January 2015 - discuss - mailman3.cels.anl.gov

Re: [mpich-discuss] How to make a non-MPI process to become a MPI process of MPI_COMM_WORLD?
by haozi 25 Jan '15

25 Jan '15

Thanks, Balaji. server_port_name.txt is built by the example, not by user. At 2015-01-25 19:24:28, "Balaji, Pavan" <balaji(a)anl.gov> wrote: > >> On Jan 25, 2015, at 5:04 AM, haozi <yidanyiji(a)163.com> wrote: >> Hello, Balaji. >> >> The example from OpenMPI is attached. > >server_port_name.txt is missing. > >-- >Pavan Balaji ✉️ >http://www.mcs.anl.gov/~balaji > >_______________________________________________ >discuss mailing list discuss(a)mpich.org >To manage subscription options or unsubscribe: >https://lists.mpich.org/mailman/listinfo/discuss _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

1 0

Re: [mpich-discuss] How to make a non-MPI process to become a MPI process of MPI_COMM_WORLD?
by Balaji, Pavan 25 Jan '15

25 Jan '15

> On Jan 25, 2015, at 5:04 AM, haozi <yidanyiji(a)163.com> wrote: > Hello, Balaji. > > The example from OpenMPI is attached. server_port_name.txt is missing. -- Pavan Balaji ✉️ http://www.mcs.anl.gov/~balaji _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

1 0

Re: [mpich-discuss] How to make a non-MPI process to become a MPI process of MPI_COMM_WORLD?
by haozi 25 Jan '15

