diff --git a/maint/gen_coll.py b/maint/gen_coll.py
index afd46f026e6..cf77d6fea5a 100644
--- a/maint/gen_coll.py
+++ b/maint/gen_coll.py
@@ -165,6 +165,8 @@ def dump_allcomm_sched_auto(name):
     dump_split(0, "int MPIR_%s_allcomm_sched_auto(%s)" % (Name, func_params))
     dump_open('{')
     G.out.append("int mpi_errno = MPI_SUCCESS;")
+    if re.match(r'Ineighbor_', Name):
+        G.out.append("int coll_group = MPIR_SUBGROUP_NONE;")
     G.out.append("")
 
     # -- Csel_search
@@ -367,6 +369,8 @@ def dump_cases(commkind):
     dump_split(0, "int MPIR_%s_sched_impl(%s)" % (Name, func_params))
     dump_open('{')
     G.out.append("int mpi_errno = MPI_SUCCESS;")
+    if re.match(r'Ineighbor_', Name):
+        G.out.append("int coll_group = MPIR_SUBGROUP_NONE;")
     G.out.append("")
 
     dump_open("if (comm_ptr->comm_kind == MPIR_COMM_KIND__INTRACOMM) {")
diff --git a/src/include/mpir_nbc.h b/src/include/mpir_nbc.h
index 9320bf5b633..710521d9a5d 100644
--- a/src/include/mpir_nbc.h
+++ b/src/include/mpir_nbc.h
@@ -45,7 +45,7 @@
 /* Open question: should tag allocation be rolled into Sched_start?  Keeping it
  * separate potentially allows more parallelism in the future, but it also
  * pushes more work onto the clients of this interface. */
-int MPIR_Sched_next_tag(MPIR_Comm * comm_ptr, int *tag);
+int MPIR_Sched_next_tag(MPIR_Comm * comm_ptr, int coll_group, int *tag);
 void MPIR_Sched_set_tag(MPIR_Sched_t s, int tag);
 
 /* the device must provide a typedef for MPIR_Sched_t in mpidpre.h */
diff --git a/src/mpi/coll/allreduce/allreduce_intra_ring.c b/src/mpi/coll/allreduce/allreduce_intra_ring.c
index 933d095c27c..f225255acaa 100644
--- a/src/mpi/coll/allreduce/allreduce_intra_ring.c
+++ b/src/mpi/coll/allreduce/allreduce_intra_ring.c
@@ -74,7 +74,7 @@ int MPIR_Allreduce_intra_ring(const void *sendbuf, void *recvbuf, MPI_Aint count
         send_rank = (nranks + rank - 1 - i) % nranks;
 
         /* get a new tag to prevent out of order messages */
-        mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+        mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
         MPIR_ERR_CHECK(mpi_errno);
 
         mpi_errno =
diff --git a/src/mpi/coll/allreduce/allreduce_intra_tree.c b/src/mpi/coll/allreduce/allreduce_intra_tree.c
index 63fe72f3b86..a1614ea8be0 100644
--- a/src/mpi/coll/allreduce/allreduce_intra_tree.c
+++ b/src/mpi/coll/allreduce/allreduce_intra_tree.c
@@ -139,7 +139,7 @@ int MPIR_Allreduce_intra_tree(const void *sendbuf,
         void *reduce_address = (char *) reduce_buffer + offset * extent;
         MPIR_ERR_CHKANDJUMP(!reduce_address, mpi_errno, MPI_ERR_OTHER, "**nomem");
 
-        mpi_errno = MPIR_Sched_next_tag(comm_ptr, &tag);
+        mpi_errno = MPIR_Sched_next_tag(comm_ptr, coll_group, &tag);
         MPIR_ERR_CHECK(mpi_errno);
 
         for (i = 0; i < num_children; i++) {
diff --git a/src/mpi/coll/iallgather/iallgather_tsp_brucks.c b/src/mpi/coll/iallgather/iallgather_tsp_brucks.c
index 9cc062802f0..41fd1395eb2 100644
--- a/src/mpi/coll/iallgather/iallgather_tsp_brucks.c
+++ b/src/mpi/coll/iallgather/iallgather_tsp_brucks.c
@@ -41,7 +41,7 @@ MPIR_TSP_Iallgather_sched_intra_brucks(const void *sendbuf, MPI_Aint sendcount,
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     MPIR_FUNC_ENTER;
diff --git a/src/mpi/coll/iallgather/iallgather_tsp_recexch.c b/src/mpi/coll/iallgather/iallgather_tsp_recexch.c
index 7ce9eddf7f1..ffea3a4ca53 100644
--- a/src/mpi/coll/iallgather/iallgather_tsp_recexch.c
+++ b/src/mpi/coll/iallgather/iallgather_tsp_recexch.c
@@ -250,7 +250,7 @@ int MPIR_TSP_Iallgather_sched_intra_recexch(const void *sendbuf, MPI_Aint sendco
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     is_inplace = (sendbuf == MPI_IN_PLACE);
diff --git a/src/mpi/coll/iallgather/iallgather_tsp_ring.c b/src/mpi/coll/iallgather/iallgather_tsp_ring.c
index 3799f005a0b..f4402cfdd73 100644
--- a/src/mpi/coll/iallgather/iallgather_tsp_ring.c
+++ b/src/mpi/coll/iallgather/iallgather_tsp_ring.c
@@ -84,7 +84,7 @@ int MPIR_TSP_Iallgather_sched_intra_ring(const void *sendbuf, MPI_Aint sendcount
     int recv_id[3] = { 0 };     /* warning fix: icc: maybe used before set */
     for (i = 0; i < size - 1; i++) {
         /* Get new tag for each cycle so that the send-recv pairs are matched correctly */
-        mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+        mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
         MPIR_ERR_CHECK(mpi_errno);
 
