Add host API for enqueuing barrier on given stream (#274)

* add host API for enqueuing barrier on given stream

src/host/host.cpp

@@ -324,6 +324,16 @@ __host__ void HostInterface::barrier_all(WindowInfo* window_info) {
   return;
 }
 
+__host__ void HostInterface::barrier_all_on_stream(hipStream_t stream) {
+  // launch kernel to do barrier with given stream, if non, use default stream
+  if (stream == nullptr) {
+    stream = hipStreamDefault;
+  }
+
+  rocshmem_barrier_all_kernel<<<1, 1, 0,  stream>>>();
+}
+
+
 __host__ void HostInterface::barrier_for_sync() {
   if (host_comm_world_ != MPI_COMM_NULL) {
     mpilib_ftable_.Barrier(host_comm_world_);

src/host/host.hpp

@@ -193,6 +193,8 @@ class HostInterface {
   __host__ void quiet(WindowInfo* window_info);
 
   __host__ void barrier_all(WindowInfo* window_info);
+  
+  __host__ void barrier_all_on_stream(hipStream_t stream);
 
   __host__ void barrier_for_sync();