diff --git a/Grid/algorithms/iterative/ConjugateGradient.h b/Grid/algorithms/iterative/ConjugateGradient.h
index e0fa7f83..3308d8fe 100644
--- a/Grid/algorithms/iterative/ConjugateGradient.h
+++ b/Grid/algorithms/iterative/ConjugateGradient.h
@@ -191,7 +191,7 @@ public:
 	std::cout << GridLogMessage << "\tAxpyNorm   " << AxpyNormTimer.Elapsed() <<std::endl;
 	std::cout << GridLogMessage << "\tLinearComb " << LinearCombTimer.Elapsed() <<std::endl;
 
-	std::cout << GridLogMessage << "\tMobius flop rate " << DwfFlops/ usecs<< " Gflops " <<std::endl;
+	std::cout << GridLogDebug << "\tMobius flop rate " << DwfFlops/ usecs<< " Gflops " <<std::endl;
 
         if (ErrorOnNoConverge) assert(true_residual / Tolerance < 10000.0);
 
diff --git a/Grid/lattice/Lattice_reduction_gpu.h b/Grid/lattice/Lattice_reduction_gpu.h
index b1139434..ecf90d19 100644
--- a/Grid/lattice/Lattice_reduction_gpu.h
+++ b/Grid/lattice/Lattice_reduction_gpu.h
@@ -219,14 +219,14 @@ inline typename vobj::scalar_objectD sumD_gpu_small(const vobj *lat, Integer osi
   commVector<sobj> buffer(numBlocks);
   sobj *buffer_v = &buffer[0];
   sobj result;
-  reduceKernel<<< numBlocks, numThreads, smemSize >>>(lat, buffer_v, size);
+  reduceKernel<<< numBlocks, numThreads, smemSize, computeStream >>>(lat, buffer_v, size);
   accelerator_barrier();
   acceleratorCopyFromDevice(buffer_v,&result,sizeof(result));
 #else
   Vector<sobj> buffer(numBlocks);
   sobj *buffer_v = &buffer[0];
   sobj result;
-  reduceKernel<<< numBlocks, numThreads, smemSize >>>(lat, buffer_v, size);
+  reduceKernel<<< numBlocks, numThreads, smemSize, computeStream >>>(lat, buffer_v, size);
   accelerator_barrier();
   result = *buffer_v;
 #endif