         int vtcs[3], nvtcs;
diff --git a/src/mpi/coll/iallgatherv/iallgatherv_tsp_brucks.c b/src/mpi/coll/iallgatherv/iallgatherv_tsp_brucks.c
index 6493eb83aa4..7286a803b8a 100644
--- a/src/mpi/coll/iallgatherv/iallgatherv_tsp_brucks.c
+++ b/src/mpi/coll/iallgatherv/iallgatherv_tsp_brucks.c
@@ -65,7 +65,7 @@ MPIR_TSP_Iallgatherv_sched_intra_brucks(const void *sendbuf, MPI_Aint sendcount,
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     is_inplace = (sendbuf == MPI_IN_PLACE);
diff --git a/src/mpi/coll/iallgatherv/iallgatherv_tsp_recexch.c b/src/mpi/coll/iallgatherv/iallgatherv_tsp_recexch.c
index c468cde9855..74bc261cd22 100644
--- a/src/mpi/coll/iallgatherv/iallgatherv_tsp_recexch.c
+++ b/src/mpi/coll/iallgatherv/iallgatherv_tsp_recexch.c
@@ -269,7 +269,7 @@ int MPIR_TSP_Iallgatherv_sched_intra_recexch(const void *sendbuf, MPI_Aint sendc
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     /* get the neighbors, the function allocates the required memory */
diff --git a/src/mpi/coll/iallgatherv/iallgatherv_tsp_ring.c b/src/mpi/coll/iallgatherv/iallgatherv_tsp_ring.c
index ea5a1f5835c..81ab8373634 100644
--- a/src/mpi/coll/iallgatherv/iallgatherv_tsp_ring.c
+++ b/src/mpi/coll/iallgatherv/iallgatherv_tsp_ring.c
@@ -85,7 +85,7 @@ int MPIR_TSP_Iallgatherv_sched_intra_ring(const void *sendbuf, MPI_Aint sendcoun
         send_rank = (rank - i + nranks) % nranks;       /* Rank whose data you're sending */
 
         /* New tag for each send-recv pair. */
-        mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+        mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
         MPIR_ERR_CHECK(mpi_errno);
 
         int nvtcs, vtcs[3];
diff --git a/src/mpi/coll/iallreduce/iallreduce_tsp_recexch.c b/src/mpi/coll/iallreduce/iallreduce_tsp_recexch.c
index d84906159a8..ad21f497054 100644
--- a/src/mpi/coll/iallreduce/iallreduce_tsp_recexch.c
+++ b/src/mpi/coll/iallreduce/iallreduce_tsp_recexch.c
@@ -51,7 +51,7 @@ int MPIR_TSP_Iallreduce_sched_intra_recexch(const void *sendbuf, void *recvbuf,
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
 
     /* get the neighbors, the function allocates the required memory */
     MPII_Recexchalgo_get_neighbors(rank, nranks, &k, &step1_sendto,
diff --git a/src/mpi/coll/iallreduce/iallreduce_tsp_recexch_reduce_scatter_recexch_allgatherv.c b/src/mpi/coll/iallreduce/iallreduce_tsp_recexch_reduce_scatter_recexch_allgatherv.c
index a05a2bc3d8d..ab54068be6e 100644
--- a/src/mpi/coll/iallreduce/iallreduce_tsp_recexch_reduce_scatter_recexch_allgatherv.c
+++ b/src/mpi/coll/iallreduce/iallreduce_tsp_recexch_reduce_scatter_recexch_allgatherv.c
@@ -55,7 +55,7 @@ int MPIR_TSP_Iallreduce_sched_intra_recexch_reduce_scatter_recexch_allgatherv(co
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     /* get the neighbors, the function allocates the required memory */
diff --git a/src/mpi/coll/iallreduce/iallreduce_tsp_ring.c b/src/mpi/coll/iallreduce/iallreduce_tsp_ring.c
index 7c9fc4c6e24..9675e3e5770 100644
--- a/src/mpi/coll/iallreduce/iallreduce_tsp_ring.c
+++ b/src/mpi/coll/iallreduce/iallreduce_tsp_ring.c
@@ -82,7 +82,7 @@ int MPIR_TSP_Iallreduce_sched_intra_ring(const void *sendbuf, void *recvbuf, MPI
         send_rank = (nranks + rank - 1 - i) % nranks;
 
         /* get a new tag to prevent out of order messages */
-        mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+        mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
         MPIR_ERR_CHECK(mpi_errno);
 
         nvtcs = (i == 0) ? 0 : 1;
diff --git a/src/mpi/coll/iallreduce/iallreduce_tsp_tree.c b/src/mpi/coll/iallreduce/iallreduce_tsp_tree.c
index cb7125bd209..55b65fad562 100644
--- a/src/mpi/coll/iallreduce/iallreduce_tsp_tree.c
+++ b/src/mpi/coll/iallreduce/iallreduce_tsp_tree.c
@@ -118,7 +118,7 @@ int MPIR_TSP_Iallreduce_sched_intra_tree(const void *sendbuf, void *recvbuf, MPI
 