25 Jan '15

Hello, Balaji. The example from OpenMPI is attached. Thank you very much. At 2015-01-25 03:16:35, "Balaji, Pavan" <balaji(a)anl.gov> wrote: > >You are missing some files needed to run your program. Also, can you please attach files instead of copy-pasting code into the email? > > -- Pavan > >> On Jan 24, 2015, at 7:57 AM, haozi <yidanyiji(a)163.com> wrote: >> >> Hi. >> The following server/client example is selected from OpenMPI. >> I compile and run the example by using OpenMPI, everything is OK. >> BUT, I compile and run it by mpich(3.1.3), which is BLOCKED as my example! >> >> This is a bug? >> >> #include <stdio.h> >> #include <stdlib.h> >> #include <string.h> >> #include <errno.h> >> #include <unistd.h> >> #include <mpi.h> >> >> /* >> >> LOGIC: >> >> - the 'server' opens a port and write the info to a file >> - the 'clients' open the file and connect to the port >> - after each accept, the server and client do a merge to >> convert the intercomm to an intracomm >> >> DETAIL STEPS: >> >> - server open port >> - server does accept >> - client #1 does connect >> - server and client #1 do merge >> - server does accept >> - client #2 does connect >> - server, client #1 and client #2 do merge >> - server does accept >> - client #3 does connect >> - server, client #1, client #2 and client #3 do merge >> >> */ >> >> #define TAG 0 >> >> #define CHK(code) do \ >> { \ >> int retval = code ; \ >> if (retval != MPI_SUCCESS) \ >> { \ >> fprintf(stderr, "Error: " #code "\n") ; \ >> exit(1) ; \ >> } \ >> } while(0) >> >> int main(int argc, char *argv[]) >> { >> char hostname[255] ; >> char buff[255] ; >> >> int role ; >> int num_clients ; >> int size, rank ; >> >> FILE *fp ; >> char server_port_name[MPI_MAX_PORT_NAME] ; >> >> MPI_Comm intercomm, intracomm ; >> MPI_Status status ; >> int msg_count ; >> int i ; >> >> /* sanity check the args */ >> if(argc != 3) >> { >> fprintf(stderr, "usage %s <num clients> <1:server | 0:client>\n", argv[0]) ; >> exit(1) ; >> } >> >> num_clients = atoi(argv[1]) ; >> role = atoi(argv[2]) ; >> >> if (num_clients <= 0 || (role != 0 && role != 1)) >> { >> fprintf(stderr, "usage %s <num clients> <1:server | 0:client>\n", argv[0]) ; >> exit(1) ; >> } >> >> /* initialize MPI */ >> CHK(MPI_Init(&argc, &argv)) ; >> >> /* get the node name */ >> { >> int retval = gethostname(hostname, 255) ; >> if(retval == -1) >> { >> fprintf(stderr, "gethostname failed: %s\n", strerror(errno)) ; >> exit(1) ; >> } >> } >> >> /* server */ >> if(role == 1) >> { >> printf("SERVER: on node '%s'\n", hostname) ; >> >> /* open port to establish connections */ >> CHK(MPI_Open_port(MPI_INFO_NULL, server_port_name)) ; >> >> printf("SERVER: opened port=%s\n", server_port_name) ; >> >> /* store the port name */ >> fp = fopen("server_port_name.txt", "w") ; >> if(fp == NULL) >> { >> fprintf(stderr, "fopen failed: %s\n", strerror(errno)) ; >> exit(1) ; >> } >> fprintf(fp, "%s", server_port_name) ; >> fclose(fp) ; >> >> /* the server accepts connections from all the clients */ >> for(i = 0 ; i < num_clients ; i++ ) >> { >> /* accept connections at this port */ >> CHK(MPI_Comm_accept(server_port_name, MPI_INFO_NULL, 0, >> i == 0 ? MPI_COMM_WORLD : intracomm, >> &intercomm)) ; >> >> printf("SERVER: accepted connection from client %d\n", i+1) ; >> >> /* merge, to form one intra communicator */ >> CHK(MPI_Intercomm_merge(intercomm, 0, &intracomm)) ; >> >> printf("SERVER: merged with client %d\n", i+1) ; >> >> CHK(MPI_Comm_size(intracomm, &size)) ; >> CHK(MPI_Comm_rank(intracomm, &rank)) ; >> >> printf("SERVER: after merging with client %d: size=%d rank=%d\n", i+1, size, rank) ; >> } >> } /* end server */ >> >> /* client */ >> if(role == 0) >> { >> printf("CLIENT: on node '%s'\n", hostname) ; >> >> fp = fopen("server_port_name.txt", "r") ; >> if(fp == NULL) >> { >> fprintf(stderr, "fopen failed: %s\n", strerror(errno)) ; >> exit(1) ; >> } >> fscanf(fp, "%s", server_port_name) ; >> fclose(fp) ; >> >> printf("CLIENT: attempting to connect to server on port=%s\n", server_port_name) ; >> >> /* connect to the server */ >> CHK(MPI_Comm_connect (server_port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD, &intercomm)) ; >> >> printf("CLIENT: connected to server on port\n") ; >> >> /* merge the server and client to one intra communicator */ >> CHK(MPI_Intercomm_merge(intercomm, 1, &intracomm)) ; >> >> printf("CLIENT: merged with existing intracomm\n") ; >> >> CHK(MPI_Comm_size(intracomm, &size)) ; >> CHK(MPI_Comm_rank(intracomm, &rank)) ; >> >> printf("CLIENT: after merging, new comm: size=%d rank=%d\n", size, rank) ; >> >> for (i = rank ; i < num_clients ; i++) >> { >> /* client performs a collective accept */ >> CHK(MPI_Comm_accept(server_port_name, MPI_INFO_NULL, 0, intracomm, &intercomm)) ; >> >> printf("CLIENT: connected to server on port\n") ; >> >> /* merge the two intra comms back to one communicator */ >> CHK(MPI_Intercomm_merge(intercomm, 0, &intracomm)) ; >> >> printf("CLIENT: merged with existing members\n") ; >> >> CHK(MPI_Comm_size(intracomm, &size)) ; >> CHK(MPI_Comm_rank(intracomm, &rank)) ; >> >> printf("CLIENT: new size after merging with existing members: size=%d rank=%d\n", size, rank) ; >> } >> >> } /* end client */ >> >> CHK(MPI_Comm_size(intracomm, &size)) ; >> CHK(MPI_Comm_rank(intracomm, &rank)) ; >> >> printf("After fusion: size=%d rank=%d\n", size, rank) ; >> >> if(rank == 0) >> { >> msg_count = num_clients ; >> >> while(msg_count) >> { >> CHK(MPI_Recv(buff, 255, MPI_CHAR, MPI_ANY_SOURCE, >> MPI_ANY_TAG, intracomm, &status)) ; >> >> printf("Received hello msg from '%s'\n", buff) ; >> msg_count-- ; >> } >> } >> else >> { >> /* all ranks > 0 */ >> >> CHK(MPI_Send(hostname, strlen(hostname) + 1, MPI_CHAR, 0, TAG, intracomm)) ; >> } >> >> CHK(MPI_Finalize()) ; >> >> fprintf(stderr, "Rank %d is exiting\n", rank); >> return 0 ; >> } >> >> >> At 2015-01-24 10:43:07, "haozi" <yidanyiji(a)163.com> wrote: >> Thanks, Bland and Lu. >> >> You are right. >> These functions (such as MPI_Comm_accept, MPI_Comm_connect, MPI_Intercomm_merge) can help me to get a new intra-communicator which contains ALL MPI processes. >> >> Now, I have a more complicated example: >> I have a server and a client. >> After they merge into an intra-communicator by using connect/accept/merge fucntions, another client would plant to join them, too. >> I thought that the code is simular, but the code CAN'T work: Second client CAN'T connect. The new comm CAN'T accept. They all BLOCK. >> >> As you see, the second client BLOCKs at MPI_Comm_connect, and processes of newcomm BLOCK at MPI_Comm_accept. >> What's Wrong with my code? >> >> //server >> #include "mpi.h" >> int main(int argc, char *argv[]) >> { >> MPI_Comm client, client2, newcomm, newcomm2; >> MPI_Status status; >> char port_name[MPI_MAX_PORT_NAME]; >> char port_name2[MPI_MAX_PORT_NAME]; >> int size, again, rank, myrank; >> >> MPI_Init(&argc, &argv); >> MPI_Open_port(MPI_INFO_NULL, port_name);//OK >> >> MPI_Comm_accept(port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD,&client);//OK >> >> MPI_Intercomm_merge(client,11,&newcomm);//OK >> >> MPI_Barrier(newcomm);//OK >> >> MPI_Open_port(MPI_INFO_NULL, port_name2);// OK >> >> MPI_Comm_accept(port_name2, MPI_INFO_NULL, 0, newcomm,&client2);// BLOCK here, Wath's wrong? >> >> MPI_Intercomm_merge(client2,12,&newcomm2); >> >> MPI_Barrier(newcomm2); >> >> MPI_Close_port(port_name); >> MPI_Comm_disconnect(&client); >> MPI_Close_port(port_name2); >> MPI_Comm_disconnect(&client2); >> >> MPI_Finalize(); >> return 0; >> } >> >> //first client >> #include "mpi.h" >> int main( int argc, char **argv ) >> { >> MPI_Comm server,newcomm,newcomm2,client2; >> char port_name[MPI_MAX_PORT_NAME]; >> char port_name2[MPI_MAX_PORT_NAME]; >> int size,rank; >> >> MPI_Init( &argc, &argv ); >> >> strcpy( port_name, argv[1] ); >> MPI_Comm_connect( port_name, MPI_INFO_NULL, 0, MPI_COMM_SELF,&server );//OK >> >> MPI_Intercomm_merge(server,11,&newcomm);//OK >> >> MPI_Barrier(newcomm);//OK >> >> MPI_Open_port(MPI_INFO_NULL, port_name2);//OK >> >> MPI_Comm_accept(port_name2, MPI_INFO_NULL, 0, newcomm,&client2);// BLOCK here, Wath's wrong? >> >> MPI_Intercomm_merge(client2,12,&newcomm2); >> >> MPI_Barrier(newcomm2); >> >> MPI_Close_port(port_name2); >> MPI_Comm_disconnect(&client2); >> MPI_Comm_disconnect( &server ); >> MPI_Finalize(); >> return 0; >> } >> >> //second client >> #include "mpi.h" >> int main( int argc, char **argv ) >> { >> MPI_Comm server,newcomm; >> char port_name[MPI_MAX_PORT_NAME]; >> int size,rank; >> >> MPI_Init( &argc, &argv ); >> >> strcpy( port_name, argv[1] );//OK >> >> MPI_Comm_connect( port_name, MPI_INFO_NULL, 0, MPI_COMM_SELF,&server );// BLOCK here, Wath's wrong? >> >> MPI_Comm_size(MPI_COMM_WORLD, &size); >> >> MPI_Intercomm_merge(server,11,&newcomm); >> >> MPI_Barrier(newcomm); >> >> MPI_Comm_disconnect( &server ); >> MPI_Finalize(); >> return 0; >> } >> >> >> >> >> At 2015-01-23 23:14:06, "Wesley Bland" <wbland(a)anl.gov> wrote: >> The size of MPI_COMM_WORLD will never change. That communicator is set at initialization time and is not ever modified. However, when you finish connect/accept, you get a new communicator which you can merge into an intra-communicator which functions exactly like MPI_COMM_WORLD. Don’t be attached to the variable name MPI_COMM_WORLD. It’s just a variable and you can use a different one with very little extra work. >> >> Thanks, >> Wesley >> >> >> On Thu, Jan 22, 2015 at 6:14 PM, haozi <yidanyiji(a)163.com> wrote: >> Thanks, Lu. >> My simple code is as following. >> //server >> #include "mpi.h" >> >> int main(int argc, char *argv[]) >> { >> MPI_Comm client; >> MPI_Status status; >> char port_name[MPI_MAX_PORT_NAME]; >> int size, again; >> >> MPI_Init(&argc, &argv); >> MPI_Comm_size(MPI_COMM_WORLD, &size); >> MPI_Open_port(MPI_INFO_NULL, port_name); >> printf("server port_name is %s\n\n", port_name); >> MPI_Comm_accept(port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD,&client); >> MPI_Comm_size(MPI_COMM_WORLD, &size); >> printf("At server, comm_size=%d @ MPI_COMM_WORLD=%x, Client_World=%x\n",size,MPI_COMM_WORLD, client); >> MPI_Comm_size(client, &size); >> printf("At server, client_size=%d @ MPI_COMM_WORLD=%x, Client_World=%x\n",size,MPI_COMM_WORLD, client); >> >> MPI_Comm_disconnect(&client); >> MPI_Finalize(); >> return 0; >> } >> >> //client >> #include "mpi.h" >> >> int main( int argc, char **argv ) >> { >> MPI_Comm server; >> char port_name[MPI_MAX_PORT_NAME]; >> int size; >> >> MPI_Init( &argc, &argv ); >> strcpy( port_name, argv[1] ); >> printf("server port name:%s\n",port_name); >> MPI_Comm_connect( port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD,&server ); >> MPI_Comm_size(MPI_COMM_WORLD, &size); >> printf("At client, comm_size=%d @ MPI_COMM_WORLD=%x, Server_World=%x\n",size,MPI_COMM_WORLD,server); >> MPI_Comm_size(server, &size); >> printf("At client, server_size=%d @ MPI_COMM_WORLD=%x, Server_World=%x\n",size,MPI_COMM_WORLD,server); >> >> MPI_Comm_disconnect( &server ); >> MPI_Finalize(); >> return 0; >> } >> >> The run command is as following. >> mpiexec -n 1 ./server >> mpiexec -n 1 ./client >> >> BUT, the SIZE is 1, NOT 2. >> >> My question is as before: How does the size of MPI_COMM_WORLD change to 2 ? >> >> >> At 2015-01-23 00:30:43, "Huiwei Lu" <huiweilu(a)mcs.anl.gov> wrote: >> You may take a look at MPI_Comm_accept and MPI_Comm_connect, which will connect a new client process to a server process. See Chap. 10 of MPI 3.0 standard (www.mpi-forum.org/docs/mpi-3.0/mpi30-report.pdf) for a detail example. >> >> -- >> Huiwei Lu >> Postdoc Appointee >> Mathematics and Computer Science Division >> Argonne National Laboratory >> http://www.mcs.anl.gov/~huiweilu/ >> >> On Thu, Jan 22, 2015 at 9:41 AM, haozi <yidanyiji(a)163.com> wrote: >> Hi, guys. >> >> This web page (http://wiki.mpich.org/mpich/index.php/PMI_v2_Design_Thoughts) says: >> Singleton init. This is the process by which a program that was not started with mpiexec can become an MPI process and make use of all MPI features, including MPI_Comm_spawn, needs to be designed and documented, with particular attention to the disposition of standard I/O. Not all process managers will want to or even be able to create a new mpiexec process, so this needs to be negotiated. Similarly, the dispostion of stdio needs to be negotiated between the singleton process and the process manager. To address these issues, a new singleton init protocol has been implemented and tested with the gforker process manager. >> >> I am very interested in this function. >> Can this function solve the following question: >> At beginning, the MPI job uses the mpiexec commond to start three MPI processes. That is to say, there are three MPI processes in MPI_COMM_WORLD. At some time, the job find itself to need another MPI process to cooperate the three MPI processes. So the question is: Could PMI help an non-MPI process to become a MPI process of the current MPI_COMM_WORLD? That is to say, Could the non-MPI process use the PMI function to become a member process of the current MPI job which would have FOUR MPI processes in MPI_COMM_WORLD? >> >> Is there some method to solve question? >> Is anybody have some example? >> >> Thandks!!! >> >> >> >> _______________________________________________ >> discuss mailing list discuss(a)mpich.org >> To manage subscription options or unsubscribe: >> https://lists.mpich.org/mailman/listinfo/discuss >> >> >> >> >> >> >> >> >> _______________________________________________ >> discuss mailing list discuss(a)mpich.org >> To manage subscription options or unsubscribe: >> https://lists.mpich.org/mailman/listinfo/discuss > >-- >Pavan Balaji ✉️ >http://www.mcs.anl.gov/~balaji > >_______________________________________________ >discuss mailing list discuss(a)mpich.org >To manage subscription options or unsubscribe: >https://lists.mpich.org/mailman/listinfo/discuss _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

1 0

Re: [mpich-discuss] Failed to compile MPICH3 v3.2a2-115-g93e816cc on Power755 (Power7/Linux)
by Jeff Hammond 24 Jan '15

