1 //===- ValueTrackingTest.cpp - ValueTracking tests ------------------------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 9 #include "llvm/Analysis/ValueTracking.h" 10 #include "llvm/Analysis/AssumptionCache.h" 11 #include "llvm/AsmParser/Parser.h" 12 #include "llvm/IR/ConstantRange.h" 13 #include "llvm/IR/Dominators.h" 14 #include "llvm/IR/Function.h" 15 #include "llvm/IR/InstIterator.h" 16 #include "llvm/IR/Instructions.h" 17 #include "llvm/IR/LLVMContext.h" 18 #include "llvm/IR/Module.h" 19 #include "llvm/Support/ErrorHandling.h" 20 #include "llvm/Support/KnownBits.h" 21 #include "llvm/Support/SourceMgr.h" 22 #include "gtest/gtest.h" 23 24 using namespace llvm; 25 26 namespace { 27 28 static Instruction &findInstructionByName(Function *F, StringRef Name) { 29 for (Instruction &I : instructions(F)) 30 if (I.getName() == Name) 31 return I; 32 33 llvm_unreachable("Expected value not found"); 34 } 35 36 class ValueTrackingTest : public testing::Test { 37 protected: 38 std::unique_ptr<Module> parseModule(StringRef Assembly) { 39 SMDiagnostic Error; 40 std::unique_ptr<Module> M = parseAssemblyString(Assembly, Error, Context); 41 42 std::string errMsg; 43 raw_string_ostream os(errMsg); 44 Error.print("", os); 45 EXPECT_TRUE(M) << os.str(); 46 47 return M; 48 } 49 50 void parseAssembly(StringRef Assembly) { 51 M = parseModule(Assembly); 52 ASSERT_TRUE(M); 53 54 F = M->getFunction("test"); 55 ASSERT_TRUE(F) << "Test must have a function @test"; 56 if (!F) 57 return; 58 59 A = &findInstructionByName(F, "A"); 60 ASSERT_TRUE(A) << "@test must have an instruction %A"; 61 } 62 63 LLVMContext Context; 64 std::unique_ptr<Module> M; 65 Function *F = nullptr; 66 Instruction *A = nullptr; 67 }; 68 69 class MatchSelectPatternTest : public ValueTrackingTest { 70 protected: 71 void expectPattern(const SelectPatternResult &P) { 72 Value *LHS, *RHS; 73 Instruction::CastOps CastOp; 74 SelectPatternResult R = matchSelectPattern(A, LHS, RHS, &CastOp); 75 EXPECT_EQ(P.Flavor, R.Flavor); 76 EXPECT_EQ(P.NaNBehavior, R.NaNBehavior); 77 EXPECT_EQ(P.Ordered, R.Ordered); 78 } 79 }; 80 81 class ComputeKnownBitsTest : public ValueTrackingTest { 82 protected: 83 void expectKnownBits(uint64_t Zero, uint64_t One) { 84 auto Known = computeKnownBits(A, M->getDataLayout()); 85 ASSERT_FALSE(Known.hasConflict()); 86 EXPECT_EQ(Known.One.getZExtValue(), One); 87 EXPECT_EQ(Known.Zero.getZExtValue(), Zero); 88 } 89 }; 90 91 } 92 93 TEST_F(MatchSelectPatternTest, SimpleFMin) { 94 parseAssembly( 95 "define float @test(float %a) {\n" 96 " %1 = fcmp ult float %a, 5.0\n" 97 " %A = select i1 %1, float %a, float 5.0\n" 98 " ret float %A\n" 99 "}\n"); 100 expectPattern({SPF_FMINNUM, SPNB_RETURNS_NAN, false}); 101 } 102 103 TEST_F(MatchSelectPatternTest, SimpleFMax) { 104 parseAssembly( 105 "define float @test(float %a) {\n" 106 " %1 = fcmp ogt float %a, 5.0\n" 107 " %A = select i1 %1, float %a, float 5.0\n" 108 " ret float %A\n" 109 "}\n"); 110 expectPattern({SPF_FMAXNUM, SPNB_RETURNS_OTHER, true}); 111 } 112 113 TEST_F(MatchSelectPatternTest, SwappedFMax) { 114 parseAssembly( 115 "define float @test(float %a) {\n" 116 " %1 = fcmp olt float 5.0, %a\n" 117 " %A = select i1 %1, float %a, float 5.0\n" 118 " ret float %A\n" 119 "}\n"); 120 expectPattern({SPF_FMAXNUM, SPNB_RETURNS_OTHER, false}); 121 } 122 123 TEST_F(MatchSelectPatternTest, SwappedFMax2) { 124 parseAssembly( 125 "define float @test(float %a) {\n" 126 " %1 = fcmp olt float %a, 5.0\n" 127 " %A = select i1 %1, float 5.0, float %a\n" 128 " ret float %A\n" 129 "}\n"); 130 expectPattern({SPF_FMAXNUM, SPNB_RETURNS_NAN, false}); 131 } 132 133 TEST_F(MatchSelectPatternTest, SwappedFMax3) { 134 parseAssembly( 135 "define float @test(float %a) {\n" 136 " %1 = fcmp ult float %a, 5.0\n" 137 " %A = select i1 %1, float 5.0, float %a\n" 138 " ret float %A\n" 139 "}\n"); 140 expectPattern({SPF_FMAXNUM, SPNB_RETURNS_OTHER, true}); 141 } 142 143 TEST_F(MatchSelectPatternTest, FastFMin) { 144 parseAssembly( 145 "define float @test(float %a) {\n" 146 " %1 = fcmp nnan olt float %a, 5.0\n" 147 " %A = select i1 %1, float %a, float 5.0\n" 148 " ret float %A\n" 149 "}\n"); 150 expectPattern({SPF_FMINNUM, SPNB_RETURNS_ANY, false}); 151 } 152 153 TEST_F(MatchSelectPatternTest, FMinConstantZero) { 154 parseAssembly( 155 "define float @test(float %a) {\n" 156 " %1 = fcmp ole float %a, 0.0\n" 157 " %A = select i1 %1, float %a, float 0.0\n" 158 " ret float %A\n" 159 "}\n"); 160 // This shouldn't be matched, as %a could be -0.0. 161 expectPattern({SPF_UNKNOWN, SPNB_NA, false}); 162 } 163 164 TEST_F(MatchSelectPatternTest, FMinConstantZeroNsz) { 165 parseAssembly( 166 "define float @test(float %a) {\n" 167 " %1 = fcmp nsz ole float %a, 0.0\n" 168 " %A = select i1 %1, float %a, float 0.0\n" 169 " ret float %A\n" 170 "}\n"); 171 // But this should be, because we've ignored signed zeroes. 172 expectPattern({SPF_FMINNUM, SPNB_RETURNS_OTHER, true}); 173 } 174 175 TEST_F(MatchSelectPatternTest, FMinMismatchConstantZero1) { 176 parseAssembly( 177 "define float @test(float %a) {\n" 178 " %1 = fcmp olt float -0.0, %a\n" 179 " %A = select i1 %1, float 0.0, float %a\n" 180 " ret float %A\n" 181 "}\n"); 182 // The sign of zero doesn't matter in fcmp. 183 expectPattern({SPF_FMINNUM, SPNB_RETURNS_NAN, true}); 184 } 185 186 TEST_F(MatchSelectPatternTest, FMinMismatchConstantZero2) { 187 parseAssembly( 188 "define float @test(float %a) {\n" 189 " %1 = fcmp ogt float %a, -0.0\n" 190 " %A = select i1 %1, float 0.0, float %a\n" 191 " ret float %A\n" 192 "}\n"); 193 // The sign of zero doesn't matter in fcmp. 194 expectPattern({SPF_FMINNUM, SPNB_RETURNS_NAN, false}); 195 } 196 197 TEST_F(MatchSelectPatternTest, FMinMismatchConstantZero3) { 198 parseAssembly( 199 "define float @test(float %a) {\n" 200 " %1 = fcmp olt float 0.0, %a\n" 201 " %A = select i1 %1, float -0.0, float %a\n" 202 " ret float %A\n" 203 "}\n"); 204 // The sign of zero doesn't matter in fcmp. 205 expectPattern({SPF_FMINNUM, SPNB_RETURNS_NAN, true}); 206 } 207 208 TEST_F(MatchSelectPatternTest, FMinMismatchConstantZero4) { 209 parseAssembly( 210 "define float @test(float %a) {\n" 211 " %1 = fcmp ogt float %a, 0.0\n" 212 " %A = select i1 %1, float -0.0, float %a\n" 213 " ret float %A\n" 214 "}\n"); 215 // The sign of zero doesn't matter in fcmp. 216 expectPattern({SPF_FMINNUM, SPNB_RETURNS_NAN, false}); 217 } 218 219 TEST_F(MatchSelectPatternTest, FMinMismatchConstantZero5) { 220 parseAssembly( 221 "define float @test(float %a) {\n" 222 " %1 = fcmp ogt float -0.0, %a\n" 223 " %A = select