[PATCH] D32241: Don't process debug intrinsics in InstCombine

[Dmitry Mikulin via Phabricator via llvm-commits]

dmikulin added inline comments.


================
Comment at: lib/Transforms/InstCombine/InstructionCombining.cpp:3074-3075
 
-      InstrsForInstCombineWorklist.push_back(Inst);
+      // Skip processing intrinsics in InstCombine
+      if (!isa<DbgInfoIntrinsic>(Inst))
+        InstrsForInstCombineWorklist.push_back(Inst);
----------------
dmikulin wrote:
> craig.topper wrote:
> > dmikulin wrote:
> > > davide wrote:
> > > > Please expand this comment to explain why we want to skip them.
> > >       // Skip processing debug intrinsics in InstCombine. Processing these call instructions
> > >       // consumes non-trivial amount of time and provides no value for the optimization.
> > > 
> > > Is this OK?
> > I'm really sorry to keep asking this to be moved, but is there any reason to run any of the code above here for debug intrinsics?
> I'm not sure how far up I can move the check. Top of the for loop? But won't we potentially miss DCE and constant folds above?
Any thoughts on this?



https://reviews.llvm.org/D32241