diff --git a/lib/FFT.h b/lib/FFT.h
index cccd1da5..a16e775b 100644
--- a/lib/FFT.h
+++ b/lib/FFT.h
@@ -226,12 +226,18 @@ namespace Grid {
       std::vector<int> lcoor(Nd), gcoor(Nd);
       result = source;
       for(int p=0;p<processors[dim];p++) {
-        for(int idx=0;idx<sgrid->lSites();idx++) {
-          sgrid->LocalIndexToLocalCoor(idx,lcoor);
+        PARALLEL_REGION
+        {
+          std::vector<int> cbuf(Nd);
           sobj s;
-          peekLocalSite(s,result,lcoor);
-          lcoor[dim]+=p*L;
-          pokeLocalSite(s,pgbuf,lcoor);
+          
+          PARALLEL_FOR_LOOP_INTERN
+          for(int idx=0;idx<sgrid->lSites();idx++) {
+            sgrid->LocalIndexToLocalCoor(idx,cbuf);
+            peekLocalSite(s,result,cbuf);
+            cbuf[dim]+=p*L;
+            pokeLocalSite(s,pgbuf,cbuf);
+          }
         }
         result = Cshift(result,dim,L);
       }
@@ -240,14 +246,18 @@ namespace Grid {
       int NN=pencil_g.lSites();
       GridStopWatch timer;
       timer.Start();
-      //PARALLEL_FOR_LOOP
-      for(int idx=0;idx<NN;idx++) {
-        pencil_g.LocalIndexToLocalCoor(idx,lcoor);
+      PARALLEL_REGION
+      {
+        std::vector<int> cbuf(Nd);
         
-        if ( lcoor[dim] == 0 ) {  // restricts loop to plane at lcoor[dim]==0
-          FFTW_scalar *in = (FFTW_scalar *)&pgbuf._odata[idx];
-          FFTW_scalar *out= (FFTW_scalar *)&pgbuf._odata[idx];
-          FFTW<scalar>::fftw_execute_dft(p,in,out);
+        PARALLEL_FOR_LOOP_INTERN
+        for(int idx=0;idx<NN;idx++) {
+          pencil_g.LocalIndexToLocalCoor(idx, cbuf);
+          if ( cbuf[dim] == 0 ) {  // restricts loop to plane at lcoor[dim]==0
+            FFTW_scalar *in = (FFTW_scalar *)&pgbuf._odata[idx];
+            FFTW_scalar *out= (FFTW_scalar *)&pgbuf._odata[idx];
+            FFTW<scalar>::fftw_execute_dft(p,in,out);
+          }
         }
       }
       timer.Stop();
@@ -261,14 +271,20 @@ namespace Grid {
       
       // writing out result
       int pc = processor_coor[dim];
-      for(int idx=0;idx<sgrid->lSites();idx++) {
-        sgrid->LocalIndexToLocalCoor(idx,lcoor);
-        gcoor = lcoor;
+      PARALLEL_REGION
+      {
+        std::vector<int> clbuf(Nd), cgbuf(Nd);
         sobj s;
-        gcoor[dim] = lcoor[dim]+L*pc;
-        peekLocalSite(s,pgbuf,gcoor);
-		s = s * div;
-        pokeLocalSite(s,result,lcoor);
+        
+        PARALLEL_FOR_LOOP_INTERN
+        for(int idx=0;idx<sgrid->lSites();idx++) {
+          sgrid->LocalIndexToLocalCoor(idx,clbuf);
+          cgbuf = clbuf;
+          cgbuf[dim] = clbuf[dim]+L*pc;
+          peekLocalSite(s,pgbuf,cgbuf);
+          s = s * div;
+          pokeLocalSite(s,result,clbuf);
+        }
       }
       
       // destroying plan
diff --git a/lib/Threads.h b/lib/Threads.h
index 08e5d545..2f270b73 100644
--- a/lib/Threads.h
+++ b/lib/Threads.h
@@ -38,14 +38,19 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #ifdef GRID_OMP
 #include <omp.h>
 #ifdef GRID_NUMA
-#define PARALLEL_FOR_LOOP _Pragma("omp parallel for schedule(static)")
+#define PARALLEL_FOR_LOOP        _Pragma("omp parallel for schedule(static)")
+#define PARALLEL_FOR_LOOP_INTERN _Pragma("omp for schedule(static)")
 #else
-#define PARALLEL_FOR_LOOP _Pragma("omp parallel for schedule(runtime)")
+#define PARALLEL_FOR_LOOP        _Pragma("omp parallel for schedule(runtime)")
+#define PARALLEL_FOR_LOOP_INTERN _Pragma("omp for schedule(runtime)")
 #endif
 #define PARALLEL_NESTED_LOOP2 _Pragma("omp parallel for collapse(2)")
+#define PARALLEL_REGION       _Pragma("omp parallel")
 #else
-#define PARALLEL_FOR_LOOP 
+#define PARALLEL_FOR_LOOP
+#define PARALLEL_FOR_LOOP_INTERN
 #define PARALLEL_NESTED_LOOP2
+#define PARALLEL_REGION
 #endif
 
 namespace Grid {