[PATCH] D113014: [fir] Add fir.zero_bits conversion pattern to LLVM IR dialect

[Valentin Clement via Phabricator via llvm-commits]

clementval added inline comments.


================
Comment at: flang/lib/Optimizer/CodeGen/CodeGen.cpp:188
+      return zero.emitOpError(
+          "conversion of fir.zero with aggregate type not implemented yet");
+    }
----------------
mehdi_amini wrote:
> Nit:  in general Patterns aren't supposed to emit errors directly but to wrap the message in notifyFailure(...).
> 
> The conversion may still succeed if another pattern ends up folding this op (or matching it), so having an error can be confusing.
> 
> In general if we want end-use diagnostic, it is best to use other mechanism (like using the partial conversion which has an extra argument for collecting "unconvertedOps").
> 
> 
> 
Ok I'll update that.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D113014/new/

https://reviews.llvm.org/D113014