September 2013 - discuss - mailman3.cels.anl.gov

Re: [mpich-discuss] Implementation of MPICH collectives
by Pavan Balaji 12 Sep '13

12 Sep '13

Jiri, The communication with the process manager (mpiexec) is only to lookup information about remote processes. The process manager is not involved in most MPI operations, e.g., send/recv or collectives. The process manager (Hydra, as well as other process managers in mpich and all its derivatives), essentially provide a key-value database. Each process can put some key/vals in there and lookup key/vals put by other processes. When hydra_pmi_proxy starts the MPI processes, each process puts a "business card" in the key-value space. After that, if process 0 wants to communicate with process 1, it'll look up the business card for process 1 and establish a communication channel with it (shared memory in this case). After the communication channel is setup, all communication happens directly between the two processes without involving the process manager. That's of course a very rough layout of how things happen. In practice, there are several optimizations to make this more efficient, with the distributed proxies exchanging and caching information for efficiency. Also, the exchange of business card information with the process manager is eager in some cases (e.g., for shared memory) and lazy in other cases (e.g., for network connections). -- Pavan On Sep 12, 2013, at 12:36 PM, Jiri Simsa wrote: > Hello, > > I have been trying to understand how MPICH implements collective operations. To do so, I have been reading the MPICH source code and stepping through mpiexec executions. > > For the sake of this discussion, let's assume that all MPI processes are executed on the same computer using: mpiexec -n <n> <mpi_program> > > This is my current abstract understanding of MPICH: > > - mpiexec spawns a hydra_pmi_proxy process, which in turn spawns <n> instances of <mpi_program> > - hydra_pmi_proxy process uses socket pairs to communicate with the instances of <mpi_program> > > I am not quite sure though what happens under the hoods when a collective operation, such as MPI_Allreduce, is executed. I have noticed that instances of <mpi_program> create and listen on a socket in the course of executing MPI_Allreduce but I am not sure who connects to these sockets. Any chance someone could describe the data flow inside of MPICH when a collective operation, such as MPI_Allreduce, is executed? Thanks! > > Best, > > --Jiri Simsa > _______________________________________________ > 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

1 0

Implementation of MPICH collectives
by Jiri Simsa 12 Sep '13

12 Sep '13

Hello, I have been trying to understand how MPICH implements collective operations. To do so, I have been reading the MPICH source code and stepping through mpiexec executions. For the sake of this discussion, let's assume that all MPI processes are executed on the same computer using: mpiexec -n <n> <mpi_program> This is my current abstract understanding of MPICH: - mpiexec spawns a hydra_pmi_proxy process, which in turn spawns <n> instances of <mpi_program> - hydra_pmi_proxy process uses socket pairs to communicate with the instances of <mpi_program> I am not quite sure though what happens under the hoods when a collective operation, such as MPI_Allreduce, is executed. I have noticed that instances of <mpi_program> create and listen on a socket in the course of executing MPI_Allreduce but I am not sure who connects to these sockets. Any chance someone could describe the data flow inside of MPICH when a collective operation, such as MPI_Allreduce, is executed? Thanks! Best, --Jiri Simsa

1 0

Re: [mpich-discuss] MPI and bordered block system with LU decomposition
by Mauro Crociara 12 Sep '13

