- mpi-comments - mailman3.cels.anl.gov

Re: [Mpi-comments] non-blocking communication deadlock possible?
by mustafa mohamed 18 Dec '14

18 Dec '14

Hello there ..i already do lab for mpich2 and connect all computers i have and i do cpi program and its work correctly .. but when i execute my program its give a message : the process stop working because one process exiting without calling MPI_Init .. what i can do to make it works.. ?? On Wednesday, December 10, 2014 4:33 PM, Rolf Rabenseifner <rabenseifner(a)hlrs.de> wrote: Dear Mr. Hehn, as long as you start your example with exactly two MPI processes, your example is guaranteed to run without deadlock based on the wording on nonblocking progress MPI-3.0 page 56, lines 30-35. Therefore, it is still deadlock-free if you Substitute your > MPI_Waitall(2,r,s); by MPI_Wait(&r[0],s[0]); MPI_Wait(&r[1],s[1]); Thank you for your request on whether the MPI standard is completely describing the nonblocking functionality. Best regards Rolf ----- Original Message ----- > From: "Andreas Hehn" <hehn(a)phys.ethz.ch> > To: mpi-comments(a)lists.mpi-forum.org > Sent: Wednesday, December 10, 2014 12:15:28 PM > Subject: [Mpi-comments] non-blocking communication deadlock possible? > > Dear MPI comments readers, > > I am not 100% sure if the following piece of code could cause a deadlock > in a standard-compliant MPI implementation or not. > > #include <mpi.h> > > int main(int argc, char** argv) { > MPI_Status s[2]; > int num; > > MPI_Init(&argc, &argv); > MPI_Comm_rank(MPI_COMM_WORLD,&num); > > double ds=5.4; // to send > double dr; // to receive > int tag=99; > > MPI_Request r[2]; > if(num==0) { > MPI_Isend(&ds,1,MPI_DOUBLE,1,tag,MPI_COMM_WORLD,&r[0]); > MPI_Irecv (&dr,1,MPI_DOUBLE,1,tag,MPI_COMM_WORLD,&r[1]); > } > else { > MPI_Isend(&ds,1,MPI_DOUBLE,0,tag,MPI_COMM_WORLD,&r[0]); > MPI_Irecv (&dr,1,MPI_DOUBLE,0,tag,MPI_COMM_WORLD,&r[1]); > } > > MPI_Waitall(2,r,s); > > MPI_Finalize(); > return 0; > } > > It is not entirely clear to me if the order of MPI_Isend and MPI_Irecv > matters (as it does for blocking communication). Both function calls > should return immediately according to section 3.7 of the standard - > even if MPI_Isend uses synchronous send. > So at this point there shouldn't be any problem. > However, will MPI_Waitall() return for any order of the filed requests? > The standard says it is equivalent to MPI_Wait() for the number of > requests in "some arbitrary order" (line 28 of page 59). > I suppose this means MPI_Waitall() is not supposed to mind the order of > the requests in any way. Therefore it could not deadlock. > Is this correct and is the line of argument sufficient to exclude the > possibility of a deadlock? > > Best regards, > > Andreas Hehn > _______________________________________________ > mpi-comments mailing list > mpi-comments(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-comments > -- Dr. Rolf Rabenseifner . . . . . . . . . .. email rabenseifner(a)hlrs.de High Performance Computing Center (HLRS) . phone ++49(0)711/685-65530 University of Stuttgart . . . . . . . . .. fax ++49(0)711 / 685-65832 Head of Dpmt Parallel Computing . . . www.hlrs.de/people/rabenseifner Nobelstr. 19, D-70550 Stuttgart, Germany . . . . (Office: Room 1.307) _______________________________________________ mpi-comments mailing list mpi-comments(a)lists.mpi-forum.org http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-comments

1 0

Re: [Mpi-comments] non-blocking communication deadlock possible?
by Rolf Rabenseifner 10 Dec '14