         /* For correctness, transport based collectives need to get the
          * tag from the same pool as schedule based collectives */
-        mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+        mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
         MPIR_ERR_CHECK(mpi_errno);
 
         for (i = 0; i < num_children; i++) {
diff --git a/src/mpi/coll/ialltoall/ialltoall_tsp_brucks.c b/src/mpi/coll/ialltoall/ialltoall_tsp_brucks.c
index 8364e6fa70a..443afb9fac6 100644
--- a/src/mpi/coll/ialltoall/ialltoall_tsp_brucks.c
+++ b/src/mpi/coll/ialltoall/ialltoall_tsp_brucks.c
@@ -144,7 +144,7 @@ MPIR_TSP_Ialltoall_sched_intra_brucks(const void *sendbuf, MPI_Aint sendcount,
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     MPIR_CHKLMEM_MALLOC(pack_invtcs, int *, sizeof(int) * k, mpi_errno, "pack_invtcs",
diff --git a/src/mpi/coll/ialltoall/ialltoall_tsp_ring.c b/src/mpi/coll/ialltoall/ialltoall_tsp_ring.c
index abc685a5456..93e6e8bc22d 100644
--- a/src/mpi/coll/ialltoall/ialltoall_tsp_ring.c
+++ b/src/mpi/coll/ialltoall/ialltoall_tsp_ring.c
@@ -118,7 +118,7 @@ int MPIR_TSP_Ialltoall_sched_intra_ring(const void *sendbuf, MPI_Aint sendcount,
     for (i = 0; i < size - 1; i++) {
         /* For correctness, transport based collectives need to get the
          * tag from the same pool as schedule based collectives */
-        mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+        mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
         MPIR_ERR_CHECK(mpi_errno);
 
         int vtcs[3], nvtcs;
diff --git a/src/mpi/coll/ialltoall/ialltoall_tsp_scattered.c b/src/mpi/coll/ialltoall/ialltoall_tsp_scattered.c
index 2a29f404d4f..75229a33c7b 100644
--- a/src/mpi/coll/ialltoall/ialltoall_tsp_scattered.c
+++ b/src/mpi/coll/ialltoall/ialltoall_tsp_scattered.c
@@ -59,7 +59,7 @@ int MPIR_TSP_Ialltoall_sched_intra_scattered(const void *sendbuf, MPI_Aint sendc
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     MPIR_COLL_RANK_SIZE(comm, coll_group, rank, size);
diff --git a/src/mpi/coll/ialltoallv/ialltoallv_tsp_blocked.c b/src/mpi/coll/ialltoallv/ialltoallv_tsp_blocked.c
index f44f40a1ac0..65a3f70b482 100644
--- a/src/mpi/coll/ialltoallv/ialltoallv_tsp_blocked.c
+++ b/src/mpi/coll/ialltoallv/ialltoallv_tsp_blocked.c
@@ -29,7 +29,7 @@ int MPIR_TSP_Ialltoallv_sched_intra_blocked(const void *sendbuf, const MPI_Aint
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     MPIR_COLL_RANK_SIZE(comm, coll_group, rank, nranks);
diff --git a/src/mpi/coll/ialltoallv/ialltoallv_tsp_inplace.c b/src/mpi/coll/ialltoallv/ialltoallv_tsp_inplace.c
index bcc3757809c..da4383a2401 100644
--- a/src/mpi/coll/ialltoallv/ialltoallv_tsp_inplace.c
+++ b/src/mpi/coll/ialltoallv/ialltoallv_tsp_inplace.c
@@ -28,7 +28,7 @@ int MPIR_TSP_Ialltoallv_sched_intra_inplace(const void *sendbuf, const MPI_Aint
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     MPIR_COLL_RANK_SIZE(comm, coll_group, rank, nranks);
diff --git a/src/mpi/coll/ialltoallv/ialltoallv_tsp_scattered.c b/src/mpi/coll/ialltoallv/ialltoallv_tsp_scattered.c
index 7f98810c224..8f6b8449481 100644
--- a/src/mpi/coll/ialltoallv/ialltoallv_tsp_scattered.c
+++ b/src/mpi/coll/ialltoallv/ialltoallv_tsp_scattered.c
@@ -56,7 +56,7 @@ int MPIR_TSP_Ialltoallv_sched_intra_scattered(const void *sendbuf, const MPI_Ain
     MPIR_Type_get_true_extent_impl(sendtype, &sendtype_lb, &sendtype_true_extent);
     sendtype_extent = MPL_MAX(sendtype_extent, sendtype_true_extent);
 
-    mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     /* First, post bblock number of sends/recvs */
diff --git a/src/mpi/coll/ialltoallw/ialltoallw_tsp_blocked.c b/src/mpi/coll/ialltoallw/ialltoallw_tsp_blocked.c
index 16dc9291189..5957f2c9dbc 100644
--- a/src/mpi/coll/ialltoallw/ialltoallw_tsp_blocked.c
+++ b/src/mpi/coll/ialltoallw/ialltoallw_tsp_blocked.c
@@ -33,7 +33,7 @@ int MPIR_TSP_Ialltoallw_sched_intra_blocked(const void *sendbuf, const MPI_Aint
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     /* post only bblock isends/irecvs at a time as suggested by Tony Ladd */
diff --git a/src/mpi/coll/ialltoallw/ialltoallw_tsp_inplace.c b/src/mpi/coll/ialltoallw/ialltoallw_tsp_inplace.c
index c78237b6d71..fb2b2d51482 100644
--- a/src/mpi/coll/ialltoallw/ialltoallw_tsp_inplace.c
+++ b/src/mpi/coll/ialltoallw/ialltoallw_tsp_inplace.c
@@ -32,7 +32,7 @@ int MPIR_TSP_Ialltoallw_sched_intra_inplace(const void *sendbuf, const MPI_Aint
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     /* FIXME: Here we allocate tmp_buf using extent and send/recv with datatype directly,
diff --git a/src/mpi/coll/ibarrier/ibarrier_intra_tsp_dissem.c b/src/mpi/coll/ibarrier/ibarrier_intra_tsp_dissem.c
index fef0a6fb852..1162d2d46c7 100644
--- a/src/mpi/coll/ibarrier/ibarrier_intra_tsp_dissem.c
+++ b/src/mpi/coll/ibarrier/ibarrier_intra_tsp_dissem.c
@@ -23,7 +23,7 @@ int MPIR_TSP_Ibarrier_sched_intra_k_dissemination(MPIR_Comm * comm, int coll_gro
 
     MPIR_COLL_RANK_SIZE(comm, coll_group, rank, nranks);
 
-    mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
     if (mpi_errno)
         MPIR_ERR_POP(mpi_errno);
 
diff --git a/src/mpi/coll/ibcast/ibcast_tsp_scatterv_allgatherv.c b/src/mpi/coll/ibcast/ibcast_tsp_scatterv_allgatherv.c
index 10da1d1e20d..e8c1a311179 100644
--- a/src/mpi/coll/ibcast/ibcast_tsp_scatterv_allgatherv.c
+++ b/src/mpi/coll/ibcast/ibcast_tsp_scatterv_allgatherv.c
@@ -33,7 +33,7 @@ int MPIR_TSP_Ibcast_sched_intra_scatterv_allgatherv(void *buffer, MPI_Aint count
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     MPIR_FUNC_ENTER;
diff --git a/src/mpi/coll/ibcast/ibcast_tsp_tree.c b/src/mpi/coll/ibcast/ibcast_tsp_tree.c
index 84234a52d01..08e5d132e0d 100644
--- a/src/mpi/coll/ibcast/ibcast_tsp_tree.c
+++ b/src/mpi/coll/ibcast/ibcast_tsp_tree.c
@@ -62,7 +62,7 @@ int MPIR_TSP_Ibcast_sched_intra_tree(void *buffer, MPI_Aint count, MPI_Datatype
 
         /* For correctness, transport based collectives need to get the
          * tag from the same pool as schedule based collectives */
-        mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+        mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
         MPIR_ERR_CHECK(mpi_errno);
 
         /* Receive message from parent */
diff --git a/src/mpi/coll/igather/igather_tsp_tree.c b/src/mpi/coll/igather/igather_tsp_tree.c
index 2d6c9fd3155..43547c59ac1 100644
--- a/src/mpi/coll/igather/igather_tsp_tree.c
+++ b/src/mpi/coll/igather/igather_tsp_tree.c
@@ -46,7 +46,7 @@ int MPIR_TSP_Igather_sched_intra_tree(const void *sendbuf, MPI_Aint sendcount,
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     if (rank == root && is_inplace) {
diff --git a/src/mpi/coll/igatherv/igatherv_tsp_linear.c b/src/mpi/coll/igatherv/igatherv_tsp_linear.c
index 72144384b1d..eebe8e2c9ea 100644
--- a/src/mpi/coll/igatherv/igatherv_tsp_linear.c
+++ b/src/mpi/coll/igatherv/igatherv_tsp_linear.c
@@ -39,7 +39,7 @@ int MPIR_TSP_Igatherv_sched_allcomm_linear(const void *sendbuf, MPI_Aint sendcou
         comm_size = comm_ptr->remote_size;
     }
 
-    mpi_errno = MPIR_Sched_next_tag(comm_ptr, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm_ptr, coll_group, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     /* If rank == root, then I recv lots, otherwise I send */
diff --git a/src/mpi/coll/include/coll_impl.h b/src/mpi/coll/include/coll_impl.h
index b5b576b1b33..cef96e59aeb 100644
--- a/src/mpi/coll/include/coll_impl.h
+++ b/src/mpi/coll/include/coll_impl.h
@@ -75,7 +75,7 @@ int MPII_Coll_finalize(void);
         mpi_errno = MPIR_Sched_create(&s, sched_kind); \
         MPIR_ERR_CHECK(mpi_errno); \
         int tag = -1; \
-        mpi_errno = MPIR_Sched_next_tag(comm_ptr, &tag); \
+        mpi_errno = MPIR_Sched_next_tag(comm_ptr, coll_group, &tag); \
         MPIR_ERR_CHECK(mpi_errno); \
         MPIR_Sched_set_tag(s, tag); \
         *sched_type_p = MPIR_SCHED_NORMAL; \
diff --git a/src/mpi/coll/ineighbor_allgather/ineighbor_allgather_tsp_linear.c b/src/mpi/coll/ineighbor_allgather/ineighbor_allgather_tsp_linear.c
index 4e8e37f9722..2b36dded1b4 100644
--- a/src/mpi/coll/ineighbor_allgather/ineighbor_allgather_tsp_linear.c
+++ b/src/mpi/coll/ineighbor_allgather/ineighbor_allgather_tsp_linear.c
@@ -39,7 +39,7 @@ int MPIR_TSP_Ineighbor_allgather_sched_allcomm_linear(const void *sendbuf, MPI_A
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm_ptr, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm_ptr, MPIR_SUBGROUP_NONE, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     for (k = 0; k < outdegree; ++k) {
diff --git a/src/mpi/coll/ineighbor_allgatherv/ineighbor_allgatherv_tsp_linear.c b/src/mpi/coll/ineighbor_allgatherv/ineighbor_allgatherv_tsp_linear.c
index 57c74bbafcb..4c274394b17 100644
--- a/src/mpi/coll/ineighbor_allgatherv/ineighbor_allgatherv_tsp_linear.c
+++ b/src/mpi/coll/ineighbor_allgatherv/ineighbor_allgatherv_tsp_linear.c
@@ -40,7 +40,7 @@ int MPIR_TSP_Ineighbor_allgatherv_sched_allcomm_linear(const void *sendbuf, MPI_
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm_ptr, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm_ptr, MPIR_SUBGROUP_NONE, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     for (k = 0; k < outdegree; ++k) {
diff --git a/src/mpi/coll/ineighbor_alltoall/ineighbor_alltoall_tsp_linear.c b/src/mpi/coll/ineighbor_alltoall/ineighbor_alltoall_tsp_linear.c
index 0fe1f75d3fa..b48e7e1f431 100644
--- a/src/mpi/coll/ineighbor_alltoall/ineighbor_alltoall_tsp_linear.c
+++ b/src/mpi/coll/ineighbor_alltoall/ineighbor_alltoall_tsp_linear.c
@@ -40,7 +40,7 @@ int MPIR_TSP_Ineighbor_alltoall_sched_allcomm_linear(const void *sendbuf, MPI_Ai
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm_ptr, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm_ptr, MPIR_SUBGROUP_NONE, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
 
