diff --git a/codes/codes.h b/codes/codes.h
index 002c66eb4ad420dfa7e703d4879e64481c3d227b..a65ce2c6b7273984e31cc44049c9ff9e65a11948 100644
--- a/codes/codes.h
+++ b/codes/codes.h
@@ -97,6 +97,8 @@ static inline void codes_local_latency_reverse(tw_lp *lp)
     return;
 }
 
+void codes_comm_update();
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/network-workloads/conf/modelnet-mpi-test-dfly-amg-1728.conf b/src/network-workloads/conf/modelnet-mpi-test-dfly-amg-1728.conf
index ed57d432bc8a3ad88a2ff4b34129977d7ed5fbc3..f393cbf2db73bbdcd2c55f6ae5ce812b38d14f0d 100644
--- a/src/network-workloads/conf/modelnet-mpi-test-dfly-amg-1728.conf
+++ b/src/network-workloads/conf/modelnet-mpi-test-dfly-amg-1728.conf
@@ -23,6 +23,6 @@ PARAMS
    local_bandwidth="5.25";
    global_bandwidth="4.7";
    cn_bandwidth="5.25";
-   message_size="592";
+   message_size="608";
    routing="adaptive";
 }
diff --git a/src/network-workloads/model-net-dumpi-traces-dump.c b/src/network-workloads/model-net-dumpi-traces-dump.c
index 80fcf981f6b0436cd07d68f0b4a817cbaf390476..c1b3944c56ae8f970c0fb6236b1df575093e090e 100644
--- a/src/network-workloads/model-net-dumpi-traces-dump.c
+++ b/src/network-workloads/model-net-dumpi-traces-dump.c
@@ -414,6 +414,11 @@ int main( int argc, char** argv )
   workload_type[0]='\0';
   tw_opt_add(app_opt);
   tw_init(&argc, &argv);
+#ifdef USE_RDAMARIS
+    if(g_st_ross_rank)
+    { // keep damaris ranks from running code between here up until tw_end()
+#endif
+  codes_comm_update();
 
   if(strlen(workload_file) == 0)
     {
@@ -423,10 +428,10 @@ int main( int argc, char** argv )
 	return -1;
     }
 
-    MPI_Comm_rank(MPI_COMM_WORLD, &rank);
-    MPI_Comm_size(MPI_COMM_WORLD, &nprocs);
+    MPI_Comm_rank(MPI_COMM_CODES, &rank);
+    MPI_Comm_size(MPI_COMM_CODES, &nprocs);
 
-   configuration_load(argv[2], MPI_COMM_WORLD, &config);
+   configuration_load(argv[2], MPI_COMM_CODES, &config);
 
    nw_add_lp_type();
 
@@ -446,20 +451,20 @@ int main( int argc, char** argv )
     double total_avg_comp_time;
     long overall_sends, overall_recvs, overall_waits, overall_cols;
 	
