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

Re: [mpich-discuss] MPI Benchmark Suite
by Pavan Balaji 05 Jul '13

05 Jul '13

On 07/05/2013 10:38 AM, Jiri Simsa wrote: > Could anyone point me in the direction of programs that are > representative of MPI programs? I am looking for something like PARSEC > (http://parsec.cs.princeton.edu/) for MPI. In other words, I am > interested in macro-bechnmarks not micro-benchmarks. Thank you. You could try the Sequoia suite. There's also the NAS parallel benchmarks and Graph500 benchmarks. -- Pavan -- Pavan Balaji http://www.mcs.anl.gov/~balaji

1 0

Re: [mpich-discuss] beginner for remote mem access (sufeng)
by Pavan Balaji 05 Jul '13

05 Jul '13

On 07/05/2013 11:18 AM, Sufeng Niu wrote: > for hybrid RMA issue, did you mean in each rank, remote get the data > into a shared memory, then threads in each rank load it by thread-level > operation? All threads are in the same process, so they all already share memory. Perhaps its best if you try to learn how threads work before attempting to use them with MPI. -- Pavan -- Pavan Balaji http://www.mcs.anl.gov/~balaji

1 0

Re: [mpich-discuss] beginner for remote mem access (sufeng)
by Sufeng Niu 05 Jul '13

05 Jul '13

Hi Pavan, Thanks a lot! Now I make sense for MPI_Win_fence. for hybrid RMA issue, did you mean in each rank, remote get the data into a shared memory, then threads in each rank load it by thread-level operation? Thank you! Sufeng On Fri, Jul 5, 2013 at 10:36 AM, Pavan Balaji <balaji(a)mcs.anl.gov> wrote: > > On 07/05/2013 10:28 AM, Sufeng Niu wrote: > >> 1. >> MPI_Win_fence is not similar to MPI_Barrier in some cases. The >> MPICH implementation can turn some calls into no-ops and others into >> MPI_Reduce_scatter. >> I am not quite sure about "turn some calls into no-ops and others into >> MPI_Reduce_scatter" Could you please give an example if possible? >> Another thing is if a process create a window for other process, but the >> data access is available after some operations. Should I use >> MPI_Win_fence or MPI_Barrier to sync? or other methods? >> > > MPI_WIN_FENCE will make sure that all PUT/GET/ACCUMULATE operations are > complete locally and remotely. This is probably what you need. > > MPI_BARRIER is useless in this case since PUT/GET/ACCUMULATE can be > nonblocking. > > > 2. >> Regarding "use MPI window as thread level", I really don't >> understand your thinking at all. MPI RMA is not shared memory nor is >> MPI a threading model. >> Sorry for uncleared statements, right now I would like to do >> multithreads and MPI hybrid programming. let me give an example: I have >> 3 processes, each one has 8 threads. thread 0 in process 0 creates a RMA >> window. if I would like all other threads to access it, should I use >> thread 0 in process 1, thread 0 in process 2 to MPI_Get the data from >> window, then use shared memory for internal threads to load the data? i >> am not sure what is the proper way for RMA in hybrid model. >> > > All threads within the OS process are part of the same MPI rank. Any of > them can access the window as the same rank. > > -- Pavan > > -- > Pavan Balaji > http://www.mcs.anl.gov/~balaji > -- Best Regards, Sufeng Niu ECASP lab, ECE department, Illinois Institute of Technology Tel: 312-731-7219

1 0

Re: [mpich-discuss] beginner for remote mem access (sufeng)
by Jim Dinan 05 Jul '13

