guards.ll 15.4 KB
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528
; RUN: opt -S -early-cse < %s | FileCheck %s
; RUN: opt < %s -S -basicaa -early-cse-memssa | FileCheck %s

declare void @llvm.experimental.guard(i1,...)

declare void @llvm.assume(i1)

define i32 @test0(i32* %ptr, i1 %cond) {
; We can do store to load forwarding over a guard, since it does not
; clobber memory

; CHECK-LABEL: @test0(
; CHECK-NEXT:  store i32 40, i32* %ptr
; CHECK-NEXT:  call void (i1, ...) @llvm.experimental.guard(i1 %cond) [ "deopt"() ]
; CHECK-NEXT:  ret i32 40

  store i32 40, i32* %ptr
  call void(i1,...) @llvm.experimental.guard(i1 %cond) [ "deopt"() ]
  %rval = load i32, i32* %ptr
  ret i32 %rval
}

define i32 @test1(i32* %val, i1 %cond) {
; We can CSE loads over a guard, since it does not clobber memory

; CHECK-LABEL: @test1(
; CHECK-NEXT:  %val0 = load i32, i32* %val
; CHECK-NEXT:  call void (i1, ...) @llvm.experimental.guard(i1 %cond) [ "deopt"() ]
; CHECK-NEXT:  ret i32 0

  %val0 = load i32, i32* %val
  call void(i1,...) @llvm.experimental.guard(i1 %cond) [ "deopt"() ]
  %val1 = load i32, i32* %val
  %rval = sub i32 %val0, %val1
  ret i32 %rval
}

define i32 @test2() {
; Guards on "true" get removed

; CHECK-LABEL: @test2(
; CHECK-NEXT: ret i32 0
  call void(i1, ...) @llvm.experimental.guard(i1 true) [ "deopt"() ]
  ret i32 0
}

define i32 @test3(i32 %val) {
; After a guard has executed the condition it was guarding is known to
; be true.

; CHECK-LABEL: @test3(
; CHECK-NEXT:  %cond0 = icmp slt i32 %val, 40
; CHECK-NEXT:  call void (i1, ...) @llvm.experimental.guard(i1 %cond0) [ "deopt"() ]
; CHECK-NEXT:  ret i32 -1

  %cond0 = icmp slt i32 %val, 40
  call void(i1,...) @llvm.experimental.guard(i1 %cond0) [ "deopt"() ]
  %cond1 = icmp slt i32 %val, 40
  call void(i1,...) @llvm.experimental.guard(i1 %cond1) [ "deopt"() ]

  %cond2 = icmp slt i32 %val, 40
  %rval = sext i1 %cond2 to i32
  ret i32 %rval
}

define i32 @test3.unhandled(i32 %val) {
; After a guard has executed the condition it was guarding is known to
; be true.

; CHECK-LABEL: @test3.unhandled(
; CHECK-NEXT:  %cond0 = icmp slt i32 %val, 40
; CHECK-NEXT:  call void (i1, ...) @llvm.experimental.guard(i1 %cond0) [ "deopt"() ]
; CHECK-NEXT:  %cond1 = icmp sge i32 %val, 40
; CHECK-NEXT:  call void (i1, ...) @llvm.experimental.guard(i1 %cond1) [ "deopt"() ]
; CHECK-NEXT:  ret i32 0

; Demonstrates a case we do not yet handle (it is legal to fold %cond2
; to false)
  %cond0 = icmp slt i32 %val, 40
  call void(i1,...) @llvm.experimental.guard(i1 %cond0) [ "deopt"() ]
  %cond1 = icmp sge i32 %val, 40
  call void(i1,...) @llvm.experimental.guard(i1 %cond1) [ "deopt"() ]
  ret i32 0
}

define i32 @test4(i32 %val, i1 %c) {
; Same as test3, but with some control flow involved.

; CHECK-LABEL: @test4(
; CHECK: entry:
; CHECK-NEXT:  %cond0 = icmp slt i32 %val, 40
; CHECK-NEXT:  call void (i1, ...) @llvm.experimental.guard(i1 %cond0
; CHECK-NEXT:  br label %bb0

; CHECK:     bb0:
; CHECK-NEXT:  %cond2 = icmp ult i32 %val, 200
; CHECK-NEXT:  call void (i1, ...) @llvm.experimental.guard(i1 %cond2
; CHECK-NEXT:  br i1 %c, label %left, label %right

