQ(expected_mul, t.output_mul);
        CHECK_EQ(expected_mul, t.output_mul2);
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
TEST(min_max_nan) {
  CcTest::InitializeVM();
  Isolate* isolate = CcTest::i_isolate();
  HandleScope scope(isolate);
  MacroAssembler assembler(isolate, nullptr, 0,
                           v8::internal::CodeObjectRequired::kYes);
  MacroAssembler* masm = &assembler;

  struct TestFloat {
    double a;
    double b;
    double c;
    double d;
    float e;
    float f;
    float g;
    float h;
  };

  TestFloat test;
  const double dnan = std::numeric_limits<double>::quiet_NaN();
  const double dinf = std::numeric_limits<double>::infinity();
  const double dminf = -std::numeric_limits<double>::infinity();
  const float fnan = std::numeric_limits<float>::quiet_NaN();
  const float finf = std::numeric_limits<float>::infinity();
  const float fminf = std::numeric_limits<float>::infinity();
  const int kTableLength = 13;

  double inputsa[kTableLength] = {2.0,  3.0,  -0.0, 0.0,  42.0, dinf, dminf,
                                  dinf, dnan, 3.0,  dinf, dnan, dnan};
  double inputsb[kTableLength] = {3.0,   2.0, 0.0,  -0.0, dinf, 42.0, dinf,
                                  dminf, 3.0, dnan, dnan, dinf, dnan};
  double outputsdmin[kTableLength] = {2.0,  2.0,   -0.0,  -0.0, 42.0,
                                      42.0, dminf, dminf, dnan, dnan,
                                      dnan, dnan,  dnan};
  double outputsdmax[kTableLength] = {3.0,  3.0,  0.0,  0.0,  dinf, dinf, dinf,
                                      dinf, dnan, dnan, dnan, dnan, dnan};

  float inputse[kTableLength] = {2.0,  3.0,  -0.0, 0.0,  42.0, finf, fminf,
                                 finf, fnan, 3.0,  finf, fnan, fnan};
  float inputsf[kTableLength] = {3.0,   2.0, 0.0,  -0.0, finf, 42.0, finf,
                                 fminf, 3.0, fnan, fnan, finf, fnan};
  float outputsfmin[kTableLength] = {2.0,   2.0,  -0.0, -0.0, 42.0, 42.0, fminf,
                                     fminf, fnan, fnan, fnan, fnan, fnan};
  float outputsfmax[kTableLength] = {3.0,  3.0,  0.0,  0.0,  finf, finf, finf,
                                     finf, fnan, fnan, fnan, fnan, fnan};

  auto handle_dnan = [masm](FPURegister dst, Label* nan, Label* back) {
    __ bind(nan);
706 707
    __ LoadRoot(t8, Heap::kNanValueRootIndex);
    __ Ldc1(dst, FieldMemOperand(t8, HeapNumber::kValueOffset));
708 709 710 711 712 713 714 715 716 717 718 719 720 721
    __ Branch(back);
  };

  auto handle_snan = [masm, fnan](FPURegister dst, Label* nan, Label* back) {
    __ bind(nan);
    __ Move(dst, fnan);
    __ Branch(back);
  };

  Label handle_mind_nan, handle_maxd_nan, handle_mins_nan, handle_maxs_nan;
  Label back_mind_nan, back_maxd_nan, back_mins_nan, back_maxs_nan;

  __ push(s6);
  __ InitializeRootRegister();
722 723
  __ Ldc1(f4, MemOperand(a0, offsetof(TestFloat, a)));
  __ Ldc1(f8, MemOperand(a0, offsetof(TestFloat, b)));
724 725
  __ lwc1(f2, MemOperand(a0, offsetof(TestFloat, e)));
  __ lwc1(f6, MemOperand(a0, offsetof(TestFloat, f)));
726
  __ Float64Min(f10, f4, f8, &handle_mind_nan);
727
  __ bind(&back_mind_nan);
728
  __ Float64Max(f12, f4, f8, &handle_maxd_nan);
729
  __ bind(&back_maxd_nan);
730
  __ Float32Min(f14, f2, f6, &handle_mins_nan);
731
  __ bind(&back_mins_nan);
732
  __ Float32Max(f16, f2, f6, &handle_maxs_nan);
733
  __ bind(&back_maxs_nan);
734 735
  __ Sdc1(f10, MemOperand(a0, offsetof(TestFloat, c)));
  __ Sdc1(f12, MemOperand(a0, offsetof(TestFloat, d)));
736 737 738 739 740 741 742 743 744 745 746 747
  __ swc1(f14, MemOperand(a0, offsetof(TestFloat, g)));
  __ swc1(f16, MemOperand(a0, offsetof(TestFloat, h)));
  __ pop(s6);
  __ jr(ra);
  __ nop();

