[llvm] [NVPTX] Do not run the NVVMReflect pass as part of the normal pipeline (PR #121834)

[Joseph Huber via llvm-commits]

jhuber6 wrote:

> I think the decision where the reflect pass should run should be tied to the earliest point where the reflect inputs get set. For CUDA, it's the beginning of the pipeline. For openMP and stand-alone compilation it's probably somewhere closer to the back-end (or wherever we may link in with libdevice, or do LTO, or other point where we finally know what we're actually compiling for).

Yeah, the point is to defer something until the backend knows what the actual target is. The optimizations that run on the initial compile are usually more generic so I wouldn't think this would fire until the backend.

https://github.com/llvm/llvm-project/pull/121834