Add the new @llvm.nacl.atomic.fence.all intrinsic

lib/IR/NaClAtomicIntrinsics.cpp

Index: lib/IR/NaClAtomicIntrinsics.cpp
diff --git a/lib/IR/NaClAtomicIntrinsics.cpp b/lib/IR/NaClAtomicIntrinsics.cpp
index 5f463380c477e09769738082c071386e6b93e7c3..02fe295fad11cfae5aca5c050b04cc7e20ce7ffa 100644
--- a/lib/IR/NaClAtomicIntrinsics.cpp
+++ b/lib/IR/NaClAtomicIntrinsics.cpp
@@ -55,6 +55,7 @@ AtomicIntrinsics::AtomicIntrinsics(LLVMContext &C) {
   INIT(RMW, Ptr, Int, Mem, NoP, rmw);
   INIT(Ptr, Int, Int, Mem, Mem, cmpxchg);
   INIT(Mem, NoP, NoP, NoP, NoP, fence);

[jvoung (off chromium), 2013/08/07 19:38:40]

I think I brought this up before -- it seems a bit wasteful to create versions
of fence and fence_all for each overload type, even though there aren't
overloads.

Doesn't it mean that the transform pass will loop through each call in a
function 3x more than needed for each of the fence and fence_all?


The loop:

NaCl::AtomicIntrinsics::View V = AI.allIntrinsicsAndOverloads();
for (NaCl::AtomicIntrinsics::View::iterator I = V.begin(), E = V.end(); I != E;
++I) {
    AtomicCallResolver AtomicResolver(F, I);
    Changed |= visitCalls(AtomicResolver);
}

[JF, 2013/08/07 19:55:21]

I can make a "default overload type" for the functions with no overload, and
then remove the duplicates from this table.

I'd rather do this in another CL, though: I'd like to get this and others in
before the branch. If it's OK with you I'll file an issue and punt to M31.

[jvoung (off chromium), 2013/08/07 22:32:11]

ok that's fine

[JF, 2013/08/07 22:47:41]

https://code.google.com/p/nativeclient/issues/detail?id=3612

+  INIT(NoP, NoP, NoP, NoP, NoP, fence_all);
 }
 
 AtomicIntrinsics::View AtomicIntrinsics::allIntrinsicsAndOverloads() const {