10 Dec '14

Dear Mr. Hehn, as long as you start your example with exactly two MPI processes, your example is guaranteed to run without deadlock based on the wording on nonblocking progress MPI-3.0 page 56, lines 30-35. Therefore, it is still deadlock-free if you Substitute your > MPI_Waitall(2,r,s); by MPI_Wait(&r[0],s[0]); MPI_Wait(&r[1],s[1]); Thank you for your request on whether the MPI standard is completely describing the nonblocking functionality. Best regards Rolf ----- Original Message ----- > From: "Andreas Hehn" <hehn(a)phys.ethz.ch> > To: mpi-comments(a)lists.mpi-forum.org > Sent: Wednesday, December 10, 2014 12:15:28 PM > Subject: [Mpi-comments] non-blocking communication deadlock possible? > > Dear MPI comments readers, > > I am not 100% sure if the following piece of code could cause a deadlock > in a standard-compliant MPI implementation or not. > > #include <mpi.h> > > int main(int argc, char** argv) { > MPI_Status s[2]; > int num; > > MPI_Init(&argc, &argv); > MPI_Comm_rank(MPI_COMM_WORLD,&num); > > double ds=5.4; // to send > double dr; // to receive > int tag=99; > > MPI_Request r[2]; > if(num==0) { > MPI_Isend(&ds,1,MPI_DOUBLE,1,tag,MPI_COMM_WORLD,&r[0]); > MPI_Irecv (&dr,1,MPI_DOUBLE,1,tag,MPI_COMM_WORLD,&r[1]); > } > else { > MPI_Isend(&ds,1,MPI_DOUBLE,0,tag,MPI_COMM_WORLD,&r[0]); > MPI_Irecv (&dr,1,MPI_DOUBLE,0,tag,MPI_COMM_WORLD,&r[1]); > } > > MPI_Waitall(2,r,s); > > MPI_Finalize(); > return 0; > } > > It is not entirely clear to me if the order of MPI_Isend and MPI_Irecv > matters (as it does for blocking communication). Both function calls > should return immediately according to section 3.7 of the standard - > even if MPI_Isend uses synchronous send. > So at this point there shouldn't be any problem. > However, will MPI_Waitall() return for any order of the filed requests? > The standard says it is equivalent to MPI_Wait() for the number of > requests in "some arbitrary order" (line 28 of page 59). > I suppose this means MPI_Waitall() is not supposed to mind the order of > the requests in any way. Therefore it could not deadlock. > Is this correct and is the line of argument sufficient to exclude the > possibility of a deadlock? > > Best regards, > > Andreas Hehn > _______________________________________________ > mpi-comments mailing list > mpi-comments(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-comments > -- Dr. Rolf Rabenseifner . . . . . . . . . .. email rabenseifner(a)hlrs.de High Performance Computing Center (HLRS) . phone ++49(0)711/685-65530 University of Stuttgart . . . . . . . . .. fax ++49(0)711 / 685-65832 Head of Dpmt Parallel Computing . . . www.hlrs.de/people/rabenseifner Nobelstr. 19, D-70550 Stuttgart, Germany . . . . (Office: Room 1.307)

1 0

non-blocking communication deadlock possible?
by Andreas Hehn 10 Dec '14

10 Dec '14

Dear MPI comments readers, I am not 100% sure if the following piece of code could cause a deadlock in a standard-compliant MPI implementation or not. #include <mpi.h> int main(int argc, char** argv) { MPI_Status s[2]; int num; MPI_Init(&argc, &argv); MPI_Comm_rank(MPI_COMM_WORLD,&num); double ds=5.4; // to send double dr; // to receive int tag=99; MPI_Request r[2]; if(num==0) { MPI_Isend(&ds,1,MPI_DOUBLE,1,tag,MPI_COMM_WORLD,&r[0]); MPI_Irecv (&dr,1,MPI_DOUBLE,1,tag,MPI_COMM_WORLD,&r[1]); } else { MPI_Isend(&ds,1,MPI_DOUBLE,0,tag,MPI_COMM_WORLD,&r[0]); MPI_Irecv (&dr,1,MPI_DOUBLE,0,tag,MPI_COMM_WORLD,&r[1]); } MPI_Waitall(2,r,s); MPI_Finalize(); return 0; } It is not entirely clear to me if the order of MPI_Isend and MPI_Irecv matters (as it does for blocking communication). Both function calls should return immediately according to section 3.7 of the standard - even if MPI_Isend uses synchronous send. So at this point there shouldn't be any problem. However, will MPI_Waitall() return for any order of the filed requests? The standard says it is equivalent to MPI_Wait() for the number of requests in "some arbitrary order" (line 28 of page 59). I suppose this means MPI_Waitall() is not supposed to mind the order of the requests in any way. Therefore it could not deadlock. Is this correct and is the line of argument sufficient to exclude the possibility of a deadlock? Best regards, Andreas Hehn