  handle_dnan(f10, &handle_mind_nan, &back_mind_nan);
  handle_dnan(f12, &handle_maxd_nan, &back_maxd_nan);
  handle_snan(f14, &handle_mins_nan, &back_mins_nan);
  handle_snan(f16, &handle_maxs_nan, &back_maxs_nan);

  CodeDesc desc;
748
  masm->GetCode(isolate, &desc);
749 750
  Handle<Code> code =
      isolate->factory()->NewCode(desc, Code::STUB, Handle<Code>());
751
  auto f = GeneratedCode<F3>::FromCode(*code);
752 753 754 755 756 757
  for (int i = 0; i < kTableLength; i++) {
    test.a = inputsa[i];
    test.b = inputsb[i];
    test.e = inputse[i];
    test.f = inputsf[i];

758
    f.Call(&test, 0, 0, 0, 0);
759 760 761 762 763 764 765 766

    CHECK_EQ(0, memcmp(&test.c, &outputsdmin[i], sizeof(test.c)));
    CHECK_EQ(0, memcmp(&test.d, &outputsdmax[i], sizeof(test.d)));
    CHECK_EQ(0, memcmp(&test.g, &outputsfmin[i], sizeof(test.g)));
    CHECK_EQ(0, memcmp(&test.h, &outputsfmax[i], sizeof(test.h)));
  }
}

767 768 769
template <typename IN_TYPE, typename Func>
bool run_Unaligned(char* memory_buffer, int32_t in_offset, int32_t out_offset,
                   IN_TYPE value, Func GenerateUnalignedInstructionFunc) {
770
  typedef int32_t(F_CVT)(char* x0, int x1, int x2, int x3, int x4);
771 772 773 774 775 776 777 778 779 780 781 782 783

  Isolate* isolate = CcTest::i_isolate();
  HandleScope scope(isolate);
  MacroAssembler assm(isolate, nullptr, 0,
                      v8::internal::CodeObjectRequired::kYes);
  MacroAssembler* masm = &assm;
  IN_TYPE res;

  GenerateUnalignedInstructionFunc(masm, in_offset, out_offset);
  __ jr(ra);
  __ nop();

  CodeDesc desc;
784
  assm.GetCode(isolate, &desc);
785 786
  Handle<Code> code =
      isolate->factory()->NewCode(desc, Code::STUB, Handle<Code>());
787

788
  auto f = GeneratedCode<F_CVT>::FromCode(*code);
789 790

  MemCopy(memory_buffer + in_offset, &value, sizeof(IN_TYPE));
791
  f.Call(memory_buffer, 0, 0, 0, 0);
792 793 794 795 796 797 798
  MemCopy(&res, memory_buffer + out_offset, sizeof(IN_TYPE));

  return res == value;
}

static const std::vector<uint64_t> unsigned_test_values() {
  static const uint64_t kValues[] = {
799 800
      0x2180F18A06384414, 0x000A714532102277, 0xBC1ACCCF180649F0,
      0x8000000080008000, 0x0000000000000001, 0xFFFFFFFFFFFFFFFF,
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
  };
  return std::vector<uint64_t>(&kValues[0], &kValues[arraysize(kValues)]);
}

static const std::vector<int32_t> unsigned_test_offset() {
  static const int32_t kValues[] = {// value, offset
                                    -132 * KB, -21 * KB, 0, 19 * KB, 135 * KB};
  return std::vector<int32_t>(&kValues[0], &kValues[arraysize(kValues)]);
}

static const std::vector<int32_t> unsigned_test_offset_increment() {
  static const int32_t kValues[] = {-5, -4, -3, -2, -1, 0, 1, 2, 3, 4, 5};
  return std::vector<int32_t>(&kValues[0], &kValues[arraysize(kValues)]);
}