diff --git a/src/mpi/coll/ineighbor_alltoallv/ineighbor_alltoallv_tsp_linear.c b/src/mpi/coll/ineighbor_alltoallv/ineighbor_alltoallv_tsp_linear.c
index 40a20ca96f1..eb14bd062b3 100644
--- a/src/mpi/coll/ineighbor_alltoallv/ineighbor_alltoallv_tsp_linear.c
+++ b/src/mpi/coll/ineighbor_alltoallv/ineighbor_alltoallv_tsp_linear.c
@@ -44,7 +44,7 @@ int MPIR_TSP_Ineighbor_alltoallv_sched_allcomm_linear(const void *sendbuf,
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm_ptr, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm_ptr, MPIR_SUBGROUP_NONE, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     for (k = 0; k < outdegree; ++k) {
diff --git a/src/mpi/coll/ineighbor_alltoallw/ineighbor_alltoallw_tsp_linear.c b/src/mpi/coll/ineighbor_alltoallw/ineighbor_alltoallw_tsp_linear.c
index 56908d5cb70..b5de2debd49 100644
--- a/src/mpi/coll/ineighbor_alltoallw/ineighbor_alltoallw_tsp_linear.c
+++ b/src/mpi/coll/ineighbor_alltoallw/ineighbor_alltoallw_tsp_linear.c
@@ -40,7 +40,7 @@ int MPIR_TSP_Ineighbor_alltoallw_sched_allcomm_linear(const void *sendbuf,
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm_ptr, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm_ptr, MPIR_SUBGROUP_NONE, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     for (k = 0; k < outdegree; ++k) {
diff --git a/src/mpi/coll/ireduce/ireduce_tsp_tree.c b/src/mpi/coll/ireduce/ireduce_tsp_tree.c
index 69c582d45ba..8663a58939c 100644
--- a/src/mpi/coll/ireduce/ireduce_tsp_tree.c
+++ b/src/mpi/coll/ireduce/ireduce_tsp_tree.c
@@ -194,7 +194,7 @@ int MPIR_TSP_Ireduce_sched_intra_tree(const void *sendbuf, void *recvbuf, MPI_Ai
 
         /* For correctness, transport based collectives need to get the
          * tag from the same pool as schedule based collectives */
-        mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+        mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
         MPIR_ERR_CHECK(mpi_errno);
 
         for (i = 0; i < num_children; i++) {
diff --git a/src/mpi/coll/ireduce_scatter/ireduce_scatter_tsp_recexch.c b/src/mpi/coll/ireduce_scatter/ireduce_scatter_tsp_recexch.c
index 74ddff69a4f..bcf345527a8 100644
--- a/src/mpi/coll/ireduce_scatter/ireduce_scatter_tsp_recexch.c
+++ b/src/mpi/coll/ireduce_scatter/ireduce_scatter_tsp_recexch.c
@@ -159,7 +159,7 @@ int MPIR_TSP_Ireduce_scatter_sched_intra_recexch(const void *sendbuf, void *recv
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
 
     is_inplace = (sendbuf == MPI_IN_PLACE);
     MPIR_COLL_RANK_SIZE(comm, coll_group, rank, nranks);
diff --git a/src/mpi/coll/ireduce_scatter_block/ireduce_scatter_block_tsp_recexch.c b/src/mpi/coll/ireduce_scatter_block/ireduce_scatter_block_tsp_recexch.c
index e7530c89318..d892bc47ea4 100644
--- a/src/mpi/coll/ireduce_scatter_block/ireduce_scatter_block_tsp_recexch.c
+++ b/src/mpi/coll/ireduce_scatter_block/ireduce_scatter_block_tsp_recexch.c
@@ -33,7 +33,7 @@ int MPIR_TSP_Ireduce_scatter_block_sched_intra_recexch(const void *sendbuf, void
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
 
