[PATCH] D118426: [InstCombine] Remove side effect of replaced constrained intrinsics

[Eli Friedman via Phabricator via llvm-commits]

efriedma added inline comments.


================
Comment at: llvm/include/llvm/Analysis/InstructionSimplify.h:305
+/// without such. It happens if simplification code deduces that side effect
+/// actually absents.
 Value *SimplifyCall(CallBase *Call, const SimplifyQuery &Q);
----------------
"is actually absent".


================
Comment at: llvm/lib/Transforms/InstCombine/InstructionCombining.cpp:4134
+            // was found for it, assume that side effect is actually absent and
+            // the instruction can be removed.
+            isa<ConstrainedFPIntrinsic>(I)) {
----------------
It's not clear to me this is the only reason a constrained FP intrinsic can be returned.  And even if it were, it's a fragile invariant.

Would it make sense to just have isInstructionTriviallyDead() do the right thing here? We already do something similar with isMathLibCallNoop().


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D118426/new/

https://reviews.llvm.org/D118426