diff --git a/resolve/random/RandSketchingCountSketch.cpp b/resolve/random/RandSketchingCountSketch.cpp
index 78f69779..e12f512a 100644
--- a/resolve/random/RandSketchingCountSketch.cpp
+++ b/resolve/random/RandSketchingCountSketch.cpp
@@ -58,14 +58,28 @@ namespace ReSolve
    */
   int RandSketchingCountSketch::Theta(vector_type* input, vector_type* output)
   {
-    mem_.deviceSynchronize();
-    count_sketch_theta(n_,
-                       k_rand_,
-                       d_labels_,
-                       d_flip_,
-                       input->getData(memspace_),
-                       output->getData(memspace_));
-    mem_.deviceSynchronize();
+    using namespace memory;
+   
+    switch (memspace_) {
+      case DEVICE:
+        mem_.deviceSynchronize();
+        count_sketch_theta(n_,
+                           k_rand_,
+                           d_labels_,
+                           d_flip_,
+                           input->getData(memspace_), 
+                           output->getData(memspace_));
+        mem_.deviceSynchronize();
+        break;
+      case HOST:
+        count_sketch_theta(n_,
+                           k_rand_,
+                           h_labels_,
+                           h_flip_,
+                           input->getData(memspace_),
+                           output->getData(memspace_));
+        break;
+    }
     return 0;
   }
 
diff --git a/resolve/random/RandSketchingFWHT.cpp b/resolve/random/RandSketchingFWHT.cpp
index 40065fec..8daf542c 100644
--- a/resolve/random/RandSketchingFWHT.cpp
+++ b/resolve/random/RandSketchingFWHT.cpp
@@ -83,27 +83,36 @@ namespace ReSolve
     switch (memspace_) {
       case DEVICE:
         mem_.setZeroArrayOnDevice(d_aux_, N_);
+        FWHT_scaleByD(n_, 
+                      d_D_,
+                      input->getData(memspace_), 
+                      d_aux_);  
+
+        mem_.deviceSynchronize();
+        FWHT(1, log2N_, d_aux_);
+
+        mem_.deviceSynchronize();
+        FWHT_select(k_rand_, 
+                    d_perm_, 
+                    d_aux_, 
+                    output->getData(memspace_)); 
+        mem_.deviceSynchronize();
         break; // remember - scaling is the solver's problem 
       case HOST:
-        // std::memset(d_aux_, 0.0, static_cast<size_t>(N_) * sizeof(real_type));
-        mem_.setZeroArrayOnHost(d_aux_, N_);
+        std::memset(d_aux_, 0.0, static_cast<size_t>(N_) * sizeof(real_type));
+        FWHT_scaleByD(n_, 
+                      h_D_,
+                      input->getData(memspace_),
+                      d_aux_);  
+
+        FWHT(1, log2N_, d_aux_);
+
+        FWHT_select(k_rand_, 
+                    h_perm_, 
+                    d_aux_, 
+                    output->getData(memspace_));
         break;
     }
-
-    FWHT_scaleByD(n_, 
-                  d_D_,
-                  input->getData(memspace_), 
-                  d_aux_);  
-
-    mem_.deviceSynchronize();
-    FWHT(1, log2N_, d_aux_);
-
-    mem_.deviceSynchronize();
-    FWHT_select(k_rand_, 
-                d_perm_, 
-                d_aux_, 
-                output->getData(memspace_)); 
-    mem_.deviceSynchronize();
     return 0;
   }