[libc] [llvm] [libc] Build the GPU during the projects setup like libc-hdrgen (PR #84667)

Joseph Huber via llvm-commits llvm-commits at lists.llvm.org
Mon Mar 11 06:58:09 PDT 2024


================
@@ -56,11 +37,10 @@ elseif(TARGET nvptx-loader AND LIBC_TARGET_ARCHITECTURE_IS_NVPTX)
   )
 endif()
 
-if(TARGET libc.utils.gpu.loader)
-  get_target_property(gpu_loader_tgt libc.utils.gpu.loader "TARGET")
+foreach(gpu_loader_tgt amdhsa-loader nvptx-loader)
   if(gpu_loader_tgt)
----------------
jhuber6 wrote:

So, we have a lot of other tools that depend on CUDA / HSA that are installed unconditionally (see amdgpu-arch) but those use `dlopen`. I could do the same for my loader utilities but that's a lot of work for what is essentially a test tool.

https://github.com/llvm/llvm-project/pull/84667


More information about the llvm-commits mailing list