12 Sep '13

Yes, thanks for the support and your time, I really appreciated. Best regards, Mauro Crociara 2013/9/11 Jeff Hammond <jeff.science(a)gmail.com>: > Did you fix the problem with your not calling collectives collectively too? > > > On Wed, Sep 11, 2013 at 12:56 PM, Mauro Crociara <mauro.crociara(a)gmail.com> > wrote: >> >> Ok thanks for the support, but i got the same problem with MPI_Reduce >> (d_tmp, b_tmp, LD * N, MPI_DOUBLE, MPI_SUM, MASTER, MPI_COMM_WORLD); >> >> Best >> Mauro Crociara >> >> 2013/9/11 Jeff Hammond <jeff.science(a)gmail.com>: >> > I answered on stackoverflow. >> > >> > Best, >> > >> > Jeff >> > >> > >> > On Wed, Sep 11, 2013 at 11:20 AM, Mauro Crociara >> > <mauro.crociara(a)gmail.com> >> > wrote: >> >> >> >> Hi guys, i'm new with MPI and i got some problem to transform a serial >> >> code into parallel code, please see here for thecode >> >> >> >> >> >> http://stackoverflow.com/questions/18743721/mpi-bordered-block-system-matri… >> >> >> >> Here the GDB error: http://bpaste.net/show/tDDBNTyqVpUjxFUF3qSr/ >> >> >> >> I'm sorry for the Italian comments, please ignore it. Btw the code >> >> should be clear. Thanks so much in advance. >> >> >> >> Have a nice day guys! >> >> -- >> >> >> >> Best Regards, >> >> Mauro Crociara >> >> >> >> _______________________________________________ >> >> discuss mailing list discuss(a)mpich.org >> >> To manage subscription options or unsubscribe: >> >> https://lists.mpich.org/mailman/listinfo/discuss >> > >> > >> > >> > >> > -- >> > Jeff Hammond >> > jeff.science(a)gmail.com >> > >> > _______________________________________________ >> > discuss mailing list discuss(a)mpich.org >> > To manage subscription options or unsubscribe: >> > https://lists.mpich.org/mailman/listinfo/discuss >> >> >> >> -- >> >> Cordiali saluti, >> Mauro Crociara >> _______________________________________________ >> discuss mailing list discuss(a)mpich.org >> To manage subscription options or unsubscribe: >> https://lists.mpich.org/mailman/listinfo/discuss > > > > > -- > Jeff Hammond > jeff.science(a)gmail.com > > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss -- Cordiali saluti, Mauro Crociara

1 0

Re: [mpich-discuss] MPI and bordered block system with LU decomposition
by Jeff Hammond 11 Sep '13

11 Sep '13

Did you fix the problem with your not calling collectives collectively too? On Wed, Sep 11, 2013 at 12:56 PM, Mauro Crociara <mauro.crociara(a)gmail.com>wrote: > Ok thanks for the support, but i got the same problem with MPI_Reduce > (d_tmp, b_tmp, LD * N, MPI_DOUBLE, MPI_SUM, MASTER, MPI_COMM_WORLD); > > Best > Mauro Crociara > > 2013/9/11 Jeff Hammond <jeff.science(a)gmail.com>: > > I answered on stackoverflow. > > > > Best, > > > > Jeff > > > > > > On Wed, Sep 11, 2013 at 11:20 AM, Mauro Crociara < > mauro.crociara(a)gmail.com> > > wrote: > >> > >> Hi guys, i'm new with MPI and i got some problem to transform a serial > >> code into parallel code, please see here for thecode > >> > >> > http://stackoverflow.com/questions/18743721/mpi-bordered-block-system-matri… > >> > >> Here the GDB error: http://bpaste.net/show/tDDBNTyqVpUjxFUF3qSr/ > >> > >> I'm sorry for the Italian comments, please ignore it. Btw the code > >> should be clear. Thanks so much in advance. > >> > >> Have a nice day guys! > >> -- > >> > >> Best Regards, > >> Mauro Crociara > >> > >> _______________________________________________ > >> discuss mailing list discuss(a)mpich.org > >> To manage subscription options or unsubscribe: > >> https://lists.mpich.org/mailman/listinfo/discuss > > > > > > > > > > -- > > Jeff Hammond > > jeff.science(a)gmail.com > > > > _______________________________________________ > > discuss mailing list discuss(a)mpich.org > > To manage subscription options or unsubscribe: > > https://lists.mpich.org/mailman/listinfo/discuss > > > > -- > > Cordiali saluti, > Mauro Crociara > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss > -- Jeff Hammond jeff.science(a)gmail.com

1 0

Re: [mpich-discuss] MPI and bordered block system with LU decomposition
by Mauro Crociara 11 Sep '13

11 Sep '13

Ok thanks for the support, but i got the same problem with MPI_Reduce (d_tmp, b_tmp, LD * N, MPI_DOUBLE, MPI_SUM, MASTER, MPI_COMM_WORLD); Best Mauro Crociara 2013/9/11 Jeff Hammond <jeff.science(a)gmail.com>: > I answered on stackoverflow. > > Best, > > Jeff > > > On Wed, Sep 11, 2013 at 11:20 AM, Mauro Crociara <mauro.crociara(a)gmail.com> > wrote: >> >> Hi guys, i'm new with MPI and i got some problem to transform a serial >> code into parallel code, please see here for thecode >> >> http://stackoverflow.com/questions/18743721/mpi-bordered-block-system-matri… >> >> Here the GDB error: http://bpaste.net/show/tDDBNTyqVpUjxFUF3qSr/ >> >> I'm sorry for the Italian comments, please ignore it. Btw the code >> should be clear. Thanks so much in advance. >> >> Have a nice day guys! >> -- >> >> Best Regards, >> Mauro Crociara >> >> _______________________________________________ >> discuss mailing list discuss(a)mpich.org >> To manage subscription options or unsubscribe: >> https://lists.mpich.org/mailman/listinfo/discuss > > > > > -- > Jeff Hammond > jeff.science(a)gmail.com > > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss -- Cordiali saluti, Mauro Crociara

