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

Question on MPICH rpm for redat 5.6
by McLeish, Allan 22 Jul '13

22 Jul '13

Hello, I'm currently having some issues installing MPICH on to our RHEL Linux system due to missing dependencies. I want to note that we use Redhat 5.6 and we don't use YUM so our install is through the use of rpm's. I initially tried installing the latest version of MPICH "mpich2-1.5-3.fc20.x86_64.rpm" but multiple dependencies where required so I tried going to an earliest version of MPICH available in the downloads "mpich2-1.0.8-3.fc11.x86_64.rpm" but this also was missing multiple dependencies. MPICH was a 3rd party app required for another program that we were requiring so I contacted there support and they suggested we used MPICH 1.4.1 as it worked the best with there program. When I try to run the RPM for 1.4.1 is asks for multiple dependencies and each of those dependencies have dependencies, is there a tar with all the required dependancies for redhat 5.6 or are there any recommendations on how to go about this? rpm -ivh mpich2-1.4.1p1-9.fc18.x86_64.rpm error: Failed dependencies: /usr/bin/bash is needed by mpich2-1.4.1p1-9.fc18.x86_64 environment-modules is needed by mpich2-1.4.1p1-9.fc18.x86_64 libc.so.6(GLIBC_2.14)(64bit) is needed by mpich2-1.4.1p1-9.fc18.x86_64 libc.so.6(GLIBC_2.15)(64bit) is needed by mpich2-1.4.1p1-9.fc18.x86_64 libc.so.6(GLIBC_2.7)(64bit) is needed by mpich2-1.4.1p1-9.fc18.x86_64 libhwloc.so.5()(64bit) is needed by mpich2-1.4.1p1-9.fc18.x86_64 rpmlib(FileDigests) <= 4.6.0-1 is needed by mpich2-1.4.1p1-9.fc18.x86_64 rpmlib(PayloadIsXz) <= 5.2-1 is needed by mpich2-1.4.1p1-9.fc18.x86_64 Many Thanks Allan Maersk Oil North Sea UK Limited, registered in England and Wales No. 03682299. Registered office Maersk House, Braham Street, London E1 8EP. This e-mail and any files transmitted with it are confidential and intended solely for the use of the individual or entity to which they are addressed. If you have received this e-mail in error please notify the system manager at hotline(a)maerskoil.com.

1 0

Re: [mpich-discuss] MPI_Win_free failed
by Pavan Balaji 19 Jul '13

19 Jul '13

On 07/19/2013 05:26 PM, Sufeng Niu wrote: > MPI_Win_create(image_buff, 2*image_info->image_size * sizeof(uint16), > sizeof(uint16), MPI_INFO_NULL, *MPI_FOR_WORLD, win); // MPI_FOR_WORLD > is custom defined comm Check if *MPI_FOR_WORLD is valid (or COMM_NULL). > I just free it as: MPI_Win_free(&win); Check if win is valid (or WIN_NULL). In general, these are errors a debugger can catch. Please try that. -- Pavan -- Pavan Balaji http://www.mcs.anl.gov/~balaji

1 0

MPI_Win_free failed
by Sufeng Niu 19 Jul '13

19 Jul '13

Hello, I am doing RMA operation, but my MPI_Win_free got the following messages: Fatal error in MPI_Win_free: Invalid MPI_Win, error stack: MPI_Win_free(134): MPI_Win_free(win=0x7fffb1893028) failed MPI_Win_free(66).: Invalid MPI_Win =================================================================================== = BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES = EXIT CODE: 1 = CLEANING UP REMAINING PROCESSES = YOU CAN IGNORE THE BELOW CLEANUP MESSAGES =================================================================================== my rma operations shows below: MPI_Win_create(image_buff, 2*image_info->image_size * sizeof(uint16), sizeof(uint16), MPI_INFO_NULL, *MPI_FOR_WORLD, win); // MPI_FOR_WORLD is custom defined comm I just free it as: MPI_Win_free(&win); Do any one knows why? another weird thing is at the end of the program, I wrote: MPI_Comm_free(&MPI_FOR_WORLD); it also gave error: Fatal error in PMPI_Comm_free: Invalid communicator, error stack: PMPI_Comm_free(141): MPI_Comm_free(comm=0x7fff44a1535c) failed PMPI_Comm_free(91).: Null communicator Any suggestions are appreciate Thank you very much! -- Best Regards, Sufeng Niu ECASP lab, ECE department, Illinois Institute of Technology Tel: 312-731-7219