TEST(Ulh) {
  CcTest::InitializeVM();

  static const int kBufferSize = 300 * KB;
  char memory_buffer[kBufferSize];
  char* buffer_middle = memory_buffer + (kBufferSize / 2);

  FOR_UINT64_INPUTS(i, unsigned_test_values) {
    FOR_INT32_INPUTS2(j1, j2, unsigned_test_offset) {
      FOR_INT32_INPUTS2(k1, k2, unsigned_test_offset_increment) {
        uint16_t value = static_cast<uint64_t>(*i & 0xFFFF);
        int32_t in_offset = *j1 + *k1;
        int32_t out_offset = *j2 + *k2;

830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862
        auto fn_1 = [](MacroAssembler* masm, int32_t in_offset,
                       int32_t out_offset) {
          __ Ulh(v0, MemOperand(a0, in_offset));
          __ Ush(v0, MemOperand(a0, out_offset), v0);
        };
        CHECK_EQ(true, run_Unaligned<uint16_t>(buffer_middle, in_offset,
                                               out_offset, value, fn_1));

        auto fn_2 = [](MacroAssembler* masm, int32_t in_offset,
                       int32_t out_offset) {
          __ mov(t0, a0);
          __ Ulh(a0, MemOperand(a0, in_offset));
          __ Ush(a0, MemOperand(t0, out_offset), v0);
        };
        CHECK_EQ(true, run_Unaligned<uint16_t>(buffer_middle, in_offset,
                                               out_offset, value, fn_2));

        auto fn_3 = [](MacroAssembler* masm, int32_t in_offset,
                       int32_t out_offset) {
          __ mov(t0, a0);
          __ Ulhu(a0, MemOperand(a0, in_offset));
          __ Ush(a0, MemOperand(t0, out_offset), t1);
        };
        CHECK_EQ(true, run_Unaligned<uint16_t>(buffer_middle, in_offset,
                                               out_offset, value, fn_3));

        auto fn_4 = [](MacroAssembler* masm, int32_t in_offset,
                       int32_t out_offset) {
          __ Ulhu(v0, MemOperand(a0, in_offset));
          __ Ush(v0, MemOperand(a0, out_offset), t1);
        };
        CHECK_EQ(true, run_Unaligned<uint16_t>(buffer_middle, in_offset,
                                               out_offset, value, fn_4));
863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881
      }
    }
  }
}

TEST(Ulh_bitextension) {
  CcTest::InitializeVM();

  static const int kBufferSize = 300 * KB;
  char memory_buffer[kBufferSize];
  char* buffer_middle = memory_buffer + (kBufferSize / 2);

  FOR_UINT64_INPUTS(i, unsigned_test_values) {
    FOR_INT32_INPUTS2(j1, j2, unsigned_test_offset) {
      FOR_INT32_INPUTS2(k1, k2, unsigned_test_offset_increment) {
        uint16_t value = static_cast<uint64_t>(*i & 0xFFFF);
        int32_t in_offset = *j1 + *k1;
        int32_t out_offset = *j2 + *k2;

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
        auto fn = [](MacroAssembler* masm, int32_t in_offset,
                     int32_t out_offset) {
          Label success, fail, end, different;
          __ Ulh(t0, MemOperand(a0, in_offset));
          __ Ulhu(t1, MemOperand(a0, in_offset));
          __ Branch(&different, ne, t0, Operand(t1));

          // If signed and unsigned values are same, check
          // the upper bits to see if they are zero
          __ sra(t0, t0, 15);
          __ Branch(&success, eq, t0, Operand(zero_reg));
          __ Branch(&fail);

          // If signed and unsigned values are different,
          // check that the upper bits are complementary
          __ bind(&different);
          __ sra(t1, t1, 15);
          __ Branch(&fail, ne, t1, Operand(1));
          __ sra(t0, t0, 15);
          __ addiu(t0, t0, 1);
          __ Branch(&fail, ne, t0, Operand(zero_reg));
          // Fall through to success

