[SCEV] Do not bother creating separate SCEVUnknown for unreachable nodes

Currently, SCEV creates SCEVUnknown for every node of unreachable code. If we have a huge amounts of such code, we will be littering SE with these nodes. We could just state that they all are undef and save some memory. Differential Revision: https://reviews.llvm.org/D57567 Reviewed By: sanjoy llvm-svn: 353017
llvm · Feb 4, 2019 · 437ee05 · 437ee05
1 parent b5e945c
commit 437ee05
Showing 2 changed files with 2 additions and 2 deletions.
diff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp
@@ -6144,7 +6144,7 @@ const SCEV *ScalarEvolution::createSCEV(Value *V) {
     // to obey basic rules for definitions dominating uses which this
     // analysis depends on.
     if (!DT.isReachableFromEntry(I->getParent()))
-      return getUnknown(V);
+      return getUnknown(UndefValue::get(V->getType()));
   } else if (ConstantInt *CI = dyn_cast<ConstantInt>(V))
     return getConstant(CI);
   else if (isa<ConstantPointerNull>(V))

diff --git a/llvm/test/Analysis/ScalarEvolution/unreachable-code.ll b/llvm/test/Analysis/ScalarEvolution/unreachable-code.ll
@@ -1,7 +1,7 @@
 ; RUN: opt < %s -analyze -scalar-evolution | FileCheck %s
 
 ; CHECK: %t = add i64 %t, 1
-; CHECK: -->  %t
+; CHECK: -->  undef
 
 define void @foo() {
 entry: