diff --git a/benchmarks/Benchmark_comms.cc b/benchmarks/Benchmark_comms.cc
index 304a09fc..6d95bbe2 100644
--- a/benchmarks/Benchmark_comms.cc
+++ b/benchmarks/Benchmark_comms.cc
@@ -450,7 +450,7 @@ int main (int argc, char ** argv)
   }    
 
 
-
+#ifdef GRID_OMP
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   std::cout<<GridLogMessage << "= Benchmarking threaded STENCIL halo exchange in "<<nmu<<" dimensions"<<std::endl;
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
@@ -537,7 +537,7 @@ int main (int argc, char ** argv)
  
     }
   }    
-
+#endif
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   std::cout<<GridLogMessage << "= All done; Bye Bye"<<std::endl;
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
diff --git a/lib/qcd/action/fermion/WilsonCompressor.h b/lib/qcd/action/fermion/WilsonCompressor.h
index b47700ac..6ec2ede9 100644
--- a/lib/qcd/action/fermion/WilsonCompressor.h
+++ b/lib/qcd/action/fermion/WilsonCompressor.h
@@ -69,39 +69,47 @@ class WilsonCompressorTemplate< _HCspinor, _Hspinor, _Spinor, projector,
   /*****************************************************/
   /* Compress includes precision change if mpi data is not same */
   /*****************************************************/
-  inline void Compress(SiteHalfSpinor *buf,Integer o,const SiteSpinor &in) {
-    projector::Proj(buf[o],in,mu,dag);
+  inline void Compress(SiteHalfSpinor * __restrict__ buf,Integer o,const SiteSpinor &in) {
+    SiteHalfSpinor tmp;
+    projector::Proj(tmp,in,mu,dag);
+    vstream(buf[o],tmp);
   }
 
   /*****************************************************/
   /* Exchange includes precision change if mpi data is not same */
   /*****************************************************/
-  inline void Exchange(SiteHalfSpinor *mp,
-                       SiteHalfSpinor *vp0,
-                       SiteHalfSpinor *vp1,
+  inline void Exchange(SiteHalfSpinor * __restrict__ mp,
+                       const SiteHalfSpinor * __restrict__ vp0,
+                       const SiteHalfSpinor * __restrict__ vp1,
 		       Integer type,Integer o){
-    exchange(mp[2*o],mp[2*o+1],vp0[o],vp1[o],type);
+    SiteHalfSpinor tmp1;
+    SiteHalfSpinor tmp2;
+    exchange(tmp1,tmp2,vp0[o],vp1[o],type);
+    vstream(mp[2*o  ],tmp1);
+    vstream(mp[2*o+1],tmp2);
   }
 
   /*****************************************************/
   /* Have a decompression step if mpi data is not same */
   /*****************************************************/
-  inline void Decompress(SiteHalfSpinor *out,
-			 SiteHalfSpinor *in, Integer o) {    
+  inline void Decompress(SiteHalfSpinor * __restrict__ out,
+			 SiteHalfSpinor * __restrict__ in, Integer o) {    
     assert(0);
   }
 
   /*****************************************************/
   /* Compress Exchange                                 */
   /*****************************************************/
-  inline void CompressExchange(SiteHalfSpinor *out0,
-			       SiteHalfSpinor *out1,
-			       const SiteSpinor *in,
+  inline void CompressExchange(SiteHalfSpinor * __restrict__ out0,
+			       SiteHalfSpinor * __restrict__ out1,
+			       const SiteSpinor * __restrict__ in,
 			       Integer j,Integer k, Integer m,Integer type){
     SiteHalfSpinor temp1, temp2,temp3,temp4;
     projector::Proj(temp1,in[k],mu,dag);
     projector::Proj(temp2,in[m],mu,dag);
-    exchange(out0[j],out1[j],temp1,temp2,type);
+    exchange(temp3,temp4,temp1,temp2,type);
+    vstream(out0[j],temp3);
+    vstream(out1[j],temp4);
   }
 
   /*****************************************************/
diff --git a/lib/util/Init.cc b/lib/util/Init.cc
index 45a37a02..7dc8230e 100644
--- a/lib/util/Init.cc
+++ b/lib/util/Init.cc
@@ -385,6 +385,7 @@ void Grid_init(int *argc,char ***argv)
   if( GridCmdOptionExists(*argv,*argv+*argc,"--comms-threads") ){
     arg= GridCmdOptionPayload(*argv,*argv+*argc,"--comms-threads");
     GridCmdOptionInt(arg,CartesianCommunicator::nCommThreads);
+    assert(CartesianCommunicator::nCommThreads > 0);
   }
   if( GridCmdOptionExists(*argv,*argv+*argc,"--cacheblocking") ){
     arg= GridCmdOptionPayload(*argv,*argv+*argc,"--cacheblocking");
diff --git a/tests/Test_compressed_lanczos_hot_start.cc b/tests/Test_compressed_lanczos_hot_start.cc
index 3276d0f8..65f69c54 100644
--- a/tests/Test_compressed_lanczos_hot_start.cc
+++ b/tests/Test_compressed_lanczos_hot_start.cc
@@ -167,7 +167,7 @@ int main (int argc, char ** argv) {
   RealD mass = Params.mass;
   RealD M5   = Params.M5;
   std::vector<int> blockSize = Params.blockSize;
-  std::vector<int> latt({16,16,16,16});
+  std::vector<int> latt({32,32,32,32});
   uint64_t     vol = Ls*latt[0]*latt[1]*latt[2]*latt[3];
   double   mat_flop= 2.0*1320.0*vol;    
   // Grids