05 Jul '13

On Fri, Jul 5, 2013 at 11:28 AM, Sufeng Niu <sniu(a)hawk.iit.edu> wrote: > 1. > MPI_Win_fence is not similar to MPI_Barrier in some cases. The > MPICH implementation can turn some calls into no-ops and others into > MPI_Reduce_scatter. > I am not quite sure about "turn some calls into no-ops and others into > MPI_Reduce_scatter" Could you please give an example if possible? Another > thing is if a process create a window for other process, but the data > access is available after some operations. Should I use MPI_Win_fence or > MPI_Barrier to sync? or other methods? > When you call MPI_Win_fence, the MPI implementation does not need to wait for all processes to call the function before proceeding. It does require that all RMA operations in a given process' access epoch are complete, and that all RMA accesses in its exposure epoch (i.e., targeting the given process) are also complete. If, for example, there is no preceding epoch, the call to MPI_Win_fence is a no op. Generally speaking, you should synchronize all accesses to data exposed in a window using RMA synchronization operations. Barrier does not provide any consistency for RMA operations. > 2. > Regarding "use MPI window as thread level", I really don't > understand your thinking at all. MPI RMA is not shared memory nor is > MPI a threading model. > Sorry for uncleared statements, right now I would like to do multithreads > and MPI hybrid programming. let me give an example: I have 3 processes, > each one has 8 threads. thread 0 in process 0 creates a RMA window. if I > would like all other threads to access it, should I use thread 0 in process > 1, thread 0 in process 2 to MPI_Get the data from window, then use shared > memory for internal threads to load the data? i am not sure what is the > proper way for RMA in hybrid model. > The threads can all access the data in the window, provided that the MPI process synchronizes accesses by threads in the process correctly with respect to what other processes are doing. Note that data consistency across threads is outside of the MPI specification, and you should use e.g. Pthreads or OpenMP primitives to ensure that data is consistent across threads. ~Jim.

1 0

MPI Benchmark Suite
by Jiri Simsa 05 Jul '13

05 Jul '13

Hello, Could anyone point me in the direction of programs that are representative of MPI programs? I am looking for something like PARSEC ( http://parsec.cs.princeton.edu/) for MPI. In other words, I am interested in macro-bechnmarks not micro-benchmarks. Thank you. Best, --Jiri Simsa

1 0

Re: [mpich-discuss] beginner for remote mem access (sufeng)
by Pavan Balaji 05 Jul '13

05 Jul '13

On 07/05/2013 10:28 AM, Sufeng Niu wrote: > 1. > MPI_Win_fence is not similar to MPI_Barrier in some cases. The > MPICH implementation can turn some calls into no-ops and others into > MPI_Reduce_scatter. > I am not quite sure about "turn some calls into no-ops and others into > MPI_Reduce_scatter" Could you please give an example if possible? > Another thing is if a process create a window for other process, but the > data access is available after some operations. Should I use > MPI_Win_fence or MPI_Barrier to sync? or other methods? MPI_WIN_FENCE will make sure that all PUT/GET/ACCUMULATE operations are complete locally and remotely. This is probably what you need. MPI_BARRIER is useless in this case since PUT/GET/ACCUMULATE can be nonblocking. > 2. > Regarding "use MPI window as thread level", I really don't > understand your thinking at all. MPI RMA is not shared memory nor is > MPI a threading model. > Sorry for uncleared statements, right now I would like to do > multithreads and MPI hybrid programming. let me give an example: I have > 3 processes, each one has 8 threads. thread 0 in process 0 creates a RMA > window. if I would like all other threads to access it, should I use > thread 0 in process 1, thread 0 in process 2 to MPI_Get the data from > window, then use shared memory for internal threads to load the data? i > am not sure what is the proper way for RMA in hybrid model. All threads within the OS process are part of the same MPI rank. Any of them can access the window as the same rank. -- Pavan -- Pavan Balaji http://www.mcs.anl.gov/~balaji

1 0

Re: [mpich-discuss] beginner for remote mem access (sufeng)
by Sufeng Niu 05 Jul '13

