Karl --
Can you force the use of just the shared memory transport -- i.e., disable the
TCP BTL? For example:
mpirun -np 2 --mca btl sm,self hello_c
If that also hangs, can you attach a debugger and see *where* it is hanging
inside MPI_Init()? (In OMPI, MPI::Init() simply invokes MPI_Init())
On Nov 27, 2013, at 2:56 PM, "Meredith, Karl" <karl.mered...@fmglobal.com>
wrote:
> /opt/trunk/apple-only/bin/ompi_info --param oob tcp --level 9
> MCA oob: parameter "oob_tcp_verbose" (current value: "0",
> data source: default, level: 9 dev/all, type: int)
> Verbose level for the OOB tcp component
> MCA oob: parameter "oob_tcp_peer_limit" (current value: "-1",
> data source: default, level: 9 dev/all, type: int)
> Maximum number of peer connections to simultaneously
> maintain (-1 = infinite)
> MCA oob: parameter "oob_tcp_peer_retries" (current value:
> "60", data source: default, level: 9 dev/all, type: int)
> Number of times to try shutting down a connection
> before giving up
> MCA oob: parameter "oob_tcp_debug" (current value: "0", data
> source: default, level: 9 dev/all, type: int)
> Enable (1) / disable (0) debugging output for this
> component
> MCA oob: parameter "oob_tcp_sndbuf" (current value: "131072",
> data source: default, level: 9 dev/all, type: int)
> TCP socket send buffering size (in bytes)
> MCA oob: parameter "oob_tcp_rcvbuf" (current value: "131072",
> data source: default, level: 9 dev/all, type: int)
> TCP socket receive buffering size (in bytes)
> MCA oob: parameter "oob_tcp_if_include" (current value: "",
> data source: default, level: 9 dev/all, type: string, synonyms:
> oob_tcp_include)
> Comma-delimited list of devices and/or CIDR notation
> of networks to use for Open MPI bootstrap communication (e.g.,
> "eth0,192.168.0.0/16"). Mutually exclusive with oob_tcp_if_exclude.
> MCA oob: parameter "oob_tcp_if_exclude" (current value: "",
> data source: default, level: 9 dev/all, type: string, synonyms:
> oob_tcp_exclude)
> Comma-delimited list of devices and/or CIDR notation
> of networks to NOT use for Open MPI bootstrap communication -- all devices
> not matching these specifications will be used (e.g., "eth0,192.168.0.0/16").
> If set to a non-default value, it is mutually exclusive with
> oob_tcp_if_include.
> MCA oob: parameter "oob_tcp_connect_sleep" (current value:
> "1", data source: default, level: 9 dev/all, type: int)
> Enable (1) / disable (0) random sleep for connection
> wireup.
> MCA oob: parameter "oob_tcp_listen_mode" (current value:
> "event", data source: default, level: 9 dev/all, type: int)
> Mode for HNP to accept incoming connections: event,
> listen_thread.
> Valid values: 0:"event", 1:"listen_thread"
> MCA oob: parameter "oob_tcp_listen_thread_max_queue" (current
> value: "10", data source: default, level: 9 dev/all, type: int)
> High water mark for queued accepted socket list
> size. Used only when listen_mode is listen_thread.
> MCA oob: parameter "oob_tcp_listen_thread_wait_time" (current
> value: "10", data source: default, level: 9 dev/all, type: int)
> Time in milliseconds to wait before actively
> checking for new connections when listen_mode is listen_thread.
> MCA oob: parameter "oob_tcp_static_ports" (current value: "",
> data source: default, level: 9 dev/all, type: string)
> Static ports for daemons and procs (IPv4)
> MCA oob: parameter "oob_tcp_dynamic_ports" (current value:
> "", data source: default, level: 9 dev/all, type: string)
> Range of ports to be dynamically used by daemons and
> procs (IPv4)
> MCA oob: parameter "oob_tcp_disable_family" (current value:
> "none", data source: default, level: 9 dev/all, type: int)
> Disable IPv4 (4) or IPv6 (6)
> Valid values: 0:"none", 4:"IPv4", 6:"IPv6"
>
> /opt/trunk/apple-only/bin/ompi_info --param btl tcp --level 9
> MCA btl: parameter "btl_tcp_links" (current value: "1", data
> source: default, level: 4 tuner/basic, type: unsigned)
> MCA btl: parameter "btl_tcp_if_include" (current value: "",
> data source: default, level: 1 user/basic, type: string)
> Comma-delimited list of devices and/or CIDR notation
> of networks to use for MPI communication (e.g., "eth0,192.168.0.0/16").
> Mutually exclusive with btl_tcp_if_exclude.
> MCA btl: parameter "btl_tcp_if_exclude" (current value:
> "127.0.0.1/8,sppp", data source: default, level: 1 user/basic, type: string)
> Comma-delimited list of devices and/or CIDR notation
> of networks to NOT use for MPI communication -- all devices not matching
> these specifications will be used (e.g., "eth0,192.168.0.0/16"). If set to a
> non-default value, it is mutually exclusive with btl_tcp_if_include.