     is_inplace = (sendbuf == MPI_IN_PLACE);
     MPIR_COLL_RANK_SIZE(comm, coll_group, rank, nranks);
diff --git a/src/mpi/coll/iscan/iscan_tsp_recursive_doubling.c b/src/mpi/coll/iscan/iscan_tsp_recursive_doubling.c
index a8977551823..f17ae5bbf31 100644
--- a/src/mpi/coll/iscan/iscan_tsp_recursive_doubling.c
+++ b/src/mpi/coll/iscan/iscan_tsp_recursive_doubling.c
@@ -29,7 +29,7 @@ int MPIR_TSP_Iscan_sched_intra_recursive_doubling(const void *sendbuf, void *rec
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     MPIR_COLL_RANK_SIZE(comm, coll_group, rank, nranks);
diff --git a/src/mpi/coll/iscatter/iscatter_tsp_tree.c b/src/mpi/coll/iscatter/iscatter_tsp_tree.c
index 3ee4696a049..8703cce266d 100644
--- a/src/mpi/coll/iscatter/iscatter_tsp_tree.c
+++ b/src/mpi/coll/iscatter/iscatter_tsp_tree.c
@@ -48,7 +48,7 @@ int MPIR_TSP_Iscatter_sched_intra_tree(const void *sendbuf, MPI_Aint sendcount,
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm, coll_group, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     if (rank == root && is_inplace) {
diff --git a/src/mpi/coll/iscatterv/iscatterv_tsp_linear.c b/src/mpi/coll/iscatterv/iscatterv_tsp_linear.c
index 96572536842..7152737a5ea 100644
--- a/src/mpi/coll/iscatterv/iscatterv_tsp_linear.c
+++ b/src/mpi/coll/iscatterv/iscatterv_tsp_linear.c
@@ -33,7 +33,7 @@ int MPIR_TSP_Iscatterv_sched_allcomm_linear(const void *sendbuf, const MPI_Aint
 
     /* For correctness, transport based collectives need to get the
      * tag from the same pool as schedule based collectives */
-    mpi_errno = MPIR_Sched_next_tag(comm_ptr, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm_ptr, coll_group, &tag);
     MPIR_ERR_CHECK(mpi_errno);
 
     /* If I'm the root, then scatter */
diff --git a/src/mpi/comm/contextid.c b/src/mpi/comm/contextid.c
index d1d2833a435..d10b9833d6f 100644
--- a/src/mpi/comm/contextid.c
+++ b/src/mpi/comm/contextid.c
@@ -762,7 +762,7 @@ static int sched_cb_gcn_allocate_cid(MPIR_Comm * comm, int tag, void *state)
                  *       are not necessarily completed in the same order as they are issued, also on the
                  *       same communicator. To avoid deadlocks, we cannot add the elements to the
                  *       list bevfore the first iallreduce is completed. The "tag" is created for the
-                 *       scheduling - by calling  MPIR_Sched_next_tag(comm_ptr, &tag) - and the same
+                 *       scheduling - by calling  MPIR_Sched_next_tag(comm_ptr, MPIR_SUBGROUP_NONE, &tag) - and the same
                  *       for a idup operation on all processes. So we use it here. */
                 /* FIXME I'm not sure if there can be an overflows for this tag */
                 st->tag = (uint64_t) tag + MPIR_Process.attrs.tag_ub;
@@ -945,7 +945,7 @@ int MPIR_Get_contextid_nonblock(MPIR_Comm * comm_ptr, MPIR_Comm * newcommp, MPIR
     MPIR_FUNC_ENTER;
 
     /* now create a schedule */
-    mpi_errno = MPIR_Sched_next_tag(comm_ptr, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm_ptr, MPIR_SUBGROUP_NONE, &tag);
     MPIR_ERR_CHECK(mpi_errno);
     mpi_errno = MPIR_Sched_create(&s, MPIR_SCHED_KIND_GENERALIZED);
     MPIR_ERR_CHECK(mpi_errno);
@@ -986,7 +986,7 @@ int MPIR_Get_intercomm_contextid_nonblock(MPIR_Comm * comm_ptr, MPIR_Comm * newc
     }
 
     /* now create a schedule */
-    mpi_errno = MPIR_Sched_next_tag(comm_ptr, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm_ptr, MPIR_SUBGROUP_NONE, &tag);
     MPIR_ERR_CHECK(mpi_errno);
     mpi_errno = MPIR_Sched_create(&s, MPIR_SCHED_KIND_GENERALIZED);
     MPIR_ERR_CHECK(mpi_errno);
diff --git a/src/mpid/ch4/netmod/ofi/coll/ofi_coll_util.h b/src/mpid/ch4/netmod/ofi/coll/ofi_coll_util.h
index f0bc022a9db..65ad84d638b 100644
--- a/src/mpid/ch4/netmod/ofi/coll/ofi_coll_util.h
+++ b/src/mpid/ch4/netmod/ofi/coll/ofi_coll_util.h
@@ -90,7 +90,7 @@ MPL_STATIC_INLINE_PREFIX int MPIDI_OFI_Ibcast_knomial_triggered_tagged(void *buf
                                                sizeof(struct fi_deferred_work), MPL_MEM_BUFFER);
     MPIR_ERR_CHKANDSTMT(*works == NULL, mpi_errno, MPI_ERR_NO_MEM, goto fn_fail, "**nomem");
 
-    mpi_errno = MPIR_Sched_next_tag(comm_ptr, &rtr_tag);
+    mpi_errno = MPIR_Sched_next_tag(comm_ptr, coll_group, &rtr_tag);
     if (mpi_errno)
         MPIR_ERR_POP(mpi_errno);
 