1 0

create group and communicator failed for RMA
by Sufeng Niu 18 Jul '13

18 Jul '13

Hello, in my mpi program, I create 3 groups and communicators: forward_rank = (int *)malloc((numprocs-1) * sizeof(int)); backward_rank = (int *)malloc((numprocs-1) * sizeof(int)); comp_rank = (int *)malloc((numprocs-2) * sizeof(int)); for (i=0; i<numprocs-1; i++){ backward_rank[i] = i+1; forward_rank[i] = i; } for (i=0; i<numprocs-2; i++){ comp_rank[i] = i+1; } MPI_Group global_group, fgroup, bgroup, cgroup; // forward group, backward group, computation group MPI_Comm MPI_FOR_WORLD, MPI_BACK_WORLD, MPI_COMP_WORLD; // forward, backward, computation communicator MPI_Comm_group(MPI_COMM_WORLD, &global_group); MPI_Group_incl(global_group, numprocs-1, forward_rank, &bgroup); MPI_Comm_create(MPI_COMM_WORLD, bgroup, &MPI_BACK_WORLD); MPI_Group_incl(global_group, numprocs-1, backward_rank, &fgroup); MPI_Comm_create(MPI_COMM_WORLD, fgroup, &MPI_FOR_WORLD); MPI_Group_incl(global_group, numprocs-2, comp_rank, &cgroup); MPI_Comm_create(MPI_COMM_WORLD, cgroup, &MPI_COMP_WORLD); And I used them for my RMA operation So in my main function transfer new communicator to another c file as: udp_sub(win, &MPI_FOR_WORLD); image_sub(rank, win, strip_buff, &MPI_FOR_WORLD, &MPI_BACK_WORLD, &MPI_COMP_WORLD); in udp_sub and image_sub function: I called: MPI_Win_create(image_buff, 2*image_info->image_size*sizeof(uint16), sizeof(uint16), MPI_INFO_NULL, *MPI_FOR_WORLD, win); but I got the error when I execute them: MPI_Win_create(189): MPI_Win_create(base=0x7fe50c06c010, size=9216000, disp_unit=2, MPI_INFO_NULL, MPI_COMM_NULL, win=0x1a5e940) failed MPI_Win_create(116): Null communicator I already transfer the communicator address to it, but I don't understand why communicator is null. or I did in a wrong way? Any suggestions are appreciate. Thank you! -- Best Regards, Sufeng Niu ECASP lab, ECE department, Illinois Institute of Technology Tel: 312-731-7219