24 Jan '15

Try building MPICH with GCC, for which I know a POWER7 version exists. GCC supports inline asm properly and you can use the resulting library with XLC codes, thanks to the beauty of C ABI interoperability. For Fortran, use CC=gcc, FC=xlf_r. Best, Jeff On Friday, January 23, 2015, "Antonio J. Peña" <apenya(a)mcs.anl.gov> wrote: > > Hi Victor, > > Most likely your hardware or compiler does not support atomic memory > access. Unfortunately, our IB netmod uses these builtin functions. If you > have Mellanox InfiniBand, you can give a try to the MXM netmod. > > Best, > Antonio > > > On 01/23/2015 05:04 AM, Victor Vysotskiy wrote: > > Hi, > > > I am trying to compile the latest MPICH3 snapshot on Power7/Linux > platform by using the XLF/XLC compiler suite v14.1. The cluster is equipped > with common IB hardware. Unfortunately, the compilation failed with the > following error message: > > > lib/.libs/libmpi.a(lib_libmpi_la-ib_poll.o): In function > `MPID_nem_ib_drain_scq': > > src/mpid/ch3/channels/nemesis/netmod/ib/ib_poll.c:(.text+0x5bc4): > undefined reference to `__builtin_ctz' > > src/mpid/ch3/channels/nemesis/netmod/ib/ib_poll.c:(.text+0x5e90): > undefined reference to `__builtin_ctz' > > lib/.libs/libmpi.a(lib_libmpi_la-ib_poll.o): In function > `MPID_nem_ib_cm_drain_scq$AF227_75': > > src/mpid/ch3/channels/nemesis/netmod/ib/ib_poll.c:(.text+0x7d7c): > undefined reference to `__builtin_ctz' > > src/mpid/ch3/channels/nemesis/netmod/ib/ib_poll.c:(.text+0x7e14): > undefined reference to `__builtin_ctz' > > src/mpid/ch3/channels/nemesis/netmod/ib/ib_poll.c:(.text+0x8040): > undefined reference to `__builtin_ctz' > > lib/.libs/libmpi.a(lib_libmpi_la-ib_poll.o):src/mpid/ch3/channels/nemesis/netmod/ib/ib_poll.c:(.text+0x814c): > more undefined references to `__builtin_ctz' follow > > > For the configure's options used please check the attached > 'config.power7.linux.log' file. > > > In addition, the IBM compiler produces few warnings: > > > #1 > > > "src/mpid/ch3/channels/nemesis/netmod/ib/ib_ibcom.h", line 169.39: > 1506-1147 (W) The asm constraint "a" is not supported. > > "src/mpid/ch3/channels/nemesis/netmod/ib/ib_ibcom.h", line 169.49: > 1506-1147 (W) The asm constraint "d" is not supported. > > "src/mpid/ch3/channels/nemesis/netmod/ib/ib_reg_mr.c", line 89.12: > 1506-1332 (W) A function with return type "void" may not return a value of > type "void". > > 1500-064: (W) The register name %rax is unknown. > > 1500-064: (W) The register name %rbx is unknown. > > 1500-064: (W) The register name %rcx is unknown. > > 1500-064: (W) The register name %rdx is unknown. > > #2 > > > "/usr/include/bits/mman.h", line 45.9: 1506-236 (W) Macro name > MAP_SHARED has been redefined. > > "/usr/include/bits/mman.h", line 45.9: 1506-358 (I) "MAP_SHARED" is > defined on line 17 of /usr/include/asm-generic/mman-common.h. > > "/usr/include/bits/mman.h", line 46.9: 1506-236 (W) Macro name MAP_PRIVATE > has been redefined. > > "/usr/include/bits/mman.h", line 46.9: 1506-358 (I) "MAP_PRIVATE" is > defined on line 18 of /usr/include/asm-generic/mman-common.h. > > "/usr/include/bits/mman.h", line 48.10: 1506-236 (W) Macro name MAP_TYPE > has been redefined. > > "/usr/include/bits/mman.h", line 48.10: 1506-358 (I) "MAP_TYPE" is defined > on line 19 of /usr/include/asm-generic/mman-common.h. > > "/usr/include/bits/mman.h", line 52.9: 1506-236 (W) Macro name MAP_FIXED > has been redefined. > > "/usr/include/bits/mman.h", line 52.9: 1506-358 (I) "MAP_FIXED" is defined > on line 20 of /usr/include/asm-generic/mman-common.h. > > "/usr/include/bits/mman.h", line 54.10: 1506-236 (W) Macro name MAP_FILE > has been redefined. > > "/usr/include/bits/mman.h", line 54.10: 1506-358 (I) "MAP_FILE" is defined > on line 43 of /usr/include/asm-generic/mman-common.h. > > "/usr/include/bits/mman.h", line 55.10: 1506-236 (W) Macro name > MAP_ANONYMOUS has been redefined. > > "/usr/include/bits/mman.h", line 55.10: 1506-358 (I) "MAP_ANONYMOUS" is > defined on line 21 of /usr/include/asm-generic/mman-common.h. > > "/usr/include/bits/mman.h", line 61.10: 1506-236 (W) Macro name > MAP_GROWSDOWN has been redefined. > > "/usr/include/bits/mman.h", line 61.10: 1506-358 (I) "MAP_GROWSDOWN" is > defined on line 19 of /usr/include/asm/mman.h. > > "/usr/include/bits/mman.h", line 62.10: 1506-236 (W) Macro name > MAP_DENYWRITE has been redefined. > > "/usr/include/bits/mman.h", line 62.10: 1506-358 (I) "MAP_DENYWRITE" is > defined on line 20 of /usr/include/asm/mman.h. > > "/usr/include/bits/mman.h", line 63.10: 1506-236 (W) Macro name > MAP_EXECUTABLE has been redefined. > > "/usr/include/bits/mman.h", line 63.10: 1506-358 (I) "MAP_EXECUTABLE" is > defined on line 21 of /usr/include/asm/mman.h. > > "/usr/include/bits/mman.h", line 64.10: 1506-236 (W) Macro name MAP_LOCKED > has been redefined. > > "/usr/include/bits/mman.h", line 64.10: 1506-358 (I) "MAP_LOCKED" is > defined on line 17 of /usr/include/asm/mman.h. > > "/usr/include/bits/mman.h", line 65.10: 1506-236 (W) Macro name > MAP_NORESERVE has been redefined. > > "/usr/include/bits/mman.h", line 65.10: 1506-358 (I) "MAP_NORESERVE" is > defined on line 16 of /usr/include/asm/mman.h. > > "/usr/include/bits/mman.h", line 66.10: 1506-236 (W) Macro name > MAP_POPULATE has been redefined. > > "/usr/include/bits/mman.h", line 66.10: 1506-358 (I) "MAP_POPULATE" is > defined on line 26 of /usr/include/asm/mman.h. > > > Is the error message related to the warnings observed? Or, I simply > should avoid using the nemesis:ib channel? > > > With best regards, > > Victor. > > > > > _______________________________________________ > discuss mailing list discuss(a)mpich.org <javascript:_e(%7B%7D,'cvml','discuss(a)mpich.org');> > To manage subscription options or unsubscribe:https://lists.mpich.org/mailman/listinfo/discuss > > > > -- > Antonio J. Peña > Postdoctoral Appointee > Mathematics and Computer Science Division > Argonne National Laboratory > 9700 South Cass Avenue, Bldg. 240, Of. 3148 > Argonne, IL 60439-4847apenya(a)mcs.anl.gov <javascript:_e(%7B%7D,'cvml','apenya(a)mcs.anl.gov');>www.mcs.anl.gov/~apenya > > -- Jeff Hammond jeff.science(a)gmail.com http://jeffhammond.github.io/ _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

1 0

Re: [mpich-discuss] How to make a non-MPI process to become a MPI process of MPI_COMM_WORLD?
by Balaji, Pavan 24 Jan '15

