[PATCH] D148186: ValueTracking: Handle insertelement in computeKnownFPClass

[Matt Arsenault via Phabricator via llvm-commits]

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


================
Comment at: llvm/lib/Analysis/ValueTracking.cpp:4128
                          const TargetLibraryInfo *TLI) {
+  assert(Known.isUnknown() && "should not be called with known information");
+
----------------
foad wrote:
> Wouldn't it be simpler to make all these computeKnownFPClass functions return KnownFPClass?
Yes. I started by copying the raw signature for computeKnownBits. Probably should do a cleanup and replace the arguments at some point. The ordering doesn't make much sense (e.g. Depth before the others seems like a shortcut taken years ago). Plus TLI not being part of Query also doesn't make much sense 


================
Comment at: llvm/lib/Analysis/ValueTracking.cpp:4606-4607
   auto *FVTy = dyn_cast<FixedVectorType>(V->getType());
   APInt DemandedElts =
       FVTy ? APInt::getAllOnes(FVTy->getNumElements()) : APInt(1, 1);
   ::computeKnownFPClass(V, DemandedElts, InterestedClasses, Known, Depth,
----------------
foad wrote:
> Could call your new computeKnownFPClass overload rather than repeating this logic?
Yes, I think so 


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D148186/new/

https://reviews.llvm.org/D148186