1 0

Re: [mpich-discuss] Question on MPI_Pack/Unpack
by Matthieu Dorier 18 Jul '13

18 Jul '13

Ok thanks for all your answers. I'm going to keep the send/recv code, just to be consistent with the standard. Matthieu ----- Mail original ----- > De: "Pavan Balaji" <balaji(a)mcs.anl.gov> > À: discuss(a)mpich.org > Envoyé: Jeudi 18 Juillet 2013 10:52:56 > Objet: Re: [mpich-discuss] Question on MPI_Pack/Unpack > > > MPI is allowed to do any form of encoding inside MPI_Pack. You can't > rely on what the data format would be. > > We don't do that in MPICH (or any other MPI implementation, AFAICT), > but > the standard allows it. > > -- Pavan > > On 07/18/2013 10:29 AM, Dries Kimpe wrote: > > There's another function for MPI_Pack. > > > > It deals with encoding for non-homogeneous systems. > > > > For example, it might convert between different representations of > > double > > floating point numbers. > > > > Normally, this happens automatically (assuming a proper datatype > > was > > given) in MPI_Send/MPI_Recv. > > > > MPI_Pack does this while it encodes into a contiguous buffer, > > which is why you're supposed to use the MPI_PACKED datatype when > > sending > > packed data, so the implementation knows not to convert the data. > > > > Dries > > > > * Antonio J. Peña <apenya(a)mcs.anl.gov> [2013-07-18 10:16:04]: > > > > > >> Hi Matthieu, > > > >> These functions only copy all the data of the datatype into a > >> contiguous > >> memory region. They don't do any further action. I'd say these are > >> a more > >> natural and efficient solution for your problem than the self > >> send-receive > >> approach. Note that the send is internally performing the pack > >> before > >> sending, so you'd effectively do the same but avoiding the > >> send/receive > >> step. > > > >> Antonio > > > > > >> On Thursday, July 18, 2013 04:35:28 PM Matthieu Dorier wrote: > > > > > >> Hi, > > > > > >> I'd like some clarification on MPI_Pack/Unpack: the standard only > >> says that > >> "some communication libraries provide pack/unpack functions for > >> sending > >> non-contiguous data", but it's not clear to me what > >> MPI_Pack/Unpack are > >> supposed to do. For example, if I have a MPI_Datatype representing > >> a > >> structure with holes between fields, is MPI_Pack just supposed to > >> put the > >> elements of the structure close to each other so that there is no > >> more > >> holes, or can it do something different (compressing, adding > >> metadata, > >> etc.)? > > > > > >> The reason I'm asking this is because I'm designing a code to > >> extract > >> chunks of multi-dimentional arrays (basically a big array has > >> ghost zones > >> and I want to extract the non-ghost data from this big array into > >> a smaller > >> array, contiguous in memory). This extraction is local and does > >> not involve > >> communication so right now after using MPI_Type_create_subarray to > >> create a datatype adapted to my need, I post an MPI_Irecv with a > >> contiguous array of MPI_CHAR as type, and then an MPI_Send with > >> one > >> element of the type I just created. I was wondering if MPI_Pack > >> would be > >> equivalent (and maybe more efficient)? > > > > > >> Thanks > > > > > > > >> Matthieu Dorier > > > >> PhD student at ENS Cachan Brittany and IRISA > > > >> http://people.irisa.fr/Matthieu.Dorier[1] > > > > > >> -------- > >> [1] http://people.irisa.fr/Matthieu.Dorier > > > >> _______________________________________________ > >> 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] Question on MPI_Pack/Unpack
by Pavan Balaji 18 Jul '13

18 Jul '13

MPI is allowed to do any form of encoding inside MPI_Pack. You can't rely on what the data format would be. We don't do that in MPICH (or any other MPI implementation, AFAICT), but the standard allows it. -- Pavan On 07/18/2013 10:29 AM, Dries Kimpe wrote: > There's another function for MPI_Pack. > > It deals with encoding for non-homogeneous systems. > > For example, it might convert between different representations of double > floating point numbers. > > Normally, this happens automatically (assuming a proper datatype was > given) in MPI_Send/MPI_Recv. > > MPI_Pack does this while it encodes into a contiguous buffer, > which is why you're supposed to use the MPI_PACKED datatype when sending > packed data, so the implementation knows not to convert the data. > > Dries > > * Antonio J. Peï¿½a <apenya(a)mcs.anl.gov> [2013-07-18 10:16:04]: > > >> Hi Matthieu, > >> These functions only copy all the data of the datatype into a contiguous >> memory region. They don't do any further action. I'd say these are a more >> natural and efficient solution for your problem than the self send-receive >> approach. Note that the send is internally performing the pack before >> sending, so you'd effectively do the same but avoiding the send/receive >> step. > >> Antonio > > >> On Thursday, July 18, 2013 04:35:28 PM Matthieu Dorier wrote: > > >> Hi, > > >> I'd like some clarification on MPI_Pack/Unpack: the standard only says that >> "some communication libraries provide pack/unpack functions for sending >> non-contiguous data", but it's not clear to me what MPI_Pack/Unpack are >> supposed to do. For example, if I have a MPI_Datatype representing a >> structure with holes between fields, is MPI_Pack just supposed to put the >> elements of the structure close to each other so that there is no more >> holes, or can it do something different (compressing, adding metadata, >> etc.)? > > >> The reason I'm asking this is because I'm designing a code to extract >> chunks of multi-dimentional arrays (basically a big array has ghost zones >> and I want to extract the non-ghost data from this big array into a smaller >> array, contiguous in memory). This extraction is local and does not involve >> communication so right now after using MPI_Type_create_subarray to >> create a datatype adapted to my need, I post an MPI_Irecv with a >> contiguous array of MPI_CHAR as type, and then an MPI_Send with one >> element of the type I just created. I was wondering if MPI_Pack would be >> equivalent (and maybe more efficient)? > > >> Thanks > > > >> Matthieu Dorier > >> PhD student at ENS Cachan Brittany and IRISA > >> http://people.irisa.fr/Matthieu.Dorier[1] > > >> -------- >> [1] http://people.irisa.fr/Matthieu.Dorier > >> _______________________________________________ >> 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

