[llvm] [Offload] Add CMake cache to be used in AMDGPU bot (PR #119369)

[Shilei Tian via llvm-commits]

================
@@ -0,0 +1,13 @@
+set(LLVM_ENABLE_PROJECTS "clang;lld" CACHE STRING "")
+set(LLVM_ENABLE_RUNTIMES "compiler-rt;openmp;offload" CACHE STRING "")
+set(LLVM_ENABLE_PER_TARGET_RUNTIME_DIR ON CACHE BOOL "")
+set(LLVM_ENABLE_ASSERTIONS ON CACHE BOOL "")
+set(LLVM_LIT_ARGS "-v --show-unsupported --timeout 100 --show-xfail -j 32" CACHE STRING "")
+set(LLVM_TARGETS_TO_BUILD "X86;AMDGPU" CACHE STRING "")
----------------
shiltian wrote:

even though we don't support ARM yet, it still might be better to be a little bit future proof by using `host`.

https://github.com/llvm/llvm-project/pull/119369