-    MPI_Reduce(&num_bytes_sent, &total_bytes_sent, 1, MPI_LONG_LONG, MPI_SUM, 0, MPI_COMM_WORLD);
-    MPI_Reduce(&num_bytes_recvd, &total_bytes_recvd, 1, MPI_LONG_LONG, MPI_SUM, 0, MPI_COMM_WORLD);
-   MPI_Reduce(&avg_time, &avg_run_time, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_WORLD);
-
-   MPI_Reduce(&avg_recv_time, &total_avg_recv_time, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_WORLD);
-   MPI_Reduce(&avg_comm_time, &avg_comm_run_time, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_WORLD);
-   MPI_Reduce(&avg_col_time, &avg_col_run_time, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_WORLD);
-  MPI_Reduce(&avg_wait_time, &total_avg_wait_time, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_WORLD);
-   MPI_Reduce(&avg_send_time, &total_avg_send_time, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_WORLD);
-   MPI_Reduce(&avg_compute_time, &total_avg_comp_time, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_WORLD);
-   MPI_Reduce(&total_sends, &overall_sends, 1, MPI_LONG, MPI_SUM, 0, MPI_COMM_WORLD);
-   MPI_Reduce(&total_recvs, &overall_recvs, 1, MPI_LONG, MPI_SUM, 0, MPI_COMM_WORLD);
-   MPI_Reduce(&total_waits, &overall_waits, 1, MPI_LONG, MPI_SUM, 0, MPI_COMM_WORLD);
-   MPI_Reduce(&total_collectives, &overall_cols, 1, MPI_LONG, MPI_SUM, 0, MPI_COMM_WORLD);
+    MPI_Reduce(&num_bytes_sent, &total_bytes_sent, 1, MPI_LONG_LONG, MPI_SUM, 0, MPI_COMM_CODES);
+    MPI_Reduce(&num_bytes_recvd, &total_bytes_recvd, 1, MPI_LONG_LONG, MPI_SUM, 0, MPI_COMM_CODES);
+   MPI_Reduce(&avg_time, &avg_run_time, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_CODES);
+
+   MPI_Reduce(&avg_recv_time, &total_avg_recv_time, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_CODES);
+   MPI_Reduce(&avg_comm_time, &avg_comm_run_time, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_CODES);
+   MPI_Reduce(&avg_col_time, &avg_col_run_time, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_CODES);
+  MPI_Reduce(&avg_wait_time, &total_avg_wait_time, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_CODES);
+   MPI_Reduce(&avg_send_time, &total_avg_send_time, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_CODES);
+   MPI_Reduce(&avg_compute_time, &total_avg_comp_time, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_CODES);
+   MPI_Reduce(&total_sends, &overall_sends, 1, MPI_LONG, MPI_SUM, 0, MPI_COMM_CODES);
+   MPI_Reduce(&total_recvs, &overall_recvs, 1, MPI_LONG, MPI_SUM, 0, MPI_COMM_CODES);
+   MPI_Reduce(&total_waits, &overall_waits, 1, MPI_LONG, MPI_SUM, 0, MPI_COMM_CODES);
+   MPI_Reduce(&total_collectives, &overall_cols, 1, MPI_LONG, MPI_SUM, 0, MPI_COMM_CODES);
 
    if(!g_tw_mynode)
 	printf("\n Total bytes sent %lld recvd %lld \n avg runtime %lf \n avg comm time %lf avg compute time %lf \n avg send time %lf \n avg recv time %lf \n avg wait time %lf \n total sends %ld total recvs %ld total waits %ld total collectives %ld ", total_bytes_sent, total_bytes_recvd, 
@@ -470,6 +475,9 @@ int main( int argc, char** argv )
 			total_avg_recv_time/num_net_lps,
 			total_avg_wait_time/num_net_lps,
 			overall_sends, overall_recvs, overall_waits, overall_cols);
+#ifdef USE_RDAMARIS
+    } // end if(g_st_ross_rank)
+#endif
    tw_end();
   
   return 0;
diff --git a/src/network-workloads/model-net-mpi-replay.c b/src/network-workloads/model-net-mpi-replay.c
index 4053829f1e1532e7813d0ba48bbca289dc4e631f..801ebcb8732e945e425e5c1762c80bb4319557b2 100644
--- a/src/network-workloads/model-net-mpi-replay.c
+++ b/src/network-workloads/model-net-mpi-replay.c
@@ -2626,6 +2626,11 @@ int modelnet_mpi_replay(MPI_Comm comm, int* argc, char*** argv )
   workload_type[0]='\0';
   tw_opt_add(app_opt);
   tw_init(argc, argv);
+#ifdef USE_RDAMARIS
+    if(g_st_ross_rank)
+    { // keep damaris ranks from running code between here up until tw_end()
+#endif
+  codes_comm_update();
 
   if(strcmp(workload_type, "dumpi") != 0)
     {
@@ -2793,7 +2798,7 @@ int modelnet_mpi_replay(MPI_Comm comm, int* argc, char*** argv )
    MPI_Reduce(&max_recv_time, &total_max_recv_time, 1, MPI_DOUBLE, MPI_MAX, 0, MPI_COMM_CODES);
    MPI_Reduce(&avg_wait_time, &total_avg_wait_time, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_CODES);
    MPI_Reduce(&avg_send_time, &total_avg_send_time, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_CODES);
-   MPI_Reduce(&total_syn_data, &g_total_syn_data, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_WORLD);  
+   MPI_Reduce(&total_syn_data, &g_total_syn_data, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_CODES);  
 
    assert(num_net_traces);
 
