[PATCH] D70075: [InstCombine] Skip scalable vectors in combineLoadToOperationType

[Diana Picus via Phabricator via llvm-commits]

rovka marked 2 inline comments as done.
rovka added inline comments.


================
Comment at: llvm/test/Transforms/InstCombine/load.ll:240
 
+define void @test16-vect(i8* %x, i8* %a, i8* %b, i8* %c) {
+; CHECK-LABEL: @test16-vect(
----------------
sdesmalen wrote:
> Is this just adding a missing test-case for existing code, or are you trying to test something else?
I'm just trying to make sure we're not accidentally breaking anything for fixed-sized vectors. It's essentially the same test-case as test16, but with vectors instead of float (the code should do the same thing for any type of that size). It should work the same before and after this patch, so I could commit it separately if you think it makes more sense that way. Personally I think it fits well in the context of this patch, since we're not trying to exhaustively test every possible type.


================
Comment at: llvm/test/Transforms/InstCombine/load.ll:340
+
+; Make sure we don't canonicalize accesses to scalable vectors.
+define void @test20(<vscale x 4 x i8>* %addr) {
----------------
sdesmalen wrote:
> Does this need a CHECK line to test for non-assert builds?
Oops! Of course it does, I just forgot to re-run update_test_checks after getting it to pass. Thanks for catching it.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D70075/new/

https://reviews.llvm.org/D70075