          __ bind(&success);
          __ Ulh(t0, MemOperand(a0, in_offset));
          __ Ush(t0, MemOperand(a0, out_offset), v0);
          __ Branch(&end);
          __ bind(&fail);
          __ Ush(zero_reg, MemOperand(a0, out_offset), v0);
          __ bind(&end);
        };
        CHECK_EQ(true, run_Unaligned<uint16_t>(buffer_middle, in_offset,
                                               out_offset, value, fn));
915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933
      }
    }
  }
}

TEST(Ulw) {
  CcTest::InitializeVM();

  static const int kBufferSize = 300 * KB;
  char memory_buffer[kBufferSize];
  char* buffer_middle = memory_buffer + (kBufferSize / 2);

  FOR_UINT64_INPUTS(i, unsigned_test_values) {
    FOR_INT32_INPUTS2(j1, j2, unsigned_test_offset) {
      FOR_INT32_INPUTS2(k1, k2, unsigned_test_offset_increment) {
        uint32_t value = static_cast<uint32_t>(*i & 0xFFFFFFFF);
        int32_t in_offset = *j1 + *k1;
        int32_t out_offset = *j2 + *k2;

934 935 936 937 938 939 940 941 942 943 944 945 946 947
        auto fn_1 = [](MacroAssembler* masm, int32_t in_offset,
                       int32_t out_offset) {
          __ Ulw(v0, MemOperand(a0, in_offset));
          __ Usw(v0, MemOperand(a0, out_offset));
        };
        CHECK_EQ(true, run_Unaligned<uint32_t>(buffer_middle, in_offset,
                                               out_offset, value, fn_1));

        auto fn_2 = [](MacroAssembler* masm, int32_t in_offset,
                       int32_t out_offset) {
          __ mov(t0, a0);
          __ Ulw(a0, MemOperand(a0, in_offset));
          __ Usw(a0, MemOperand(t0, out_offset));
        };
948
        CHECK_EQ(true,
949 950
                 run_Unaligned<uint32_t>(buffer_middle, in_offset, out_offset,
                                         (uint32_t)value, fn_2));
951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969
      }
    }
  }
}

TEST(Ulwc1) {
  CcTest::InitializeVM();

  static const int kBufferSize = 300 * KB;
  char memory_buffer[kBufferSize];
  char* buffer_middle = memory_buffer + (kBufferSize / 2);

  FOR_UINT64_INPUTS(i, unsigned_test_values) {
    FOR_INT32_INPUTS2(j1, j2, unsigned_test_offset) {
      FOR_INT32_INPUTS2(k1, k2, unsigned_test_offset_increment) {
        float value = static_cast<float>(*i & 0xFFFFFFFF);
        int32_t in_offset = *j1 + *k1;
        int32_t out_offset = *j2 + *k2;

970 971 972 973 974 975 976
        auto fn = [](MacroAssembler* masm, int32_t in_offset,
                     int32_t out_offset) {
          __ Ulwc1(f0, MemOperand(a0, in_offset), t0);
          __ Uswc1(f0, MemOperand(a0, out_offset), t0);
        };
        CHECK_EQ(true, run_Unaligned<float>(buffer_middle, in_offset,
                                            out_offset, value, fn));
977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995
      }
    }
  }
}

TEST(Uldc1) {
  CcTest::InitializeVM();

  static const int kBufferSize = 300 * KB;
  char memory_buffer[kBufferSize];
  char* buffer_middle = memory_buffer + (kBufferSize / 2);

  FOR_UINT64_INPUTS(i, unsigned_test_values) {
    FOR_INT32_INPUTS2(j1, j2, unsigned_test_offset) {
      FOR_INT32_INPUTS2(k1, k2, unsigned_test_offset_increment) {
        double value = static_cast<double>(*i);
        int32_t in_offset = *j1 + *k1;
        int32_t out_offset = *j2 + *k2;

996 997 998 999 1000 1001 1002
        auto fn = [](MacroAssembler* masm, int32_t in_offset,
                     int32_t out_offset) {
          __ Uldc1(f0, MemOperand(a0, in_offset), t0);
          __ Usdc1(f0, MemOperand(a0, out_offset), t0);
        };
        CHECK_EQ(true, run_Unaligned<double>(buffer_middle, in_offset,
                                             out_offset, value, fn));
1003 1004 1005 1006 1007
      }
    }
  }
}

