Cache the result of Function::getIntrinsicID() in a DenseMap attached to the LLVMContext.

This reduces the time actually spent doing string to ID conversion and shows a 10% improvement in compile time for a particularly bad case that involves ARM Neon intrinsics (these have many overloads).

Patch by Jean-Luc Duprat!

llvm-svn: 176365
diff --git a/llvm/lib/IR/LLVMContextImpl.h b/llvm/lib/IR/LLVMContextImpl.h
index 7353dc0..0c659b8 100644
--- a/llvm/lib/IR/LLVMContextImpl.h
+++ b/llvm/lib/IR/LLVMContextImpl.h
@@ -318,7 +318,7 @@
 
   /// ValueHandles - This map keeps track of all of the value handles that are
   /// watching a Value*.  The Value::HasValueHandle bit is used to know
-  // whether or not a value has an entry in this map.
+  /// whether or not a value has an entry in this map.
   typedef DenseMap<Value*, ValueHandleBase*> ValueHandlesTy;
   ValueHandlesTy ValueHandles;
   
@@ -350,6 +350,11 @@
   /// to date.
   std::vector<std::pair<DebugRecVH, DebugRecVH> > ScopeInlinedAtRecords;
   
+  /// IntrinsicIDCache - Cache of intrinsic name (string) to numeric ID mappings
+  /// requested in this context
+  typedef DenseMap<const Function*, unsigned> IntrinsicIDCacheTy;
+  IntrinsicIDCacheTy IntrinsicIDCache;
+
   int getOrAddScopeRecordIdxEntry(MDNode *N, int ExistingIdx);
   int getOrAddScopeInlinedAtIdxEntry(MDNode *Scope, MDNode *IA,int ExistingIdx);