05 Jul '13

Hi, Jeff Thanks a lot for your reply. Sorry about subject issue. 1. MPI_Win_fence is not similar to MPI_Barrier in some cases. The MPICH implementation can turn some calls into no-ops and others into MPI_Reduce_scatter. I am not quite sure about "turn some calls into no-ops and others into MPI_Reduce_scatter" Could you please give an example if possible? Another thing is if a process create a window for other process, but the data access is available after some operations. Should I use MPI_Win_fence or MPI_Barrier to sync? or other methods? 2. Regarding "use MPI window as thread level", I really don't understand your thinking at all. MPI RMA is not shared memory nor is MPI a threading model. Sorry for uncleared statements, right now I would like to do multithreads and MPI hybrid programming. let me give an example: I have 3 processes, each one has 8 threads. thread 0 in process 0 creates a RMA window. if I would like all other threads to access it, should I use thread 0 in process 1, thread 0 in process 2 to MPI_Get the data from window, then use shared memory for internal threads to load the data? i am not sure what is the proper way for RMA in hybrid model. Thank you so much! Sufeng On Fri, Jul 5, 2013 at 6:15 AM, <discuss-request(a)mpich.org> wrote: > Send discuss mailing list submissions to > discuss(a)mpich.org > > To subscribe or unsubscribe via the World Wide Web, visit > https://lists.mpich.org/mailman/listinfo/discuss > or, via email, send a message with subject or body 'help' to > discuss-request(a)mpich.org > > You can reach the person managing the list at > discuss-owner(a)mpich.org > > When replying, please edit your Subject line so it is more specific > than "Re: Contents of discuss digest..." > > > Today's Topics: > > 1. Re: beginner for remote mem access (Jeff Hammond) > 2. Running configure on Suse Enterprise 10 > (anton.s.murfin(a)sellafieldsites.com) > > > ---------------------------------------------------------------------- > > Message: 1 > Date: Wed, 3 Jul 2013 14:50:44 -0500 > From: Jeff Hammond <jeff.science(a)gmail.com> > To: discuss(a)mpich.org > Subject: Re: [mpich-discuss] beginner for remote mem access > Message-ID: > <CAGKz=uJN2iH3Pxjxyq=YzrP= > XG5Tp2+huxmBFTh+Lr7nVdLU4w(a)mail.gmail.com> > Content-Type: text/plain; charset=ISO-8859-1 > > 1) When you reply to the digest emails, change the subject to that of > the thread you care to respond to. That will make it much easier for > others to track the thread. > 2) MPI_Win_fence is not similar to MPI_Barrier in some cases. The > MPICH implementation can turn some calls into no-ops and others into > MPI_Reduce_scatter. > 3) Regarding "use MPI window as thread level", I really don't > understand your thinking at all. MPI RMA is not shared memory nor is > MPI a threading model. > > Jeff > > On Wed, Jul 3, 2013 at 2:46 PM, Sufeng Niu <sniu(a)hawk.iit.edu> wrote: > > > > Thanks a lot for your guys reply. I should initial b array, I figure out > the > > MPI_Win_fence should be called by all processes, which will solve the > > assertion failed issue. > > > > it seems that MPI_Win_fence is very similar to MPI_Barrier, Can I say > > MPI_Win_fence is the RMA version of MPI_Barrier? > > > > sorry about so many questions. Can I use MPI window as thread level? > let's > > say I have 4 processes, each process has 8 threads. if only one thread is > > used to create mem access window, other process copy it and internal 8 > > threads use shared memory to perform some operations. Is that ok? > > > > Thanks a lot! > > > > > > > > On Wed, Jul 3, 2013 at 1:28 PM, <discuss-request(a)mpich.org> wrote: > >> > >> Send discuss mailing list submissions to > >> discuss(a)mpich.org > >> > >> To subscribe or unsubscribe via the World Wide Web, visit > >> https://lists.mpich.org/mailman/listinfo/discuss > >> or, via email, send a message