24 Jan '15

You are missing some files needed to run your program. Also, can you please attach files instead of copy-pasting code into the email? -- Pavan > On Jan 24, 2015, at 7:57 AM, haozi <yidanyiji(a)163.com> wrote: > > Hi. > The following server/client example is selected from OpenMPI. > I compile and run the example by using OpenMPI, everything is OK. > BUT, I compile and run it by mpich(3.1.3), which is BLOCKED as my example! > > This is a bug? > > #include <stdio.h> > #include <stdlib.h> > #include <string.h> > #include <errno.h> > #include <unistd.h> > #include <mpi.h> > > /* > > LOGIC: > > - the 'server' opens a port and write the info to a file > - the 'clients' open the file and connect to the port > - after each accept, the server and client do a merge to > convert the intercomm to an intracomm > > DETAIL STEPS: > > - server open port > - server does accept > - client #1 does connect > - server and client #1 do merge > - server does accept > - client #2 does connect > - server, client #1 and client #2 do merge > - server does accept > - client #3 does connect > - server, client #1, client #2 and client #3 do merge > > */ > > #define TAG 0 > > #define CHK(code) do \ > { \ > int retval = code ; \ > if (retval != MPI_SUCCESS) \ > { \ > fprintf(stderr, "Error: " #code "\n") ; \ > exit(1) ; \ > } \ > } while(0) > > int main(int argc, char *argv[]) > { > char hostname[255] ; > char buff[255] ; > > int role ; > int num_clients ; > int size, rank ; > > FILE *fp ; > char server_port_name[MPI_MAX_PORT_NAME] ; > > MPI_Comm intercomm, intracomm ; > MPI_Status status ; > int msg_count ; > int i ; > > /* sanity check the args */ > if(argc != 3) > { > fprintf(stderr, "usage %s <num clients> <1:server | 0:client>\n", argv[0]) ; > exit(1) ; > } > > num_clients = atoi(argv[1]) ; > role = atoi(argv[2]) ; > > if (num_clients <= 0 || (role != 0 && role != 1)) > { > fprintf(stderr, "usage %s <num clients> <1:server | 0:client>\n", argv[0]) ; > exit(1) ; > } > > /* initialize MPI */ > CHK(MPI_Init(&argc, &argv)) ; > > /* get the node name */ > { > int retval = gethostname(hostname, 255) ; > if(retval == -1) > { > fprintf(stderr, "gethostname failed: %s\n", strerror(errno)) ; > exit(1) ; > } > } > > /* server */ > if(role == 1) > { > printf("SERVER: on node '%s'\n", hostname) ; > > /* open port to establish connections */ > CHK(MPI_Open_port(MPI_INFO_NULL, server_port_name)) ; > > printf("SERVER: opened port=%s\n", server_port_name) ; > > /* store the port name */ > fp = fopen("server_port_name.txt", "w") ; > if(fp == NULL) > { > fprintf(stderr, "fopen failed: %s\n", strerror(errno)) ; > exit(1) ; > } > fprintf(fp, "%s", server_port_name) ; > fclose(fp) ; > > /* the server accepts connections from all the clients */ > for(i = 0 ; i < num_clients ; i++ ) > { > /* accept connections at this port */ > CHK(MPI_Comm_accept(server_port_name, MPI_INFO_NULL, 0, > i == 0 ? MPI_COMM_WORLD : intracomm, > &intercomm)) ; > > printf("SERVER: accepted connection from client %d\n", i+1) ; > > /* merge, to form one intra communicator */ > CHK(MPI_Intercomm_merge(intercomm, 0, &intracomm)) ; > > printf("SERVER: merged with client %d\n", i+1) ; > > CHK(MPI_Comm_size(intracomm, &size)) ; > CHK(MPI_Comm_rank(intracomm, &rank)) ; > > printf("SERVER: after merging with client %d: size=%d rank=%d\n", i+1, size, rank) ; > } > } /* end server */ > > /* client */ > if(role == 0) > { > printf("CLIENT: on node '%s'\n", hostname) ; > > fp = fopen("server_port_name.txt", "r") ; > if(fp == NULL) > { > fprintf(stderr, "fopen failed: %s\n", strerror(errno)) ; > exit(1) ; > } > fscanf(fp, "%s", server_port_name) ; > fclose(fp) ; > > printf("CLIENT: attempting to connect to server on port=%s\n", server_port_name) ; > > /* connect to the server */ > CHK(MPI_Comm_connect (server_port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD, &intercomm)) ; > > printf("CLIENT: connected to server on port\n") ; > > /* merge the server and client to one intra communicator */ > CHK(MPI_Intercomm_merge(intercomm, 1, &intracomm)) ; > > printf("CLIENT: merged with existing intracomm\n") ; > > CHK(MPI_Comm_size(intracomm, &size)) ; > CHK(MPI_Comm_rank(intracomm, &rank)) ; > > printf("CLIENT: after merging, new comm: size=%d rank=%d\n", size, rank) ; > > for (i = rank ; i < num_clients ; i++) > { > /* client performs a collective accept */ > CHK(MPI_Comm_accept(server_port_name, MPI_INFO_NULL, 0, intracomm, &intercomm)) ; > > printf("CLIENT: connected to server on port\n") ; > > /* merge the two intra comms back to one communicator */ > CHK(MPI_Intercomm_merge(intercomm, 0, &intracomm)) ; > > printf("CLIENT: merged with existing members\n") ; > > CHK(MPI_Comm_size(intracomm, &size)) ; > CHK(MPI_Comm_rank(intracomm, &rank)) ; > > printf("CLIENT: new size after merging with existing members: size=%d rank=%d\n", size, rank) ; > } > > } /* end client */ > > CHK(MPI_Comm_size(intracomm, &size)) ; > CHK(MPI_Comm_rank(intracomm, &rank)) ; > > printf("After fusion: size=%d rank=%d\n", size, rank) ; > > if(rank == 0) > { > msg_count = num_clients ; > > while(msg_count) > { > CHK(MPI_Recv(buff, 255, MPI_CHAR, MPI_ANY_SOURCE, > MPI_ANY_TAG, intracomm, &status)) ; > > printf("Received hello msg from '%s'\n", buff) ; > msg_count-- ; > } > } > else > { > /* all ranks > 0 */ > > CHK(MPI_Send(hostname, strlen(hostname) + 1, MPI_CHAR, 0, TAG, intracomm)) ; > } > > CHK(MPI_Finalize()) ; > > fprintf(stderr, "Rank %d is exiting\n", rank); > return 0 ; > } > > > At 2015-01-24 10:43:07, "haozi" <yidanyiji(a)163.com> wrote: > Thanks, Bland and Lu. > > You are right. > These functions (such as MPI_Comm_accept, MPI_Comm_connect, MPI_Intercomm_merge) can help me to get a new intra-communicator which contains ALL MPI processes. > > Now, I have a more complicated example: > I have a server and a client. > After they merge into an intra-communicator by using connect/accept/merge fucntions, another client would plant to join them, too. > I thought that the code is simular, but the code CAN'T work: Second client CAN'T connect. The new comm CAN'T accept. They all BLOCK. > > As you see, the second client BLOCKs at MPI_Comm_connect, and processes of newcomm BLOCK at MPI_Comm_accept. > What's Wrong with my code? > > //server > #include "mpi.h" > int main(int argc, char *argv[]) > { > MPI_Comm client, client2, newcomm, newcomm2; > MPI_Status status; > char port_name[MPI_MAX_PORT_NAME]; > char port_name2[MPI_MAX_PORT_NAME]; > int size, again, rank, myrank; > > MPI_Init(&argc, &argv); > MPI_Open_port(MPI_INFO_NULL, port_name);//OK > > MPI_Comm_accept(port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD,&client);//OK > > MPI_Intercomm_merge(client,11,&newcomm);//OK > > MPI_Barrier(newcomm);//OK > > MPI_Open_port(MPI_INFO_NULL, port_name2);// OK > > MPI_Comm_accept(port_name2, MPI_INFO_NULL, 0, newcomm,&client2);// BLOCK here, Wath's wrong? > > MPI_Intercomm_merge(client2,12,&newcomm2); > > MPI_Barrier(newcomm2); > > MPI_Close_port(port_name); > MPI_Comm_disconnect(&client); > MPI_Close_port(port_name2); > MPI_Comm_disconnect(&client2); > > MPI_Finalize(); > return 0; > } > > //first client > #include "mpi.h" > int main( int argc, char **argv ) > { > MPI_Comm server,newcomm,newcomm2,client2; > char port_name[MPI_MAX_PORT_NAME]; > char port_name2[MPI_MAX_PORT_NAME]; > int size,rank; > > MPI_Init( &argc, &argv ); > > strcpy( port_name, argv[1] ); > MPI_Comm_connect( port_name, MPI_INFO_NULL, 0, MPI_COMM_SELF,&server );//OK > > MPI_Intercomm_merge(server,11,&newcomm);//OK > > MPI_Barrier(newcomm);//OK > > MPI_Open_port(MPI_INFO_NULL, port_name2);//OK > > MPI_Comm_accept(port_name2, MPI_INFO_NULL, 0, newcomm,&client2);// BLOCK here, Wath's wrong? > > MPI_Intercomm_merge(client2,12,&newcomm2); > > MPI_Barrier(newcomm2); > > MPI_Close_port(port_name2); > MPI_Comm_disconnect(&client2); > MPI_Comm_disconnect( &server ); > MPI_Finalize(); > return 0; > } > > //second client > #include "mpi.h" > int main( int argc, char **argv ) > { > MPI_Comm server,newcomm; > char port_name[MPI_MAX_PORT_NAME]; > int size,rank; > > MPI_Init( &argc, &argv ); > > strcpy( port_name, argv[1] );//OK > > MPI_Comm_connect( port_name, MPI_INFO_NULL, 0, MPI_COMM_SELF,&server );// BLOCK here, Wath's wrong? > > MPI_Comm_size(MPI_COMM_WORLD, &size); > > MPI_Intercomm_merge(server,11,&newcomm); > > MPI_Barrier(newcomm); > > MPI_Comm_disconnect( &server ); > MPI_Finalize(); > return 0; > } > > > > > At 2015-01-23 23:14:06, "Wesley Bland" <wbland(a)anl.gov> wrote: > The size of MPI_COMM_WORLD will never change. That communicator is set at initialization time and is not ever modified. However, when you finish connect/accept, you get a new communicator which you can merge into an intra-communicator which functions exactly like MPI_COMM_WORLD. Don’t be attached to the variable name MPI_COMM_WORLD. It’s just a variable and you can use a different one with very little extra work. > > Thanks, > Wesley > > > On Thu, Jan 22, 2015 at 6:14 PM, haozi <yidanyiji(a)163.com> wrote: > Thanks, Lu. > My simple code is as following. > //server > #include "mpi.h" > > int main(int argc, char *argv[]) > { > MPI_Comm client; > MPI_Status status; > char port_name[MPI_MAX_PORT_NAME]; > int size, again; > > MPI_Init(&argc, &argv); > MPI_Comm_size(MPI_COMM_WORLD, &size); > MPI_Open_port(MPI_INFO_NULL, port_name); > printf("server port_name is %s\n\n", port_name); > MPI_Comm_accept(port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD,&client); > MPI_Comm_size(MPI_COMM_WORLD, &size); > printf("At server, comm_size=%d @ MPI_COMM_WORLD=%x, Client_World=%x\n",size,MPI_COMM_WORLD, client); > MPI_Comm_size(client, &size); > printf("At server, client_size=%d @ MPI_COMM_WORLD=%x, Client_World=%x\n",size,MPI_COMM_WORLD, client); > > MPI_Comm_disconnect(&client); > MPI_Finalize(); > return 0; > } > > //client > #include "mpi.h" > > int main( int argc, char **argv ) > { > MPI_Comm server; > char port_name[MPI_MAX_PORT_NAME]; > int size; > > MPI_Init( &argc, &argv ); > strcpy( port_name, argv[1] ); > printf("server port name:%s\n",port_name); > MPI_Comm_connect( port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD,&server ); > MPI_Comm_size(MPI_COMM_WORLD, &size); > printf("At client, comm_size=%d @ MPI_COMM_WORLD=%x, Server_World=%x\n",size,MPI_COMM_WORLD,server); > MPI_Comm_size(server, &size); > printf("At client, server_size=%d @ MPI_COMM_WORLD=%x, Server_World=%x\n",size,MPI_COMM_WORLD,server); > > MPI_Comm_disconnect( &server ); > MPI_Finalize(); > return 0; > } > > The run command is as following. > mpiexec -n 1 ./server > mpiexec -n 1 ./client > > BUT, the SIZE is 1, NOT 2. > > My question is as before: How does the size of MPI_COMM_WORLD change to 2 ? > > > At 2015-01-23 00:30:43, "Huiwei Lu" <huiweilu(a)mcs.anl.gov> wrote: > You may take a look at MPI_Comm_accept and MPI_Comm_connect, which will connect a new client process to a server process. See Chap. 10 of MPI 3.0 standard (www.mpi-forum.org/docs/mpi-3.0/mpi30-report.pdf) for a detail example. > > -- > Huiwei Lu > Postdoc Appointee > Mathematics and Computer Science Division > Argonne National Laboratory > http://www.mcs.anl.gov/~huiweilu/ > > On Thu, Jan 22, 2015 at 9:41 AM, haozi <yidanyiji(a)163.com> wrote: > Hi, guys. > > This web page (http://wiki.mpich.org/mpich/index.php/PMI_v2_Design_Thoughts) says: > Singleton init. This is the process by which a program that was not started with mpiexec can become an MPI process and make use of all MPI features, including MPI_Comm_spawn, needs to be designed and documented, with particular attention to the disposition of standard I/O. Not all process managers will want to or even be able to create a new mpiexec process, so this needs to be negotiated. Similarly, the dispostion of stdio needs to be negotiated between the singleton process and the process manager. To address these issues, a new singleton init protocol has been implemented and tested with the gforker process manager. > > I am very interested in this function. > Can this function solve the following question: > At beginning, the MPI job uses the mpiexec commond to start three MPI processes. That is to say, there are three MPI processes in MPI_COMM_WORLD. At some time, the job find itself to need another MPI process to cooperate the three MPI processes. So the question is: Could PMI help an non-MPI process to become a MPI process of the current MPI_COMM_WORLD? That is to say, Could the non-MPI process use the PMI function to become a member process of the current MPI job which would have FOUR MPI processes in MPI_COMM_WORLD? > > Is there some method to solve question? > Is anybody have some example? > > Thandks!!! > > > > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss > > > > > > > > > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss -- Pavan Balaji ✉️ http://www.mcs.anl.gov/~balaji _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

