Re: [O-MPI users] Thread-safety and daemon initiation

2005-02-15 Thread George Bosilca
Neil, Open-MPI support all thread models defined by the MPI standard if they are available on the target system. Few years ago I did some work with MPI and OpenMP and locking mechanism of existing MPI implementations was the performance killer. We have work hard to remove this bottleneck on O

Re: [O-MPI users] Questions about Open MPI

2005-05-05 Thread George Bosilca
On May 5, 2005, at 7:58 AM, atarpley wrote: 1) When will the final Open MPI be released (non development)? As soon as everybody is happy with the stability and the features of a version. And like for most of the HPC software, SC05 seems like a reasonable deadline. Meanwhile, a beta vers

Re: [O-MPI users] Collective communications with derived datatypes

2005-08-15 Thread George Bosilca
Joel, I took a look at your code and found the error. Basically, it's just a datatype problem. The datatype as described in your program does not correspond to the one you expect to see in practice. Actually you forget to set the correct extent. Let me show you the problem. Let's suppose

Re: [O-MPI users] How pass an array??

2005-08-22 Thread George Bosilca
Read the section about datatypes in order to figure out how to describe the memory layout that you want to move between processors. george. On Mon, 22 Aug 2005, Raul Mosquera wrote: HI. I just started working with MPI's. I'm been reading some documentation about MPI's but I have not found

Re: [O-MPI users] How pass an structure

2005-08-25 Thread George Bosilca
First of all with your approach you're not sending all the structure. You will just send the num_Rocs and the elements, missing the num_Cols member off the structure. The problem come from the fact that lena[0] is set to 1 not 2. Another problem (which I don't think it can happens with this

Re: [O-MPI users] How pass an structure

2005-08-26 Thread George Bosilca
First of all with your approach you're not sending all the structure. You will just send the num_Rocs and the elements, missing the num_Cols member off the structure. The problem come from the fact that lena[0] is set to 1 not 2. Another problem (which I don't think it can happens with this

Re: [O-MPI users] MIPS64 - The Long-Awaited Config File

2005-09-09 Thread George Bosilca
The configure is complaining about the missing atomic directives for your processor. We have the MIPS atomic calls but not the MIPS64. We just have to add them in the opal/asm/base. I dont have access to a MIPS64 node. Maybe Brian ? Thanks, george. On Fri, 9 Sep 2005, Jonathan Day wrote:

Re: [O-MPI users] Building 1.0rc4 failed on a 5-node Beowulf cluster running RedHat Linux 7.3

2005-10-21 Thread George Bosilca
Ken, Please apply the following patch (from your /home/mighell/pkg/ompi/ openmpi-1.0rc4/ base directory). Index: opal/runtime/opal_init.c === --- opal/runtime/opal_init.c(revision 7831) +++ opal/runtime/opal_init.c(working

Re: [O-MPI users] TCP problems with 1.0rc4

2005-10-31 Thread George Bosilca
Mike, If your nodes have more than one network interface it can happens that we do not select the right one. There is a simple way to insure that this does not happens. Create a directory named .openmpi in your home area. In this directory edit the file mca-params.conf. This file is loade

Re: [O-MPI users] Questions about pmpi_wtick and pmpi_wtime

2005-11-08 Thread George Bosilca
We did discover this problem on MAC OS X yesterday. There was a patch on the trunk but I don't think it get into the stable branch. It will be in the tarball by tomorrow. Sorry about that, george. On Nov 8, 2005, at 1:17 PM, Charles Williams wrote: Hi Brian, Thanks for working on th

Re: [O-MPI users] HPL and TCP

2005-11-14 Thread George Bosilca
Allan, If there are 2 Ethernet cards it's better if you can point to the one you want to use. For that you can modify the .openmpi/mca-params.conf file in your home directory. All of the options can go in this file so you will not have to specify them on the mpirun command every time. I give

Re: [O-MPI users] Minor issue: Failthrough of MCA components.

2005-11-17 Thread George Bosilca
The MX library call exit if the error handler is not set before the initialization. This error get fixed, it will get into the tarball shortly. Meanwhile you can use the btl_base_exclude=mx,gm in order to force them to be skipped. Thanks, george. On Thu, 17 Nov 2005, Troy Telford wrot

Re: [O-MPI users] failure to build openmpi-1.0.1rc5 on Solaris10 Opteron

2005-12-06 Thread George Bosilca
Pierre, The problem seems to come from the fact that we do not detect how to generate the assembly code for our atomic operations. As a result we fall back on the gcc mode for 32 bits architectures. Here is the corresponding output from the configure script: checking if cc supports GCC inline as