with subject or body 'help' to > >> discuss-request(a)mpich.org > >> > >> You can reach the person managing the list at > >> discuss-owner(a)mpich.org > >> > >> When replying, please edit your Subject line so it is more specific > >> than "Re: Contents of discuss digest..." > >> > >> > >> Today's Topics: > >> > >> 1. beginner for remote mem access (Sufeng Niu) > >> 2. Re: beginner for remote mem access (Yi Gu) > >> 3. Re: beginner for remote mem access (Yi Gu) > >> 4. Re: beginner for remote mem access (Jeff Hammond) > >> > >> > >> ---------------------------------------------------------------------- > >> > >> Message: 1 > >> Date: Wed, 3 Jul 2013 13:12:46 -0500 > >> From: Sufeng Niu <sniu(a)hawk.iit.edu> > >> To: discuss(a)mpich.org > >> Subject: [mpich-discuss] beginner for remote mem access > >> Message-ID: > >> > >> <CAFNNHkwxqXUB_+b8_tZ=L5K7VcDVhsz4ZMDDMGF55v-FAhgeqg(a)mail.gmail.com> > >> Content-Type: text/plain; charset="iso-8859-1" > >> > >> Hi, > >> > >> I am a beginner and just try to use remote memory access, and I wrote a > >> simple program to test it: > >> > >> #include "mpi.h" > >> #include <stdio.h> > >> #define SIZE 8 > >> > >> int main(int argc, char *argv[]) > >> { > >> int numtasks, rank, source=0, dest, tag=1, i; > >> float a[64] = > >> { > >> 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, > >> 12.0, 13.0, 14.0, 15.0, 16.0, > >> 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, > >> 12.0, 13.0, 14.0, 15.0, 16.0, > >> 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, > >> 12.0, 13.0, 14.0, 15.0, 16.0, > >> 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, > >> 12.0, 13.0, 14.0, 15.0, 16.0, > >> }; > >> float b[SIZE]; > >> > >> MPI_Status stat; > >> > >> MPI_Init(&argc,&argv); > >> MPI_Comm_rank(MPI_COMM_WORLD, &rank); > >> MPI_Comm_size(MPI_COMM_WORLD, &numtasks); > >> > >> MPI_Win win; > >> > >> // check processor rank > >> char processor_name[MPI_MAX_PROCESSOR_NAME]; > >> int name_len; > >> MPI_Get_processor_name(processor_name, &name_len); > >> printf("-- processor %s, rank %d out of %d processors\n", > >> processor_name, rank, numtasks); > >> > >> MPI_Barrier(MPI_COMM_WORLD); > >> > >> if (numtasks == 4) { > >> if (rank == 0) { > >> printf("create window \n"); > >> MPI_Win_create(a, 8*sizeof(float), > sizeof(float), > >> MPI_INFO_NULL, MPI_COMM_WORLD, &win); > >> > >> } > >> else { > >> MPI_Win_create(MPI_BOTTOM, 0, sizeof(float), > >> MPI_INFO_NULL, MPI_COMM_WORLD, &win); > >> } > >> > >> MPI_Win_fence(0, win); > >> > >> if (rank == 1){ > >> MPI_Get(b, SIZE, MPI_FLOAT, 0, 8, SIZE, > MPI_FLOAT, > >> win); > >> > >> MPI_Win_fence(0, win); > >> } > >> > >> printf("rank= %d b= %3.1f %3.1f %3.1f %3.1f %3.1f %3.1f > >> %3.1f %3.1f\n", rank,b[0],b[1],b[2],b[3],b[4],b[5],b[6],b[7]); > >> } > >> else > >> printf("Must specify %d processors. > Terminating.\n",SIZE); > >> > >> MPI_Win_free(&win); > >> MPI_Finalize(); > >> } > >> > >> However the terminal gives some odd results: > >> rank= 0 b= 0.0 0.0 0.0 0.0 0.0 0.0 -71847793475452928.0 0.0 > >> rank= 2 b= 0.0 0.0 0.0 0.0 0.0 0.0 222086852849451401216.0 0.0 > >> rank= 3 b= 0.0 0.0 0.0 0.0 0.0 0.0 -74882.4 0.0 > >> rank= 1 b= 9.0 10.0 11.0 12.0 13.0 14.0 15.0 16.0 > >> the rank 1 is correct results, but others should be all zero. > >> terminal also give some comments: "Assertion failed in file > >> src/mpid/ch3/src/ch3u_rma_sync.c at line 5061: win_ptr->my_counter >= 0 > >> internal ABORT - process 0" > >> > >> another question is if I use remote memory access. all process which > does > >> not create window for share must add additional line: > >> MPI_Win_create(MPI_BOTTOM, 0, data_type, MPI_INFO_NULL, MPI_COMM_WORLD, > >> &win); correct? > >> > >> Thanks a lot! > >> > >> -- > >> Best Regards, > >> Sufeng Niu > >> ECASP lab, ECE department, Illinois Institute of Technology > >> Tel: 312-731-7219 > >>