; CHECK:     left:
; CHECK-NEXT:  ret i32 0

; CHECK:     right:
; CHECK-NEXT:  ret i32 20

entry:
  %cond0 = icmp slt i32 %val, 40
  call void(i1,...) @llvm.experimental.guard(i1 %cond0) [ "deopt"() ]
  %cond1 = icmp slt i32 %val, 40
  call void(i1,...) @llvm.experimental.guard(i1 %cond1) [ "deopt"() ]
  br label %bb0

bb0:
  %cond2 = icmp ult i32 %val, 200
  call void(i1,...) @llvm.experimental.guard(i1 %cond2) [ "deopt"() ]
  br i1 %c, label %left, label %right

left:
  %cond3 = icmp ult i32 %val, 200
  call void(i1,...) @llvm.experimental.guard(i1 %cond3) [ "deopt"() ]
  ret i32 0

right:
 ret i32 20
}

define i32 @test5(i32 %val, i1 %c) {
; Same as test4, but the %left block has mutliple predecessors.

; CHECK-LABEL: @test5(

; CHECK: entry:
; CHECK-NEXT:  %cond0 = icmp slt i32 %val, 40
; CHECK-NEXT:  call void (i1, ...) @llvm.experimental.guard(i1 %cond0
; CHECK-NEXT:  br label %bb0

; CHECK: bb0:
; CHECK-NEXT:  %cond2 = icmp ult i32 %val, 200
; CHECK-NEXT:  call void (i1, ...) @llvm.experimental.guard(i1 %cond2
; CHECK-NEXT:  br i1 %c, label %left, label %right

; CHECK: left:
; CHECK-NEXT:  br label %right

; CHECK: right:
; CHECK-NEXT:  br label %left

entry:
  %cond0 = icmp slt i32 %val, 40
  call void(i1,...) @llvm.experimental.guard(i1 %cond0) [ "deopt"() ]
  %cond1 = icmp slt i32 %val, 40
  call void(i1,...) @llvm.experimental.guard(i1 %cond1) [ "deopt"() ]
  br label %bb0

bb0:
  %cond2 = icmp ult i32 %val, 200
  call void(i1,...) @llvm.experimental.guard(i1 %cond2) [ "deopt"() ]
  br i1 %c, label %left, label %right

left:
  %cond3 = icmp ult i32 %val, 200
  call void(i1,...) @llvm.experimental.guard(i1 %cond3) [ "deopt"() ]
  br label %right

right:
  br label %left
}

define void @test6(i1 %c, i32* %ptr) {
; Check that we do not DSE over calls to @llvm.experimental.guard.
; Guard intrinsics do _read_ memory, so th call to guard below needs
; to see the store of 500 to %ptr

; CHECK-LABEL: @test6(
; CHECK-NEXT:  store i32 500, i32* %ptr
; CHECK-NEXT:  call void (i1, ...) @llvm.experimental.guard(i1 %c) [ "deopt"() ]
; CHECK-NEXT:  store i32 600, i32* %ptr


  store i32 500, i32* %ptr
  call void(i1,...) @llvm.experimental.guard(i1 %c) [ "deopt"() ]
  store i32 600, i32* %ptr
  ret void
}

define void @test07(i32 %a, i32 %b) {
; Check that we are able to remove the guards on the same condition even if the
; condition is not being recalculated.

; CHECK-LABEL: @test07(
; CHECK-NEXT:  %cmp = icmp eq i32 %a, %b
; CHECK-NEXT:  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
; CHECK-NEXT:  ret void

  %cmp = icmp eq i32 %a, %b
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  ret void
}

define void @test08(i32 %a, i32 %b, i32* %ptr) {
; Check that we deal correctly with stores when removing guards in the same
; block in case when the condition is not recalculated.

; CHECK-LABEL: @test08(
; CHECK-NEXT:  %cmp = icmp eq i32 %a, %b
; CHECK-NEXT:  store i32 100, i32* %ptr
; CHECK-NEXT:  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
; CHECK-NEXT:  store i32 400, i32* %ptr
; CHECK-NEXT:  ret void

  %cmp = icmp eq i32 %a, %b
  store i32 100, i32* %ptr
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  store i32 200, i32* %ptr
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  store i32 300, i32* %ptr
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  store i32 400, i32* %ptr
  ret void
}

define void @test09(i32 %a, i32 %b, i1 %c, i32* %ptr) {
; Similar to test08, but with more control flow.
; TODO: Can we get rid of the store in the end of entry given that it is
; post-dominated by other stores?

