[PATCH] D58232: GlobalISel: Don't materialize immarg arguments to intrinsics

[Matt Arsenault via Phabricator via llvm-commits]

arsenm marked an inline comment as done.
arsenm added inline comments.


================
Comment at: test/TableGen/immarg.td:27-31
+// FIXME: This should not crash, but should it work or be an error?
+// def : Pat <
+//   (int_mytarget_sleep1 timm:$src),
+//   (SLEEP1 imm:$src)
+// >;
----------------
paquette wrote:
> Does this still crash?
> 
> Would it be possible to make it just error for now, saying it's currently unsupported?
Yes. It hits this:

```
InstructionMatcher &
RuleMatcher::getInstructionMatcher(StringRef SymbolicName) const {
  for (const auto &I : InsnVariableIDs)
    if (I.first->getSymbolicName() == SymbolicName)
      return *I.first;
  llvm_unreachable(
      ("Failed to lookup instruction " + SymbolicName).str().c_str());
}
```

I'm not sure how exactly it ends up concluding this is an instruction of some kind


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D58232/new/

https://reviews.llvm.org/D58232