1008 1009
static const std::vector<uint32_t> sltu_test_values() {
  static const uint32_t kValues[] = {
1010 1011 1012
      0,          1,          0x7FFE,     0x7FFF,     0x8000,
      0x8001,     0xFFFE,     0xFFFF,     0xFFFF7FFE, 0xFFFF7FFF,
      0xFFFF8000, 0xFFFF8001, 0xFFFFFFFE, 0xFFFFFFFF,
1013 1014 1015 1016 1017 1018
  };
  return std::vector<uint32_t>(&kValues[0], &kValues[arraysize(kValues)]);
}

template <typename Func>
bool run_Sltu(uint32_t rs, uint32_t rd, Func GenerateSltuInstructionFunc) {
1019
  typedef int32_t(F_CVT)(uint32_t x0, uint32_t x1, int x2, int x3, int x4);
1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031

  Isolate* isolate = CcTest::i_isolate();
  HandleScope scope(isolate);
  MacroAssembler assm(isolate, nullptr, 0,
                      v8::internal::CodeObjectRequired::kYes);
  MacroAssembler* masm = &assm;

  GenerateSltuInstructionFunc(masm, rd);
  __ jr(ra);
  __ nop();

  CodeDesc desc;
1032
  assm.GetCode(isolate, &desc);
1033 1034
  Handle<Code> code =
      isolate->factory()->NewCode(desc, Code::STUB, Handle<Code>());
1035

1036 1037
  auto f = GeneratedCode<F_CVT>::FromCode(*code);
  int32_t res = reinterpret_cast<int32_t>(f.Call(rs, rd, 0, 0, 0));
1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048
  return res == 1;
}

TEST(Sltu) {
  CcTest::InitializeVM();

  FOR_UINT32_INPUTS(i, sltu_test_values) {
    FOR_UINT32_INPUTS(j, sltu_test_values) {
      uint32_t rs = *i;
      uint32_t rd = *j;

1049 1050 1051 1052 1053 1054 1055 1056 1057
      auto fn_1 = [](MacroAssembler* masm, uint32_t imm) {
        __ Sltu(v0, a0, Operand(imm));
      };
      CHECK_EQ(rs < rd, run_Sltu(rs, rd, fn_1));

      auto fn_2 = [](MacroAssembler* masm, uint32_t imm) {
        __ Sltu(v0, a0, a1);
      };
      CHECK_EQ(rs < rd, run_Sltu(rs, rd, fn_2));
1058 1059 1060 1061
    }
  }
}

1062
template <typename T, typename Inputs, typename Results>
1063
static GeneratedCode<F4> GenerateMacroFloat32MinMax(MacroAssembler* masm) {
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
  T a = T::from_code(4);  // f4
  T b = T::from_code(6);  // f6
  T c = T::from_code(8);  // f8

  Label ool_min_abc, ool_min_aab, ool_min_aba;
  Label ool_max_abc, ool_max_aab, ool_max_aba;

  Label done_min_abc, done_min_aab, done_min_aba;
  Label done_max_abc, done_max_aab, done_max_aba;

#define FLOAT_MIN_MAX(fminmax, res, x, y, done, ool, res_field) \
  __ lwc1(x, MemOperand(a0, offsetof(Inputs, src1_)));          \
  __ lwc1(y, MemOperand(a0, offsetof(Inputs, src2_)));          \
  __ fminmax(res, x, y, &ool);                                  \
  __ bind(&done);                                               \
  __ swc1(a, MemOperand(a1, offsetof(Results, res_field)))

  // a = min(b, c);
  FLOAT_MIN_MAX(Float32Min, a, b, c, done_min_abc, ool_min_abc, min_abc_);
  // a = min(a, b);
  FLOAT_MIN_MAX(Float32Min, a, a, b, done_min_aab, ool_min_aab, min_aab_);
  // a = min(b, a);
  FLOAT_MIN_MAX(Float32Min, a, b, a, done_min_aba, ool_min_aba, min_aba_);