1 0

Re: [mpich-discuss] Running configure on Suse Enterprise 10
by anton.s.murfin＠sellafieldsites.com 05 Jul '13

05 Jul '13

1 0

Re: [mpich-discuss] Running configure on Suse Enterprise 10
by William Gropp 05 Jul '13

05 Jul '13

Can you also send that config.log file that was created - this contains more details; in particular, I need to see why configure thought that the Fortran compiler worked but then was unable to use it to determine properties such as the size of a Fortran INTEGER. Bill William Gropp Director, Parallel Computing Institute Deputy Director for Research Institute for Advanced Computing Applications and Technologies Thomas M. Siebel Chair in Computer Science University of Illinois Urbana-Champaign On Jul 5, 2013, at 6:15 AM, <anton.s.murfin(a)sellafieldsites.com> wrote: > > Dear MPICH developers, > > I have downloaded and extracted mpich-3.0.4.tar.gz to /home/ris3/asm7/mpich-3.0.4/ > > I then created a build directory as /home/ris3/asm7/mpich-3.0.4_build/ > > I went back to /home/ris3/asm7/mpich-3.0.4/ and executed the .configure script detailed below > > ./configure --prefix = /home/ris3/asm7/mpich-3.0.4_build 2>&1 | tee c.txt (I confirmed I was using bash shell, using echo $SHELL) > > The output from this c.txt is attached and there is only one error right at the bottom although there are some warnings. > > No Makefile is created so the next step make doesn't work. > > I just wanted to see if you can see what the issue is. > > Many thanks > > Anton > > > "The information contained in this email may be commercially sensitive and/or legally privileged. It is intended solely for the person(s) to whom it is addressed. If you are not a named recipient, you are on notice of its status. Please notify the sender immediately by reply e-mail and then delete this message from your system. You must not disclose it to any other person, copy or distribute it or use it for any purpose. > > Views expressed in this email are not necessarily those of Sellafield Ltd. > > Sellafield Ltd, a company owned by Nuclear Management Partners Ltd, is registered in England and Wales, Company number 1002607. The registered office is situated at Booths Park, Chelford Road, Knutsford, Cheshire WA16 8QZ." > > <c.txt>_______________________________________________ > 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] Running configure on Suse Enterprise 10
by Jeff Hammond 05 Jul '13

05 Jul '13

Disable Fortran if you don't have a Fortran compiler. The options are shown with the help invocation. Jeff Sent from my iPhone On Jul 5, 2013, at 6:15 AM, "anton.s.murfin(a)sellafieldsites.com" < anton.s.murfin(a)sellafieldsites.com> wrote: Dear MPICH developers, I have downloaded and extracted mpich-3.0.4.tar.gz to /home/ris3/asm7/mpich-3.0.4/ I then created a build directory as /home/ris3/asm7/mpich-3.0.4_build/ I went back to /home/ris3/asm7/mpich-3.0.4/ and executed the .configure script detailed below *./configure --prefix = /home/ris3/asm7/mpich-3.0.4_build 2>&1 | tee c.txt* (I confirmed I was using bash shell, using echo $SHELL) The output from this c.txt is attached and there is only one error right at the bottom although there are some warnings. No Makefile is created so the next step make doesn't work. I just wanted to see if you can see what the issue is. Many thanks Anton "The information contained in this email may be commercially sensitive and/or legally privileged. It is intended solely for the person(s) to whom it is addressed. If you are not a named recipient, you are on notice of its status. Please notify the sender immediately by reply e-mail and then delete this message from your system. You must not disclose it to any other person, copy or distribute it or use it for any purpose. Views expressed in this email are not necessarily those of Sellafield Ltd. Sellafield Ltd, a company owned by Nuclear Management Partners Ltd, is registered in England and Wales, Company number 1002607. The registered office is situated at Booths Park, Chelford Road, Knutsford, Cheshire WA16 8QZ." <c.txt> _______________________________________________ discuss mailing list discuss(a)mpich.org To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

1 0