1 0

Re: [mpich-discuss] Error running examples after install
by Balaji, Pavan 24 Jan '15

24 Jan '15

Sounds like a network setup issue, such as a firewall or /etc/host file description. Did you look through the FAQ entry on this? http://wiki.mpich.org/mpich/index.php/Frequently_Asked_Questions#Q:_My_MPI_… -- Pavan > On Jan 24, 2015, at 12:58 PM, Tiago dos Santos <santos.tmd(a)gmail.com> wrote: > > Hello everyone, > > After installing mpich, I ran the examples and keep getting this stack error: > > tds@ubuntu:~/Downloads/mpich-3.1.3$ mpiexec -f machinefile -n 2 ./examples/cpi > Warning: Permanently added the ECDSA host key for IP address '192.168.201.138' to the list of known hosts. > Process 0 of 2 is on ubuntu > Fatal error in PMPI_Reduce: Unknown error class, error stack: > PMPI_Reduce(1263)...............: MPI_Reduce(sbuf=0x7fff3f86da00, rbuf=0x7fff3f86da08, count=1, MPI_DOUBLE, MPI_SUM, root=0, MPI_COMM_WORLD) failed > MPIR_Reduce_impl(1075)..........: > MPIR_Reduce_intra(881)..........: > MPIR_Reduce_binomial(188).......: > MPIDI_CH3U_Recvq_FDU_or_AEP(636): Communication error with rank 1 > > =================================================================================== > = BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES > = PID 7563 RUNNING AT ubuntu > = EXIT CODE: 1 > = CLEANING UP REMAINING PROCESSES > = YOU CAN IGNORE THE BELOW CLEANUP MESSAGES > =================================================================================== > [proxy:0:1@ubuntu-clone] HYD_pmcd_pmip_control_cmd_cb (pm/pmiserv/pmip_cb.c:885): assert (!closed) failed > [proxy:0:1@ubuntu-clone] HYDT_dmxu_poll_wait_for_event (tools/demux/demux_poll.c:76): callback returned error status > [proxy:0:1@ubuntu-clone] main (pm/pmiserv/pmip.c:206): demux engine error waiting for event > [mpiexec@ubuntu] HYDT_bscu_wait_for_completion (tools/bootstrap/utils/bscu_wait.c:76): one of the processes terminated badly; aborting > [mpiexec@ubuntu] HYDT_bsci_wait_for_completion (tools/bootstrap/src/bsci_wait.c:23): launcher returned error waiting for completion > [mpiexec@ubuntu] HYD_pmci_wait_for_completion (pm/pmiserv/pmiserv_pmci.c:218): launcher returned error waiting for completion > [mpiexec@ubuntu] main (ui/mpich/mpiexec.c:344): process manager error waiting for completion > > > Since I’m pretty new to the MPI world, I kinda can’t get what did I did wrong - Did I do something wrong with ssh? Was it something else? > > System Specifications: > - Ubuntu 14.04 64 bits > - gcc version 4.8.2 > - While installing, fortran support was disable > - This system is running on a virtual machine > > > Network Specification: > - Two machines with the specifications above in a private virtual network > - One machine is called ubuntu and the other one is ubuntu-clone > > Host Files: > - ubuntu > - ubuntu > - ubuntu-clone > - ubuntu-clone > - ubuntu-clone > - ubuntu > > > As you can see, the stack trace is from a command running on host ubuntu. The same error is razed when I run the same command on host ubuntu-clone. > Can anyone help me getting where I messed up? > > Thanks in advance > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss -- Pavan Balaji ✉️ http://www.mcs.anl.gov/~balaji _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

1 0

Error running examples after install
by Tiago dos Santos 24 Jan '15

24 Jan '15

Hello everyone, After installing mpich, I ran the examples and keep getting this stack error: tds@ubuntu:~/Downloads/mpich-3.1.3$ mpiexec -f machinefile -n 2 ./examples/cpi Warning: Permanently added the ECDSA host key for IP address '192.168.201.138' to the list of known hosts. Process 0 of 2 is on ubuntu Fatal error in PMPI_Reduce: Unknown error class, error stack: PMPI_Reduce(1263)...............: MPI_Reduce(sbuf=0x7fff3f86da00, rbuf=0x7fff3f86da08, count=1, MPI_DOUBLE, MPI_SUM, root=0, MPI_COMM_WORLD) failed MPIR_Reduce_impl(1075)..........: MPIR_Reduce_intra(881)..........: MPIR_Reduce_binomial(188).......: MPIDI_CH3U_Recvq_FDU_or_AEP(636): Communication error with rank 1 =================================================================================== = BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES = PID 7563 RUNNING AT ubuntu = EXIT CODE: 1 = CLEANING UP REMAINING PROCESSES = YOU CAN IGNORE THE BELOW CLEANUP MESSAGES =================================================================================== [proxy:0:1@ubuntu-clone] HYD_pmcd_pmip_control_cmd_cb (pm/pmiserv/pmip_cb.c:885): assert (!closed) failed [proxy:0:1@ubuntu-clone] HYDT_dmxu_poll_wait_for_event (tools/demux/demux_poll.c:76): callback returned error status [proxy:0:1@ubuntu-clone] main (pm/pmiserv/pmip.c:206): demux engine error waiting for event [mpiexec@ubuntu] HYDT_bscu_wait_for_completion (tools/bootstrap/utils/bscu_wait.c:76): one of the processes terminated badly; aborting [mpiexec@ubuntu] HYDT_bsci_wait_for_completion (tools/bootstrap/src/bsci_wait.c:23): launcher returned error waiting for completion [mpiexec@ubuntu] HYD_pmci_wait_for_completion (pm/pmiserv/pmiserv_pmci.c:218): launcher returned error waiting for completion [mpiexec@ubuntu] main (ui/mpich/mpiexec.c:344): process manager error waiting for completion Since I’m pretty new to the MPI world, I kinda can’t get what did I did wrong - Did I do something wrong with ssh? Was it something else? System Specifications: - Ubuntu 14.04 64 bits - gcc version 4.8.2 - While installing, fortran support was disable - This system is running on a virtual machine Network Specification: - Two machines with the specifications above in a private virtual network - One machine is called ubuntu and the other one is ubuntu-clone Host Files: - ubuntu - ubuntu - ubuntu-clone - ubuntu-clone - ubuntu-clone - ubuntu As you can see, the stack trace is from a command running on host ubuntu. The same error is razed when I run the same command on host ubuntu-clone. Can anyone help me getting where I messed up? Thanks in advance _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

