diff --git a/projects/rocshmem/src/gda/context_gda_tmpl_device.hpp b/projects/rocshmem/src/gda/context_gda_tmpl_device.hpp
index 8fbb33af01..2c48cd5f8b 100644
--- a/projects/rocshmem/src/gda/context_gda_tmpl_device.hpp
+++ b/projects/rocshmem/src/gda/context_gda_tmpl_device.hpp
@@ -669,6 +669,7 @@ __device__ void GDAContext::alltoall_linear_thread_puts(rocshmem_team_t team, T
 
   // wait until everyone has obtained their designated data
   internal_direct_barrier_wg_thread_puts(my_pe, pe_start, stride, pe_size, pSync);
+  __syncthreads();
 }
 
 template <typename T>