  // a = max(b, c);
  FLOAT_MIN_MAX(Float32Max, a, b, c, done_max_abc, ool_max_abc, max_abc_);
  // a = max(a, b);
  FLOAT_MIN_MAX(Float32Max, a, a, b, done_max_aab, ool_max_aab, max_aab_);
  // a = max(b, a);
  FLOAT_MIN_MAX(Float32Max, a, b, a, done_max_aba, ool_max_aba, max_aba_);

#undef FLOAT_MIN_MAX

  __ jr(ra);
  __ nop();

  // Generate out-of-line cases.
  __ bind(&ool_min_abc);
  __ Float32MinOutOfLine(a, b, c);
1103
  __ Branch(&done_min_abc);
1104 1105 1106

  __ bind(&ool_min_aab);
  __ Float32MinOutOfLine(a, a, b);
1107
  __ Branch(&done_min_aab);
1108 1109 1110

  __ bind(&ool_min_aba);
  __ Float32MinOutOfLine(a, b, a);
1111
  __ Branch(&done_min_aba);
1112 1113 1114

  __ bind(&ool_max_abc);
  __ Float32MaxOutOfLine(a, b, c);
1115
  __ Branch(&done_max_abc);
1116 1117 1118

  __ bind(&ool_max_aab);
  __ Float32MaxOutOfLine(a, a, b);
1119
  __ Branch(&done_max_aab);
1120 1121 1122

  __ bind(&ool_max_aba);
  __ Float32MaxOutOfLine(a, b, a);
1123
  __ Branch(&done_max_aba);
1124 1125

  CodeDesc desc;
1126
  masm->GetCode(masm->isolate(), &desc);
1127 1128
  Handle<Code> code =
      masm->isolate()->factory()->NewCode(desc, Code::STUB, Handle<Code>());
1129
#ifdef DEBUG
1130
  StdoutStream os;
1131
  code->Print(os);
1132
#endif
1133
  return GeneratedCode<F4>::FromCode(*code);
1134 1135 1136 1137 1138 1139 1140 1141
}

TEST(macro_float_minmax_f32) {
  // Test the Float32Min and Float32Max macros.
  CcTest::InitializeVM();
  Isolate* isolate = CcTest::i_isolate();
  HandleScope scope(isolate);

1142
  MacroAssembler assembler(isolate, nullptr, 0,
1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161
                           v8::internal::CodeObjectRequired::kYes);
  MacroAssembler* masm = &assembler;

  struct Inputs {
    float src1_;
    float src2_;
  };

  struct Results {
    // Check all register aliasing possibilities in order to exercise all
    // code-paths in the macro assembler.
    float min_abc_;
    float min_aab_;
    float min_aba_;
    float max_abc_;
    float max_aab_;
    float max_aba_;
  };

1162 1163
  GeneratedCode<F4> f =
      GenerateMacroFloat32MinMax<FPURegister, Inputs, Results>(masm);
1164 1165 1166 1167 1168

#define CHECK_MINMAX(src1, src2, min, max)                                   \
  do {                                                                       \
    Inputs inputs = {src1, src2};                                            \
    Results results;                                                         \
1169
    f.Call(&inputs, &results, 0, 0, 0);                                      \
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
    CHECK_EQ(bit_cast<uint32_t>(min), bit_cast<uint32_t>(results.min_abc_)); \
    CHECK_EQ(bit_cast<uint32_t>(min), bit_cast<uint32_t>(results.min_aab_)); \
    CHECK_EQ(bit_cast<uint32_t>(min), bit_cast<uint32_t>(results.min_aba_)); \
    CHECK_EQ(bit_cast<uint32_t>(max), bit_cast<uint32_t>(results.max_abc_)); \
    CHECK_EQ(bit_cast<uint32_t>(max), bit_cast<uint32_t>(results.max_aab_)); \
    CHECK_EQ(bit_cast<uint32_t>(max), bit_cast<uint32_t>(results.max_aba_)); \
    /* Use a bit_cast to correctly identify -0.0 and NaNs. */                \
  } while (0)

  float nan_a = std::numeric_limits<float>::quiet_NaN();
  float nan_b = std::numeric_limits<float>::quiet_NaN();