1 0

Re: [mpich-discuss] How to make a non-MPI process to become a MPI process of MPI_COMM_WORLD?
by haozi 24 Jan '15

24 Jan '15

Hi. The following server/client example is selected from OpenMPI. I compile and run the example by using OpenMPI, everything is OK. BUT, I compile and run it by mpich(3.1.3), which is BLOCKED as my example! This is a bug? #include <stdio.h> #include <stdlib.h> #include <string.h> #include <errno.h> #include <unistd.h> #include <mpi.h> /* LOGIC: - the 'server' opens a port and write the info to a file - the 'clients' open the file and connect to the port - after each accept, the server and client do a merge to convert the intercomm to an intracomm DETAIL STEPS: - server open port - server does accept - client #1 does connect - server and client #1 do merge - server does accept - client #2 does connect - server, client #1 and client #2 do merge - server does accept - client #3 does connect - server, client #1, client #2 and client #3 do merge */ #define TAG 0 #define CHK(code) do \ { \ int retval = code ; \ if (retval != MPI_SUCCESS) \ { \ fprintf(stderr, "Error: " #code "\n") ; \ exit(1) ; \ } \ } while(0) int main(int argc, char *argv[]) { char hostname[255] ; char buff[255] ; int role ; int num_clients ; int size, rank ; FILE *fp ; char server_port_name[MPI_MAX_PORT_NAME] ; MPI_Comm intercomm, intracomm ; MPI_Status status ; int msg_count ; int i ; /* sanity check the args */ if(argc != 3) { fprintf(stderr, "usage %s <num clients> <1:server | 0:client>\n", argv[0]) ; exit(1) ; } num_clients = atoi(argv[1]) ; role = atoi(argv[2]) ; if (num_clients <= 0 || (role != 0 && role != 1)) { fprintf(stderr, "usage %s <num clients> <1:server | 0:client>\n", argv[0]) ; exit(1) ; } /* initialize MPI */ CHK(MPI_Init(&argc, &argv)) ; /* get the node name */ { int retval = gethostname(hostname, 255) ; if(retval == -1) { fprintf(stderr, "gethostname failed: %s\n", strerror(errno)) ; exit(1) ; } } /* server */ if(role == 1) { printf("SERVER: on node '%s'\n", hostname) ; /* open port to establish connections */ CHK(MPI_Open_port(MPI_INFO_NULL, server_port_name)) ; printf("SERVER: opened port=%s\n", server_port_name) ; /* store the port name */ fp = fopen("server_port_name.txt", "w") ; if(fp == NULL) { fprintf(stderr, "fopen failed: %s\n", strerror(errno)) ; exit(1) ; } fprintf(fp, "%s", server_port_name) ; fclose(fp) ; /* the server accepts connections from all the clients */ for(i = 0 ; i < num_clients ; i++ ) { /* accept connections at this port */ CHK(MPI_Comm_accept(server_port_name, MPI_INFO_NULL, 0, i == 0 ? MPI_COMM_WORLD : intracomm, &intercomm)) ; printf("SERVER: accepted connection from client %d\n", i+1) ; /* merge, to form one intra communicator */ CHK(MPI_Intercomm_merge(intercomm, 0, &intracomm)) ; printf("SERVER: merged with client %d\n", i+1) ; CHK(MPI_Comm_size(intracomm, &size)) ; CHK(MPI_Comm_rank(intracomm, &rank)) ; printf("SERVER: after merging with client %d: size=%d rank=%d\n", i+1, size, rank) ; } } /* end server */ /* client */ if(role == 0) { printf("CLIENT: on node '%s'\n", hostname) ; fp = fopen("server_port_name.txt", "r") ; if(fp == NULL) { fprintf(stderr, "fopen failed: %s\n", strerror(errno)) ; exit(1) ; } fscanf(fp, "%s", server_port_name) ; fclose(fp) ; printf("CLIENT: attempting to connect to server on port=%s\n", server_port_name) ; /* connect to the server */ CHK(MPI_Comm_connect (server_port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD, &intercomm)) ; printf("CLIENT: connected to server on port\n") ; /* merge the server and client to one intra communicator */ CHK(MPI_Intercomm_merge(intercomm, 1, &intracomm)) ; printf("CLIENT: merged with existing intracomm\n") ; CHK(MPI_Comm_size(intracomm, &size)) ; CHK(MPI_Comm_rank(intracomm, &rank)) ; printf("CLIENT: after merging, new comm: size=%d rank=%d\n", size, rank) ; for (i = rank ; i < num_clients ; i++) { /* client performs a collective accept */ CHK(MPI_Comm_accept(server_port_name, MPI_INFO_NULL, 0, intracomm, &intercomm)) ; printf("CLIENT: connected to server on port\n") ; /* merge the two intra comms back to one communicator */ CHK(MPI_Intercomm_merge(intercomm, 0, &intracomm)) ; printf("CLIENT: merged with existing members\n") ; CHK(MPI_Comm_size(intracomm, &size)) ; CHK(MPI_Comm_rank(intracomm, &rank)) ; printf("CLIENT: new size after merging with existing members: size=%d rank=%d\n", size, rank) ; } } /* end client */ CHK(MPI_Comm_size(intracomm, &size)) ; CHK(MPI_Comm_rank(intracomm, &rank)) ; printf("After fusion: size=%d rank=%d\n", size, rank) ; if(rank == 0) { msg_count = num_clients ; while(msg_count) { CHK(MPI_Recv(buff, 255, MPI_CHAR, MPI_ANY_SOURCE, MPI_ANY_TAG, intracomm, &status)) ; printf("Received hello msg from '%s'\n", buff) ; msg_count-- ; } } else { /* all ranks > 0 */ CHK(MPI_Send(hostname, strlen(hostname) + 1, MPI_CHAR, 0, TAG, intracomm)) ; } CHK(MPI_Finalize()) ; fprintf(stderr, "Rank %d is exiting\n", rank); return 0 ; } At 2015-01-24 10:43:07, "haozi" <yidanyiji(a)163.com> wrote: Thanks, Bland and Lu. You are right. These functions (such as MPI_Comm_accept, MPI_Comm_connect, MPI_Intercomm_merge) can help me to get a new intra-communicator which contains ALL MPI processes. Now, I have a more complicated example: I have a server and a client. After they merge into an intra-communicator by using connect/accept/merge fucntions, another client would plant to join them, too. I thought that the code is simular, but the code CAN'T work: Second client CAN'T connect. The new comm CAN'T accept. They all BLOCK. As you see, the second client BLOCKs at MPI_Comm_connect, and processes of newcomm BLOCK at MPI_Comm_accept. What's Wrong with my code? //server #include "mpi.h" int main(int argc, char *argv[]) { MPI_Comm client, client2, newcomm, newcomm2; MPI_Status status; char port_name[MPI_MAX_PORT_NAME]; char port_name2[MPI_MAX_PORT_NAME]; int size, again, rank, myrank; MPI_Init(&argc, &argv); MPI_Open_port(MPI_INFO_NULL, port_name);//OK MPI_Comm_accept(port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD,&client);//OK MPI_Intercomm_merge(client,11,&newcomm);//OK MPI_Barrier(newcomm);//OK MPI_Open_port(MPI_INFO_NULL, port_name2);// OK MPI_Comm_accept(port_name2, MPI_INFO_NULL, 0, newcomm,&client2);// BLOCK here, Wath's wrong? MPI_Intercomm_merge(client2,12,&newcomm2); MPI_Barrier(newcomm2); MPI_Close_port(port_name); MPI_Comm_disconnect(&client); MPI_Close_port(port_name2); MPI_Comm_disconnect(&client2); MPI_Finalize(); return 0; } //first client #include "mpi.h" int main( int argc, char **argv ) { MPI_Comm server,newcomm,newcomm2,client2; char port_name[MPI_MAX_PORT_NAME]; char port_name2[MPI_MAX_PORT_NAME]; int size,rank; MPI_Init( &argc, &argv ); strcpy( port_name, argv[1] ); MPI_Comm_connect( port_name, MPI_INFO_NULL, 0, MPI_COMM_SELF,&server );//OK MPI_Intercomm_merge(server,11,&newcomm);//OK MPI_Barrier(newcomm);//OK MPI_Open_port(MPI_INFO_NULL, port_name2);//OK MPI_Comm_accept(port_name2, MPI_INFO_NULL, 0, newcomm,&client2);// BLOCK here, Wath's wrong? MPI_Intercomm_merge(client2,12,&newcomm2); MPI_Barrier(newcomm2); MPI_Close_port(port_name2); MPI_Comm_disconnect(&client2); MPI_Comm_disconnect( &server ); MPI_Finalize(); return 0; } //second client #include "mpi.h" int main( int argc, char **argv ) { MPI_Comm server,newcomm; char port_name[MPI_MAX_PORT_NAME]; int size,rank; MPI_Init( &argc, &argv ); strcpy( port_name, argv[1] );//OK MPI_Comm_connect( port_name, MPI_INFO_NULL, 0, MPI_COMM_SELF,&server );// BLOCK here, Wath's wrong? MPI_Comm_size(MPI_COMM_WORLD, &size); MPI_Intercomm_merge(server,11,&newcomm); MPI_Barrier(newcomm); MPI_Comm_disconnect( &server ); MPI_Finalize(); return 0; } At 2015-01-23 23:14:06, "Wesley Bland" <wbland(a)anl.gov> wrote: The size of MPI_COMM_WORLD will never change. That communicator is set at initialization time and is not ever modified. However, when you finish connect/accept, you get a new communicator which you can merge into an intra-communicator which functions exactly like MPI_COMM_WORLD. Don’t be attached to the variable name MPI_COMM_WORLD. It’s just a variable and you can use a different one with very little extra work. Thanks, Wesley On Thu, Jan 22, 2015 at 6:14 PM, haozi <yidanyiji(a)163.com> wrote: Thanks, Lu. My simple code is as following. //server #include "mpi.h" int main(int argc, char *argv[]) { MPI_Comm client; MPI_Status status; char port_name[MPI_MAX_PORT_NAME]; int size, again; MPI_Init(&argc, &argv); MPI_Comm_size(MPI_COMM_WORLD, &size); MPI_Open_port(MPI_INFO_NULL, port_name); printf("server port_name is %s\n\n", port_name); MPI_Comm_accept(port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD,&client); MPI_Comm_size(MPI_COMM_WORLD, &size); printf("At server, comm_size=%d @ MPI_COMM_WORLD=%x, Client_World=%x\n",size,MPI_COMM_WORLD, client); MPI_Comm_size(client, &size); printf("At server, client_size=%d @ MPI_COMM_WORLD=%x, Client_World=%x\n",size,MPI_COMM_WORLD, client); MPI_Comm_disconnect(&client); MPI_Finalize(); return 0; } //client #include "mpi.h" int main( int argc, char **argv ) { MPI_Comm server; char port_name[MPI_MAX_PORT_NAME]; int size; MPI_Init( &argc, &argv ); strcpy( port_name, argv[1] ); printf("server port name:%s\n",port_name); MPI_Comm_connect( port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD,&server ); MPI_Comm_size(MPI_COMM_WORLD, &size); printf("At client, comm_size=%d @ MPI_COMM_WORLD=%x, Server_World=%x\n",size,MPI_COMM_WORLD,server); MPI_Comm_size(server, &size); printf("At client, server_size=%d @ MPI_COMM_WORLD=%x, Server_World=%x\n",size,MPI_COMM_WORLD,server); MPI_Comm_disconnect( &server ); MPI_Finalize(); return 0; } The run command is as following. mpiexec -n 1 ./server mpiexec -n 1 ./client BUT, the SIZE is 1, NOT 2. My question is as before: How does the size of MPI_COMM_WORLD change to 2 ? At 2015-01-23 00:30:43, "Huiwei Lu" <huiweilu(a)mcs.anl.gov> wrote: You may take a look at MPI_Comm_accept and MPI_Comm_connect, which will connect a new client process to a server process. See Chap. 10 of MPI 3.0 standard (www.mpi-forum.org/docs/mpi-3.0/mpi30-report.pdf) for a detail example. -- Huiwei Lu Postdoc Appointee Mathematics and Computer Science Division Argonne National Laboratory http://www.mcs.anl.gov/~huiweilu/ On Thu, Jan 22, 2015 at 9:41 AM, haozi <yidanyiji(a)163.com> wrote: Hi, guys. This web page (http://wiki.mpich.org/mpich/index.php/PMI_v2_Design_Thoughts) says: Singleton init. This is the process by which a program that was not started with mpiexec can become an MPI process and make use of all MPI features, including MPI_Comm_spawn, needs to be designed and documented, with particular attention to the disposition of standard I/O. Not all process managers will want to or even be able to create a new mpiexec process, so this needs to be negotiated. Similarly, the dispostion of stdio needs to be negotiated between the singleton process and the process manager. To address these issues, a new singleton init protocol has been implemented and tested with the gforker process manager. I am very interested in this function. Can this function solve the following question: At beginning, the MPI job uses the mpiexec commond to start three MPI processes. That is to say, there are three MPI processes in MPI_COMM_WORLD. At some time, the job find itself to need another MPI process to cooperate the three MPI processes. So the question is: Could PMI help an non-MPI process to become a MPI process of the current MPI_COMM_WORLD? That is to say, Could the non-MPI process use the PMI function to become a member process of the current MPI job which would have FOUR MPI processes in MPI_COMM_WORLD? Is there some method to solve question? Is anybody have some example? Thandks!!! _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