@@ -110,7 +110,7 @@ MPL_STATIC_INLINE_PREFIX int MPIDI_OFI_Ibcast_knomial_triggered_tagged(void *buf
     }
 
     i = i + j;
-    mpi_errno = MPIR_Sched_next_tag(comm_ptr, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm_ptr, coll_group, &tag);
     if (mpi_errno)
         MPIR_ERR_POP(mpi_errno);
 
@@ -240,7 +240,7 @@ MPL_STATIC_INLINE_PREFIX int MPIDI_OFI_Ibcast_kary_triggered_tagged(void *buffer
     MPIR_ERR_CHKANDSTMT(*works == NULL, mpi_errno, MPI_ERR_NO_MEM, goto fn_fail, "**nomem");
 
 
-    mpi_errno = MPIR_Sched_next_tag(comm_ptr, &rtr_tag);
+    mpi_errno = MPIR_Sched_next_tag(comm_ptr, coll_group, &rtr_tag);
     if (mpi_errno)
         MPIR_ERR_POP(mpi_errno);
 
@@ -261,7 +261,7 @@ MPL_STATIC_INLINE_PREFIX int MPIDI_OFI_Ibcast_kary_triggered_tagged(void *buffer
     }
 
     i = i + j;
-    mpi_errno = MPIR_Sched_next_tag(comm_ptr, &tag);
+    mpi_errno = MPIR_Sched_next_tag(comm_ptr, coll_group, &tag);
     if (mpi_errno)
         MPIR_ERR_POP(mpi_errno);
 
@@ -421,7 +421,7 @@ MPL_STATIC_INLINE_PREFIX int MPIDI_OFI_Ibcast_knomial_triggered_rma(void *buffer
     MPIR_ERR_CHKANDJUMP1(*works == NULL, mpi_errno, MPI_ERR_OTHER, "**nomem", "**nomem %s",
                          "Triggered bcast deferred work alloc");
 
-    mpi_errno = MPIR_Sched_next_tag(comm_ptr, &rtr_tag);
+    mpi_errno = MPIR_Sched_next_tag(comm_ptr, coll_group, &rtr_tag);
     if (mpi_errno)
         MPIR_ERR_POP(mpi_errno);
 
@@ -567,7 +567,7 @@ MPL_STATIC_INLINE_PREFIX int MPIDI_OFI_Ibcast_kary_triggered_rma(void *buffer, i
                                                sizeof(struct fi_deferred_work), MPL_MEM_BUFFER);
     MPIR_ERR_CHKANDSTMT(*works == NULL, mpi_errno, MPI_ERR_NO_MEM, goto fn_fail, "**nomem");
 
-    mpi_errno = MPIR_Sched_next_tag(comm_ptr, &rtr_tag);
+    mpi_errno = MPIR_Sched_next_tag(comm_ptr, coll_group, &rtr_tag);
     if (mpi_errno)
         MPIR_ERR_POP(mpi_errno);
 
diff --git a/src/mpid/ch4/netmod/ofi/ofi_events.c b/src/mpid/ch4/netmod/ofi/ofi_events.c
index 84046946d11..b2ef6c8ebdf 100644
--- a/src/mpid/ch4/netmod/ofi/ofi_events.c
+++ b/src/mpid/ch4/netmod/ofi/ofi_events.c
@@ -189,6 +189,7 @@ static int pipeline_recv_event(struct fi_cq_tagged_entry *wc, MPIR_Request * r,
                     chunk_req->buf = host_buf;
                     int ret = 0;
                     if (!MPIDI_OFI_global.gpu_recv_queue && host_buf) {
+                        /* FIXME: error handling */
                         ret = fi_trecv
                             (MPIDI_OFI_global.ctx
                              [MPIDI_OFI_REQUEST(rreq, pipeline_info.ctx_idx)].rx,
diff --git a/src/mpid/ch4/netmod/ofi/ofi_recv.h b/src/mpid/ch4/netmod/ofi/ofi_recv.h
index ffd66c98c35..a33018e3bcb 100644
--- a/src/mpid/ch4/netmod/ofi/ofi_recv.h
+++ b/src/mpid/ch4/netmod/ofi/ofi_recv.h
@@ -271,6 +271,7 @@ MPL_STATIC_INLINE_PREFIX int MPIDI_OFI_do_irecv(void *buf,
             chunk_req->parent = rreq;
             chunk_req->buf = host_buf;
             int ret = 0;
+            /* FIXME: handle error */
             ret = fi_trecv(MPIDI_OFI_global.ctx[ctx_idx].rx,
                            host_buf,
                            MPIR_CVAR_CH4_OFI_GPU_PIPELINE_BUFFER_SZ,
diff --git a/src/mpid/ch4/shm/posix/release_gather/nb_bcast_release_gather.h b/src/mpid/ch4/shm/posix/release_gather/nb_bcast_release_gather.h
index a1a1b6b38dc..e3f73f9ac3b 100644
--- a/src/mpid/ch4/shm/posix/release_gather/nb_bcast_release_gather.h
+++ b/src/mpid/ch4/shm/posix/release_gather/nb_bcast_release_gather.h
@@ -355,6 +355,7 @@ MPL_STATIC_INLINE_PREFIX int MPIDI_POSIX_nb_release_gather_ibcast_impl(void *loc
     MPI_Aint type_size, nbytes, true_lb, true_extent;
     void *ori_local_buf = local_buf;
     MPI_Datatype ori_datatype = datatype;
+    int coll_group = MPIR_SUBGROUP_NONE;
 
