Fix a major source of compile-time slowness at -O0 -g by optimizing

the storage of !dbg metadata kinds in the instruction themselves.
The on-the-side hash table works great for metadata that not-all
instructions get, or for metadata that only exists when optimizing.
But when compile-time is everything, it isn't great.

I'm not super thrilled with the fact that this plops a TrackingVH in
Instruction, because it grows it by 3 words.  I'm investigating 
alternatives, but this should be a step in the right direction in any
case.

llvm-svn: 99957

1 files changed, 5 insertions, 1 deletions

diff --git a/llvm/lib/VMCore/LLVMContext.cpp b/llvm/lib/VMCore/LLVMContext.cpp
index f9d3fe0021d..2a870ec6cfd 100644
--- a/llvm/lib/VMCore/LLVMContext.cpp
+++ b/llvm/lib/VMCore/LLVMContext.cpp
@@ -26,7 +26,11 @@ LLVMContext& llvm::getGlobalContext() {
   return *GlobalContext;
 }
 
-LLVMContext::LLVMContext() : pImpl(new LLVMContextImpl(*this)) { }
+LLVMContext::LLVMContext() : pImpl(new LLVMContextImpl(*this)) {
+  // Create the first metadata kind, which is always 'dbg'.
+  unsigned DbgID = getMDKindID("dbg");
+  assert(DbgID == MD_dbg && "dbg kind id drifted"); (void)DbgID;
+}
 LLVMContext::~LLVMContext() { delete pImpl; }