Integer variants are implemented as atomicrmw or cmpxchg instructions.
Atomic add for floating point (__nvvm_atom_add_gen_f()) is implemented
as a call to an overloaded @llvm.nvvm.atomic.load.add.f32.* LVVM
intrinsic.

Differential Revision: http://reviews.llvm.org/D10666

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@240669 91177308-0d34-0410-b5e6-96231b3b80d8