[PATCH] D32344: InstCombine/AMDGPU: Fix constant folding of llvm.amdgcn.{icmp, fcmp}

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: lib/Transforms/InstCombine/InstCombineCalls.cpp:3406
+                                     ConstantExpr::getSExt(CCmp, II->getType()));
+        } else {
+          Value *NewF = Intrinsic::getDeclaration(II->getModule(),
----------------
No else


================
Comment at: lib/Transforms/InstCombine/InstCombineCalls.cpp:3410-3413
+          Metadata *MDArgs[] = { MDString::get(II->getContext(), "exec") };
+          MDNode *MD = MDNode::get(II->getContext(), MDArgs);
+          Value *Args[] = { MetadataAsValue::get(II->getContext(), MD) };
+          CallInst *NewCall = Builder->CreateCall(NewF, Args);
----------------
Should have a comment about why this is done


================
Comment at: test/Transforms/InstCombine/amdgcn-intrinsics.ll:1538
   ret i64 %result
 }
----------------
Check that attributes 4 contains convergent?


https://reviews.llvm.org/D32344