[clang] [Clang] Introduce scoped variants of GNU atomic functions (PR #72280)

Joseph Huber via cfe-commits cfe-commits at lists.llvm.org
Thu Nov 16 13:04:24 PST 2023


jhuber6 wrote:

> I'm a little wary of adding these without actually going through any sort of standardization process; if other vendors don't support the same interface, we just have more variations. (See also https://clang.llvm.org/get_involved.html#criteria )
> 
> How consistent are the various scopes across targets? How likely is it that some target will need additional scopes that you haven't specified?

I figured we can just treat these as `clang` extensions for the time being. We already have two variants that are more or less redundant for specific use-cases, (OpenCL and HIP), which should be able to be removed after this. Predicting all kinds of scopes is hard. The easy solution is to just number this or something since it's hierarchical. But @Artem-B has already pointed out that Nvidia has a scope between "device" and "blocks". Pretty much every system is going to have a conception of "device" and "system" and "single threaded" however.

https://github.com/llvm/llvm-project/pull/72280


More information about the cfe-commits mailing list