1 0

Re: [Mpi-comments] Error in MPI_Neighbor_alltoallw example?
by Timothy Brown 05 Dec '14

05 Dec '14

Ah, perfect thanks! I didn't know about the trac/ticket system. Tim > On Dec 5, 2014, at 9:04 AM, Rajeev Thakur <thakur(a)mcs.anl.gov> wrote: > > Yes, see https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/420 > > Rajeev > > > On Dec 5, 2014, at 9:41 AM, Timothy Brown <Timothy.Brown-1(a)Colorado.EDU> wrote: > >> Hello. >> >> I was trying to use MPI_Neighbor_alltoallw() and was following the example on page 333 of the MPI-3.0 standard. However I believe the send and receive counts are incorrect. Lines 21 to 25: >> >> DO i=1,num_neigh >> sndcounts(i) = 100 >> rcvcounts(i) = 100 >> rcvtypes(i) = sndtypes(i) >> END DO >> >> Since the send/receive counts for 1 and 2 are vectors types of length 100, we should be sending/receiving only 1 type/element. As follows: >> >> sndcounts(1) = 1 >> sndcounts(2) = 1 >> sndcounts(3) = 100 >> sndcounts(4) = 100 >> >> rcvcounts = sndcounts >> >> Indeed if I compile the example as per the standard, I get data corruption. >> >> I couldn't find this in the errata published on May 14, 2014. However if it is a known mistake please ignore my comment. >> >> Regards >> Timothy Brown >> _______________________________________________ >> mpi-comments mailing list >> mpi-comments(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-comments > > _______________________________________________ > mpi-comments mailing list > mpi-comments(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-comments

1 0

Re: [Mpi-comments] Error in MPI_Neighbor_alltoallw example?
by Rajeev Thakur 05 Dec '14

05 Dec '14

Yes, see https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/420 Rajeev On Dec 5, 2014, at 9:41 AM, Timothy Brown <Timothy.Brown-1(a)Colorado.EDU> wrote: > Hello. > > I was trying to use MPI_Neighbor_alltoallw() and was following the example on page 333 of the MPI-3.0 standard. However I believe the send and receive counts are incorrect. Lines 21 to 25: > > DO i=1,num_neigh > sndcounts(i) = 100 > rcvcounts(i) = 100 > rcvtypes(i) = sndtypes(i) > END DO > > Since the send/receive counts for 1 and 2 are vectors types of length 100, we should be sending/receiving only 1 type/element. As follows: > > sndcounts(1) = 1 > sndcounts(2) = 1 > sndcounts(3) = 100 > sndcounts(4) = 100 > > rcvcounts = sndcounts > > Indeed if I compile the example as per the standard, I get data corruption. > > I couldn't find this in the errata published on May 14, 2014. However if it is a known mistake please ignore my comment. > > Regards > Timothy Brown > _______________________________________________ > mpi-comments mailing list > mpi-comments(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-comments

