Missed Optimization: Fold x - umin(x,13) < 2 into single compare x < 15

[Leo0506]

define i1 @src(i64 %arg0) {
  %v0 = tail call i64 @llvm.umin.i64(i64 %arg0, i64 13)
  %v1 = sub nsw i64 %arg0, %v0
  %v2 = icmp slt i64 %v1, 2
  ret i1 %v2
}

define i1 @tgt(i64 %arg0) {
  %v0 = icmp slt i64 %arg0, 15
  ret i1 %v0
}

alive2: https://alive2.llvm.org/ce/z/pyUQAP
godbolt: https://godbolt.org/z/TaYWM9GTe
Pattern found in: https://github.com/dtcxzyw/llvm-opt-benchmark/blob/main/bench/ffmpeg/optimized/pictordec.ll

[ParkHanbum]

I think implementing this pattern would be a good idea. What do you think? @dtcxzyw

actually, this pattern already implemented. see 8320b97#diff-a9b6075c524d5b5a214ea651f0c91612280e7d52401931c022a86b6996ce8206

but something is missed

[dtcxzyw]

I think implementing this pattern would be a good idea. What do you think? @dtcxzyw

Yes this pattern is common across multiple codebases. See Grep issue167055 dtcxzyw/llvm-opt-benchmark#3050

actually, this pattern already implemented. see 8320b97#diff-a9b6075c524d5b5a214ea651f0c91612280e7d52401931c022a86b6996ce8206

but something is missed

You may need to implement another helper function to handle all the variants.