[libc-commits] [libc] [llvm] [LLVM] Port 'llvm-gpu-loader' to use LLVMOffload (PR #162739)

[Joseph Huber via libc-commits]

jhuber6 wrote:

@Artem-B I've updated this and will merge it soon. The CUDA builder bot will need to have `libLLVMOffload.so` somewhere in the library path because we open this dynamically. You should be able to build this standalone with something as simple as this. Assuming you are inside a build directory which serves as the LLVM root.

```
$ cmake ../runtimes -DCMAKE_BUILD_TYPE=Release -DLLVM_ENABLE_RUNTIMES=offload -DLLVM_BINARY_DIR=$PWD
```

https://github.com/llvm/llvm-project/pull/162739