; CHECK-LABEL: @test09(
; CHECK:       entry:
; CHECK-NEXT:    %cmp = icmp eq i32 %a, %b
; CHECK-NEXT:    store i32 100, i32* %ptr
; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
; CHECK-NEXT:    store i32 400, i32* %ptr
; CHECK-NEXT:    br i1 %c, label %if.true, label %if.false
; CHECK:       if.true:
; CHECK-NEXT:    store i32 500, i32* %ptr
; CHECK-NEXT:    br label %merge
; CHECK:       if.false:
; CHECK-NEXT:    store i32 600, i32* %ptr
; CHECK-NEXT:    br label %merge
; CHECK:       merge:
; CHECK-NEXT:    ret void

entry:
  %cmp = icmp eq i32 %a, %b
  store i32 100, i32* %ptr
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  store i32 200, i32* %ptr
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  store i32 300, i32* %ptr
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  store i32 400, i32* %ptr
  br i1 %c, label %if.true, label %if.false

if.true:
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  store i32 500, i32* %ptr
  br label %merge

if.false:
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  store i32 600, i32* %ptr
  br label %merge

merge:
  ret void
}

define void @test10(i32 %a, i32 %b, i1 %c, i32* %ptr) {
; Make sure that non-dominating guards do not cause other guards removal.

; CHECK-LABEL: @test10(
; CHECK:       entry:
; CHECK-NEXT:    %cmp = icmp eq i32 %a, %b
; CHECK-NEXT:    br i1 %c, label %if.true, label %if.false
; CHECK:       if.true:
; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
; CHECK-NEXT:    store i32 100, i32* %ptr
; CHECK-NEXT:    br label %merge
; CHECK:       if.false:
; CHECK-NEXT:    store i32 200, i32* %ptr
; CHECK-NEXT:    br label %merge
; CHECK:       merge:
; CHECK-NEXT:    store i32 300, i32* %ptr
; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
; CHECK-NEXT:    store i32 400, i32* %ptr
; CHECK-NEXT:    ret void

entry:
  %cmp = icmp eq i32 %a, %b
  br i1 %c, label %if.true, label %if.false

if.true:
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  store i32 100, i32* %ptr
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  br label %merge

if.false:
  store i32 200, i32* %ptr
  br label %merge

merge:
  store i32 300, i32* %ptr
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  store i32 400, i32* %ptr
  ret void
}

define void @test11(i32 %a, i32 %b, i32* %ptr) {
; Make sure that branching condition is applied to guards.

; CHECK-LABEL: @test11(
; CHECK:       entry:
; CHECK-NEXT:    %cmp = icmp eq i32 %a, %b
; CHECK-NEXT:    br i1 %cmp, label %if.true, label %if.false
; CHECK:       if.true:
; CHECK-NEXT:    br label %merge
; CHECK:       if.false:
; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 false) [ "deopt"() ]
; CHECK-NEXT:    br label %merge
; CHECK:       merge:
; CHECK-NEXT:    ret void

entry:
  %cmp = icmp eq i32 %a, %b
  br i1 %cmp, label %if.true, label %if.false

if.true:
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  br label %merge

if.false:
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  br label %merge

merge:
  ret void
}

define void @test12(i32 %a, i32 %b) {
; Check that the assume marks its condition as being true (and thus allows to
; eliminate the dominated guards).

; CHECK-LABEL: @test12(
; CHECK-NEXT:  %cmp = icmp eq i32 %a, %b
; CHECK-NEXT:  call void @llvm.assume(i1 %cmp)
; CHECK-NEXT:  ret void

  %cmp = icmp eq i32 %a, %b
  call void @llvm.assume(i1 %cmp)
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  ret void
}

define void @test13(i32 %a, i32 %b, i32* %ptr) {
; Check that we deal correctly with stores when removing guards due to assume.

; CHECK-LABEL: @test13(
; CHECK-NEXT:  %cmp = icmp eq i32 %a, %b
; CHECK-NEXT:  call void @llvm.assume(i1 %cmp)
; CHECK-NEXT:  store i32 400, i32* %ptr
; CHECK-NEXT:  ret void

  %cmp = icmp eq i32 %a, %b
  call void @llvm.assume(i1 %cmp)
  store i32 100, i32* %ptr
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  store i32 200, i32* %ptr
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  store i32 300, i32* %ptr
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  store i32 400, i32* %ptr
  ret void
}

define void @test14(i32 %a, i32 %b, i1 %c, i32* %ptr) {
; Similar to test13, but with more control flow.
; TODO: Can we get rid of the store in the end of entry given that it is
; post-dominated by other stores?