@@ -2822,6 +2827,9 @@ int modelnet_mpi_replay(MPI_Comm comm, int* argc, char*** argv )
    if(alloc_spec)
        codes_jobmap_destroy(jobmap_ctx);
 
+#ifdef USE_RDAMARIS
+    } // end if(g_st_ross_rank)
+#endif
    tw_end();
 
   return 0;
diff --git a/src/network-workloads/model-net-synthetic-custom-dfly.c b/src/network-workloads/model-net-synthetic-custom-dfly.c
index 43212c1bcb24d69153b9767ef8b4acc6cb90c8e1..a07ebedc4459a7e1a208448e4454b8f6f798a660 100644
--- a/src/network-workloads/model-net-synthetic-custom-dfly.c
+++ b/src/network-workloads/model-net-synthetic-custom-dfly.c
@@ -407,6 +407,11 @@ int main(
 
     tw_opt_add(app_opt);
     tw_init(&argc, &argv);
+#ifdef USE_RDAMARIS
+    if(g_st_ross_rank)
+    { // keep damaris ranks from running code between here up until tw_end()
+#endif
+    codes_comm_update();
 
     if(argc < 2)
     {
@@ -470,6 +475,9 @@ int main(
         assert(ret == 0 || !"lp_io_flush failure");
     }
     model_net_report_stats(net_id);
+#ifdef USE_RDAMARIS
+    } // end if(g_st_ross_rank)
+#endif
     tw_end();
     return 0;
 }
diff --git a/src/network-workloads/model-net-synthetic-fattree.c b/src/network-workloads/model-net-synthetic-fattree.c
index a4283068a07aa4499d5f1f758d25546667d80f0e..1fde570d3d1a1405aa11cd53599fac6c270a2931 100644
--- a/src/network-workloads/model-net-synthetic-fattree.c
+++ b/src/network-workloads/model-net-synthetic-fattree.c
@@ -436,6 +436,11 @@ int main(
     tw_opt_add(app_opt);
 
     tw_init(&argc, &argv);
+#ifdef USE_RDAMARIS
+    if(g_st_ross_rank)
+    { // keep damaris ranks from running code between here up until tw_end()
+#endif
+    codes_comm_update();
 
     offset = 1;
 
@@ -446,10 +451,10 @@ int main(
             return 0;
     }
 
-    MPI_Comm_rank(MPI_COMM_WORLD, &rank);
-    MPI_Comm_size(MPI_COMM_WORLD, &nprocs);
+    MPI_Comm_rank(MPI_COMM_CODES, &rank);
+    MPI_Comm_size(MPI_COMM_CODES, &nprocs);
 
-    configuration_load(argv[2], MPI_COMM_WORLD, &config);
+    configuration_load(argv[2], MPI_COMM_CODES, &config);
 
     model_net_register();
 
@@ -484,7 +489,7 @@ int main(
 
     printf("num_nodes:%d \n",num_nodes);
 
-    if(lp_io_prepare("modelnet-test", LP_IO_UNIQ_SUFFIX, &handle, MPI_COMM_WORLD) < 0)
+    if(lp_io_prepare("modelnet-test", LP_IO_UNIQ_SUFFIX, &handle, MPI_COMM_CODES) < 0)
     {
         return(-1);
     }
@@ -516,11 +521,13 @@ int main(
     }
 #endif
 
-    if(lp_io_flush(handle, MPI_COMM_WORLD) < 0)
+    if(lp_io_flush(handle, MPI_COMM_CODES) < 0)
     {
         return(-1);
     }
-
+#ifdef USE_RDAMARIS
+    } // end if(g_st_ross_rank)
+#endif
     tw_end();
 
