[PATCH] D55956: [WebAssembly] Fix invalid machine instrs in -O0, verify in tests

[Heejin Ahn via Phabricator via llvm-commits]

aheejin added inline comments.


================
Comment at: test/CodeGen/WebAssembly/simd-noopt.ll:9
+
+define i64 @foo(<2 x i64> %vec) #0 {
+entry:
----------------
aheejin wrote:
> tlively wrote:
> > aheejin wrote:
> > > tlively wrote:
> > > > aheejin wrote:
> > > > > No check lines?
> > > > I'm mostly just interested in this not crashing, but would be happy to add some if you think it would be better.
> > > Not crashing meaning, not crashing llvm regression test or V8? If you mean the latter, I guess it's better to check that v2i64 types are not generated by one or two `CHECK-NOT` lines.
> > Oh no, sorry, I meant not crashing LLVM. It was previously crashing because it was generating MachineInstrs that used a register that had never been defined.
> Then I guess it's better add a comment to this test what this is supposed to test, because if someone does not know the context that some types are enabled only with a flag (which is not included in this case) and this test is supposed to test the bahavior of *not* generating those types in case of -O0.
Sorry the comment ended prematurely, what I wanted to write is "because if someone does not know ..., it would be hard to understand what this test is doing."


Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D55956/new/

https://reviews.llvm.org/D55956