diff --git a/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h b/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
index 90bee389..90defc54 100644
--- a/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
@@ -462,6 +462,7 @@ void WilsonKernels<Impl>::DhopKernel(int Opt,StencilImpl &st,  DoubledGaugeField
     autoView(st_v , st,AcceleratorRead);
 
    if( interior && exterior ) {
+     acceleratorFenceComputeStream();
      if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSite); return;}
      if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSite);    return;}
 #ifndef GRID_CUDA
@@ -495,6 +496,7 @@ void WilsonKernels<Impl>::DhopKernel(int Opt,StencilImpl &st,  DoubledGaugeField
     autoView(st_v ,st,AcceleratorRead);
 
    if( interior && exterior ) {
+     acceleratorFenceComputeStream();
      if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSiteDag); return;}
      if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSiteDag);    return;}
 #ifndef GRID_CUDA