afonso360 opened PR #6624 from afonso360:riscv-simd-icmp-opts
to bytecodealliance:main
:
:wave: Hey,
This is a follow up PR to #6609. Here we implement some of the inverted rules. I've only implemented the cases where we can directly use another instruction. There are still a few more improvements to do, but I've filed those as a follow up in #6623
I got these mostly by feeding in the equivalent IR to LLVM and checking which inverted instructions they produce. That being said, these cases all appear correct to me.
afonso360 requested jameysharp for a review on PR #6624.
afonso360 requested wasmtime-compiler-reviewers for a review on PR #6624.
afonso360 edited PR #6624:
:wave: Hey,
This is a follow up PR to #6609. Here we implement some of the inverted rules. I've only implemented the cases where we can directly use another instruction. There are still a few more improvements to do, but I've filed those as a follow up in #6623 since I'm not planning on working on them right now.
I got these mostly by feeding in the equivalent IR to LLVM and checking which inverted instructions they produce. That being said, these cases all appear correct to me.
alexcrichton submitted PR review.
afonso360 merged PR #6624.
Last updated: Nov 22 2024 at 17:03 UTC