[flang-commits] [flang] [Flang] Adding atomicadd as a cudadevice intrinsic and converting it LLVM dialect (PR #123840)
via flang-commits
flang-commits at lists.llvm.org
Wed Jan 22 01:45:55 PST 2025
================
@@ -147,6 +147,10 @@ static constexpr IntrinsicHandler handlers[]{
{"atan2pi", &I::genAtanpi},
{"atand", &I::genAtand},
{"atanpi", &I::genAtanpi},
+ {"atomicaddd", &I::genAtomAdd, {{{"addr", asAddr}, {"v", asValue}}}, false},
----------------
jeanPerier wrote:
At some point, we should maybe move the CUF intrinsic module procedures into there own table (a bit like what is done for ppc intrinsics).
This would clarify what are regular Fortran intrinsics expected to be intercepted here, and what are CUF extensions (and if well done, could speed-up the lookup by only looking into the right table based on the intrinsic module).
You can leave that alone, I will try to split the table later (maybe ieee proc could also be taken out).
https://github.com/llvm/llvm-project/pull/123840
More information about the flang-commits
mailing list