 #if PARAMS_LOG
diff --git a/src/network-workloads/model-net-synthetic-slimfly.c b/src/network-workloads/model-net-synthetic-slimfly.c
index 098bccbf7d92a877c8e30a9cd42fc1b0a394178f..ab1f1af0a197dbb3c5d70a2900a3e12204f461d3 100644
--- a/src/network-workloads/model-net-synthetic-slimfly.c
+++ b/src/network-workloads/model-net-synthetic-slimfly.c
@@ -494,6 +494,11 @@ int main(
  
     tw_opt_add(app_opt);
     tw_init(&argc, &argv);
+#ifdef USE_RDAMARIS
+    if(g_st_ross_rank)
+    { // keep damaris ranks from running code between here up until tw_end()
+#endif
+    codes_comm_update();
 
     if(argc < 2)
     {
@@ -502,10 +507,10 @@ int main(
             return 0;
     }
 
-    MPI_Comm_rank(MPI_COMM_WORLD, &rank);
-    MPI_Comm_size(MPI_COMM_WORLD, &nprocs);
+    MPI_Comm_rank(MPI_COMM_CODES, &rank);
+    MPI_Comm_size(MPI_COMM_CODES, &nprocs);
 
-    configuration_load(argv[2], MPI_COMM_WORLD, &config);
+    configuration_load(argv[2], MPI_COMM_CODES, &config);
     model_net_register();
     svr_add_lp_type();
     
@@ -526,7 +531,7 @@ int main(
     num_nodes_per_grp = num_routers_per_grp * num_servers_per_rep;
     total_routers = num_routers_per_grp * num_routers_per_grp * 2;
 
-/*    if(lp_io_prepare("modelnet-test", LP_IO_UNIQ_SUFFIX, &handle, MPI_COMM_WORLD) < 0)
+/*    if(lp_io_prepare("modelnet-test", LP_IO_UNIQ_SUFFIX, &handle, MPI_COMM_CODES) < 0)
     {
         return(-1);
     }
@@ -536,7 +541,7 @@ int main(
     {
         do_lp_io = 1;
         int flags = lp_io_use_suffix ? LP_IO_UNIQ_SUFFIX : 0;
-        int ret = lp_io_prepare(lp_io_dir, flags, &io_handle, MPI_COMM_WORLD);
+        int ret = lp_io_prepare(lp_io_dir, flags, &io_handle, MPI_COMM_CODES);
         assert(ret == 0 || !"lp_io_prepare failure");
     }
 
@@ -558,7 +563,7 @@ int main(
 
  
    if (do_lp_io){
-       int ret = lp_io_flush(io_handle, MPI_COMM_WORLD);
+       int ret = lp_io_flush(io_handle, MPI_COMM_CODES);
        assert(ret == 0 || !"lp_io_flush failure");
    }
 
@@ -578,12 +583,15 @@ int main(
 #endif
     }
 
-/*    if(lp_io_flush(handle, MPI_COMM_WORLD) < 0)
+/*    if(lp_io_flush(handle, MPI_COMM_CODES) < 0)
     {
         assert(ret == 0 || !"lp_io_flush failure");
         return(-1);
     }
 */
+#ifdef USE_RDAMARIS
+    } // end if(g_st_ross_rank)
+#endif
     tw_end();
 
     if(rank == 0)
diff --git a/src/network-workloads/model-net-synthetic.c b/src/network-workloads/model-net-synthetic.c
index 65b95a6108f875a7c7352abad2b1f39f7f00cb7b..17728894710f189d6efb2b795fe5c909b1221fd6 100644
--- a/src/network-workloads/model-net-synthetic.c
+++ b/src/network-workloads/model-net-synthetic.c
@@ -407,6 +407,11 @@ int main(
 
     tw_opt_add(app_opt);
     tw_init(&argc, &argv);
+#ifdef USE_RDAMARIS
+    if(g_st_ross_rank)
+    { // keep damaris ranks from running code between here up until tw_end()
+#endif
+    codes_comm_update();
 
     if(argc < 2)
     {
@@ -415,10 +420,10 @@ int main(
             return 0;
     }
 
