[PATCH] D142379: [RISCV] Don't explicitly select X0 in select branch comparisons

[Alex Bradbury via Phabricator via llvm-commits]

asb added inline comments.


================
Comment at: llvm/test/CodeGen/RISCV/double-convert.ll:141
 ; RV32I-NEXT:    call __unorddf2 at plt
 ; RV32I-NEXT:    snez a0, a0
 ; RV32I-NEXT:    addi a0, a0, -1
----------------
craig.topper wrote:
> There used to be a branch here when I wrote the original patch. It was made branchless sometime later. Perhaps the coalescer wasn't improved and the test cases don't cover it anymore?
Unfortunately, I think you're right and that's what's going on here. Seems to be a load of regressions with inputs such as the GCC torture suite.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D142379/new/

https://reviews.llvm.org/D142379