[clang] [CIR][AMDGPU] Add support for AMDGCN cos builtins (PR #197806)

[Ayokunle Amodu via cfe-commits]

================
@@ -0,0 +1,26 @@
+#include "../CodeGenCUDA/Inputs/cuda.h"
----------------
ayokunle321 wrote:

Hi @arsenm,

I included the header because I was following the existing tests, e.g. https://github.com/llvm/llvm-project/blob/816e2924786a0d7d300aa9a2ff80e5a2a082e6fa/clang/test/CIR/CodeGenHIP/builtins-amdgcn.hip#L1.

I'll switch it to use #define __device__ __attribute__((device)).

As for the file names, I was mistaken when following the patches in the incubator. I had been looking at HIP tests in the OpenCL test directory, which use different naming conventions. I've found the correct directory and will rename the files accordingly.

https://github.com/llvm/llvm-project/pull/197806