1 0

Re: [mpich-discuss] How to make a non-MPI process to become a MPI process of MPI_COMM_WORLD?
by haozi 23 Jan '15

23 Jan '15

Thanks, Bland and Lu. You are right. These functions (such as MPI_Comm_accept, MPI_Comm_connect, MPI_Intercomm_merge) can help me to get a new intra-communicator which contains ALL MPI processes. Now, I have a more complicated example: I have a server and a client. After they merge into an intra-communicator by using connect/accept/merge fucntions, another client would plant to join them, too. I thought that the code is simular, but the code CAN'T work: Second client CAN'T connect. The new comm CAN'T accept. They all BLOCK. As you see, the second client BLOCKs at MPI_Comm_connect, and processes of newcomm BLOCK at MPI_Comm_accept. What's Wrong with my code? //server #include "mpi.h" int main(int argc, char *argv[]) { MPI_Comm client, client2, newcomm, newcomm2; MPI_Status status; char port_name[MPI_MAX_PORT_NAME]; char port_name2[MPI_MAX_PORT_NAME]; int size, again, rank, myrank; MPI_Init(&argc, &argv); MPI_Open_port(MPI_INFO_NULL, port_name);//OK MPI_Comm_accept(port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD,&client);//OK MPI_Intercomm_merge(client,11,&newcomm);//OK MPI_Barrier(newcomm);//OK MPI_Open_port(MPI_INFO_NULL, port_name2);// OK MPI_Comm_accept(port_name2, MPI_INFO_NULL, 0, newcomm,&client2);// BLOCK here, Wath's wrong? MPI_Intercomm_merge(client2,12,&newcomm2); MPI_Barrier(newcomm2); MPI_Close_port(port_name); MPI_Comm_disconnect(&client); MPI_Close_port(port_name2); MPI_Comm_disconnect(&client2); MPI_Finalize(); return 0; } //first client #include "mpi.h" int main( int argc, char **argv ) { MPI_Comm server,newcomm,newcomm2,client2; char port_name[MPI_MAX_PORT_NAME]; char port_name2[MPI_MAX_PORT_NAME]; int size,rank; MPI_Init( &argc, &argv ); strcpy( port_name, argv[1] ); MPI_Comm_connect( port_name, MPI_INFO_NULL, 0, MPI_COMM_SELF,&server );//OK MPI_Intercomm_merge(server,11,&newcomm);//OK MPI_Barrier(newcomm);//OK MPI_Open_port(MPI_INFO_NULL, port_name2);//OK MPI_Comm_accept(port_name2, MPI_INFO_NULL, 0, newcomm,&client2);// BLOCK here, Wath's wrong? MPI_Intercomm_merge(client2,12,&newcomm2); MPI_Barrier(newcomm2); MPI_Close_port(port_name2); MPI_Comm_disconnect(&client2); MPI_Comm_disconnect( &server ); MPI_Finalize(); return 0; } //second client #include "mpi.h" int main( int argc, char **argv ) { MPI_Comm server,newcomm; char port_name[MPI_MAX_PORT_NAME]; int size,rank; MPI_Init( &argc, &argv ); strcpy( port_name, argv[1] );//OK MPI_Comm_connect( port_name, MPI_INFO_NULL, 0, MPI_COMM_SELF,&server );// BLOCK here, Wath's wrong? MPI_Comm_size(MPI_COMM_WORLD, &size); MPI_Intercomm_merge(server,11,&newcomm); MPI_Barrier(newcomm); MPI_Comm_disconnect( &server ); MPI_Finalize(); return 0; } At 2015-01-23 23:14:06, "Wesley Bland" <wbland(a)anl.gov> wrote: The size of MPI_COMM_WORLD will never change. That communicator is set at initialization time and is not ever modified. However, when you finish connect/accept, you get a new communicator which you can merge into an intra-communicator which functions exactly like MPI_COMM_WORLD. Don’t be attached to the variable name MPI_COMM_WORLD. It’s just a variable and you can use a different one with very little extra work. Thanks, Wesley On Thu, Jan 22, 2015 at 6:14 PM, haozi <yidanyiji(a)163.com> wrote: Thanks, Lu. My simple code is as following. //server #include "mpi.h" int main(int argc, char *argv[]) { MPI_Comm client; MPI_Status status; char port_name[MPI_MAX_PORT_NAME]; int size, again; MPI_Init(&argc, &argv); MPI_Comm_size(MPI_COMM_WORLD, &size); MPI_Open_port(MPI_INFO_NULL, port_name); printf("server port_name is %s\n\n", port_name); MPI_Comm_accept(port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD,&client); MPI_Comm_size(MPI_COMM_WORLD, &size); printf("At server, comm_size=%d @ MPI_COMM_WORLD=%x, Client_World=%x\n",size,MPI_COMM_WORLD, client); MPI_Comm_size(client, &size); printf("At server, client_size=%d @ MPI_COMM_WORLD=%x, Client_World=%x\n",size,MPI_COMM_WORLD, client); MPI_Comm_disconnect(&client); MPI_Finalize(); return 0; } //client #include "mpi.h" int main( int argc, char **argv ) { MPI_Comm server; char port_name[MPI_MAX_PORT_NAME]; int size; MPI_Init( &argc, &argv ); strcpy( port_name, argv[1] ); printf("server port name:%s\n",port_name); MPI_Comm_connect( port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD,&server ); MPI_Comm_size(MPI_COMM_WORLD, &size); printf("At client, comm_size=%d @ MPI_COMM_WORLD=%x, Server_World=%x\n",size,MPI_COMM_WORLD,server); MPI_Comm_size(server, &size); printf("At client, server_size=%d @ MPI_COMM_WORLD=%x, Server_World=%x\n",size,MPI_COMM_WORLD,server); MPI_Comm_disconnect( &server ); MPI_Finalize(); return 0; } The run command is as following. mpiexec -n 1 ./server mpiexec -n 1 ./client BUT, the SIZE is 1, NOT 2. My question is as before: How does the size of MPI_COMM_WORLD change to 2 ? At 2015-01-23 00:30:43, "Huiwei Lu" <huiweilu(a)mcs.anl.gov> wrote: You may take a look at MPI_Comm_accept and MPI_Comm_connect, which will connect a new client process to a server process. See Chap. 10 of MPI 3.0 standard (www.mpi-forum.org/docs/mpi-3.0/mpi30-report.pdf) for a detail example. -- Huiwei Lu Postdoc Appointee Mathematics and Computer Science Division Argonne National Laboratory http://www.mcs.anl.gov/~huiweilu/ On Thu, Jan 22, 2015 at 9:41 AM, haozi <yidanyiji(a)163.com> wrote: Hi, guys. This web page (http://wiki.mpich.org/mpich/index.php/PMI_v2_Design_Thoughts) says: Singleton init. This is the process by which a program that was not started with mpiexec can become an MPI process and make use of all MPI features, including MPI_Comm_spawn, needs to be designed and documented, with particular attention to the disposition of standard I/O. Not all process managers will want to or even be able to create a new mpiexec process, so this needs to be negotiated. Similarly, the dispostion of stdio needs to be negotiated between the singleton process and the process manager. To address these issues, a new singleton init protocol has been implemented and tested with the gforker process manager. I am very interested in this function. Can this function solve the following question: At beginning, the MPI job uses the mpiexec commond to start three MPI processes. That is to say, there are three MPI processes in MPI_COMM_WORLD. At some time, the job find itself to need another MPI process to cooperate the three MPI processes. So the question is: Could PMI help an non-MPI process to become a MPI process of the current MPI_COMM_WORLD? That is to say, Could the non-MPI process use the PMI function to become a member process of the current MPI job which would have FOUR MPI processes in MPI_COMM_WORLD? Is there some method to solve question? Is anybody have some example? Thandks!!! _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

1 0

Re: [mpich-discuss] How to make a non-MPI process to become a MPI process of MPI_COMM_WORLD?
by Huiwei Lu 23 Jan '15

23 Jan '15

Based on my understanding, MPI_COMM_WORLD is a pre-defined intra-communicator. I don't think you can change the size of it. What MPI_Comm_accept and MPI_Comm_connect create is a inter-communicator, which consist of two disjoint group of processors. (See Chap 6.6 of MPI 3.0 standard.) If you need a intra-communicator, you may use MPI_Intercomm_merge to create one. -- Huiwei On Thu, Jan 22, 2015 at 6:14 PM, haozi <yidanyiji(a)163.com> wrote: > Thanks, Lu. > My simple code is as following. > //server > #include "mpi.h" > > int main(int argc, char *argv[]) > { > MPI_Comm client; > MPI_Status status; > char port_name[MPI_MAX_PORT_NAME]; > int size, again; > > MPI_Init(&argc, &argv); > MPI_Comm_size(MPI_COMM_WORLD, &size); > MPI_Open_port(MPI_INFO_NULL, port_name); > printf("server port_name is %s\n\n", port_name); > MPI_Comm_accept(port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD,&client); > MPI_Comm_size(MPI_COMM_WORLD, &size); > printf("At server, comm_size=%d @ MPI_COMM_WORLD=%x, > Client_World=%x\n",size,MPI_COMM_WORLD, client); > MPI_Comm_size(client, &size); > printf("At server, client_size=%d @ MPI_COMM_WORLD=%x, > Client_World=%x\n",size,MPI_COMM_WORLD, client); > > MPI_Comm_disconnect(&client); > MPI_Finalize(); > return 0; > } > > //client > #include "mpi.h" > > int main( int argc, char **argv ) > { > MPI_Comm server; > char port_name[MPI_MAX_PORT_NAME]; > int size; > > MPI_Init( &argc, &argv ); > strcpy( port_name, argv[1] ); > printf("server port name:%s\n",port_name); > MPI_Comm_connect( port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD,&server ); > MPI_Comm_size(MPI_COMM_WORLD, &size); > printf("At client, comm_size=%d @ MPI_COMM_WORLD=%x, > Server_World=%x\n",size,MPI_COMM_WORLD,server); > MPI_Comm_size(server, &size); > printf("At client, server_size=%d @ MPI_COMM_WORLD=%x, > Server_World=%x\n",size,MPI_COMM_WORLD,server); > > MPI_Comm_disconnect( &server ); > MPI_Finalize(); > return 0; > } > > The run command is as following. > mpiexec -n 1 ./server > mpiexec -n 1 ./client > > BUT, the SIZE is 1, NOT 2. > > My question is as before: How does the size of MPI_COMM_WORLD change to 2 ? > > > At 2015-01-23 00:30:43, "Huiwei Lu" <huiweilu(a)mcs.anl.gov> wrote: > > You may take a look at MPI_Comm_accept and MPI_Comm_connect, which will > connect a new client process to a server process. See Chap. 10 of MPI 3.0 > standard (www.mpi-forum.org/docs/mpi-3.0/mpi30-report.pdf) for a detail > example. > > -- > Huiwei Lu > Postdoc Appointee > Mathematics and Computer Science Division > Argonne National Laboratory > http://www.mcs.anl.gov/~huiweilu/ > > On Thu, Jan 22, 2015 at 9:41 AM, haozi <yidanyiji(a)163.com> wrote: > >> Hi, guys. >> >> This web page ( >> http://wiki.mpich.org/mpich/index.php/PMI_v2_Design_Thoughts) says: >> *Singleton init.* This is the process by which a program >> that was not started with mpiexec can *become an MPI process *and make >> use of all MPI features, including MPI_Comm_spawn, needs to be designed and >> documented, with particular attention to the disposition of standard I/O. >> Not all process managers will want to or even be able to create a new >> mpiexec process, so this needs to be negotiated. Similarly, the dispostion >> of stdio needs to be negotiated between the singleton process and the >> process manager. To address these issues, a new singleton init protocol has >> been implemented and tested with the gforker process manager. >> >> I am very interested in this function. >> Can this function solve the following question: >> At beginning, the MPI job uses the mpiexec commond to start >> three MPI processes. That is to say, there are three MPI processes in >> MPI_COMM_WORLD. At some time, the job find itself to need another MPI >> process to cooperate the three MPI processes. So the question is: Could PMI >> help an non-MPI process to become a MPI process of the current >> MPI_COMM_WORLD? That is to say, Could the non-MPI process use the PMI >> function to become a member process of the current MPI job which would have >> FOUR MPI processes in MPI_COMM_WORLD? >> >> Is there some method to solve question? >> Is anybody have some example? >> >> Thandks!!! >> >> >> >> _______________________________________________ >> discuss mailing list discuss(a)mpich.org >> To manage subscription options or unsubscribe: >> https://lists.mpich.org/mailman/listinfo/discuss >> > > > > > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss > _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

1 0