tra created this revision.
tra added a reviewer: timshen.
Herald added subscribers: sanjoy.google, bixia.

vote.ballot instruction is gone in recent CUDA versions and
vote.sync.ballot can not be used because it needs a thread mask parameter.
Fortunately PTX 6.2 (introduced with CUDA-9.2) provides activemask.b32
instruction for this.


https://reviews.llvm.org/D66665

Files:
  clang/lib/Headers/__clang_cuda_intrinsics.h


Index: clang/lib/Headers/__clang_cuda_intrinsics.h
===================================================================
--- clang/lib/Headers/__clang_cuda_intrinsics.h
+++ clang/lib/Headers/__clang_cuda_intrinsics.h
@@ -211,7 +211,15 @@
   return __nvvm_vote_ballot_sync(mask, pred);
 }
 
-inline __device__ unsigned int __activemask() { return __nvvm_vote_ballot(1); }
+inline __device__ unsigned int __activemask() {
+#if CUDA_VERSION < 9020
+  return __nvvm_vote_ballot(1);
+#else
+  unsigned int mask;
+  asm volatile("activemask.b32 %0;" : "=r"(mask));
+  return mask;
+#endif
+}
 
 inline __device__ unsigned int __fns(unsigned mask, unsigned base, int offset) 
{
   return __nvvm_fns(mask, base, offset);


Index: clang/lib/Headers/__clang_cuda_intrinsics.h
===================================================================
--- clang/lib/Headers/__clang_cuda_intrinsics.h
+++ clang/lib/Headers/__clang_cuda_intrinsics.h
@@ -211,7 +211,15 @@
   return __nvvm_vote_ballot_sync(mask, pred);
 }
 
-inline __device__ unsigned int __activemask() { return __nvvm_vote_ballot(1); }
+inline __device__ unsigned int __activemask() {
+#if CUDA_VERSION < 9020
+  return __nvvm_vote_ballot(1);
+#else
+  unsigned int mask;
+  asm volatile("activemask.b32 %0;" : "=r"(mask));
+  return mask;
+#endif
+}
 
 inline __device__ unsigned int __fns(unsigned mask, unsigned base, int offset) {
   return __nvvm_fns(mask, base, offset);
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
  • [PATCH] D66665: [CUDA] Use ... Artem Belevich via Phabricator via cfe-commits

Reply via email to