Re: [O-MPI users] Performance of all-to-all on Gbit Ethernet

2005-12-19 Thread George Bosilca
Carsten, In the Open MPI source code directory there is a collective component called tuned (ompi/mca/coll/tuned). This component is not enabled by default right now, but usually it give better performances than the basic one. You should give it a try (go inside and remove the .ompi_ignor

Re: [O-MPI users] Performance of all-to-all on Gbit Ethernet

2005-12-20 Thread George Bosilca
On Dec 20, 2005, at 3:19 AM, Carsten Kutzner wrote: I don't see how you deduct that adding barriers increase the congestion ? It increase the latency for the all-to-all but for me When I do an all-to-all a lot of times, I see that the time for a single all-to-all varies a lot. My time meas

Re: [O-MPI users] Mac OS X Open-MPI with Myrinet MX

2006-01-27 Thread George Bosilca
Please remove the lines 109 to 111 from the ompi/mca/btl/mx/ btl_mx_component.c file. They get in by accident, they belong to some other patch from the version 1.0.2 of Open MPI. It was not supposed to get back-ported to the 1.0.1. Sorry about the inconvenience, george. On Jan 27, 200

Re: [O-MPI users] A few benchmarks

2006-02-02 Thread George Bosilca
Glen, Thanks for the spending time benchmarking OpenMPI and for sending us the feedback. We know we have some issues on the 1.0.2 version, more precisely with the collective communications. We just look inside the CMAQ code, and there are a lot of reduce and Allreduce. As it look like the collecti

Re: [O-MPI users] direct openib btl and latency

2006-02-08 Thread George Bosilca
On Feb 8, 2006, at 7:06 PM, Jean-Christophe Hugly wrote: But should I understand from all this that the "direct" mode will never actually work ? It seems that if you need at least two transports, then none of them can be the hardwired unique one, right ? Unless there's a built-in switch bet

Re: [OMPI users] [O-MPI users] problem running Migrate with open-MPI

2006-02-10 Thread George Bosilca
There are 2 things that have to be done in order to be able to run a Open MPI application. First the runtime environment need access to some of the files in the bin directory so you have to add the Open MPI bin directory to your path. And second, as we use shared libraries the OS need to kn

Re: [OMPI users] Bug in OMPI 1.0.1 using MPI_Recv with indexed datatypes

2006-02-10 Thread George Bosilca
Yvan, I'm looking into this one. So far I cannot reproduce it with the current version from the trunk. I will look into the stable versions. Until I figure out what's wrong, can you please use the nightly builds to run your test. Once the problem get fixed it will be included in the 1.0.2

Re: [OMPI users] [O-MPI users] "alltoall" vs "alltoallv"

2006-02-10 Thread George Bosilca
Konstantin, The all2all scheduling works only because we know they will all send the same amount of data, so the communications will take "nearly" the same time. Therefore, we can predict how to schedule the communications to get the best out of the network. But this approach can lead to

Re: [OMPI users] Bug in OMPI 1.0.1 using MPI_Recv with indexed datatypes

2006-02-12 Thread George Bosilca
est stable. Thanks, george. On Fri, 10 Feb 2006, George Bosilca wrote: Yvan, I'm looking into this one. So far I cannot reproduce it with the current version from the trunk. I will look into the stable versions. Until I figure out what's wrong, can you please use the nightly builds to run

Re: [OMPI users] Cannonical ring program and Mac OSX 10.4.4

2006-02-13 Thread George Bosilca
James, I not 100% sure but I think I might know what's wrong. I can reproduce something similar (oddly it does not happens all the time) if I activate my firewall and let all the trafic through (ie. accept all connections). In few words, I think the firewall (even when disabled) introduce some

[OMPI users] EuroMPI/ASIA 2014: CFP

2014-03-21 Thread George Bosilca
Tennessee Program co-chairs: Yutaka Ishikawa, The University of Tokyo Atsushi Hori, Riken AICS Workshop chair: Yuichi Tsujita, Riken AICS Program Committee: Ahmad Afsahi, Queen's University Pavan Balaji, Argonne National Laboratory Siegfried Benkner, University of Vienna Gil Bloch, Mellanox Tech

Re: [OMPI users] Waitall never returns

2014-04-04 Thread George Bosilca
Ross, I’m not familiar with the R implementation you are using, but bear with me and I will explain how you can all Open MPI about the list of all pending requests on a process. Disclosure: This is Open MPI deep voodoo, an extreme way to debug applications that might save you quite some time.

