[PATCH] D85516: [SVE][CodeGen] Fix issues with EXTRACT_SUBVECTOR when using scalable FP vectors

[David Sherwood via Phabricator via llvm-commits]

david-arm added a comment.

Hi @efriedma, I'm happy to try and fix up during the legalisation if you think that EXTRACT_SUBVECTOR should accept MVT::i32? However, I'm not sure how I would fix this during legalisation - would this be in DAGTypeLegalizer::PromoteIntOp_EXTRACT_SUBVECTOR? I originally tried adding extract_subvector patterns similar to how it's done for AArch64 and X86, i.e. using a iPTR or i64 type for the index, but the patterns didn't match because the SVE get/set tuple intrinsics were passing MVT::i32. I had a look at every other instance in the codebase where we create EXTRACT_SUBVECTOR operations and they all either call getIntPtrConstant() or getVectorIdxConstant() to create the index type. So it just seemed like this was the way we're expected to do this. Thanks!


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D85516/new/

https://reviews.llvm.org/D85516