[Openmp-commits] [PATCH] D66851: [libomptarget] Refactor activemask macro to inline function

[Jon Chesterfield via Phabricator via Openmp-commits]

JonChesterfield marked an inline comment as done.
JonChesterfield added inline comments.


================
Comment at: openmp/libomptarget/deviceRTLs/nvptx/src/target_impl.h:46
+// __ballot(1) in Cuda 8.0 is replaced with __activemask().
+INLINE __kmpc_impl_lanemask_t __kmpc_impl_activemask() {
+#if CUDA_VERSION >= 9000
----------------
ABataev wrote:
> I would suggest to change the types of all variables used as mask to __kmpc_impl_lanemask_t, just like for this function.
That sounds good. I'll rebase this on master (that'll catch syncwarp) and look through the codebase for int/int32_t/uint32_t that are obviously lane masks.

This means the lanemask type will be in the extern API - nothing will break as it's still an i32, and that'll be helpful when targeting a 64 bit warp from clang.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D66851/new/

https://reviews.llvm.org/D66851