[PATCH] D143646: [RISCV] Return false from shouldFormOverflowOp

[Liao Chunyu via Phabricator via llvm-commits]

liaolucy updated this revision to Diff 498742.
liaolucy added a comment.

Solving the regression of uaddo1_math_overflow_used.

the regression testcase.

  define i64 @uaddo1_math_overflow_used(i64 %a, i64 %b, ptr %res) nounwind ssp {
    %add = add i64 %b, %a
    %cmp = icmp ult i64 %add, %a ----------%a
    %Q = select i1 %cmp, i64 %b, i64 42
    store i64 %add, ptr %res
    ret i64 %Q
  }

No regression, the testcase from the following of file

  define i64 @uaddo2_math_overflow_used(i64 %a, i64 %b, ptr %res) nounwind ssp {
    %add = add i64 %b, %a
    %cmp = icmp ult i64 %add, %b -----------%b
    %Q = select i1 %cmp, i64 %b, i64 42
    store i64 %add, ptr %res
    ret i64 %Q
  }

So I think it is good to modify it to:

  define i64 @uaddo1_math_overflow_used(i64 %a, i64 %b, ptr %res) nounwind ssp {
    %add = add i64 %a, %b    --------a and b exchange positions
    %cmp = icmp ult i64 %add, %a 
    %Q = select i1 %cmp, i64 %b, i64 42
    store i64 %add, ptr %res
    ret i64 %Q
  }


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D143646/new/

https://reviews.llvm.org/D143646

Files:
  llvm/lib/Target/RISCV/RISCVISelLowering.cpp
  llvm/lib/Target/RISCV/RISCVISelLowering.h
  llvm/test/CodeGen/RISCV/overflow-intrinsics.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D143646.498742.patch
Type: text/x-patch
Size: 7196 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20230220/767fffdd/attachment.bin>