1 0

Re: [mpich-discuss] MPI and bordered block system with LU decomposition
by Jeff Hammond 11 Sep '13

11 Sep '13

I answered on stackoverflow. Best, Jeff On Wed, Sep 11, 2013 at 11:20 AM, Mauro Crociara <mauro.crociara(a)gmail.com>wrote: > Hi guys, i'm new with MPI and i got some problem to transform a serial > code into parallel code, please see here for thecode > > http://stackoverflow.com/questions/18743721/mpi-bordered-block-system-matri… > > Here the GDB error: http://bpaste.net/show/tDDBNTyqVpUjxFUF3qSr/ > > I'm sorry for the Italian comments, please ignore it. Btw the code > should be clear. Thanks so much in advance. > > Have a nice day guys! > -- > > Best Regards, > Mauro Crociara > > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss > -- Jeff Hammond jeff.science(a)gmail.com

1 0

MPI and bordered block system with LU decomposition
by Mauro Crociara 11 Sep '13

11 Sep '13

Hi guys, i'm new with MPI and i got some problem to transform a serial code into parallel code, please see here for thecode http://stackoverflow.com/questions/18743721/mpi-bordered-block-system-matri… Here the GDB error: http://bpaste.net/show/tDDBNTyqVpUjxFUF3qSr/ I'm sorry for the Italian comments, please ignore it. Btw the code should be clear. Thanks so much in advance. Have a nice day guys! -- Best Regards, Mauro Crociara

1 0

Re: [mpich-discuss] MPI 3 on Win7(64-bit)
by Wesley Bland 11 Sep '13

11 Sep '13

Unfortunately, we haven't supported Windows since MPICH version 1.4.1p1. You can find out more at the FAQ page: http://wiki.mpich.org/mpich/index.php/Frequently_Asked_Questions#Q:_Why_can… Thanks, Wesley On Sep 11, 2013, at 8:13 AM, Jeff Hammond <jeff.science(a)gmail.com> wrote: > There's a frequently quoted entry in the MPICH FAQ about Windows support that you may wish to consult. Or peruse list archives. > > Jeff > > Sent from my iPhone > > On Sep 11, 2013, at 5:55 AM, "Michael.Rachner(a)dlr.de" <Michael.Rachner(a)dlr.de> wrote: > >> Dear developers of MPICH, >> >> Can anyone tell me, when a stable MPI 3 version (binary) of MPICH will be available for Win7(64-bit) operating system? >> The new feature of MPI 3 of accessing a shared memory data window from the processes on the same node could become very helpful for my Ftn-code. >> >> Thank You >> Michael Rachner >> _______________________________________________ >> 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] MPI 3 on Win7(64-bit)
by Jeff Hammond 11 Sep '13

11 Sep '13

There's a frequently quoted entry in the MPICH FAQ about Windows support that you may wish to consult. Or peruse list archives. Jeff Sent from my iPhone On Sep 11, 2013, at 5:55 AM, "Michael.Rachner(a)dlr.de" < Michael.Rachner(a)dlr.de> wrote: Dear developers of MPICH, Can anyone tell me, when a stable MPI 3 version (binary) of MPICH will be available for Win7(64-bit) operating system? The new feature of MPI 3 of accessing a shared memory data window from the processes on the same node could become very helpful for my Ftn-code. Thank You Michael Rachner _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

1 0

MPI 3 on Win7(64-bit)
by Michael.Rachner＠dlr.de 11 Sep '13

11 Sep '13

Dear developers of MPICH, Can anyone tell me, when a stable MPI 3 version (binary) of MPICH will be available for Win7(64-bit) operating system? The new feature of MPI 3 of accessing a shared memory data window from the processes on the same node could become very helpful for my Ftn-code. Thank You Michael Rachner

1 0