jhuber6 wrote:
> I think the correct place is the clang resource directory Tough to say, the compiler resource directory is for things the compiler emits as a part of the language, yet we don't store `libomp` there. I'd say it makes sense, if you had `lib/clang/21/lib/amdgcn-amd-amdhsa/libclang_rt.clc.a` it would be in-line with things like builtins and instrumentation. > Regarding triples, libclc doesn't currently have a `amdgcn-amd-amdhsa` > target, per se. It's got `amdgcn--` or `amdgcn--amdhsa`. Again I don't know > how the runtimes triples work and how strict the triple matching would be or > if libclc would have to massage triples into its own internal set. If a > triple isn't found, like `x86_64-unknown-linux-gnu` is it silently ignored, > warned, errored? Yeah, `amdgcn-amd-amdhsa` should be roughly equivalent to `amdgcn--`. The one problem I see is that if we do the per-target install directories, then `amdgcn--` and `amdgcn-amd-amdhsa` will have different directories. I wonder if we could just unify those somehow, or just accept `amdgcn-amd-amdhsa` in `libclc` as the same thing. > Oh, also for context, there was #124709 a few weeks ago. So it appears that > there are other people interested in moving libclc to the runtimes > infrastructure. Yep, part of why I'm bringing it up. So, the runtimes build is basically a CMake [ExternalProject](https://cmake.org/cmake/help/latest/module/ExternalProject.html) that points to the project. The file `./runtimes/CMakeLists.txt` is the entry point for that build. So, the `llvm/runtimes/CMakeListst.txt` basically creates sub-jobs by pointing them to `./runtimes/CMakeLists.txt`. This is mostly helpful because it guarantees you're going to be using an up-to-date clang compiler for the build. Right now what `libclc` does would be pretty easy to port to a normal `LLVM_ENABLE_RUNTIMES=libclc` build, because we currently create a few jobs and compile them with `--target=`. So, that should likely still work. You can also do cross-compiling through the runtimes interface, which is my request. So likely we'd have `LLVM_ENABLE_RUNTIMES=libclc` work as it does now, But also for `-DRUNTIMES_amdgcn-amd-amdhsa_LLVM_ENABLE_RUNTIMES=libclc` to basically just enable `amdgcn--` currently. I have an unfortunate amount of experience mangling the runtimes builds, so if you need help feel free to ask. https://github.com/llvm/llvm-project/pull/126078 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