i1 %1, float %a, float 0.0\n" 224 " ret float %A\n" 225 "}\n"); 226 // The sign of zero doesn't matter in fcmp. 227 expectPattern({SPF_FMINNUM, SPNB_RETURNS_OTHER, false}); 228 } 229 230 TEST_F(MatchSelectPatternTest, FMinMismatchConstantZero6) { 231 parseAssembly( 232 "define float @test(float %a) {\n" 233 " %1 = fcmp olt float %a, -0.0\n" 234 " %A = select i1 %1, float %a, float 0.0\n" 235 " ret float %A\n" 236 "}\n"); 237 // The sign of zero doesn't matter in fcmp. 238 expectPattern({SPF_FMINNUM, SPNB_RETURNS_OTHER, true}); 239 } 240 241 TEST_F(MatchSelectPatternTest, FMinMismatchConstantZero7) { 242 parseAssembly( 243 "define float @test(float %a) {\n" 244 " %1 = fcmp ogt float 0.0, %a\n" 245 " %A = select i1 %1, float %a, float -0.0\n" 246 " ret float %A\n" 247 "}\n"); 248 // The sign of zero doesn't matter in fcmp. 249 expectPattern({SPF_FMINNUM, SPNB_RETURNS_OTHER, false}); 250 } 251 252 TEST_F(MatchSelectPatternTest, FMinMismatchConstantZero8) { 253 parseAssembly( 254 "define float @test(float %a) {\n" 255 " %1 = fcmp olt float %a, 0.0\n" 256 " %A = select i1 %1, float %a, float -0.0\n" 257 " ret float %A\n" 258 "}\n"); 259 // The sign of zero doesn't matter in fcmp. 260 expectPattern({SPF_FMINNUM, SPNB_RETURNS_OTHER, true}); 261 } 262 263 TEST_F(MatchSelectPatternTest, FMaxMismatchConstantZero1) { 264 parseAssembly( 265 "define float @test(float %a) {\n" 266 " %1 = fcmp ogt float -0.0, %a\n" 267 " %A = select i1 %1, float 0.0, float %a\n" 268 " ret float %A\n" 269 "}\n"); 270 // The sign of zero doesn't matter in fcmp. 271 expectPattern({SPF_FMAXNUM, SPNB_RETURNS_NAN, true}); 272 } 273 274 TEST_F(MatchSelectPatternTest, FMaxMismatchConstantZero2) { 275 parseAssembly( 276 "define float @test(float %a) {\n" 277 " %1 = fcmp olt float %a, -0.0\n" 278 " %A = select i1 %1, float 0.0, float %a\n" 279 " ret float %A\n" 280 "}\n"); 281 // The sign of zero doesn't matter in fcmp. 282 expectPattern({SPF_FMAXNUM, SPNB_RETURNS_NAN, false}); 283 } 284 285 TEST_F(MatchSelectPatternTest, FMaxMismatchConstantZero3) { 286 parseAssembly( 287 "define float @test(float %a) {\n" 288 " %1 = fcmp ogt float 0.0, %a\n" 289 " %A = select i1 %1, float -0.0, float %a\n" 290 " ret float %A\n" 291 "}\n"); 292 // The sign of zero doesn't matter in fcmp. 293 expectPattern({SPF_FMAXNUM, SPNB_RETURNS_NAN, true}); 294 } 295 296 TEST_F(MatchSelectPatternTest, FMaxMismatchConstantZero4) { 297 parseAssembly( 298 "define float @test(float %a) {\n" 299 " %1 = fcmp olt float %a, 0.0\n" 300 " %A = select i1 %1, float -0.0, float %a\n" 301 " ret float %A\n" 302 "}\n"); 303 // The sign of zero doesn't matter in fcmp. 304 expectPattern({SPF_FMAXNUM, SPNB_RETURNS_NAN, false}); 305 } 306 307 TEST_F(MatchSelectPatternTest, FMaxMismatchConstantZero5) { 308 parseAssembly( 309 "define float @test(float %a) {\n" 310 " %1 = fcmp olt float -0.0, %a\n" 311 " %A = select i1 %1, float %a, float 0.0\n" 312 " ret float %A\n" 313 "}\n"); 314 // The sign of zero doesn't matter in fcmp. 315 expectPattern({SPF_FMAXNUM, SPNB_RETURNS_OTHER, false}); 316 } 317 318 TEST_F(MatchSelectPatternTest, FMaxMismatchConstantZero6) { 319 parseAssembly( 320 "define float @test(float %a) {\n" 321 " %1 = fcmp ogt float %a, -0.0\n" 322 " %A = select i1 %1, float %a, float 0.0\n" 323 " ret float %A\n" 324 "}\n"); 325 // The sign of zero doesn't matter in fcmp. 326 expectPattern({SPF_FMAXNUM, SPNB_RETURNS_OTHER, true}); 327 } 328 329 TEST_F(MatchSelectPatternTest, FMaxMismatchConstantZero7) { 330 parseAssembly( 331 "define float @test(float %a) {\n" 332 " %1 = fcmp olt float 0.0, %a\n" 333 " %A = select i1 %1, float %a, float -0.0\n" 334 " ret float %A\n" 335 "}\n"); 336 // The sign of zero doesn't matter in fcmp. 337 expectPattern({SPF_FMAXNUM, SPNB_RETURNS_OTHER, false}); 338 } 339 340 TEST_F(MatchSelectPatternTest, FMaxMismatchConstantZero8) { 341 parseAssembly( 342 "define float @test(float %a) {\n" 343 " %1 = fcmp ogt float %a, 0.0\n" 344 " %A = select i1 %1, float %a, float -0.0\n" 345 " ret float %A\n" 346 "}\n"); 347 // The sign of zero doesn't matter in fcmp. 348 expectPattern({SPF_FMAXNUM, SPNB_RETURNS_OTHER, true}); 349 } 350 351 TEST_F(MatchSelectPatternTest, FMinMismatchConstantZeroVecUndef) { 352 parseAssembly( 353 "define <2 x float> @test(<2 x float> %a) {\n" 354 " %1 = fcmp ogt <2 x float> %a, <float -0.0, float -0.0>\n" 355 " %A = select <2 x i1> %1, <2 x float> <float undef, float 0.0>, <2 x float> %a\n" 356 " ret <2 x float> %A\n" 357 "}\n"); 358 // An undef in a vector constant can not be back-propagated for this analysis. 359 expectPattern({SPF_UNKNOWN, SPNB_NA, false}); 360 } 361 362 TEST_F(MatchSelectPatternTest, FMaxMismatchConstantZeroVecUndef) { 363 parseAssembly( 364 "define <2 x float> @test(<2 x float> %a) {\n" 365 " %1 = fcmp ogt <2 x float> %a, zeroinitializer\n" 366 " %A = select <2 x i1> %1, <2 x float> %a, <2 x float> <float -0.0, float undef>\n" 367 " ret <2 x float> %A\n" 368 "}\n"); 369 // An undef in a vector constant can not be back-propagated for this analysis. 370 expectPattern({SPF_UNKNOWN, SPNB_NA, false}); 371 } 372 373 TEST_F(MatchSelectPatternTest, VectorFMinimum) { 374 parseAssembly( 375 "define <4 x float> @test(<4 x float> %a) {\n" 376 " %1 = fcmp ule <4 x float> %a, \n" 377 " <float 5.0, float 5.0, float 5.0, float 5.0>\n" 378 " %A = select <4 x i1> %1, <4 x float> %a,\n" 379 " <4 x float> <float 5.0, float 5.0, float 5.0, float 5.0>\n" 380 " ret <4 x float> %A\n" 381 "}\n"); 382 // Check that pattern matching works on vectors where each lane has the same 383 // unordered pattern. 384 expectPattern({SPF_FMINNUM, SPNB_RETURNS_NAN, false}); 385 } 386 387 TEST_F(MatchSelectPatternTest, VectorFMinOtherOrdered) { 388 parseAssembly( 389 "define <4 x float> @test(<4 x float> %a) {\n" 390 " %1 = fcmp ole <4 x float> %a, \n" 391 " <float 5.0, float 5.0, float 5.0, float 5.0>\n" 392 " %A = select <4 x i1> %1, <4 x float> %a,\n" 393 " <4 x float> <float 5.0, float 5.0, float 5.0, float 5.0>\n" 394 " ret <4 x float> %A\n" 395 "}\n"); 396 // Check that pattern matching works on vectors where each lane has the same 397 // ordered pattern. 398 expectPattern({SPF_FMINNUM, SPNB_RETURNS_OTHER, true}); 399 } 400 401 TEST_F(MatchSelectPatternTest, VectorNotFMinimum) { 402 parseAssembly( 403 "define <4 x float> @test(<4 x float> %a) {\n" 404 " %1 = fcmp ule <4 x float> %a, \n" 405 " <float 5.0, float 0x7ff8000000000000, float 5.0, float 5.0>\n" 406 " %A = select <4 x i1> %1, <4 x float> %a,\n" 407 " <4 x float> <float 5.0, float 0x7ff8000000000000, float 5.0, float " 408 "5.0>\n" 409 " ret <4 x float> %A\n" 410 "}\n"); 411 // The lane that contains a NaN (0x7ff80...) behaves like a 412 // non-NaN-propagating min and the other lines behave like a NaN-propagating 413 // min, so check that neither is returned. 