     MPIR_CHKLMEM_DECL(1);
     /* Register the vertices */
@@ -426,7 +427,7 @@ MPL_STATIC_INLINE_PREFIX int MPIDI_POSIX_nb_release_gather_ibcast_impl(void *loc
             MPIR_TSP_sched_malloc(sizeof(MPIDI_POSIX_per_call_ibcast_info_t), sched);
         MPIR_ERR_CHKANDJUMP(!data, mpi_errno, MPI_ERR_OTHER, "**nomem");
 
-        mpi_errno = MPIR_Sched_next_tag(comm_ptr, &tag);
+        mpi_errno = MPIR_Sched_next_tag(comm_ptr, coll_group, &tag);
         if (mpi_errno)
             MPIR_ERR_POP(mpi_errno);
 
diff --git a/src/mpid/ch4/shm/posix/release_gather/nb_reduce_release_gather.h b/src/mpid/ch4/shm/posix/release_gather/nb_reduce_release_gather.h
index 979589068f8..8c942238bf0 100644
--- a/src/mpid/ch4/shm/posix/release_gather/nb_reduce_release_gather.h
+++ b/src/mpid/ch4/shm/posix/release_gather/nb_reduce_release_gather.h
@@ -364,6 +364,7 @@ MPL_STATIC_INLINE_PREFIX int MPIDI_POSIX_nb_release_gather_ireduce_impl(void *se
     MPI_Aint num_chunks, chunk_count_floor, chunk_count_ceil;
     MPI_Aint true_extent, type_size, lb, extent;
     int offset = 0, is_contig;
+    int coll_group = MPIR_SUBGROUP_NONE;
 
     /* Register the vertices */
     reserve_buf_type_id = MPIR_TSP_sched_new_type(sched, MPIDI_POSIX_NB_RG_rank0_hold_buf_issue,
@@ -418,7 +419,7 @@ MPL_STATIC_INLINE_PREFIX int MPIDI_POSIX_nb_release_gather_ireduce_impl(void *se
 
         data->seq_no = MPIDI_POSIX_COMM(comm_ptr, nb_reduce_seq_no);
 
-        mpi_errno = MPIR_Sched_next_tag(comm_ptr, &tag);
+        mpi_errno = MPIR_Sched_next_tag(comm_ptr, coll_group, &tag);
         if (mpi_errno)
             MPIR_ERR_POP(mpi_errno);
 
diff --git a/src/mpid/common/sched/mpidu_sched.c b/src/mpid/common/sched/mpidu_sched.c
index 95b607fff7c..7b025929997 100644
--- a/src/mpid/common/sched/mpidu_sched.c
+++ b/src/mpid/common/sched/mpidu_sched.c
@@ -148,7 +148,7 @@ int MPIDU_Sched_are_pending(void)
     return (all_schedules.head != NULL);
 }
 
-int MPIDU_Sched_next_tag(MPIR_Comm * comm_ptr, int *tag)
+int MPIDU_Sched_next_tag(MPIR_Comm * comm_ptr, int coll_group, int *tag)
 {
     int mpi_errno = MPI_SUCCESS;
     /* TODO there should be an internal accessor/utility macro for getting the
@@ -162,6 +162,10 @@ int MPIDU_Sched_next_tag(MPIR_Comm * comm_ptr, int *tag)
     MPIR_FUNC_ENTER;
 
     *tag = comm_ptr->next_sched_tag;
+    if (coll_group != MPIR_SUBGROUP_NONE) {
+        /* subgroup collectives use the same tag within a paraent collective */
+        goto fn_exit;
+    }
     ++comm_ptr->next_sched_tag;
 
 #if defined(HAVE_ERROR_CHECKING)
@@ -191,11 +195,13 @@ int MPIDU_Sched_next_tag(MPIR_Comm * comm_ptr, int *tag)
     if (comm_ptr->next_sched_tag == tag_ub) {
         comm_ptr->next_sched_tag = MPIR_FIRST_NBC_TAG;
     }
+  fn_exit:
+    MPIR_FUNC_EXIT;
+    return mpi_errno;
 #if defined(HAVE_ERROR_CHECKING)
   fn_fail:
+    goto fn_exit;
 #endif
-    MPIR_FUNC_EXIT;
-    return mpi_errno;
 }
 
 void MPIDU_Sched_set_tag(struct MPIDU_Sched *s, int tag)
diff --git a/src/mpid/common/sched/mpidu_sched.h b/src/mpid/common/sched/mpidu_sched.h
index 11f2cd5ac2b..40179ec6257 100644
--- a/src/mpid/common/sched/mpidu_sched.h
+++ b/src/mpid/common/sched/mpidu_sched.h
@@ -135,7 +135,7 @@ struct MPIDU_Sched {
 /* prototypes */
 int MPIDU_Sched_progress(int vci, int *made_progress);
 int MPIDU_Sched_are_pending(void);
-int MPIDU_Sched_next_tag(struct MPIR_Comm *comm_ptr, int *tag);
+int MPIDU_Sched_next_tag(struct MPIR_Comm *comm_ptr, int coll_group, int *tag);
 void MPIDU_Sched_set_tag(MPIR_Sched_t s, int tag);
 int MPIDU_Sched_create(MPIR_Sched_t * sp, enum MPIR_Sched_kind kind);
 int MPIDU_Sched_clone(MPIR_Sched_t orig, MPIR_Sched_t * cloned);