; CHECK-LABEL: @test14(
; CHECK:       entry:
; CHECK-NEXT:    %cmp = icmp eq i32 %a, %b
; CHECK-NEXT:    call void @llvm.assume(i1 %cmp)
; CHECK-NEXT:    store i32 400, i32* %ptr
; CHECK-NEXT:    br i1 %c, label %if.true, label %if.false
; CHECK:       if.true:
; CHECK-NEXT:    store i32 500, i32* %ptr
; CHECK-NEXT:    br label %merge
; CHECK:       if.false:
; CHECK-NEXT:    store i32 600, i32* %ptr
; CHECK-NEXT:    br label %merge
; CHECK:       merge:
; CHECK-NEXT:    ret void

entry:
  %cmp = icmp eq i32 %a, %b
  call void @llvm.assume(i1 %cmp)
  store i32 100, i32* %ptr
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  store i32 200, i32* %ptr
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  store i32 300, i32* %ptr
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  store i32 400, i32* %ptr
  br i1 %c, label %if.true, label %if.false

if.true:
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  store i32 500, i32* %ptr
  br label %merge

if.false:
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  store i32 600, i32* %ptr
  br label %merge

merge:
  ret void
}

define void @test15(i32 %a, i32 %b, i1 %c, i32* %ptr) {
; Make sure that non-dominating assumes do not cause guards removal.

; CHECK-LABEL: @test15(
; CHECK:       entry:
; CHECK-NEXT:    %cmp = icmp eq i32 %a, %b
; CHECK-NEXT:    br i1 %c, label %if.true, label %if.false
; CHECK:       if.true:
; CHECK-NEXT:    call void @llvm.assume(i1 %cmp)
; CHECK-NEXT:    store i32 100, i32* %ptr
; CHECK-NEXT:    br label %merge
; CHECK:       if.false:
; CHECK-NEXT:    store i32 200, i32* %ptr
; CHECK-NEXT:    br label %merge
; CHECK:       merge:
; CHECK-NEXT:    store i32 300, i32* %ptr
; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
; CHECK-NEXT:    store i32 400, i32* %ptr
; CHECK-NEXT:    ret void

entry:
  %cmp = icmp eq i32 %a, %b
  br i1 %c, label %if.true, label %if.false

if.true:
  call void @llvm.assume(i1 %cmp)
  store i32 100, i32* %ptr
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  br label %merge

if.false:
  store i32 200, i32* %ptr
  br label %merge

merge:
  store i32 300, i32* %ptr
  call void (i1, ...) @llvm.experimental.guard(i1 %cmp) [ "deopt"() ]
  store i32 400, i32* %ptr
  ret void
}

define void @test16(i32 %a, i32 %b) {
; Check that we don't bother to do anything with assumes even if we know the
; condition being true.

; CHECK-LABEL: @test16(
; CHECK-NEXT:    %cmp = icmp eq i32 %a, %b
; CHECK-NEXT:    call void @llvm.assume(i1 %cmp)
; CHECK-NEXT:    call void @llvm.assume(i1 %cmp)
; CHECK-NEXT:    ret void

  %cmp = icmp eq i32 %a, %b
  call void @llvm.assume(i1 %cmp)
  call void @llvm.assume(i1 %cmp)
  ret void
}

define void @test17(i32 %a, i32 %b, i1 %c, i32* %ptr) {
; Check that we don't bother to do anything with assumes even if we know the
; condition being true or false (includes come control flow).

; CHECK-LABEL: @test17(
; CHECK:       entry:
; CHECK-NEXT:    %cmp = icmp eq i32 %a, %b
; CHECK-NEXT:    br i1 %c, label %if.true, label %if.false
; CHECK:       if.true:
; CHECK-NEXT:    call void @llvm.assume(i1 %cmp)
; CHECK-NEXT:    br label %merge
; CHECK:       if.false:
; CHECK-NEXT:    call void @llvm.assume(i1 %cmp)
; CHECK-NEXT:    br label %merge
; CHECK:       merge:
; CHECK-NEXT:    ret void

entry:
  %cmp = icmp eq i32 %a, %b
  br i1 %c, label %if.true, label %if.false

if.true:
  call void @llvm.assume(i1 %cmp)
  br label %merge

if.false:
  call void @llvm.assume(i1 %cmp)
  br label %merge

merge:
  ret void
}

define void @test18(i1 %c) {
; Check that we don't bother to do anything with assumes even if we know the
; condition being true and not being an instruction.

; CHECK-LABEL: @test18(
; CHECK-NEXT:    call void @llvm.assume(i1 %c)
; CHECK-NEXT:    call void @llvm.assume(i1 %c)
; CHECK-NEXT:    ret void

  call void @llvm.assume(i1 %c)
  call void @llvm.assume(i1 %c)
  ret void
}