-    MPI_Comm_rank(MPI_COMM_WORLD, &rank);
-    MPI_Comm_size(MPI_COMM_WORLD, &nprocs);
+    MPI_Comm_rank(MPI_COMM_CODES, &rank);
+    MPI_Comm_size(MPI_COMM_CODES, &nprocs);
 
-    configuration_load(argv[2], MPI_COMM_WORLD, &config);
+    configuration_load(argv[2], MPI_COMM_CODES, &config);
 
     model_net_register();
     svr_add_lp_type();
@@ -457,15 +462,18 @@ int main(
     {
         do_lp_io = 1;
         int flags = lp_io_use_suffix ? LP_IO_UNIQ_SUFFIX : 0;
-        int ret = lp_io_prepare(lp_io_dir, flags, &io_handle, MPI_COMM_WORLD);
+        int ret = lp_io_prepare(lp_io_dir, flags, &io_handle, MPI_COMM_CODES);
         assert(ret == 0 || !"lp_io_prepare failure");
     }
     tw_run();
     if (do_lp_io){
-        int ret = lp_io_flush(io_handle, MPI_COMM_WORLD);
+        int ret = lp_io_flush(io_handle, MPI_COMM_CODES);
         assert(ret == 0 || !"lp_io_flush failure");
     }
     model_net_report_stats(net_id);
+#ifdef USE_RDAMARIS
+    } // end if(g_st_ross_rank)
+#endif
     tw_end();
     return 0;
 }
diff --git a/src/networks/model-net/express-mesh.C b/src/networks/model-net/express-mesh.C
index 3cb247f4253098c79d9046789a6eac80bc5e8a23..b39b9d1c8796a50fadd64ceb9a314be7f6e5e850 100644
--- a/src/networks/model-net/express-mesh.C
+++ b/src/networks/model-net/express-mesh.C
@@ -435,7 +435,7 @@ static void local_read_config(const char * anno, local_param *params){
   p->total_terminals = p->total_routers * p->num_cn;
 
   int rank;
-  MPI_Comm_rank(MPI_COMM_WORLD, &rank);
+  MPI_Comm_rank(MPI_COMM_CODES, &rank);
   if(!rank) {
     printf("\n Total nodes %d routers %d radix %d \n",
         p->total_terminals, p->total_routers, p->radix);
@@ -480,22 +480,22 @@ static void local_report_stats()
   long total_gen, total_fin;
 
   MPI_Reduce( &total_hops, &avg_hops, 1, MPI_LONG_LONG, MPI_SUM, 0,
-      MPI_COMM_WORLD);
+      MPI_COMM_CODES);
   MPI_Reduce( &N_finished_packets, &total_finished_packets, 1, MPI_LONG_LONG,
-      MPI_SUM, 0, MPI_COMM_WORLD);
+      MPI_SUM, 0, MPI_COMM_CODES);
   MPI_Reduce( &N_finished_msgs, &total_finished_msgs, 1, MPI_LONG_LONG, MPI_SUM,
-      0, MPI_COMM_WORLD);
+      0, MPI_COMM_CODES);
   MPI_Reduce( &N_finished_chunks, &total_finished_chunks, 1, MPI_LONG_LONG,
-      MPI_SUM, 0, MPI_COMM_WORLD);
+      MPI_SUM, 0, MPI_COMM_CODES);
   MPI_Reduce( &total_msg_sz, &final_msg_sz, 1, MPI_LONG_LONG, MPI_SUM, 0,
-      MPI_COMM_WORLD);
+      MPI_COMM_CODES);
   MPI_Reduce( &local_total_time, &avg_time, 1,MPI_DOUBLE, MPI_SUM, 0,
-      MPI_COMM_WORLD);
+      MPI_COMM_CODES);
   MPI_Reduce( &local_max_latency, &max_time, 1, MPI_DOUBLE, MPI_MAX, 0,
-      MPI_COMM_WORLD);
+      MPI_COMM_CODES);
 
