[PATCH] D109888: Bitcast wrapped in a call obscures function attributes, pessimizing MemorySSA

[Antonio Frighetto via Phabricator via llvm-commits]

antoniofrighetto added inline comments.


================
Comment at: llvm/lib/IR/Instructions.cpp:357-358
+
+  if (isa<Function>(V))
+    return cast<Function>(V)->getAttributes().hasFnAttr(Kind);
+
----------------
lebedev.ri wrote:
> Could also use `dyn_cast`, don't know if that would be better/worse.
Since I've changed to use `dyn_cast` before, it seems reasonable to use it here too.


================
Comment at: llvm/lib/IR/Instructions.cpp:341
 /// Determine whether the argument or parameter has the given attribute.
 bool CallBase::paramHasAttr(unsigned ArgNo, Attribute::AttrKind Kind) const {
   assert(ArgNo < getNumArgOperands() && "Param index out of bounds!");
----------------
I suspect methods `CallBase::getIntrinsicID` and `CallBase::paramHasAttr` might be changed as well in the future.


================
Comment at: llvm/lib/IR/Instructions.cpp:353
   if (const Function *F = getCalledFunction())
     return F->getAttributes().hasFnAttr(Kind);
+
----------------
nikic wrote:
> You should be able to drop the getCalledFunction() part, as it is subsumed by the newly added code.
Right, forgot to drop it.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D109888/new/

https://reviews.llvm.org/D109888