Add memoization for getSCEVAtScope results for instructions

which are not analyzed with SCEV techniques, which can require
brute-forcing through a large number of instructions. This
fixes a massive compile-time issue on 400.perlbench (in
particular, the loop in MD5Transform).


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@71259 91177308-0d34-0410-b5e6-96231b3b80d8

include/llvm/Analysis/ScalarEvolution.h

@@ -271,6 +271,11 @@ namespace llvm {
     /// exit value.
     std::map<PHINode*, Constant*> ConstantEvolutionLoopExitValue;
 
+    /// ValuesAtScopes - This map contains entries for all the instructions
+    /// that we attempt to compute getSCEVAtScope information for without
+    /// using SCEV techniques, which can be expensive.
+    std::map<Instruction *, std::map<const Loop *, Constant *> > ValuesAtScopes;
+
     /// createSCEV - We know that there is no SCEV for the specified value.
     /// Analyze the expression.
     SCEVHandle createSCEV(Value *V);