[PATCH] D126893: [RISCV] simplify emitVSETVLIs handling of vsetvli xN, phi(), vtype case [NFC]

[Fraser Cormack via Phabricator via llvm-commits]

frasercrmck accepted this revision.
frasercrmck added a comment.
This revision is now accepted and ready to land.

I was wondering if this would be better represented as `Optional<VSETVLIInfo> LocalBBInfo` initialized to `None` before traversing the block. We'd update it along with `CurInfo` (`LocalBBInfo = CurInfo`). Then we'd know that if `LocalBBInfo` has been set there have been block-local updates to VL/VTYPE and thus it's not safe to look through PHIs. At least I'd find that a little to grok. It's subjective, after all.

So, LGTM.



================
Comment at: llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp:1079
+  // Track whether the prefix of the block we've scanned is transparent
+  // (meaning has not yet changing the abstract state).
+  bool PrefixTransparent = true;
----------------
`changed`?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D126893/new/

https://reviews.llvm.org/D126893