1 0

Error in MPI_Neighbor_alltoallw example?
by Timothy Brown 05 Dec '14

05 Dec '14

Hello. I was trying to use MPI_Neighbor_alltoallw() and was following the example on page 333 of the MPI-3.0 standard. However I believe the send and receive counts are incorrect. Lines 21 to 25: DO i=1,num_neigh sndcounts(i) = 100 rcvcounts(i) = 100 rcvtypes(i) = sndtypes(i) END DO Since the send/receive counts for 1 and 2 are vectors types of length 100, we should be sending/receiving only 1 type/element. As follows: sndcounts(1) = 1 sndcounts(2) = 1 sndcounts(3) = 100 sndcounts(4) = 100 rcvcounts = sndcounts Indeed if I compile the example as per the standard, I get data corruption. I couldn't find this in the errata published on May 14, 2014. However if it is a known mistake please ignore my comment. Regards Timothy Brown

1 0

Re: [Mpi-comments] Status of the Async-proc-mgmt proposal
by William Gropp 10 Oct '14

10 Oct '14

That effort is inactive at this time. One reason is that you can achieve much of the same behavior by creating a separate thread for each blocking call that you want to be asynchronous, and most environments that support MPI dynamic processes also support threads. Bill On Oct 10, 2014, at 5:12 AM, Maciej Szpindler <m.szpindler(a)icm.edu.pl> wrote: > Dear MPI-Forum, > > I would like to ask about the status of the nonblocking process > creation mechanism (aka Async-proc-mgmt) which I have found on > the MPI Forum wiki pages. I am trying to implement dynamic resource > allocation for my research work and I was thinking about nonblocking > MPI_Comm_spawn counterpart. I have googled for "MPI_Icomm_spawn" and > found the proposal details. > > I am curious if any further development exist. Is there any reference > implementation that might be usable? What does it mean in practice > that this proposal is dormant, would it be included in any predictable > future? > > I would appreciate any hints or comments on the nonblocking process > management with MPI. > > > Best Regards, > Maciej Szpindler > > ICM, University of Warsaw > _______________________________________________ > mpi-comments mailing list > mpi-comments(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-comments

1 0

Status of the Async-proc-mgmt proposal
by Maciej Szpindler 10 Oct '14

10 Oct '14

Dear MPI-Forum, I would like to ask about the status of the nonblocking process creation mechanism (aka Async-proc-mgmt) which I have found on the MPI Forum wiki pages. I am trying to implement dynamic resource allocation for my research work and I was thinking about nonblocking MPI_Comm_spawn counterpart. I have googled for "MPI_Icomm_spawn" and found the proposal details. I am curious if any further development exist. Is there any reference implementation that might be usable? What does it mean in practice that this proposal is dormant, would it be included in any predictable future? I would appreciate any hints or comments on the nonblocking process management with MPI. Best Regards, Maciej Szpindler ICM, University of Warsaw

1 0

Re: [Mpi-comments] install mpich with hydra
by Martin Schulz 05 Aug '14

05 Aug '14

Hi Chun, This list is intended for comments and feedback on the actual MPI standard, not particular implementations. We can¹t answer questions that are particular to individual MPI implementations. Please direct your question to the MPICH support lists. Thanks, Martin ________________________________________________________________________ Martin Schulz, schulzm(a)llnl.gov, http://scalability.llnl.gov/ CASC @ Lawrence Livermore National Laboratory, Livermore, USA From: Chun Meng <miracle.chun.2010(a)gmail.com> Reply-To: Comments on the MPI Standard <mpi-comments(a)lists.mpi-forum.org> Date: Monday, August 4, 2014 at 6:27 PM To: mpi-comments <mpi-comments(a)lists.mpi-forum.org> Subject: [Mpi-comments] install mpich with hydra Hi All, I have a question about the installation of mpich. Do I really need to install hydra? Someone told me the source code of hydra is already in the mpich package. But The model I'm using can not be successfully run without installing hydra individually. Any help is appreciated. Thanks a lot! Best, Chun

1 0

install mpich with hydra
by Chun Meng 04 Aug '14

04 Aug '14

Hi All, I have a question about the installation of mpich. Do I really need to install hydra? Someone told me the source code of hydra is already in the mpich package. But The model I'm using can not be successfully run without installing hydra individually. Any help is appreciated. Thanks a lot! Best, Chun

1 0