[PATCH] D68700: [ARM] Add IR intrinsics for MVE VLD[24] and VST[24].

[Dave Green via Phabricator via llvm-commits]

dmgreen added a comment.

Looks good. I think we can use this for autovec codegen too (there is a pre-isel pass that allows us to convert load+shuffle combos that the vectorizer produces into these intrinsics). As that is the case it would probably be worth making sure we have lots of test coverage of the different types. (I'm happy enough to do that later if you with, but adding them here sounds more sensible, as this is where they are being introduced).



================
Comment at: llvm/lib/Target/ARM/ARMInstrMVE.td:4291
+defm : MVE_vst24_patterns<16, v8i16>;
+defm : MVE_vst24_patterns<32, v4i32>;
+
----------------
Do we need floating point types?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D68700/new/

https://reviews.llvm.org/D68700