Re: [OMPI users] Problem in Open MPI (v1.8) Performance on 10G Ethernet

2014-04-16 Thread George Bosilca
Muhammad, Our configuration of TCP is tailored for 1Gbs networks, so it’s performance on 10G might be sub-optimal. That being said, the remaining of this email will be speculation as I do not have access to a 10G system to test it. There are two things that I would test to see if I can improve

Re: [OMPI users] Problem in Open MPI (v1.8) Performance on 10G Ethernet

2014-04-21 Thread George Bosilca
helped in achieving expected network bandwidth. Varying send and recv buffer > sizes from 128 KB to 1 MB added just 50 Mbps with maximum bandwidth achieved > on 1 MB buffer size. > Thanks for support. > > > On Thu, Apr 17, 2014 at 6:05 AM, George Bosilca wrote: > Muhammad, &g

[OMPI users] Heteropar 2014 CFP

2014-04-30 Thread George Bosilca
de de Engenharia do Porto, Portugal Olivier Beaumont, INRIA, France Paolo Bientinesi, RWTH Aachen, Germany Cristina Boeres, Universidade Federal Fluminense, Brasil George Bosilca, University of Tennessee, USA Louis-Claude Canon, Université de Franche-Comté, France Alexandre Denis, Inria, France

Re: [OMPI users] MPI_Alltoall with Vector Datatype

2014-05-07 Thread George Bosilca
Spenser, There are several issues with the code you provided. 1. You are using a 1D process grid to create a 2D block cyclic distribution. That’s just not possible. 2. You forgot to take in account the extent of the datatype. By default the extent of a vector type is starting from the first by

Re: [OMPI users] MPI_Alltoall with Vector Datatype

2014-05-07 Thread George Bosilca
Spenser, Do you mind posting your working example here on the mailing list? This might help future users understanding how to correctly use the MPI datatype. Thanks, George. On Wed, May 7, 2014 at 3:16 PM, Spenser Gilliland wrote: > George, > > Thanks for taking the time to respond to my qu

Re: [OMPI users] MPI_Alltoall with Vector Datatype

2014-05-08 Thread George Bosilca
I think the issue is with the way you define the send and receive buffer in the MPI_Alltoall. You have to keep in mind that the all-to-all pattern will overwrite the entire data in the receive buffer. Thus, starting from a relative displacement in the data (in this case matrix[wrank*wrows]), begs f

Re: [OMPI users] MPI_Alltoall with Vector Datatype

2014-05-08 Thread George Bosilca
The segfault indicates that you overwrite outside of the allocated memory (and conflicts with the ptmalloc library). I’m quite certain that you write outside the allocated array … George. On May 8, 2014, at 15:16 , Spenser Gilliland wrote: > George & Mattheiu, > >> The Alltoall should only

Re: [OMPI users] MPI_Alltoall with Vector Datatype

2014-05-08 Thread George Bosilca
Spenser, Here is basically what is happening. On the top left, I depicted the datatype resulting from the vector type. The two arrows point to the lower bound and upper bound (thus the extent) of the datatype. On the top right, the resized datatype, where the ub is now moved 2 elements after th

Re: [OMPI users] ROMIO bug reading darrays

2014-05-08 Thread George Bosilca
I read the MPICH trac ticket you pointed to and your analysis seems pertinent. The impact of my patch for “count = 0” has a similar outcome to yours: removed all references to the datatype if the count was zero, without looking fo the special markers. Let me try to come up with a fix. Thanks,

Re: [OMPI users] MPI_Alltoall with Vector Datatype

2014-05-08 Thread George Bosilca
The alltoall exchanges data from all nodes to all nodes, including the local participant. So every participant will write the same amount of data. George. On Thu, May 8, 2014 at 6:16 PM, Spenser Gilliland wrote: > George, > >> Here is basically what is happening. On the top left, I depicted t

Re: [OMPI users] False positive from valgrind in sec_basic.c

2014-05-22 Thread George Bosilca
This is more subtle that described here. It's a vectorization problem and frankly it should appear on all loop-based string operations and for most compilers (confirmed with gcc, clang and icc). The proposed patch is merely a band-aid ... More info @ https://bugs.launchpad.net/ubuntu/+source/valgr

Re: [OMPI users] OpenMPI Compilation Error

2014-06-05 Thread George Bosilca
Alan, I think we forgot to cleanup after a merge and as a result we have c_destweights and c_sourceweights defined twice. Please try the following patch and let us know if this fixes your issue. Index: ompi/mpi/fortran/mpif-h/dist_graph_create_adjacent_f.c

Re: [OMPI users] OpenMPI Compilation Error

2014-06-05 Thread George Bosilca
e this is a more pervasive problem. > > Let us look at this a bit more... > > > On Jun 5, 2014, at 10:37 AM, George Bosilca wrote: > >> Alan, >> >> I think we forgot to cleanup after a merge and as a result we have >> c_destweights and c_sourceweights defi