> MCA btl: parameter "btl_tcp_free_list_num" (current value:
> "8", data source: default, level: 5 tuner/detail, type: int)
> MCA btl: parameter "btl_tcp_free_list_max" (current value:
> "-1", data source: default, level: 5 tuner/detail, type: int)
> MCA btl: parameter "btl_tcp_free_list_inc" (current value:
> "32", data source: default, level: 5 tuner/detail, type: int)
> MCA btl: parameter "btl_tcp_sndbuf" (current value: "131072",
> data source: default, level: 4 tuner/basic, type: int)
> MCA btl: parameter "btl_tcp_rcvbuf" (current value: "131072",
> data source: default, level: 4 tuner/basic, type: int)
> MCA btl: parameter "btl_tcp_endpoint_cache" (current value:
> "30720", data source: default, level: 4 tuner/basic, type: int)
> The size of the internal cache for each TCP
> connection. This cache is used to reduce the number of syscalls, by replacing
> them with memcpy. Every read will read the expected data plus the amount of
> the endpoint_cache
> MCA btl: parameter "btl_tcp_use_nagle" (current value: "0",
> data source: default, level: 4 tuner/basic, type: int)
> Whether to use Nagle's algorithm or not (using
> Nagle's algorithm may increase short message latency)
> MCA btl: parameter "btl_tcp_port_min_v4" (current value:
> "1024", data source: default, level: 2 user/detail, type: int)
> The minimum port where the TCP BTL will try to bind
> (default 1024)
> MCA btl: parameter "btl_tcp_port_range_v4" (current value:
> "64511", data source: default, level: 2 user/detail, type: int)
> The number of ports where the TCP BTL will try to
> bind (default 64511). This parameter together with the port min, define a
> range of ports where Open MPI will open sockets.
> MCA btl: parameter "btl_tcp_exclusivity" (current value:
> "100", data source: default, level: 7 dev/basic, type: unsigned)
> BTL exclusivity (must be >= 0)
> MCA btl: parameter "btl_tcp_flags" (current value: "314",
> data source: default, level: 5 tuner/detail, type: unsigned)
> BTL bit flags (general flags: SEND=1, PUT=2, GET=4,
> SEND_INPLACE=8, RDMA_MATCHED=64, HETEROGENEOUS_RDMA=256; flags only used by
> the "dr" PML (ignored by others): ACK=16, CHECKSUM=32, RDMA_COMPLETION=128;
> flags only used by the "bfo" PML (ignored by others): FAILOVER_SUPPORT=512)
> MCA btl: parameter "btl_tcp_rndv_eager_limit" (current value:
> "65536", data source: default, level: 4 tuner/basic, type: size_t)
> Size (in bytes, including header) of "phase 1"
> fragment sent for all large messages (must be >= 0 and <= eager_limit)
> MCA btl: parameter "btl_tcp_eager_limit" (current value:
> "65536", data source: default, level: 4 tuner/basic, type: size_t)
> Maximum size (in bytes, including header) of "short"
> messages (must be >= 1).
> MCA btl: parameter "btl_tcp_max_send_size" (current value:
> "131072", data source: default, level: 4 tuner/basic, type: size_t)
> Maximum size (in bytes) of a single "phase 2"
> fragment of a long message when using the pipeline protocol (must be >= 1)
> MCA btl: parameter "btl_tcp_rdma_pipeline_send_length"
> (current value: "131072", data source: default, level: 4 tuner/basic, type:
> size_t)
> Length of the "phase 2" portion of a large message
> (in bytes) when using the pipeline protocol. This part of the message will
> be split into fragments of size max_send_size and sent using send/receive
> semantics (must be >= 0; only relevant when the PUT flag is set)
> MCA btl: parameter "btl_tcp_rdma_pipeline_frag_size" (current
> value: "2147483647", data source: default, level: 4 tuner/basic, type: size_t)
> Maximum size (in bytes) of a single "phase 3"
> fragment from a long message when using the pipeline protocol. These
> fragments will be sent using RDMA semantics (must be >= 1; only relevant when
> the PUT flag is set)
> MCA btl: parameter "btl_tcp_min_rdma_pipeline_size" (current
> value: "196608", data source: default, level: 4 tuner/basic, type: size_t)
> Messages smaller than this size (in bytes) will not
> use the RDMA pipeline protocol. Instead, they will be split into fragments
> of max_send_size and sent using send/receive semantics (must be >=0, and is
> automatically adjusted up to at least
> (eager_limit+btl_rdma_pipeline_send_length); only relevant when the PUT flag
> is set)
> MCA btl: parameter "btl_tcp_bandwidth" (current value: "100",
> data source: default, level: 5 tuner/detail, type: unsigned)
> Approximate maximum bandwidth of interconnect (0 =
> auto-detect value at run-time [not supported in all BTL modules], >= 1 =
> bandwidth in Mbps)
> MCA btl: parameter "btl_tcp_disable_family" (current value:
> "0", data source: default, level: 2 user/detail, type: int)
> MCA btl: parameter "btl_tcp_if_seq" (current value: "", data
> source: default, level: 9 dev/all, type: string)
> If specified, a comma-delimited list of TCP
> interfaces. Interfaces will be assigned, one to each MPI process, in a
> round-robin fashion on each server. For example, if the list is "eth0,eth1"
> and four MPI processes are run on a single server, then local ranks 0 and 2
> will use eth0 and local ranks 1 and 3 will use eth1.
>
>
> On Nov 27, 2013, at 2:41 PM, George Bosilca
> <bosi...@icl.utk.edu<mailto:bosi...@icl.utk.edu>> wrote:
>
> ompi_info —param oob tcp —level 9
> ompi_info —param btl tcp —level 9
>
> _______________________________________________
> users mailing list
> us...@open-mpi.org
> http://www.open-mpi.org/mailman/listinfo.cgi/users
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