414 expectPattern({SPF_UNKNOWN, SPNB_NA, false}); 415 } 416 417 TEST_F(MatchSelectPatternTest, VectorNotFMinZero) { 418 parseAssembly( 419 "define <4 x float> @test(<4 x float> %a) {\n" 420 " %1 = fcmp ule <4 x float> %a, \n" 421 " <float 5.0, float -0.0, float 5.0, float 5.0>\n" 422 " %A = select <4 x i1> %1, <4 x float> %a,\n" 423 " <4 x float> <float 5.0, float 0.0, float 5.0, float 5.0>\n" 424 " ret <4 x float> %A\n" 425 "}\n"); 426 // Always selects the second lane of %a if it is positive or negative zero, so 427 // this is stricter than a min. 428 expectPattern({SPF_UNKNOWN, SPNB_NA, false}); 429 } 430 431 TEST_F(MatchSelectPatternTest, DoubleCastU) { 432 parseAssembly( 433 "define i32 @test(i8 %a, i8 %b) {\n" 434 " %1 = icmp ult i8 %a, %b\n" 435 " %2 = zext i8 %a to i32\n" 436 " %3 = zext i8 %b to i32\n" 437 " %A = select i1 %1, i32 %2, i32 %3\n" 438 " ret i32 %A\n" 439 "}\n"); 440 // We should be able to look through the situation where we cast both operands 441 // to the select. 442 expectPattern({SPF_UMIN, SPNB_NA, false}); 443 } 444 445 TEST_F(MatchSelectPatternTest, DoubleCastS) { 446 parseAssembly( 447 "define i32 @test(i8 %a, i8 %b) {\n" 448 " %1 = icmp slt i8 %a, %b\n" 449 " %2 = sext i8 %a to i32\n" 450 " %3 = sext i8 %b to i32\n" 451 " %A = select i1 %1, i32 %2, i32 %3\n" 452 " ret i32 %A\n" 453 "}\n"); 454 // We should be able to look through the situation where we cast both operands 455 // to the select. 456 expectPattern({SPF_SMIN, SPNB_NA, false}); 457 } 458 459 TEST_F(MatchSelectPatternTest, DoubleCastBad) { 460 parseAssembly( 461 "define i32 @test(i8 %a, i8 %b) {\n" 462 " %1 = icmp ult i8 %a, %b\n" 463 " %2 = zext i8 %a to i32\n" 464 " %3 = sext i8 %b to i32\n" 465 " %A = select i1 %1, i32 %2, i32 %3\n" 466 " ret i32 %A\n" 467 "}\n"); 468 // The cast types here aren't the same, so we cannot match an UMIN. 469 expectPattern({SPF_UNKNOWN, SPNB_NA, false}); 470 } 471 472 TEST_F(MatchSelectPatternTest, NotNotSMin) { 473 parseAssembly( 474 "define i8 @test(i8 %a, i8 %b) {\n" 475 " %cmp = icmp sgt i8 %a, %b\n" 476 " %an = xor i8 %a, -1\n" 477 " %bn = xor i8 %b, -1\n" 478 " %A = select i1 %cmp, i8 %an, i8 %bn\n" 479 " ret i8 %A\n" 480 "}\n"); 481 expectPattern({SPF_SMIN, SPNB_NA, false}); 482 } 483 484 TEST_F(MatchSelectPatternTest, NotNotSMinSwap) { 485 parseAssembly( 486 "define <2 x i8> @test(<2 x i8> %a, <2 x i8> %b) {\n" 487 " %cmp = icmp slt <2 x i8> %a, %b\n" 488 " %an = xor <2 x i8> %a, <i8 -1, i8-1>\n" 489 " %bn = xor <2 x i8> %b, <i8 -1, i8-1>\n" 490 " %A = select <2 x i1> %cmp, <2 x i8> %bn, <2 x i8> %an\n" 491 " ret <2 x i8> %A\n" 492 "}\n"); 493 expectPattern({SPF_SMIN, SPNB_NA, false}); 494 } 495 496 TEST_F(MatchSelectPatternTest, NotNotSMax) { 497 parseAssembly( 498 "define i8 @test(i8 %a, i8 %b) {\n" 499 " %cmp = icmp slt i8 %a, %b\n" 500 " %an = xor i8 %a, -1\n" 501 " %bn = xor i8 %b, -1\n" 502 " %A = select i1 %cmp, i8 %an, i8 %bn\n" 503 " ret i8 %A\n" 504 "}\n"); 505 expectPattern({SPF_SMAX, SPNB_NA, false}); 506 } 507 508 TEST_F(MatchSelectPatternTest, NotNotSMaxSwap) { 509 parseAssembly( 510 "define <2 x i8> @test(<2 x i8> %a, <2 x i8> %b) {\n" 511 " %cmp = icmp sgt <2 x i8> %a, %b\n" 512 " %an = xor <2 x i8> %a, <i8 -1, i8-1>\n" 513 " %bn = xor <2 x i8> %b, <i8 -1, i8-1>\n" 514 " %A = select <2 x i1> %cmp, <2 x i8> %bn, <2 x i8> %an\n" 515 " ret <2 x i8> %A\n" 516 "}\n"); 517 expectPattern({SPF_SMAX, SPNB_NA, false}); 518 } 519 520 TEST_F(MatchSelectPatternTest, NotNotUMin) { 521 parseAssembly( 522 "define <2 x i8> @test(<2 x i8> %a, <2 x i8> %b) {\n" 523 " %cmp = icmp ugt <2 x i8> %a, %b\n" 524 " %an = xor <2 x i8> %a, <i8 -1, i8-1>\n" 525 " %bn = xor <2 x i8> %b, <i8 -1, i8-1>\n" 526 " %A = select <2 x i1> %cmp, <2 x i8> %an, <2 x i8> %bn\n" 527 " ret <2 x i8> %A\n" 528 "}\n"); 529 expectPattern({SPF_UMIN, SPNB_NA, false}); 530 } 531 532 TEST_F(MatchSelectPatternTest, NotNotUMinSwap) { 533 parseAssembly( 534 "define i8 @test(i8 %a, i8 %b) {\n" 535 " %cmp = icmp ult i8 %a, %b\n" 536 " %an = xor i8 %a, -1\n" 537 " %bn = xor i8 %b, -1\n" 538 " %A = select i1 %cmp, i8 %bn, i8 %an\n" 539 " ret i8 %A\n" 540 "}\n"); 541 expectPattern({SPF_UMIN, SPNB_NA, false}); 542 } 543 544 TEST_F(MatchSelectPatternTest, NotNotUMax) { 545 parseAssembly( 546 "define <2 x i8> @test(<2 x i8> %a, <2 x i8> %b) {\n" 547 " %cmp = icmp ult <2 x i8> %a, %b\n" 548 " %an = xor <2 x i8> %a, <i8 -1, i8-1>\n" 549 " %bn = xor <2 x i8> %b, <i8 -1, i8-1>\n" 550 " %A = select <2 x i1> %cmp, <2 x i8> %an, <2 x i8> %bn\n" 551 " ret <2 x i8> %A\n" 552 "}\n"); 553 expectPattern({SPF_UMAX, SPNB_NA, false}); 554 } 555 556 TEST_F(MatchSelectPatternTest, NotNotUMaxSwap) { 557 parseAssembly( 558 "define i8 @test(i8 %a, i8 %b) {\n" 559 " %cmp = icmp ugt i8 %a, %b\n" 560 " %an = xor i8 %a, -1\n" 561 " %bn = xor i8 %b, -1\n" 562 " %A = select i1 %cmp, i8 %bn, i8 %an\n" 563 " ret i8 %A\n" 564 "}\n"); 565 expectPattern({SPF_UMAX, SPNB_NA, false}); 566 } 567 568 TEST_F(MatchSelectPatternTest, NotNotEq) { 569 parseAssembly( 570 "define i8 @test(i8 %a, i8 %b) {\n" 571 " %cmp = icmp eq i8 %a, %b\n" 572 " %an = xor i8 %a, -1\n" 573 " %bn = xor i8 %b, -1\n" 574 " %A = select i1 %cmp, i8 %bn, i8 %an\n" 575 " ret i8 %A\n" 576 "}\n"); 577 expectPattern({SPF_UNKNOWN, SPNB_NA, false}); 578 } 579 580 TEST_F(MatchSelectPatternTest, NotNotNe) { 581 parseAssembly( 582 "define i8 @test(i8 %a, i8 %b) {\n" 583 " %cmp = icmp ne i8 %a, %b\n" 584 " %an = xor i8 %a, -1\n" 585 " %bn = xor i8 %b, -1\n" 586 " %A = select i1 %cmp, i8 %bn, i8 %an\n" 587 " ret i8 %A\n" 588 "}\n"); 589 expectPattern({SPF_UNKNOWN, SPNB_NA, false}); 590 } 591 592 TEST(ValueTracking, GuaranteedToTransferExecutionToSuccessor) { 593 StringRef Assembly = 594 "declare void @nounwind_readonly(i32*) nounwind readonly " 595 "declare void @nounwind_argmemonly(i32*) nounwind argmemonly " 596 "declare void @throws_but_readonly(i32*) readonly " 597 "declare void @throws_but_argmemonly(i32*) argmemonly " 598 "declare void @nounwind_willreturn(i32*) nounwind willreturn" 599 " " 600 "declare void @unknown(i32*) " 601 " " 602 "define void @f(i32* %p) { " 603 " call void @nounwind_readonly(i32* %p) " 604 " call void @nounwind_argmemonly(i32* %p) " 605 " call void @throws_but_readonly(i32* %p) " 606 " call void @throws_but_argmemonly(i32* %p) " 607 " call void @unknown(i32* %p) nounwind readonly " 608 " call void @unknown(i32* %p) nounwind argmemonly " 609 " call void @unknown(i32* %p) readonly " 610 " call void @unknown(i32* %p) argmemonly " 611 " call void @nounwind_willreturn(i32* %p)" 612 " ret void " 613 "} "; 614 615 LLVMContext Context; 616 SMDiagnostic Error; 617 auto M = parseAssemblyString(Assembly, Error, Context); 618 assert(M && "Bad assembly?"); 619 620 auto *F = M->getFunction("f"); 621 assert(F && "Bad assembly?"); 622 623 auto &BB = F->getEntryBlock(); 624 bool ExpectedAnswers[] = { 625 true, // call void @nounwind_readonly(i32* %p) 626 true, // call void @nounwind_argmemonly(i32* %p) 627 false, // call void @throws_but_readonly(i32* %p) 628 false, // call void @throws_but_argmemonly(i32* %p) 629 true, // call void @unknown(i32* %p) nounwind readonly 630 true, // call void @unknown(i32* %p) nounwind argmemonly 631 false, // call void @unknown(i32* %p) readonly 632 false, // call void @unknown(i32* %p) argmemonly 633 true, // call void @nounwind_willreturn(i32* %p) 634 false, // ret void 635 }; 636 637 int Index = 0; 638 for (auto &I : BB) { 639 EXPECT_EQ(isGuaranteedToTransferExecutionToSuccessor(&I), 640 ExpectedAnswers[Index]) 641 << "Incorrect answer at instruction " << Index << " = " << I; 642 Index++; 643 } 644 } 645 646 TEST_F(ValueTrackingTest, ComputeNumSignBits_PR32045) { 647 parseAssembly( 648 "define i32 @test(i32 %a) {\n" 649 " %A = ashr i32 %a, -1\n" 650 " ret i32 %A\n" 651 "}\n"); 652 EXPECT_EQ(ComputeNumSignBits(A, M->getDataLayout()), 1u); 653 } 654 655 // No guarantees for canonical IR in this analysis, so this just bails out. 656 TEST_F(ValueTrackingTest, ComputeNumSignBits_Shuffle) { 657 parseAssembly( 658 "define <2 x i32> @test() {\n" 659 " %A = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 0, i32 0>\n" 660 " ret <2 x i32> %A\n" 661 "}\n"); 662 EXPECT_EQ(ComputeNumSignBits(A, M->getDataLayout()), 1u); 663 } 664 665 // No guarantees for canonical IR in this analysis, so a shuffle element that 666 // references an undef value means this can't return any extra information. 667 TEST_F(ValueTrackingTest, ComputeNumSignBits_Shuffle2) { 668 parseAssembly( 669 "define <2 x i32> @test(<2 x i1> %x) {\n" 670 " %sext = sext <2 x i1> %x to <2 x i32>\n" 671 " %A = shufflevector <2 x i32> %sext, <2 x i32> undef, <2 x i32> <i32 0, i32 2>\n" 672 " ret <2 x i32> %A\n" 673 "}\n"); 674 EXPECT_EQ(ComputeNumSignBits(A, M->getDataLayout()), 1u); 675 } 676 677 TEST(ValueTracking, propagatesPoison) { 678 std::string AsmHead = "declare i32 @g(i32)\n" 679 "define void @f(i32 %x, i32 %y, float %fx, float %fy, " 680 "i1 %cond, i8* %p) {\n"; 681 std::string AsmTail = " ret void\n}"; 682 // (propagates poison?, IR instruction) 683 SmallVector<std::pair<bool, std::string>, 32> Data = { 684 {true, "add i32 %x, %y"}, 685 {true, "add nsw nuw i32 %x, %y"}, 686 {true, "ashr i32 %x, %y"}, 687 {true, "lshr exact i32 %x, 31"}, 688 {true, "fcmp oeq float %fx, %fy"}, 689 {true, "icmp eq i32 %x, %y"}, 690 {true, "getelementptr i8, i8* %p, i32 %x"}, 691 {true, "getelementptr inbounds i8, i8* %p, i32 %x"}, 692 {true, "bitcast float %fx to i32"}, 693 {false, "select i1 %cond, i32 %x, i32 %y"}, 694 {false, "freeze i32 %x"}, 695 {true, "udiv i32 %x, %y"}, 696 {true, "urem i32 %x, %y"}, 697 {true, "sdiv exact i32 %x, %y"}, 698 {true, "srem i32 %x, %y"}, 699 {false, "call i32 @g(i32 %x)"}}; 700 701 std::string AssemblyStr = AsmHead; 702 for (auto &Itm : Data) 703 AssemblyStr += Itm.second + "\n"; 704 AssemblyStr += AsmTail; 705 706 LLVMContext Context; 707 SMDiagnostic Error; 708 auto M = parseAssemblyString(AssemblyStr, Error, Context); 709 assert(M && "Bad assembly?"); 710 711 auto *F = M->getFunction("f"); 712 assert(F && "Bad assembly?"); 713 714 auto &BB = F->getEntryBlock(); 715 716 int Index = 0; 717 for (auto &I : BB) { 718 if (isa<ReturnInst>(&I)) 719 break; 720 EXPECT_EQ(propagatesPoison(cast<Operator>(&I)), Data[Index].first) 721 << "Incorrect answer at instruction " << Index << " = " << I; 722 Index++; 723 } 724 } 725 726 TEST_F(ValueTrackingTest, programUndefinedIfPoison) { 727 parseAssembly("declare i32 @any_num()" 728 "define void @test(i32 %mask) {\n" 729 " %A = call i32 @any_num()\n" 730 " %B = or i32 %A, %mask\n" 731 " udiv i32 1, %B" 732 " ret void\n" 733 "}\n"); 734 // If %A was poison, udiv raises UB regardless of %mask's value 735 EXPECT_EQ(programUndefinedIfPoison(A), true); 736 } 737 738 TEST_F(ValueTrackingTest, programUndefinedIfUndefOrPoison) { 739 parseAssembly("declare i32 @any_num()" 740 "define void @test(i32 %mask) {\n" 741 " %A = call i32 @any_num()\n" 742 " %B = or i32 %A, %mask\n" 743 " udiv i32 1, %B" 744 " ret void\n" 745 "}\n"); 746 // If %A was undef and %mask was 1, udiv does not raise UB 747 EXPECT_EQ(programUndefinedIfUndefOrPoison(A), false); 748 } 749 750 TEST_F(ValueTrackingTest, isGuaranteedNotToBePoison_exploitBranchCond) { 751 parseAssembly("declare i1 @any_bool()" 752 "define void @test(i1 %y) {\n" 753 " %A = call i1 @any_bool()\n" 754 " %cond = and i1 %A, %y\n" 755 " br i1 %cond, label %BB1, label %BB2\n" 756 "BB1:\n" 757 " ret void\n" 758 "BB2:\n" 759 " ret void\n" 760 "}\n"); 761 DominatorTree DT(*F); 762 for (auto &BB : *F) { 763 if (&BB == &F->getEntryBlock()) 764 continue; 765 766 EXPECT_EQ(isGuaranteedNotToBePoison(A, BB.getTerminator(), &DT), true) 767 << "isGuaranteedNotToBePoison does not hold at " << *BB.getTerminator(); 768 } 769 } 770 771 TEST_F(ValueTrackingTest, isGuaranteedNotToBePoison_phi) { 772 parseAssembly("declare i32 @any_i32(i32)" 773 "define void @test() {\n" 774 "ENTRY:\n" 775 " br label %LOOP\n" 776 "LOOP:\n" 777 " %A = phi i32 [0, %ENTRY], [%A.next, %NEXT]\n" 778 " %A.next = call i32 @any_i32(i32 %A)\n" 779 " %cond = icmp eq i32 %A.next, 0\n" 780 " br i1 %cond, label %NEXT, label %EXIT\n" 781 "NEXT:\n" 782 " br label %LOOP\n" 783 "EXIT:\n" 784 " ret void\n" 785 "}\n"); 786 DominatorTree DT(*F); 787 for (auto &BB : *F) { 788 if (BB.getName() == "LOOP") { 789 EXPECT_EQ(isGuaranteedNotToBePoison(A, A, &DT), true) 790 << "isGuaranteedNotToBePoison does not hold"; 791 } 792 } 793 } 794 795 TEST_F(ValueTrackingTest, isGuaranteedNotToBeUndefOrPoison) { 796 parseAssembly("declare void @f(i32 noundef)" 797 "define void @test(i32 %x) {\n" 798 " %A = bitcast i32 %x to i32\n" 799 " call void @f(i32 noundef %x)\n" 800 " ret void\n" 801 "}\n"); 802 EXPECT_EQ(isGuaranteedNotToBeUndefOrPoison(A), true); 803 } 804 805 TEST(ValueTracking, canCreatePoisonOrUndef) { 806 std::string AsmHead = 807 "declare i32 @g(i32)\n" 808 "define void @f(i32 %x, i32 %y, float %fx, float %fy, i1 %cond, " 809 "<4 x i32> %vx, <4 x i32> %vx2, <vscale x 4 x i32> %svx, i8* %p) {\n"; 810 std::string AsmTail = " ret void\n}"; 811 // (can create poison?, can create undef?, IR instruction) 812 SmallVector<std::pair<std::pair<bool, bool>, std::string>, 32> Data = { 813 {{false, false}, "add i32 %x, %y"}, 814 {{true, false}, "add nsw nuw i32 %x, %y"}, 815 {{true, false}, "shl i32 %x, %y"}, 816 {{true, false}, "shl <4 x i32> %vx, %vx2"}, 817 {{true, false}, "shl nsw i32 %x, %y"}, 818 {{true, false}, "shl nsw <4 x i32> %vx, <i32 0, i32 1, i32 2, i32 3>"}, 819 {{false, false}, "shl i32 %x, 31"}, 820 {{true, false}, "shl i32 %x, 32"}, 821 {{false, false}, "shl <4 x i32> %vx, <i32 0, i32 1, i32 2, i32 3>"}, 822 {{true, false}, "shl <4 x i32> %vx, <i32 0, i32 1, i32 2, i32 32>"}, 823 {{true, false}, "ashr i32 %x, %y"}, 824 {{true, false}, "ashr exact i32 %x, %y"}, 825 {{false, false}, "ashr i32 %x, 31"}, 826 {{true, false}, "ashr exact i32 %x, 31"}, 827 {{false, false}, "ashr <4 x i32> %vx, <i32 0, i32 1, i32 2, i32 3>"}, 828 {{true, false}, "ashr <4 x i32> %vx, <i32 0, i32 1, i32 2, i32 32>"}, 829 {{true, false}, "ashr exact <4 x i32> %vx, <i32 0, i32 1, i32 2, i32 3>"}, 830 {{true, false}, "lshr i32 %x, %y"}, 831 {{true, false}, "lshr exact i32 %x, 31"}, 832 {{false, false}, "udiv i32 %x, %y"}, 833 {{true, false}, "udiv exact i32 %x, %y"}, 834 {{false, false}, "getelementptr i8, i8* %p, i32 %x"}, 835 {{true, false}, "getelementptr inbounds i8, i8* %p, i32 %x"}, 836 {{true, false}, "fneg nnan float %fx"}, 837 {{false, false}, "fneg float %fx"}, 838 {{false, false}, "fadd float %fx, %fy"}, 839 {{true, false}, "fadd nnan float %fx, %fy"}, 840 {{false, false}, "urem i32 %x, %y"}, 841 {{true, false}, "fptoui float %fx to i32"}, 842 {{true, false}, "fptosi float %fx to i32"}, 843 {{false, false}, "bitcast float %fx to i32"}, 844 {{false, false}, "select i1 %cond, i32 %x, i32 %y"}, 845 {{true, false}, "select nnan i1 %cond, float %fx, float %fy"}, 846 {{true, false}, "extractelement <4 x i32> %vx, i32 %x"}, 847 {{false, false}, "extractelement <4 x i32> %vx, i32 3"}, 848 {{true, false}, "extractelement <vscale x 4 x i32> %svx, i32 4"}, 849 {{true, false}, "insertelement <4 x i32> %vx, i32 %x, i32 %y"}, 850 {{false, false}, "insertelement <4 x i32> %vx, i32 %x, i32 3"}, 851 {{true, false}, "insertelement <vscale x 4 x i32> %svx, i32 %x, i32 4"}, 852 {{false, false}, "freeze i32 %x"}, 853 {{false, false}, 854 "shufflevector <4 x i32> %vx, <4 x i32> %vx2, " 855 "<4 x i32> <i32 0, i32 1, i32 2, i32 3>"}, 856 {{false, true}, 857 "shufflevector <4 x i32> %vx, <4 x i32> %vx2, " 858 "<4 x i32> <i32 0, i32 1, i32 2, i32 undef>"}, 859 {{false, true}, 860 "shufflevector <vscale x 4 x i32> %svx, " 861 "<vscale x 4 x i32> %svx, <vscale x 4 x i32> undef"}, 862 {{true, false}, "call i32 @g(i32 %x)"}, 863 {{false, false}, "call noundef i32 @g(i32 %x)"}, 864 {{true, false}, "fcmp nnan oeq float %fx, %fy"}, 865 {{false, false}, "fcmp oeq float %fx, %fy"}}; 866 867 std::string AssemblyStr = AsmHead; 868 for (auto &Itm : Data) 869 AssemblyStr += Itm.second + "\n"; 870 AssemblyStr += AsmTail; 871 872 LLVMContext Context; 873 SMDiagnostic Error; 874 auto M = parseAssemblyString(AssemblyStr, Error, Context); 875 assert(M && "Bad assembly?"); 876 877 auto *F = M->getFunction("f"); 878 assert(F && "Bad assembly?"); 879 880 auto &BB = F->getEntryBlock(); 881 882 int Index = 0; 883 for (auto &I : BB) { 884 if (isa<ReturnInst>(&I)) 885 break; 886 bool Poison = Data[Index].first.first; 887 bool Undef = Data[Index].first.second; 888 EXPECT_EQ(canCreatePoison(cast<Operator>(&I)), Poison) 889 << "Incorrect answer of canCreatePoison at instruction " << Index 890 << " = " << I; 891 EXPECT_EQ(canCreateUndefOrPoison(cast<Operator>(&I)), Undef || Poison) 892 << "Incorrect answer of canCreateUndef at instruction " << Index 893 << " = " << I; 894 Index++; 895 } 896 } 897 898 TEST_F(ComputeKnownBitsTest, ComputeKnownBits) { 899 parseAssembly( 900 "define i32 @test(i32 %a, i32 %b) {\n" 901 " %ash = mul i32 %a, 8\n" 902 " %aad = add i32 %ash, 7\n" 903 " %aan = and i32 %aad, 4095\n" 904 " %bsh = shl i32 %b, 4\n" 905 " %bad = or i32 %bsh, 6\n" 906 " %ban = and i32 %bad, 4095\n" 907 " %A = mul i32 %aan, %ban\n" 908 " ret i32 %A\n" 909 "}\n"); 910 expectKnownBits(/*zero*/ 4278190085u, /*one*/ 10u); 911 } 912 913 TEST_F(ComputeKnownBitsTest, ComputeKnownMulBits) { 914 parseAssembly( 915 "define i32 @test(i32 %a, i32 %b) {\n" 916 " %aa = shl i32 %a, 5\n" 917 " %bb = shl i32 %b, 5\n" 918 " %aaa = or i32 %aa, 24\n" 919 " %bbb = or i32 %bb, 28\n" 920 " %A = mul i32 %aaa, %bbb\n" 921 " ret i32 %A\n" 922 "}\n"); 923 expectKnownBits(/*zero*/ 95u, /*one*/ 32u); 924 } 925 926 TEST_F(ComputeKnownBitsTest, KnownNonZeroShift) { 927 // %q is known nonzero without known bits. 928 // Because %q is nonzero, %A[0] is known to be zero. 929 parseAssembly( 930 "define i8 @test(i8 %p, i8* %pq) {\n" 931 " %q = load i8, i8* %pq, !range !0\n" 932 " %A = shl i8 %p, %q\n" 933 " ret i8 %A\n" 934 "}\n" 935 "!0 = !{ i8 1, i8 5 }\n"); 936 expectKnownBits(/*zero*/ 1u, /*one*/ 0u); 937 } 938 939 TEST_F(ComputeKnownBitsTest, ComputeKnownFshl) { 940 // fshl(....1111....0000, 00..1111........, 6) 941 // = 11....000000..11 942 parseAssembly( 943 "define i16 @test(i16 %a, i16 %b) {\n" 944 " %aa = shl i16 %a, 4\n" 945 " %bb = lshr i16 %b, 2\n" 946 " %aaa = or i16 %aa, 3840\n" 947 " %bbb = or i16 %bb, 3840\n" 948 " %A = call i16 @llvm.fshl.i16(i16 %aaa, i16 %bbb, i16 6)\n" 949 " ret i16 %A\n" 950 "}\n" 951 "declare i16 @llvm.fshl.i16(i16, i16, i16)\n"); 952 expectKnownBits(/*zero*/ 1008u, /*one*/ 49155u); 953 } 954 955 TEST_F(ComputeKnownBitsTest, ComputeKnownFshr) { 956 // fshr(....1111....0000, 00..1111........, 26) 957 // = 11....000000..11 958 parseAssembly( 959 "define i16 @test(i16 %a, i16 %b) {\n" 960 " %aa = shl i16 %a, 4\n" 961 " %bb = lshr i16 %b, 2\n" 962 " %aaa = or i16 %aa, 3840\n" 963 " %bbb = or i16 %bb, 3840\n" 964 " %A = call i16 @llvm.fshr.i16(i16 %aaa, i16 %bbb, i16 26)\n" 965 " ret i16 %A\n" 966 "}\n" 967 "declare i16 @llvm.fshr.i16(i16, i16, i16)\n"); 968 expectKnownBits(/*zero*/ 1008u, /*one*/ 49155u); 969 } 970 971 TEST_F(ComputeKnownBitsTest, ComputeKnownFshlZero) { 972 // fshl(....1111....0000, 00..1111........, 0) 973 // = ....1111....0000 974 parseAssembly( 975 "define i16 @test(i16 %a, i16 %b) {\n" 976 " %aa = shl i16 %a, 4\n" 977 " %bb = lshr i16 %b, 2\n" 978 " %aaa = or i16 %aa, 3840\n" 979 " %bbb = or i16 %bb, 3840\n" 980 " %A = call i16 @llvm.fshl.i16(i16 %aaa, i16 %bbb, i16 0)\n" 981 " ret i16 %A\n" 982 "}\n" 983 "declare i16 @llvm.fshl.i16(i16, i16, i16)\n"); 984 expectKnownBits(/*zero*/ 15u, /*one*/ 3840u); 985 } 986 987 TEST_F(ComputeKnownBitsTest, ComputeKnownUAddSatLeadingOnes) { 988 // uadd.sat(1111...1, ........) 