1 0

Re: [mpich-discuss] Question on MPI_Pack/Unpack
by Dries Kimpe 18 Jul '13

18 Jul '13

There's another function for MPI_Pack. It deals with encoding for non-homogeneous systems. For example, it might convert between different representations of double floating point numbers. Normally, this happens automatically (assuming a proper datatype was given) in MPI_Send/MPI_Recv. MPI_Pack does this while it encodes into a contiguous buffer, which is why you're supposed to use the MPI_PACKED datatype when sending packed data, so the implementation knows not to convert the data. Dries * Antonio J. Peña <apenya(a)mcs.anl.gov> [2013-07-18 10:16:04]: > Hi Matthieu, > These functions only copy all the data of the datatype into a contiguous > memory region. They don't do any further action. I'd say these are a more > natural and efficient solution for your problem than the self send-receive > approach. Note that the send is internally performing the pack before > sending, so you'd effectively do the same but avoiding the send/receive > step. > Antonio > On Thursday, July 18, 2013 04:35:28 PM Matthieu Dorier wrote: > Hi, > I'd like some clarification on MPI_Pack/Unpack: the standard only says that > "some communication libraries provide pack/unpack functions for sending > non-contiguous data", but it's not clear to me what MPI_Pack/Unpack are > supposed to do. For example, if I have a MPI_Datatype representing a > structure with holes between fields, is MPI_Pack just supposed to put the > elements of the structure close to each other so that there is no more > holes, or can it do something different (compressing, adding metadata, > etc.)? > The reason I'm asking this is because I'm designing a code to extract > chunks of multi-dimentional arrays (basically a big array has ghost zones > and I want to extract the non-ghost data from this big array into a smaller > array, contiguous in memory). This extraction is local and does not involve > communication so right now after using MPI_Type_create_subarray to > create a datatype adapted to my need, I post an MPI_Irecv with a > contiguous array of MPI_CHAR as type, and then an MPI_Send with one > element of the type I just created. I was wondering if MPI_Pack would be > equivalent (and maybe more efficient)? > Thanks > Matthieu Dorier > PhD student at ENS Cachan Brittany and IRISA > http://people.irisa.fr/Matthieu.Dorier[1] > -------- > [1] http://people.irisa.fr/Matthieu.Dorier > _______________________________________________ > 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] Question on MPI_Pack/Unpack
by Antonio J. Peña 18 Jul '13

18 Jul '13

Hi Matthieu, These functions only copy all the data of the datatype into a contiguous memory region. They don't do any further action. I'd say these are a more natural and efficient solution for your problem than the self send-receive approach. Note that the send is internally performing the pack before sending, so you'd effectively do the same but avoiding the send/receive step. Antonio On Thursday, July 18, 2013 04:35:28 PM Matthieu Dorier wrote: Hi, I'd like some clarification on MPI_Pack/Unpack: the standard only says that "some communication libraries provide pack/unpack functions for sending non-contiguous data", but it's not clear to me what MPI_Pack/Unpack are supposed to do. For example, if I have a MPI_Datatype representing a structure with holes between fields, is MPI_Pack just supposed to put the elements of the structure close to each other so that there is no more holes, or can it do something different (compressing, adding metadata, etc.)? The reason I'm asking this is because I'm designing a code to extract chunks of multi-dimentional arrays (basically a big array has ghost zones and I want to extract the non-ghost data from this big array into a smaller array, contiguous in memory). This extraction is local and does not involve communication so right now after using MPI_Type_create_subarray to create a datatype adapted to my need, I post an MPI_Irecv with a contiguous array of MPI_CHAR as type, and then an MPI_Send with one element of the type I just created. I was wondering if MPI_Pack would be equivalent (and maybe more efficient)? Thanks Matthieu Dorier PhD student at ENS Cachan Brittany and IRISA http://people.irisa.fr/Matthieu.Dorier[1] -------- [1] http://people.irisa.fr/Matthieu.Dorier

