[all-commits] [llvm/llvm-project] 32ad45: [InstCombine] Add test cases from PR62898. NFC.

[Yingwei Zheng via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 32ad45556e9c673d2b6daba4a3b7993ad4e8e9ac
      https://github.com/llvm/llvm-project/commit/32ad45556e9c673d2b6daba4a3b7993ad4e8e9ac
  Author: Yingwei Zheng <dtcxzyw2333 at gmail.com>
  Date:   2023-09-11 (Mon, 11 Sep 2023)

  Changed paths:
    M llvm/test/Transforms/InstCombine/smax-icmp.ll
    M llvm/test/Transforms/InstCombine/smin-icmp.ll
    M llvm/test/Transforms/InstCombine/umax-icmp.ll
    M llvm/test/Transforms/InstCombine/umin-icmp.ll

  Log Message:
  -----------
  [InstCombine] Add test cases from PR62898. NFC.

This patch adds some test cases from https://github.com/llvm/llvm-project/issues/62898.

As nikic noted in the issue, we should start by implementing a generalization of the fold `smin(X, Y) < Z -> X < Z` when `Y > Z` is implied by constant folds/invariants/dom conditions.
```
define i1 @src(i32 %x, i32 %y, i32 %z) {
  %cmp = icmp sgt i32 %y, %z
  br i1 %cmp, label %if, label %end
if:
  %cond = call i32 @llvm.smin.i32(i32 %x, i32 %y)
  %tobool = icmp slt i32 %cond, %z
  ret i1 %tobool
end:
  ret i1 false
}

define i1 @tgt(i32 %x, i32 %y, i32 %z) {
  %cmp = icmp sgt i32 %y, %z
  br i1 %cmp, label %if, label %end
if:
  %tobool = icmp slt i32 %x, %z
  ret i1 %tobool
end:
  ret i1 false
}

declare i32 @llvm.smin.i32(i32, i32)
```
Alive2: https://alive2.llvm.org/ce/z/dK9vXz

This patch also adds some generalized test cases like the above.

Reviewed By: goldstein.w.n

Differential Revision: https://reviews.llvm.org/D156227


  Commit: 44e5afdb91b92dedfbd3100a3e73dab27de1c9cf
      https://github.com/llvm/llvm-project/commit/44e5afdb91b92dedfbd3100a3e73dab27de1c9cf
  Author: Yingwei Zheng <dtcxzyw2333 at gmail.com>
  Date:   2023-09-11 (Mon, 11 Sep 2023)

  Changed paths:
    M llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
    M llvm/lib/Transforms/InstCombine/InstCombineInternal.h
    M llvm/test/Transforms/InstCombine/smax-icmp.ll
    M llvm/test/Transforms/InstCombine/smin-icmp.ll
    M llvm/test/Transforms/InstCombine/umax-icmp.ll
    M llvm/test/Transforms/InstCombine/umin-icmp.ll
    M llvm/test/Transforms/LoopVectorize/X86/pr23997.ll
    M llvm/test/Transforms/LoopVectorize/interleaved-accesses.ll

  Log Message:
  -----------
  [InstCombine] Generalize foldICmpWithMinMax

This patch generalizes the fold of `icmp pred min/max(X, Y), Z` to address the issue https://github.com/llvm/llvm-project/issues/62898.

For example, we can fold `smin(X, Y) < Z` into `X < Z` when `Y > Z` is implied by constant folds/invariants/dom conditions.

Alive2 (with `--disable-undef-input` due to the limitation of --smt-to=10000): https://alive2.llvm.org/ce/z/rB7qLc
You can run the standalone translation validation tool `alive-tv` locally to verify these transformations.
```
alive-tv transforms.ll --smt-to=600000 --exit-on-error
```

Reviewed By: goldstein.w.n

Differential Revision: https://reviews.llvm.org/D156238


Compare: https://github.com/llvm/llvm-project/compare/fd453e2381d6...44e5afdb91b9