-  MPI_Reduce( &packet_gen, &total_gen, 1, MPI_LONG, MPI_SUM, 0, MPI_COMM_WORLD);
-  MPI_Reduce( &packet_fin, &total_fin, 1, MPI_LONG, MPI_SUM, 0, MPI_COMM_WORLD);
+  MPI_Reduce( &packet_gen, &total_gen, 1, MPI_LONG, MPI_SUM, 0, MPI_COMM_CODES);
+  MPI_Reduce( &packet_fin, &total_fin, 1, MPI_LONG, MPI_SUM, 0, MPI_COMM_CODES);
 
   /* print statistics */
   if(!g_tw_mynode)
diff --git a/src/networks/model-net/net-template.C b/src/networks/model-net/net-template.C
index 988ba0359d37bbd4f0ee4f01a007ff18367fe74b..6a3c867c9eca62d259035a7522a155f61758b5df 100644
--- a/src/networks/model-net/net-template.C
+++ b/src/networks/model-net/net-template.C
@@ -401,22 +401,22 @@ static void local_report_stats()
   long total_gen, total_fin;
 
   MPI_Reduce( &total_hops, &avg_hops, 1, MPI_LONG_LONG, MPI_SUM, 0,
-      MPI_COMM_WORLD);
+      MPI_COMM_CODES);
   MPI_Reduce( &N_finished_packets, &total_finished_packets, 1, MPI_LONG_LONG,
-      MPI_SUM, 0, MPI_COMM_WORLD);
+      MPI_SUM, 0, MPI_COMM_CODES);
   MPI_Reduce( &N_finished_msgs, &total_finished_msgs, 1, MPI_LONG_LONG, MPI_SUM,
-      0, MPI_COMM_WORLD);
+      0, MPI_COMM_CODES);
   MPI_Reduce( &N_finished_chunks, &total_finished_chunks, 1, MPI_LONG_LONG,
-      MPI_SUM, 0, MPI_COMM_WORLD);
+      MPI_SUM, 0, MPI_COMM_CODES);
   MPI_Reduce( &total_msg_sz, &final_msg_sz, 1, MPI_LONG_LONG, MPI_SUM, 0,
-      MPI_COMM_WORLD);
+      MPI_COMM_CODES);
   MPI_Reduce( &local_total_time, &avg_time, 1,MPI_DOUBLE, MPI_SUM, 0,
-      MPI_COMM_WORLD);
+      MPI_COMM_CODES);
   MPI_Reduce( &local_max_latency, &max_time, 1, MPI_DOUBLE, MPI_MAX, 0,
-      MPI_COMM_WORLD);
+      MPI_COMM_CODES);
 
-  MPI_Reduce( &packet_gen, &total_gen, 1, MPI_LONG, MPI_SUM, 0, MPI_COMM_WORLD);
-  MPI_Reduce( &packet_fin, &total_fin, 1, MPI_LONG, MPI_SUM, 0, MPI_COMM_WORLD);
+  MPI_Reduce( &packet_gen, &total_gen, 1, MPI_LONG, MPI_SUM, 0, MPI_COMM_CODES);
+  MPI_Reduce( &packet_fin, &total_fin, 1, MPI_LONG, MPI_SUM, 0, MPI_COMM_CODES);
 
   /* print statistics */
   if(!g_tw_mynode)
diff --git a/src/networks/model-net/slimfly.c b/src/networks/model-net/slimfly.c
index 047adcb26cd6db1d9a2e2fe6ba7f4357b54d646c..79441f5e93831b199451fa3f95d9ca5749bd0ccf 100644
--- a/src/networks/model-net/slimfly.c
+++ b/src/networks/model-net/slimfly.c
@@ -638,19 +638,19 @@ static void slimfly_report_stats()
     float throughput_avg = 0.0;
     float throughput_avg2 = 0.0;
 
