[compiler-rt] [compiler-rt] Initial support for builtins on GPU targets (PR #95304)

Joseph Huber via llvm-commits llvm-commits at lists.llvm.org
Mon Jun 17 13:04:43 PDT 2024


jhuber6 wrote:

> > That way we don't need 50 different builds and we get ISA linking that's correct.
> 
> The compatibility window isn't that big, we should be able to start using the new generic targets for this too. But ultimately I don't see this as a big deal.

I'm just wondering how this would work in practice. We'd need special logic for both the build system and the linker job in clang. And like I said, building it for even a few architectures is going to add at least 10 minutes to a trivial rebuild.

The end goal is definitely to have this supported in the backend, since it should simplify a lot of stuff. But I think the initial version of this patch doesn't need to tackle the very wide problem of ISA linking on GPUs just yet.

https://github.com/llvm/llvm-project/pull/95304


More information about the llvm-commits mailing list