  CHECK_MINMAX(1.0f, -1.0f, -1.0f, 1.0f);
  CHECK_MINMAX(-1.0f, 1.0f, -1.0f, 1.0f);
  CHECK_MINMAX(0.0f, -1.0f, -1.0f, 0.0f);
  CHECK_MINMAX(-1.0f, 0.0f, -1.0f, 0.0f);
  CHECK_MINMAX(-0.0f, -1.0f, -1.0f, -0.0f);
  CHECK_MINMAX(-1.0f, -0.0f, -1.0f, -0.0f);
  CHECK_MINMAX(0.0f, 1.0f, 0.0f, 1.0f);
  CHECK_MINMAX(1.0f, 0.0f, 0.0f, 1.0f);

  CHECK_MINMAX(0.0f, 0.0f, 0.0f, 0.0f);
  CHECK_MINMAX(-0.0f, -0.0f, -0.0f, -0.0f);
  CHECK_MINMAX(-0.0f, 0.0f, -0.0f, 0.0f);
  CHECK_MINMAX(0.0f, -0.0f, -0.0f, 0.0f);

  CHECK_MINMAX(0.0f, nan_a, nan_a, nan_a);
  CHECK_MINMAX(nan_a, 0.0f, nan_a, nan_a);
  CHECK_MINMAX(nan_a, nan_b, nan_a, nan_a);
  CHECK_MINMAX(nan_b, nan_a, nan_b, nan_b);

#undef CHECK_MINMAX
}

template <typename T, typename Inputs, typename Results>
1205
static GeneratedCode<F4> GenerateMacroFloat64MinMax(MacroAssembler* masm) {
1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216
  T a = T::from_code(4);  // f4
  T b = T::from_code(6);  // f6
  T c = T::from_code(8);  // f8

  Label ool_min_abc, ool_min_aab, ool_min_aba;
  Label ool_max_abc, ool_max_aab, ool_max_aba;

  Label done_min_abc, done_min_aab, done_min_aba;
  Label done_max_abc, done_max_aab, done_max_aba;

#define FLOAT_MIN_MAX(fminmax, res, x, y, done, ool, res_field) \
1217 1218
  __ Ldc1(x, MemOperand(a0, offsetof(Inputs, src1_)));          \
  __ Ldc1(y, MemOperand(a0, offsetof(Inputs, src2_)));          \
1219 1220
  __ fminmax(res, x, y, &ool);                                  \
  __ bind(&done);                                               \
1221
  __ Sdc1(a, MemOperand(a1, offsetof(Results, res_field)))
1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244

  // a = min(b, c);
  FLOAT_MIN_MAX(Float64Min, a, b, c, done_min_abc, ool_min_abc, min_abc_);
  // a = min(a, b);
  FLOAT_MIN_MAX(Float64Min, a, a, b, done_min_aab, ool_min_aab, min_aab_);
  // a = min(b, a);
  FLOAT_MIN_MAX(Float64Min, a, b, a, done_min_aba, ool_min_aba, min_aba_);

  // a = max(b, c);
  FLOAT_MIN_MAX(Float64Max, a, b, c, done_max_abc, ool_max_abc, max_abc_);
  // a = max(a, b);
  FLOAT_MIN_MAX(Float64Max, a, a, b, done_max_aab, ool_max_aab, max_aab_);
  // a = max(b, a);
  FLOAT_MIN_MAX(Float64Max, a, b, a, done_max_aba, ool_max_aba, max_aba_);

#undef FLOAT_MIN_MAX

  __ jr(ra);
  __ nop();

  // Generate out-of-line cases.
  __ bind(&ool_min_abc);
  __ Float64MinOutOfLine(a, b, c);
1245
  __ Branch(&done_min_abc);
1246 1247 1248

  __ bind(&ool_min_aab);
  __ Float64MinOutOfLine(a, a, b);
1249
  __ Branch(&done_min_aab);
1250 1251 1252

  __ bind(&ool_min_aba);
  __ Float64MinOutOfLine(a, b, a);
1253
  __ Branch(&done_min_aba);
1254 1255 1256

  __ bind(&ool_max_abc);
  __ Float64MaxOutOfLine(a, b, c);
1257
  __ Branch(&done_max_abc);
1258 1259 1260

  __ bind(&ool_max_aab);
  __ Float64MaxOutOfLine(a, a, b);
1261
  __ Branch(&done_max_aab);
1262 1263 1264

  __ bind(&ool_max_aba);
  __ Float64MaxOutOfLine(a, b, a);
1265
  __ Branch(&done_max_aba);
1266 1267

  CodeDesc desc;
1268
  masm->GetCode(masm->isolate(), &desc);
1269 1270
  Handle<Code> code =
      masm->isolate()->factory()->NewCode(desc, Code::STUB, Handle<Code>());
1271
#ifdef DEBUG
1272
  StdoutStream os;
1273
  code->Print(os);
1274
#endif
1275
  return GeneratedCode<F4>::FromCode(*code);
1276 1277 1278 1279 1280 1281 1282 1283
}

