[PATCH] D69275: Add constrained int->FP intrinsics

Tue Oct 29 08:40:18 PDT 2019

cameron.mcinally added inline comments.

================
Comment at: llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp:1020
+                                    Node->getOperand(1).getValueType());
+    break;
   case ISD::SIGN_EXTEND_INREG: {
----------------
cameron.mcinally wrote:
> kpn wrote:
> > cameron.mcinally wrote:
> > > Should these be clustered with STRICT_LRINT/etc?
> > The regular U/SINT_TO_FP nodes are handled a couple of lines above. That makes these clustered close to the matching non-constrained versions, which seemed appropriate.
> > 
> > Is there a better place?
> I think the STRICT_LRINT/etc cases should be moved up and combined with these new cases. All the non-strict variants are clustered together. It would make sense to cluster the strict variants directly under those.
I suppose STRICT_LRINT/etc could be moved in a separate NFCI patch and then this patch just rebased. That's probably best. Sorry, just thinking aloud...

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D69275/new/

https://reviews.llvm.org/D69275