989 // = 1111.... 990 parseAssembly( 991 "define i8 @test(i8 %a, i8 %b) {\n" 992 " %aa = or i8 %a, 241\n" 993 " %A = call i8 @llvm.uadd.sat.i8(i8 %aa, i8 %b)\n" 994 " ret i8 %A\n" 995 "}\n" 996 "declare i8 @llvm.uadd.sat.i8(i8, i8)\n"); 997 expectKnownBits(/*zero*/ 0u, /*one*/ 240u); 998 } 999 1000 TEST_F(ComputeKnownBitsTest, ComputeKnownUAddSatOnesPreserved) { 1001 // uadd.sat(00...011, .1...110) 1002 // = .......1 1003 parseAssembly( 1004 "define i8 @test(i8 %a, i8 %b) {\n" 1005 " %aa = or i8 %a, 3\n" 1006 " %aaa = and i8 %aa, 59\n" 1007 " %bb = or i8 %b, 70\n" 1008 " %bbb = and i8 %bb, 254\n" 1009 " %A = call i8 @llvm.uadd.sat.i8(i8 %aaa, i8 %bbb)\n" 1010 " ret i8 %A\n" 1011 "}\n" 1012 "declare i8 @llvm.uadd.sat.i8(i8, i8)\n"); 1013 expectKnownBits(/*zero*/ 0u, /*one*/ 1u); 1014 } 1015 1016 TEST_F(ComputeKnownBitsTest, ComputeKnownUSubSatLHSLeadingZeros) { 1017 // usub.sat(0000...0, ........) 1018 // = 0000.... 1019 parseAssembly( 1020 "define i8 @test(i8 %a, i8 %b) {\n" 1021 " %aa = and i8 %a, 14\n" 1022 " %A = call i8 @llvm.usub.sat.i8(i8 %aa, i8 %b)\n" 1023 " ret i8 %A\n" 1024 "}\n" 1025 "declare i8 @llvm.usub.sat.i8(i8, i8)\n"); 1026 expectKnownBits(/*zero*/ 240u, /*one*/ 0u); 1027 } 1028 1029 TEST_F(ComputeKnownBitsTest, ComputeKnownUSubSatRHSLeadingOnes) { 1030 // usub.sat(........, 1111...1) 1031 // = 0000.... 1032 parseAssembly( 1033 "define i8 @test(i8 %a, i8 %b) {\n" 1034 " %bb = or i8 %a, 241\n" 1035 " %A = call i8 @llvm.usub.sat.i8(i8 %a, i8 %bb)\n" 1036 " ret i8 %A\n" 1037 "}\n" 1038 "declare i8 @llvm.usub.sat.i8(i8, i8)\n"); 1039 expectKnownBits(/*zero*/ 240u, /*one*/ 0u); 1040 } 1041 1042 TEST_F(ComputeKnownBitsTest, ComputeKnownUSubSatZerosPreserved) { 1043 // usub.sat(11...011, .1...110) 1044 // = ......0. 1045 parseAssembly( 1046 "define i8 @test(i8 %a, i8 %b) {\n" 1047 " %aa = or i8 %a, 195\n" 1048 " %aaa = and i8 %aa, 251\n" 1049 " %bb = or i8 %b, 70\n" 1050 " %bbb = and i8 %bb, 254\n" 1051 " %A = call i8 @llvm.usub.sat.i8(i8 %aaa, i8 %bbb)\n" 1052 " ret i8 %A\n" 1053 "}\n" 1054 "declare i8 @llvm.usub.sat.i8(i8, i8)\n"); 1055 expectKnownBits(/*zero*/ 2u, /*one*/ 0u); 1056 } 1057 1058 TEST_F(ComputeKnownBitsTest, ComputeKnownBitsPtrToIntTrunc) { 1059 // ptrtoint truncates the pointer type. 1060 parseAssembly( 1061 "define void @test(i8** %p) {\n" 1062 " %A = load i8*, i8** %p\n" 1063 " %i = ptrtoint i8* %A to i32\n" 1064 " %m = and i32 %i, 31\n" 1065 " %c = icmp eq i32 %m, 0\n" 1066 " call void @llvm.assume(i1 %c)\n" 1067 " ret void\n" 1068 "}\n" 1069 "declare void @llvm.assume(i1)\n"); 1070 AssumptionCache AC(*F); 1071 KnownBits Known = computeKnownBits( 1072 A, M->getDataLayout(), /* Depth */ 0, &AC, F->front().getTerminator()); 1073 EXPECT_EQ(Known.Zero.getZExtValue(), 31u); 1074 EXPECT_EQ(Known.One.getZExtValue(), 0u); 1075 } 1076 1077 TEST_F(ComputeKnownBitsTest, ComputeKnownBitsPtrToIntZext) { 1078 // ptrtoint zero extends the pointer type. 1079 parseAssembly( 1080 "define void @test(i8** %p) {\n" 1081 " %A = load i8*, i8** %p\n" 1082 " %i = ptrtoint i8* %A to i128\n" 1083 " %m = and i128 %i, 31\n" 1084 " %c = icmp eq i128 %m, 0\n" 1085 " call void @llvm.assume(i1 %c)\n" 1086 " ret void\n" 1087 "}\n" 1088 "declare void @llvm.assume(i1)\n"); 1089 AssumptionCache AC(*F); 1090 KnownBits Known = computeKnownBits( 1091 A, M->getDataLayout(), /* Depth */ 0, &AC, F->front().getTerminator()); 1092 EXPECT_EQ(Known.Zero.getZExtValue(), 31u); 1093 EXPECT_EQ(Known.One.getZExtValue(), 0u); 1094 } 1095 1096 TEST_F(ComputeKnownBitsTest, ComputeKnownBitsFreeze) { 1097 parseAssembly("define void @test() {\n" 1098 " %m = call i32 @any_num()\n" 1099 " %A = freeze i32 %m\n" 1100 " %n = and i32 %m, 31\n" 1101 " %c = icmp eq i32 %n, 0\n" 1102 " call void @llvm.assume(i1 %c)\n" 1103 " ret void\n" 1104 "}\n" 1105 "declare void @llvm.assume(i1)\n" 1106 "declare i32 @any_num()\n"); 1107 AssumptionCache AC(*F); 1108 KnownBits Known = computeKnownBits(A, M->getDataLayout(), /* Depth */ 0, &AC, 1109 F->front().getTerminator()); 1110 EXPECT_EQ(Known.Zero.getZExtValue(), 31u); 1111 EXPECT_EQ(Known.One.getZExtValue(), 0u); 1112 } 1113 1114 TEST_F(ComputeKnownBitsTest, ComputeKnownBitsAddWithRange) { 1115 parseAssembly("define void @test(i64* %p) {\n" 1116 " %A = load i64, i64* %p, !range !{i64 64, i64 65536}\n" 1117 " %APlus512 = add i64 %A, 512\n" 1118 " %c = icmp ugt i64 %APlus512, 523\n" 1119 " call void @llvm.assume(i1 %c)\n" 1120 " ret void\n" 1121 "}\n" 1122 "declare void @llvm.assume(i1)\n"); 1123 AssumptionCache AC(*F); 1124 KnownBits Known = computeKnownBits(A, M->getDataLayout(), /* Depth */ 0, &AC, 1125 F->front().getTerminator()); 1126 EXPECT_EQ(Known.Zero.getZExtValue(), ~(65536llu - 1)); 1127 EXPECT_EQ(Known.One.getZExtValue(), 0u); 1128 Instruction &APlus512 = findInstructionByName(F, "APlus512"); 1129 Known = computeKnownBits(&APlus512, M->getDataLayout(), /* Depth */ 0, &AC, 1130 F->front().getTerminator()); 1131 // We know of one less zero because 512 may have produced a 1 that 1132 // got carried all the way to the first trailing zero. 1133 EXPECT_EQ(Known.Zero.getZExtValue(), (~(65536llu - 1)) << 1); 1134 EXPECT_EQ(Known.One.getZExtValue(), 0u); 1135 // The known range is not precise given computeKnownBits works 1136 // with the masks of zeros and ones, not the ranges. 1137 EXPECT_EQ(Known.getMinValue(), 0u); 1138 EXPECT_EQ(Known.getMaxValue(), 131071); 1139 } 1140 1141 // 512 + [32, 64) doesn't produce overlapping bits. 1142 // Make sure we get all the individual bits properly. 1143 TEST_F(ComputeKnownBitsTest, ComputeKnownBitsAddWithRangeNoOverlap) { 1144 parseAssembly("define void @test(i64* %p) {\n" 1145 " %A = load i64, i64* %p, !range !{i64 32, i64 64}\n" 1146 " %APlus512 = add i64 %A, 512\n" 1147 " %c = icmp ugt i64 %APlus512, 523\n" 1148 " call void @llvm.assume(i1 %c)\n" 1149 " ret void\n" 1150 "}\n" 1151 "declare void @llvm.assume(i1)\n"); 1152 AssumptionCache AC(*F); 1153 KnownBits Known = computeKnownBits(A, M->getDataLayout(), /* Depth */ 0, &AC, 1154 F->front().getTerminator()); 1155 EXPECT_EQ(Known.Zero.getZExtValue(), ~(64llu - 1)); 1156 EXPECT_EQ(Known.One.getZExtValue(), 32u); 1157 Instruction &APlus512 = findInstructionByName(F, "APlus512"); 1158 Known = computeKnownBits(&APlus512, M->getDataLayout(), /* Depth */ 0, &AC, 1159 F->front().getTerminator()); 1160 EXPECT_EQ(Known.Zero.getZExtValue(), ~512llu & ~(64llu - 1)); 1161 EXPECT_EQ(Known.One.getZExtValue(), 512u | 32u); 1162 // The known range is not precise given computeKnownBits works 1163 // with the masks of zeros and ones, not the ranges. 