diff --git a/Makefile b/Makefile
index 6b529a5f965f84d634c6922264e190e6d6c92f8c..318842cba5834abd6b2d18df7a959d61c0eea0ef 100644
--- a/Makefile
+++ b/Makefile
@@ -30,7 +30,7 @@ endif
 
 #OPT += -DNVIDIA
 # perform one-side communication (suggested) instead of reduce (only if MPI is active)
-#OPT += -DONE_SIDE
+OPT += -DONE_SIDE
 # write the full 3D cube of gridded visibilities and its FFT transform
 #OPT += -DWRITE_DATA
 # write the final image
@@ -39,8 +39,8 @@ OPT += -DWRITE_IMAGE
 OPT += -DPHASE_ON
 
 
-DEPS = w-stacking.h main.c w-stacking.cu phase_correction.cu allvars.h init.c gridding.c fourier_transform.c result.c
-COBJ = w-stacking.o main.o phase_correction.o allvars.o init.o gridding.o fourier_transform.o result.o
+DEPS = w-stacking.h main.c w-stacking.cu phase_correction.cu allvars.h init.c gridding.c fourier_transform.c result.c numa.h
+COBJ = w-stacking.o main.o phase_correction.o allvars.o init.o gridding.o fourier_transform.o result.o numa.o
 
 w-stacking.c: w-stacking.cu
 	cp w-stacking.cu w-stacking.c
diff --git a/allvars.c b/allvars.c
index 95d09f53777038f4830cceb17c184a368e2908e1..8e67a3c12c5d51c0b102f1c61e060b662a447196 100644
--- a/allvars.c
+++ b/allvars.c
@@ -27,7 +27,8 @@ long * histo_send, size_of_grid;
 double * grid, *gridss, *gridss_real, *gridss_img, *gridss_w;
 
 #ifdef USE_MPI
-    MPI_Win slabwin;
+      MPI_Comm MYMPI_COMM_WORLD;
+      MPI_Win slabwin;
 #endif
 
 long **sectorarray;
diff --git a/allvars.h b/allvars.h
index 4edb09c690204193e7a01b3cd2fdda241ef595af..2f66fc12b14ddf063a573cbf284f6aab6f4bc02b 100644
--- a/allvars.h
+++ b/allvars.h
@@ -27,6 +27,8 @@
 #include <math.h>
 #include <time.h>
 #include <unistd.h>
+#include "numa.h"
+
 
 extern struct io
 {
@@ -128,6 +130,7 @@ extern long * histo_send, size_of_grid;
 extern double * grid, *gridss, *gridss_real, *gridss_img, *gridss_w;
 
 #ifdef USE_MPI
+    extern MPI_Comm MYMPI_COMM_WORLD;
     extern  MPI_Win slabwin;
 #endif
 
diff --git a/gridding.c b/gridding.c
index 1b57c94986db0e5ff906e86d03491b4f46c99de6..de4cc6e338f79e41fac112773c85bf416419839d 100644
--- a/gridding.c
+++ b/gridding.c
@@ -248,6 +248,7 @@ void gridding_data(){
        
       #ifdef ONE_SIDE
        printf("One Side communication active\n");
+       memcpy(Me.win.ptr, gridss, size_of_grid);
        MPI_Win_lock(MPI_LOCK_SHARED,target_rank,0,slabwin);
        MPI_Accumulate(gridss,size_of_grid,MPI_DOUBLE,target_rank,0,size_of_grid,MPI_DOUBLE,MPI_SUM,slabwin);
        MPI_Win_unlock(target_rank,slabwin);
diff --git a/init.c b/init.c
index 35711ccffe1a610e1f92e28d462dd7b3a2d27b91..fc9297d64f5c0f3a7c577c54ae592904adeff14d 100644
--- a/init.c
+++ b/init.c
@@ -30,6 +30,10 @@ void init(int index)
    xaxis = local_grid_size_x;
    yaxis = local_grid_size_y;
 
+   #ifdef USE_MPI
+   	init_numa( rank, size, &MYMPI_COMM_WORLD, &Me );
+   #endif
+
    clock_gettime(CLOCK_MONOTONIC, &begin);
    start = clock();
 
diff --git a/main.c b/main.c
index c2fb72ae039c02ae61b5f95b0ed78acb2042a3c8..6bf7d013f4e90f1b191aad3ab50881837b83f458 100644
--- a/main.c
+++ b/main.c
@@ -28,6 +28,7 @@ int main(int argc, char * argv[])
 		#ifdef USE_FFTW
 			fftw_mpi_init();
   		#endif
+                MPI_Comm_dup(MPI_COMM_WORLD, &MYMPI_COMM_WORLD);
 	#else
 		rank = 0;
 		size = 1;
diff --git a/proto.h b/proto.h
index 977f2f03958392f7512ba72eee1ba285b3a6f602..5b6aadc62545976abe7ccb2662158e1c91d0c1be 100644
--- a/proto.h
+++ b/proto.h
@@ -1,6 +1,10 @@
 /* function declaration */
 
 
+/* numa.c */
+int init_numa( int, int, MPI_Comm *, map_t *);
+int shutdown_numa( int, int, MPI_Comm *, map_t *);
+
 /* init.c */
 
 void init(int i);