-    MPI_Reduce( &total_hops, &avg_hops, 1, MPI_LONG_LONG, MPI_SUM, 0, MPI_COMM_WORLD);
-    MPI_Reduce( &N_finished_packets, &total_finished_packets, 1, MPI_LONG_LONG, MPI_SUM, 0, MPI_COMM_WORLD);
-    MPI_Reduce( &N_finished_msgs, &total_finished_msgs, 1, MPI_LONG_LONG, MPI_SUM, 0, MPI_COMM_WORLD);
-    MPI_Reduce( &N_finished_chunks, &total_finished_chunks, 1, MPI_LONG_LONG, MPI_SUM, 0, MPI_COMM_WORLD);
-    MPI_Reduce( &total_msg_sz, &final_msg_sz, 1, MPI_LONG_LONG, MPI_SUM, 0, MPI_COMM_WORLD);
-    MPI_Reduce( &slimfly_total_time, &avg_time, 1,MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_WORLD);
-    MPI_Reduce( &slimfly_max_latency, &max_time, 1, MPI_DOUBLE, MPI_MAX, 0, MPI_COMM_WORLD);
-
-    MPI_Reduce(&pe_throughput_percent, &throughput_avg, 1, MPI_FLOAT, MPI_SUM, 0, MPI_COMM_WORLD);
-    MPI_Reduce(&pe_throughput, &throughput_avg2, 1, MPI_FLOAT, MPI_SUM, 0, MPI_COMM_WORLD);
-
-    MPI_Reduce(&minimal_count, &total_minimal_packets, 1, MPI_INT, MPI_SUM, 0, MPI_COMM_WORLD);
-    MPI_Reduce(&nonmin_count, &total_nonmin_packets, 1, MPI_INT, MPI_SUM, 0, MPI_COMM_WORLD);
+    MPI_Reduce( &total_hops, &avg_hops, 1, MPI_LONG_LONG, MPI_SUM, 0, MPI_COMM_CODES);
+    MPI_Reduce( &N_finished_packets, &total_finished_packets, 1, MPI_LONG_LONG, MPI_SUM, 0, MPI_COMM_CODES);
+    MPI_Reduce( &N_finished_msgs, &total_finished_msgs, 1, MPI_LONG_LONG, MPI_SUM, 0, MPI_COMM_CODES);
+    MPI_Reduce( &N_finished_chunks, &total_finished_chunks, 1, MPI_LONG_LONG, MPI_SUM, 0, MPI_COMM_CODES);
+    MPI_Reduce( &total_msg_sz, &final_msg_sz, 1, MPI_LONG_LONG, MPI_SUM, 0, MPI_COMM_CODES);
+    MPI_Reduce( &slimfly_total_time, &avg_time, 1,MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_CODES);
+    MPI_Reduce( &slimfly_max_latency, &max_time, 1, MPI_DOUBLE, MPI_MAX, 0, MPI_COMM_CODES);
+
+    MPI_Reduce(&pe_throughput_percent, &throughput_avg, 1, MPI_FLOAT, MPI_SUM, 0, MPI_COMM_CODES);
+    MPI_Reduce(&pe_throughput, &throughput_avg2, 1, MPI_FLOAT, MPI_SUM, 0, MPI_COMM_CODES);
+
+    MPI_Reduce(&minimal_count, &total_minimal_packets, 1, MPI_INT, MPI_SUM, 0, MPI_COMM_CODES);
+    MPI_Reduce(&nonmin_count, &total_nonmin_packets, 1, MPI_INT, MPI_SUM, 0, MPI_COMM_CODES);
 
     /* print statistics */
     if(!g_tw_mynode)
diff --git a/src/util/codes-comm.c b/src/util/codes-comm.c
index d0bc4159df98334c8950a694462fca5e88635a4a..ef59a753b1556033b762c3d2acf3d00379c81b46 100644
--- a/src/util/codes-comm.c
+++ b/src/util/codes-comm.c
@@ -5,9 +5,19 @@
  */
 
 #include <mpi.h>
+#include <ross.h>
 
 MPI_Comm MPI_COMM_CODES = MPI_COMM_WORLD;
 
+/*
+ * Needs to be called AFTER tw_init() because in tw_init, 
+ * ROSS may split the MPI_COMM_ROSS communicator
+ */
+void codes_comm_update()
+{
+    MPI_COMM_CODES = MPI_COMM_ROSS;
+}
+
 /*
  * Local variables:
  *  c-indent-level: 4