1164 EXPECT_EQ(Known.getMinValue(), 544); 1165 EXPECT_EQ(Known.getMaxValue(), 575); 1166 } 1167 1168 class IsBytewiseValueTest : public ValueTrackingTest, 1169 public ::testing::WithParamInterface< 1170 std::pair<const char *, const char *>> { 1171 protected: 1172 }; 1173 1174 const std::pair<const char *, const char *> IsBytewiseValueTests[] = { 1175 { 1176 "i8 0", 1177 "i48* null", 1178 }, 1179 { 1180 "i8 undef", 1181 "i48* undef", 1182 }, 1183 { 1184 "i8 0", 1185 "i8 zeroinitializer", 1186 }, 1187 { 1188 "i8 0", 1189 "i8 0", 1190 }, 1191 { 1192 "i8 -86", 1193 "i8 -86", 1194 }, 1195 { 1196 "i8 -1", 1197 "i8 -1", 1198 }, 1199 { 1200 "i8 undef", 1201 "i16 undef", 1202 }, 1203 { 1204 "i8 0", 1205 "i16 0", 1206 }, 1207 { 1208 "", 1209 "i16 7", 1210 }, 1211 { 1212 "i8 -86", 1213 "i16 -21846", 1214 }, 1215 { 1216 "i8 -1", 1217 "i16 -1", 1218 }, 1219 { 1220 "i8 0", 1221 "i48 0", 1222 }, 1223 { 1224 "i8 -1", 1225 "i48 -1", 1226 }, 1227 { 1228 "i8 0", 1229 "i49 0", 1230 }, 1231 { 1232 "", 1233 "i49 -1", 1234 }, 1235 { 1236 "i8 0", 1237 "half 0xH0000", 1238 }, 1239 { 1240 "i8 -85", 1241 "half 0xHABAB", 1242 }, 1243 { 1244 "i8 0", 1245 "float 0.0", 1246 }, 1247 { 1248 "i8 -1", 1249 "float 0xFFFFFFFFE0000000", 1250 }, 1251 { 1252 "i8 0", 1253 "double 0.0", 1254 }, 1255 { 1256 "i8 -15", 1257 "double 0xF1F1F1F1F1F1F1F1", 1258 }, 1259 { 1260 "i8 undef", 1261 "i16* undef", 1262 }, 1263 { 1264 "i8 0", 1265 "i16* inttoptr (i64 0 to i16*)", 1266 }, 1267 { 1268 "i8 -1", 1269 "i16* inttoptr (i64 -1 to i16*)", 1270 }, 1271 { 1272 "i8 -86", 1273 "i16* inttoptr (i64 -6148914691236517206 to i16*)", 1274 }, 1275 { 1276 "", 1277 "i16* inttoptr (i48 -1 to i16*)", 1278 }, 1279 { 1280 "i8 -1", 1281 "i16* inttoptr (i96 -1 to i16*)", 1282 }, 1283 { 1284 "i8 undef", 1285 "[0 x i8] zeroinitializer", 1286 }, 1287 { 1288 "i8 undef", 1289 "[0 x i8] undef", 1290 }, 1291 { 1292 "i8 undef", 1293 "[5 x [0 x i8]] zeroinitializer", 1294 }, 1295 { 1296 "i8 undef", 1297 "[5 x [0 x i8]] undef", 1298 }, 1299 { 1300 "i8 0", 1301 "[6 x i8] zeroinitializer", 1302 }, 1303 { 1304 "i8 undef", 1305 "[6 x i8] undef", 1306 }, 1307 { 1308 "i8 1", 1309 "[5 x i8] [i8 1, i8 1, i8 1, i8 1, i8 1]", 1310 }, 1311 { 1312 "", 1313 "[5 x i64] [i64 1, i64 1, i64 1, i64 1, i64 1]", 1314 }, 1315 { 1316 "i8 -1", 1317 "[5 x i64] [i64 -1, i64 -1, i64 -1, i64 -1, i64 -1]", 1318 }, 1319 { 1320 "", 1321 "[4 x i8] [i8 1, i8 2, i8 1, i8 1]", 1322 }, 1323 { 1324 "i8 1", 1325 "[4 x i8] [i8 1, i8 undef, i8 1, i8 1]", 1326 }, 1327 { 1328 "i8 0", 1329 "<6 x i8> zeroinitializer", 1330 }, 1331 { 1332 "i8 undef", 1333 "<6 x i8> undef", 1334 }, 1335 { 1336 "i8 1", 1337 "<5 x i8> <i8 1, i8 1, i8 1, i8 1, i8 1>", 1338 }, 1339 { 1340 "", 1341 "<5 x i64> <i64 1, i64 1, i64 1, i64 1, i64 1>", 1342 }, 1343 { 1344 "i8 -1", 1345 "<5 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1>", 1346 }, 1347 { 1348 "", 1349 "<4 x i8> <i8 1, i8 1, i8 2, i8 1>", 1350 }, 1351 { 1352 "i8 5", 1353 "<2 x i8> < i8 5, i8 undef >", 1354 }, 1355 { 1356 "i8 0", 1357 "[2 x [2 x i16]] zeroinitializer", 1358 }, 1359 { 1360 "i8 undef", 1361 "[2 x [2 x i16]] undef", 1362 }, 1363 { 1364 "i8 -86", 1365 "[2 x [2 x i16]] [[2 x i16] [i16 -21846, i16 -21846], " 1366 "[2 x i16] [i16 -21846, i16 -21846]]", 1367 }, 1368 { 1369 "", 1370 "[2 x [2 x i16]] [[2 x i16] [i16 -21846, i16 -21846], " 1371 "[2 x i16] [i16 -21836, i16 -21846]]", 1372 }, 1373 { 1374 "i8 undef", 1375 "{ } zeroinitializer", 1376 }, 1377 { 1378 "i8 undef", 1379 "{ } undef", 1380 }, 1381 { 1382 "i8 undef", 1383 "{ {}, {} } zeroinitializer", 1384 }, 1385 { 1386 "i8 undef", 1387 "{ {}, {} } undef", 1388 }, 1389 { 1390 "i8 0", 1391 "{i8, i64, i16*} zeroinitializer", 1392 }, 1393 { 1394 "i8 undef", 1395 "{i8, i64, i16*} undef", 1396 }, 1397 { 1398 "i8 -86", 1399 "{i8, i64, i16*} {i8 -86, i64 -6148914691236517206, i16* undef}", 1400 }, 1401 { 1402 "", 1403 "{i8, i64, i16*} {i8 86, i64 -6148914691236517206, i16* undef}", 1404 }, 1405 }; 1406 1407 INSTANTIATE_TEST_CASE_P(IsBytewiseValueParamTests, IsBytewiseValueTest, 1408 ::testing::ValuesIn(IsBytewiseValueTests),); 1409 1410 TEST_P(IsBytewiseValueTest, IsBytewiseValue) { 1411 auto M = parseModule(std::string("@test = global ") + GetParam().second); 1412 GlobalVariable *GV = dyn_cast<GlobalVariable>(M->getNamedValue("test")); 1413 Value *Actual = isBytewiseValue(GV->getInitializer(), M->getDataLayout()); 1414 std::string Buff; 1415 raw_string_ostream S(Buff); 1416 if (Actual) 1417 S << *Actual; 1418 EXPECT_EQ(GetParam().first, S.str()); 1419 } 1420 1421 TEST_F(ValueTrackingTest, ComputeConstantRange) { 1422 { 1423 // Assumptions: 1424 // * stride >= 5 1425 // * stride < 10 1426 // 1427 // stride = [5, 10) 1428 auto M = parseModule(R"( 1429 declare void @llvm.assume(i1) 1430 1431 define i32 @test(i32 %stride) { 1432 %gt = icmp uge i32 %stride, 5 1433 call void @llvm.assume(i1 %gt) 1434 %lt = icmp ult i32 %stride, 10 1435 call void @llvm.assume(i1 %lt) 1436 %stride.plus.one = add nsw nuw i32 %stride, 1 1437 ret i32 %stride.plus.one 1438 })"); 1439 Function *F = M->getFunction("test"); 1440 1441 AssumptionCache AC(*F); 1442 Value *Stride = &*F->arg_begin(); 1443 ConstantRange CR1 = computeConstantRange(Stride, true, &AC, nullptr); 1444 EXPECT_TRUE(CR1.isFullSet()); 1445 1446 Instruction *I = &findInstructionByName(F, "stride.plus.one"); 1447 ConstantRange CR2 = computeConstantRange(Stride, true, &AC, I); 1448 EXPECT_EQ(5, CR2.getLower()); 1449 EXPECT_EQ(10, CR2.getUpper()); 1450 } 1451 1452 { 1453 // Assumptions: 1454 // * stride >= 5 1455 // * stride < 200 1456 // * stride == 99 1457 // 1458 // stride = [99, 100) 1459 auto M = parseModule(R"( 1460 declare void @llvm.assume(i1) 1461 1462 define i32 @test(i32 %stride) { 1463 %gt = icmp uge i32 %stride, 5 1464 call void @llvm.assume(i1 %gt) 1465 %lt = icmp ult i32 %stride, 200 1466 call void @llvm.assume(i1 %lt) 1467 %eq = icmp eq i32 %stride, 99 1468 call void @llvm.assume(i1 %eq) 1469 %stride.plus.one = add nsw nuw i32 %stride, 1 1470 ret i32 %stride.plus.one 1471 })"); 1472 Function *F = M->getFunction("test"); 1473 1474 AssumptionCache AC(*F); 1475 Value *Stride = &*F->arg_begin(); 1476 Instruction *I = &findInstructionByName(F, "stride.plus.one"); 1477 ConstantRange