[clang] [llvm] AMDGPU: Mark workitem ID intrinsics with range attribute (PR #136196)

[Matt Arsenault via llvm-commits]

arsenm wrote:

> I attempted to fix the failure in clang/test/Headers/gpuintrin_lang.c by removing the range() in the CHECK, but now some other build fails and complains that the range() is missing:
> 
> https://lab.llvm.org/buildbot/#/builders/174/builds/16421/steps/6/logs/FAIL__Clang__gpuintrin_lang_c
> 
> @arsenm can you PTAL?

Something is wrong with specific hosts, but I don't know why. The problem is the nvptx builtins don't actually annotate the callsites with the range, the NVVMIntrRange pass is running them. For some reason these hosts don't run it; is this just what happens if you run the test without the NVPTX backend built?

https://github.com/llvm/llvm-project/pull/136196