1 0

Question on MPI_Pack/Unpack
by Matthieu Dorier 18 Jul '13

18 Jul '13

Hi, I'd like some clarification on MPI_Pack/Unpack: the standard only says that "some communication libraries provide pack/unpack functions for sending non-contiguous data", but it's not clear to me what MPI_Pack/Unpack are supposed to do. For example, if I have a MPI_Datatype representing a structure with holes between fields, is MPI_Pack just supposed to put the elements of the structure close to each other so that there is no more holes, or can it do something different (compressing, adding metadata, etc.)? The reason I'm asking this is because I'm designing a code to extract chunks of multi-dimentional arrays (basically a big array has ghost zones and I want to extract the non-ghost data from this big array into a smaller array, contiguous in memory). This extraction is local and does not involve communication so right now after using MPI_Type_create_subarray to create a datatype adapted to my need, I post an MPI_Irecv with a contiguous array of MPI_CHAR as type, and then an MPI_Send with one element of the type I just created. I was wondering if MPI_Pack would be equivalent (and maybe more efficient)? Thanks Matthieu Dorier PhD student at ENS Cachan Brittany and IRISA http://people.irisa.fr/Matthieu.Dorier

1 0

Re: [mpich-discuss] Errors occurred when compiling fpi.f with g95
by Wesley Bland 16 Jul '13

16 Jul '13

http://wiki.mpich.org/mpich/index.php/Frequently_Asked_Questions#Q:_Why_can… On Jul 16, 2013, at 8:44 AM, Wen-Chung Liu <ken80135(a)gmail.com> wrote: > Dear all： > > I've simply try to compile the example file fpi.f with g95 compiler. > > My OS is Windows 7 64 bits. Therefore, I use 1.4.1p1 x86_64 version of MPICH2. > > And I got errors listed below: > > C:\Program Files\MPICH2\examples>g95 -I"C:/Program Files/MPICH2/include" -L"C:/Program Files/MPICH2/lib" fpi.f -lfmpich2g > > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.text+0x1e): undefined reference to `mpi_init__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.text+0x42): undefined reference to `mpi_comm_rank__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.text+0x66): undefined reference to `mpi_comm_size__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.text+0x248): undefined reference to `mpi_bcast__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.text+0x3a9): undefined reference to `mpi_reduce__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.text+0x465): undefined reference to `mpi_finalize__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.data+0x0): undefined reference to `mpi_comm_dup_fn__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.data+0x4): undefined reference to `mpi_comm_null_delete_fn__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.data+0x8): undefined reference to `mpi_conversion_fn_null__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.data+0xc): undefined reference to `mpi_dup_fn__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.data+0x10): undefined reference to `mpi_comm_null_copy_fn__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.data+0x14): undefined reference to `mpi_null_copy_fn__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.data+0x18): undefined reference to `mpi_null_delete_fn__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.data+0x1c): undefined reference to `mpi_type_dup_fn__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.data+0x20): undefined reference to `mpi_type_null_delete_fn__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.data+0x24): undefined reference to `mpi_win_dup_fn__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.data+0x28): undefined reference to `mpi_win_null_copy_fn__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.data+0x2c): undefined reference to `mpi_win_null_delete_fn__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.data+0x30): undefined reference to `mpi_wtime__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.data+0x34): undefined reference to `mpi_wtick__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.data+0x38): undefined reference to `pmpi_wtime__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.data+0x3c): undefined reference to `pmpi_wtick__' > C:\Users\Troy\AppData\Local\Temp/ccUROg3z.o:fpi.f:(.data+0x40): undefined reference to `mpi_type_null_copy_fn__' > > I've try to run mpi.exe with mpiexec. It works well. I've also reinstall MPICH2, but get same errors. > > What steps did I lost? Could anyone give me some advises? > > best > regards, > > Troy > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss

1 0