CR = computeConstantRange(Stride, true, &AC, I); 1478 EXPECT_EQ(99, *CR.getSingleElement()); 1479 } 1480 1481 { 1482 // Assumptions: 1483 // * stride >= 5 1484 // * stride >= 50 1485 // * stride < 100 1486 // * stride < 200 1487 // 1488 // stride = [50, 100) 1489 auto M = parseModule(R"( 1490 declare void @llvm.assume(i1) 1491 1492 define i32 @test(i32 %stride, i1 %cond) { 1493 %gt = icmp uge i32 %stride, 5 1494 call void @llvm.assume(i1 %gt) 1495 %gt.2 = icmp uge i32 %stride, 50 1496 call void @llvm.assume(i1 %gt.2) 1497 br i1 %cond, label %bb1, label %bb2 1498 1499 bb1: 1500 %lt = icmp ult i32 %stride, 200 1501 call void @llvm.assume(i1 %lt) 1502 %lt.2 = icmp ult i32 %stride, 100 1503 call void @llvm.assume(i1 %lt.2) 1504 %stride.plus.one = add nsw nuw i32 %stride, 1 1505 ret i32 %stride.plus.one 1506 1507 bb2: 1508 ret i32 0 1509 })"); 1510 Function *F = M->getFunction("test"); 1511 1512 AssumptionCache AC(*F); 1513 Value *Stride = &*F->arg_begin(); 1514 Instruction *GT2 = &findInstructionByName(F, "gt.2"); 1515 ConstantRange CR = computeConstantRange(Stride, true, &AC, GT2); 1516 EXPECT_EQ(5, CR.getLower()); 1517 EXPECT_EQ(0, CR.getUpper()); 1518 1519 Instruction *I = &findInstructionByName(F, "stride.plus.one"); 1520 ConstantRange CR2 = computeConstantRange(Stride, true, &AC, I); 1521 EXPECT_EQ(50, CR2.getLower()); 1522 EXPECT_EQ(100, CR2.getUpper()); 1523 } 1524 1525 { 1526 // Assumptions: 1527 // * stride > 5 1528 // * stride < 5 1529 // 1530 // stride = empty range, as the assumptions contradict each other. 1531 auto M = parseModule(R"( 1532 declare void @llvm.assume(i1) 1533 1534 define i32 @test(i32 %stride, i1 %cond) { 1535 %gt = icmp ugt i32 %stride, 5 1536 call void @llvm.assume(i1 %gt) 1537 %lt = icmp ult i32 %stride, 5 1538 call void @llvm.assume(i1 %lt) 1539 %stride.plus.one = add nsw nuw i32 %stride, 1 1540 ret i32 %stride.plus.one 1541 })"); 1542 Function *F = M->getFunction("test"); 1543 1544 AssumptionCache AC(*F); 1545 Value *Stride = &*F->arg_begin(); 1546 1547 Instruction *I = &findInstructionByName(F, "stride.plus.one"); 1548 ConstantRange CR = computeConstantRange(Stride, true, &AC, I); 1549 EXPECT_TRUE(CR.isEmptySet()); 1550 } 1551 1552 { 1553 // Assumptions: 1554 // * x.1 >= 5 1555 // * x.2 < x.1 1556 // 1557 // stride = [0, 5) 1558 auto M = parseModule(R"( 1559 declare void @llvm.assume(i1) 1560 1561 define i32 @test(i32 %x.1, i32 %x.2) { 1562 %gt = icmp uge i32 %x.1, 5 1563 call void @llvm.assume(i1 %gt) 1564 %lt = icmp ult i32 %x.2, %x.1 1565 call void @llvm.assume(i1 %lt) 1566 %stride.plus.one = add nsw nuw i32 %x.1, 1 1567 ret i32 %stride.plus.one 1568 })"); 1569 Function *F = M->getFunction("test"); 1570 1571 AssumptionCache AC(*F); 1572 Value *X2 = &*std::next(F->arg_begin()); 1573 1574 Instruction *I = &findInstructionByName(F, "stride.plus.one"); 1575 ConstantRange CR1 = computeConstantRange(X2, true, &AC, I); 1576 EXPECT_EQ(0, CR1.getLower()); 1577 EXPECT_EQ(5, CR1.getUpper()); 1578 1579 // Check the depth cutoff results in a conservative result (full set) by 1580 // passing Depth == MaxDepth == 6. 1581 ConstantRange CR2 = computeConstantRange(X2, true, &AC, I, 6); 1582 EXPECT_TRUE(CR2.isFullSet()); 1583 } 1584 } 1585 1586 struct FindAllocaForValueTestParams { 1587 const char *IR; 1588 bool AnyOffsetResult; 1589 bool ZeroOffsetResult; 1590 }; 1591 1592 class FindAllocaForValueTest 1593 : public ValueTrackingTest, 1594 public ::testing::WithParamInterface<FindAllocaForValueTestParams> { 1595 protected: 1596 }; 1597 1598 const FindAllocaForValueTestParams FindAllocaForValueTests[] = { 1599 {R"( 1600 define void @test() { 1601 %a = alloca i64 1602 %r = bitcast i64* %a to i32* 1603 ret void 1604 })", 1605 true, true}, 1606 1607 {R"( 1608 define void @test() { 1609 %a = alloca i32 1610 %r = getelementptr i32, i32* %a, i32 1 1611 ret void 1612 })", 1613 true, false}, 1614 1615 {R"( 1616 define void @test() { 1617 %a = alloca i32 1618 %r = getelementptr i32, i32* %a, i32 0 1619 ret void 1620 })", 1621 true, true}, 1622 1623 {R"( 1624 define void @test(i1 %cond) { 1625 entry: 1626 %a = alloca i32 1627 br label %bb1 1628 1629 bb1: 1630 %r = phi i32* [ %a, %entry ], [ %r, %bb1 ] 1631 br i1 %cond, label %bb1, label %exit 1632 1633 exit: 1634 ret void 1635 })", 1636 true, true}, 1637 1638 {R"( 1639 define void @test(i1 %cond) { 1640 %a = alloca i32 1641 %r = select i1 %cond, i32* %a, i32* %a 1642 ret void 1643 })", 1644 true, true}, 1645 1646 {R"( 1647 define void @test(i1 %cond) { 1648 %a = alloca i32 1649 %b = alloca i32 1650 %r = select i1 %cond, i32* %a, i32* %b 1651 ret void 1652 })", 1653 false, false}, 1654 1655 {R"( 1656 define void @test(i1 %cond) { 1657 entry: 1658 %a = alloca i64 1659 %a32 = bitcast i64* %a to i32* 1660 br label %bb1 1661 1662 bb1: 1663 %x = phi i32* [ %a32, %entry ], [ %x, %bb1 ] 1664 %r = getelementptr i32, i32* %x, i32 1 1665 br i1 %cond, label %bb1, label %exit 1666 1667 exit: 1668 ret void 1669 })", 1670 true, false}, 1671 1672 {R"( 1673 define void @test(i1 %cond) { 1674 entry: 1675 %a = alloca i64 1676 %a32 = bitcast i64* %a to i32* 1677 br label %bb1 1678 1679 bb1: 1680 %x = phi i32* [ %a32, %entry ], [ %r, %bb1 ] 1681 %r = getelementptr i32, i32* %x, i32 1 1682 br i1 %cond, label %bb1, label %exit 1683 1684 exit: 1685 ret void 1686 })", 1687 true, false}, 1688 1689 {R"( 1690 define void @test(i1 %cond, i64* %a) { 1691 entry: 1692 %r = bitcast i64* %a to i32* 1693 ret void 1694 })", 1695 false, false}, 1696 1697 {R"( 1698 define void @test(i1 %cond) { 1699 entry: 1700 %a = alloca i32 1701 %b = alloca i32 1702 br label %bb1 1703 1704 bb1: 1705 %r = phi i32* [ %a, %entry ], [ %b, %bb1 ] 1706 br i1 %cond, label %bb1, label %exit 1707 1708 exit: 1709 ret void 1710 })", 1711 false, false}, 1712 }; 1713 1714 TEST_P(FindAllocaForValueTest, findAllocaForValue) { 1715 auto M = parseModule(GetParam().IR); 1716 Function *F = M->getFunction("test"); 1717 Instruction *I = &findInstructionByName(F, "r"); 1718 const AllocaInst *AI = findAllocaForValue(I); 1719 EXPECT_EQ(!!AI, GetParam().AnyOffsetResult); 1720 } 1721 1722 TEST_P(FindAllocaForValueTest, findAllocaForValueZeroOffset) { 1723 auto M = parseModule(GetParam().IR); 1724 Function *F = M->getFunction("test"); 1725 Instruction *I = &findInstructionByName(F, "r"); 1726 const AllocaInst *AI = findAllocaForValue(I, true); 1727 EXPECT_EQ(!!AI, GetParam().ZeroOffsetResult); 1728 } 1729 1730 INSTANTIATE_TEST_CASE_P(FindAllocaForValueTest, FindAllocaForValueTest, 1731 ::testing::ValuesIn(FindAllocaForValueTests), ); 1732