TEST(macro_float_minmax_f64) {
  // Test the Float64Min and Float64Max macros.
  CcTest::InitializeVM();
  Isolate* isolate = CcTest::i_isolate();
  HandleScope scope(isolate);

1284
  MacroAssembler assembler(isolate, nullptr, 0,
1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303
                           v8::internal::CodeObjectRequired::kYes);
  MacroAssembler* masm = &assembler;

  struct Inputs {
    double src1_;
    double src2_;
  };

  struct Results {
    // Check all register aliasing possibilities in order to exercise all
    // code-paths in the macro assembler.
    double min_abc_;
    double min_aab_;
    double min_aba_;
    double max_abc_;
    double max_aab_;
    double max_aba_;
  };

1304 1305
  GeneratedCode<F4> f =
      GenerateMacroFloat64MinMax<DoubleRegister, Inputs, Results>(masm);
1306 1307 1308 1309 1310

#define CHECK_MINMAX(src1, src2, min, max)                                   \
  do {                                                                       \
    Inputs inputs = {src1, src2};                                            \
    Results results;                                                         \
1311
    f.Call(&inputs, &results, 0, 0, 0);                                      \
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
    CHECK_EQ(bit_cast<uint64_t>(min), bit_cast<uint64_t>(results.min_abc_)); \
    CHECK_EQ(bit_cast<uint64_t>(min), bit_cast<uint64_t>(results.min_aab_)); \
    CHECK_EQ(bit_cast<uint64_t>(min), bit_cast<uint64_t>(results.min_aba_)); \
    CHECK_EQ(bit_cast<uint64_t>(max), bit_cast<uint64_t>(results.max_abc_)); \
    CHECK_EQ(bit_cast<uint64_t>(max), bit_cast<uint64_t>(results.max_aab_)); \
    CHECK_EQ(bit_cast<uint64_t>(max), bit_cast<uint64_t>(results.max_aba_)); \
    /* Use a bit_cast to correctly identify -0.0 and NaNs. */                \
  } while (0)

  double nan_a = std::numeric_limits<double>::quiet_NaN();
  double nan_b = std::numeric_limits<double>::quiet_NaN();

  CHECK_MINMAX(1.0, -1.0, -1.0, 1.0);
  CHECK_MINMAX(-1.0, 1.0, -1.0, 1.0);
  CHECK_MINMAX(0.0, -1.0, -1.0, 0.0);
  CHECK_MINMAX(-1.0, 0.0, -1.0, 0.0);
  CHECK_MINMAX(-0.0, -1.0, -1.0, -0.0);
  CHECK_MINMAX(-1.0, -0.0, -1.0, -0.0);
  CHECK_MINMAX(0.0, 1.0, 0.0, 1.0);
  CHECK_MINMAX(1.0, 0.0, 0.0, 1.0);

  CHECK_MINMAX(0.0, 0.0, 0.0, 0.0);
  CHECK_MINMAX(-0.0, -0.0, -0.0, -0.0);
  CHECK_MINMAX(-0.0, 0.0, -0.0, 0.0);
  CHECK_MINMAX(0.0, -0.0, -0.0, 0.0);

  CHECK_MINMAX(0.0, nan_a, nan_a, nan_a);
  CHECK_MINMAX(nan_a, 0.0, nan_a, nan_a);
  CHECK_MINMAX(nan_a, nan_b, nan_a, nan_a);
  CHECK_MINMAX(nan_b, nan_a, nan_b, nan_b);

#undef CHECK_MINMAX
}

1346
#undef __
1347 1348 1349

}  // namespace internal
}  // namespace v8