This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "MPICH primary repository". The branch, master has been updated via 848a0f312ef9418aa2bca4d63464b71441c59beb (commit) via 0f9a4417e46cae4c8aec4f741b92cf3de2f54b9c (commit) via 25f3c2de82a25adde125c1f63ae067887996f593 (commit) from 35de3d16d75207b1a7270a469c954e1226f25a8c (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- http://git.mpich.org/mpich.git/commitdiff/848a0f312ef9418aa2bca4d63464b71441... commit 848a0f312ef9418aa2bca4d63464b71441c59beb Author: Wesley Bland <[email protected]> Date: Thu Nov 6 14:46:36 2014 -0600 Enable revoke_nofail in test Signed-off-by: Huiwei Lu <[email protected]> diff --git a/test/mpi/ft/testlist b/test/mpi/ft/testlist index d2706ce..503dfaf 100644 --- a/test/mpi/ft/testlist +++ b/test/mpi/ft/testlist @@ -12,7 +12,7 @@ gather 4 env=MPIR_CVAR_ENABLE_FT=1 mpiexecarg=-disable-auto-cleanup resultTest=T reduce 4 env=MPIR_CVAR_ENABLE_FT=1 mpiexecarg=-disable-auto-cleanup resultTest=TestStatusNoErrors strict=false timeLimit=10 bcast 4 env=MPIR_CVAR_ENABLE_FT=1 mpiexecarg=-disable-auto-cleanup resultTest=TestStatusNoErrors strict=false timeLimit=10 xfail=ticket1945 scatter 4 env=MPIR_CVAR_ENABLE_FT=1 mpiexecarg=-disable-auto-cleanup resultTest=TestStatusNoErrors strict=false timeLimit=10 xfail=ticket1945 -anysource 3 env=MPIR_CVAR_ENABLE_FT=1 mpiexecarg=-disable-auto-cleanup resultTest=TestStatusNoErrors strict=false timeLimit=10 xfail=ticket1945 -revoke_nofail 2 env=MPIR_CVAR_ENABLE_FT=1 mpiexecarg=-disable-auto-cleanup resultTest=TestStatusNoErrors strict=false timeLimit=10 xfail=ticket1945 +anysource 3 env=MPIR_CVAR_ENABLE_FT=1 mpiexecarg=-disable-auto-cleanup resultTest=TestStatusNoErrors strict=false timeLimit=10 +revoke_nofail 2 env=MPIR_CVAR_ENABLE_FT=1 strict=false timeLimit=10 shrink 8 env=MPIR_CVAR_ENABLE_FT=1 mpiexecarg=-disable-auto-cleanup resultTest=TestStatusNoErrors strict=false timeLimit=10 xfail=ticket1945 agree 4 env=MPIR_CVAR_ENABLE_FT=1 mpiexecarg=-disable-auto-cleanup resultTest=TestStatusNoErrors strict=false timeLimit=10 xfail=ticket1945 http://git.mpich.org/mpich.git/commitdiff/0f9a4417e46cae4c8aec4f741b92cf3de2... commit 0f9a4417e46cae4c8aec4f741b92cf3de2f54b9c Author: Wesley Bland <[email protected]> Date: Thu Nov 6 13:17:42 2014 -0600 Correctly match requests when revoking Some of the code to do the matching for requests in the posted queue was missing. This caused local collectives to hang if the communicator had been revoked. See #1945 Signed-off-by: Huiwei Lu <[email protected]> diff --git a/src/mpid/ch3/src/ch3u_recvq.c b/src/mpid/ch3/src/ch3u_recvq.c index 9290470..bf59978 100644 --- a/src/mpid/ch3/src/ch3u_recvq.c +++ b/src/mpid/ch3/src/ch3u_recvq.c @@ -1082,6 +1082,7 @@ int MPIDI_CH3U_Clean_recvq(MPID_Comm *comm_ptr) } offset = (comm_ptr->comm_kind == MPID_INTRACOMM) ? MPID_CONTEXT_INTRA_COLL : MPID_CONTEXT_INTER_COLL; + match.parts.context_id = comm_ptr->recvcontext_id + MPID_CONTEXT_INTRANODE_OFFSET + offset; if (MATCH_WITH_LEFT_RIGHT_MASK(rreq->dev.match, match, mask)) { if (MPIR_TAG_MASK_ERROR_BIT(rreq->dev.match.parts.tag) != MPIR_AGREE_TAG && @@ -1109,6 +1110,7 @@ int MPIDI_CH3U_Clean_recvq(MPID_Comm *comm_ptr) } offset = (comm_ptr->comm_kind == MPID_INTRACOMM) ? MPID_CONTEXT_INTRA_COLL : MPID_CONTEXT_INTER_COLL; + match.parts.context_id = comm_ptr->recvcontext_id + MPID_CONTEXT_INTERNODE_OFFSET + offset; if (MATCH_WITH_LEFT_RIGHT_MASK(rreq->dev.match, match, mask)) { if (MPIR_TAG_MASK_ERROR_BIT(rreq->dev.match.parts.tag) != MPIR_AGREE_TAG && http://git.mpich.org/mpich.git/commitdiff/25f3c2de82a25adde125c1f63ae0678879... commit 25f3c2de82a25adde125c1f63ae067887996f593 Author: Wesley Bland <[email protected]> Date: Thu Nov 6 12:49:24 2014 -0600 Clean up revoke function Set the counter for processes to be revoked before calling sending out the revoke notifications. Clean up some unused code. Signed-off-by: Huiwei Lu <[email protected]> diff --git a/src/mpid/ch3/src/mpid_comm_revoke.c b/src/mpid/ch3/src/mpid_comm_revoke.c index ab26cf2..1fae5da 100644 --- a/src/mpid/ch3/src/mpid_comm_revoke.c +++ b/src/mpid/ch3/src/mpid_comm_revoke.c @@ -24,7 +24,7 @@ int MPID_Comm_revoke(MPID_Comm *comm_ptr, int is_remote) MPIDI_VC_t *vc; MPID_IOV iov[MPID_IOV_LIMIT]; int mpi_errno = MPI_SUCCESS; - int i, size, my_rank, failed=0; + int i, size, my_rank; MPID_Request *request; MPIDI_CH3_Pkt_t upkt; MPIDI_CH3_Pkt_revoke_t *revoke_pkt = &upkt.revoke; @@ -38,6 +38,11 @@ int MPID_Comm_revoke(MPID_Comm *comm_ptr, int is_remote) if (comm_ptr->node_comm) comm_ptr->node_comm->revoked = 1; if (comm_ptr->node_roots_comm) comm_ptr->node_roots_comm->revoked = 1; + /* Start a counter to track how many revoke messages we've received from + * other ranks */ + comm_ptr->dev.waiting_for_revoke = comm_ptr->local_size - 1 - is_remote; /* Subtract the processes who already know about the revoke */ + MPIU_DBG_MSG_FMT(CH3_OTHER, VERBOSE, (MPIU_DBG_FDEST, "Comm %08x waiting_for_revoke: %d", comm_ptr->handle, comm_ptr->dev.waiting_for_revoke)); + /* Keep a reference to this comm so it doesn't get destroyed while * it's being revoked */ MPIR_Comm_add_ref(comm_ptr); @@ -60,7 +65,7 @@ int MPID_Comm_revoke(MPID_Comm *comm_ptr, int is_remote) MPIU_THREAD_CS_ENTER(CH3COMM, vc); mpi_errno = MPIDI_CH3_iStartMsgv(vc, iov, 1, &request); MPIU_THREAD_CS_EXIT(CH3COMM, vc); - if (mpi_errno) failed++; + if (mpi_errno) comm_ptr->dev.waiting_for_revoke--; if (NULL != request) /* We don't need to keep a reference to this request. The * progress engine will keep a reference until it completes @@ -68,11 +73,6 @@ int MPID_Comm_revoke(MPID_Comm *comm_ptr, int is_remote) MPID_Request_release(request); } - /* Start a counter to track how many revoke messages we've received from - * other ranks */ - comm_ptr->dev.waiting_for_revoke = comm_ptr->local_size - 1 - is_remote - failed; /* Subtract the processes who already know about the revoke */ - MPIU_DBG_MSG_FMT(CH3_OTHER, VERBOSE, (MPIU_DBG_FDEST, "Comm %08x waiting_for_revoke: %d", comm_ptr->handle, comm_ptr->dev.waiting_for_revoke)); - /* Check to see if we are done revoking */ if (comm_ptr->dev.waiting_for_revoke == 0) { MPIR_Comm_release(comm_ptr, 0); @@ -98,13 +98,6 @@ int MPID_Comm_revoke(MPID_Comm *comm_ptr, int is_remote) } } -fn_exit: MPIDI_FUNC_EXIT(MPID_STATE_MPID_COMM_REVOKE); return MPI_SUCCESS; -fn_fail: - if (request) { - MPIU_Object_set_ref(request, 0); - MPIDI_CH3_Request_destroy(request); - } - goto fn_exit; } ----------------------------------------------------------------------- Summary of changes: src/mpid/ch3/src/ch3u_recvq.c | 2 ++ src/mpid/ch3/src/mpid_comm_revoke.c | 21 +++++++-------------- test/mpi/ft/testlist | 4 ++-- 3 files changed, 11 insertions(+), 16 deletions(-) hooks/post-receive -- MPICH primary repository