Re: [OMPI users] importing to MPI data already in memory from another process

2014-06-27 Thread George Bosilca
The One-Sided Communications from the Chapter 11 of the MPI standard? For processes on the same node you might want to look at MPI_WIN_ALLOCATE_SHARED. George. On Fri, Jun 27, 2014 at 9:53 AM, Brock Palen wrote: > Is there a way to import/map memory from a process (data acquisition) such > th

Re: [OMPI users] importing to MPI data already in memory from another process

2014-06-27 Thread George Bosilca
George. On Fri, Jun 27, 2014 at 10:30 AM, Brock Palen wrote: > But this is within the same MPI "universe" right? > > Brock Palen > www.umich.edu/~brockp > CAEN Advanced Computing > XSEDE Campus Champion > bro...@umich.edu > (734)936-1985 > > > > On Ju

[OMPI users] EuroMPI/Asia 2014 Call for Participation

2014-07-01 Thread George Bosilca
EuroMPI/ASIA 2014 Call for participation EuroMPI/ASIA 2014 in-cooperation status with ACM and SIGHPC in Kyoto, Japan, 9th - 12th September, 2014. The prime annual meeting for researchers, developers, and students in message-passing parallel computing with MPI and related paradigms. Deadline

Re: [OMPI users] mpi_comm_spawn question

2014-07-03 Thread George Bosilca
Why are you using system() the second time ? As you want to spawn an MPI application calling MPI_Call_spawn would make everything simpler. George On Jul 3, 2014 4:34 PM, "Milan Hodoscek" wrote: > > Hi, > > I am trying to run the following setup in fortran without much > success: > > I have an MP

Re: [OMPI users] [OMPI devel] bus error with openmpi-1.8.2rc2 on Solaris 10 Sparc

2014-08-08 Thread George Bosilca
I have an extremely vague recollection about a similar issue in the datatype engine: on the SPARC architecture the 64 bits integers must be aligned on a 64bits boundary or you get a bus error. Takahiro you can confirm this by printing the value of data when signal is raised. George. On Fri, Au

Re: [OMPI users] MPI-I/O issues

2014-08-11 Thread George Bosilca
The patch related to ticket #4597 is zapping only the datatypes where the user explicitly provided a zero count. We can argue about LB and UB, but I have a hard time understanding the rationale of allowing zero count only for LB and UB. If it is required by the standard we can easily support it (t

Re: [OMPI users] MPI-I/O issues

2014-08-11 Thread George Bosilca
On Mon, Aug 11, 2014 at 10:41 AM, Rob Latham wrote: > > > On 08/11/2014 08:54 AM, George Bosilca wrote: > >> The patch related to ticket #4597 is zapping only the datatypes where >> the user explicitly provided a zero count. >> >> We can argue abo

Re: [OMPI users] 答复: Does multiple Irecv means concurrent receiving ?

2014-08-27 Thread George Bosilca
You have a physical constraint, the capacity of your links. If you are over 90% of your network bandwidth, there is little to be improved. George. On Aug 27, 2014, at 0:18, "Zhang,Lei(Ecom)" wrote: >> I'm not sure what you mean by this statement. If you add N asynchronous >> requests and the

Re: [OMPI users] core dump on MPI_Finalize in child process.

2014-09-01 Thread George Bosilca
Based on the MPI standard (MPI 3.0 section 10.5.4 page 399) there is no need to disconnect the child processes from the parent in order to cleanly finalize. From this perspective, the original example is correct, but sub-optimal as the parent processes calling MPI_Finalize might block until all con

Re: [OMPI users] Forcing OpenMPI to use Ethernet interconnect instead of InfiniBand

2014-09-09 Thread George Bosilca
Look at your ifconfig output and select the Ethernet device (instead of the IPoIB one). Traditionally the name lack any fanciness, most distributions using eth0 as a default. George. On Tue, Sep 9, 2014 at 11:24 PM, Muhammad Ansar Javed < muhammad.an...@seecs.edu.pk> wrote: > Hi, > > I am cur

Re: [OMPI users] Forcing OpenMPI to use Ethernet interconnect instead of InfiniBand

2014-09-09 Thread George Bosilca
889 errors:0 dropped:0 overruns:0 frame:0 > TX packets:66889 errors:0 dropped:0 overruns:0 carrier:0 > collisions:0 txqueuelen:0 > RX bytes:19005445 (18.1 MiB) TX bytes:19005445 (18.1 MiB) > > > > > > >> Date: Wed, 10 Sep 2014 0

Re: [OMPI users] Can "timeout" be implemented in OpenMPI?

2014-09-19 Thread George Bosilca
1. It is extremely unlike to have a broken MPI communication pipe. Use a parallel debugger to validate your communication pattern is correct. I would suspect a deadlock due to an incomplete communication pattern more than a broken communication pipe. 2. Nope, you cant set timeouts on MPI calls. Th

Re: [OMPI users] SENDRECV + MPI_TYPE_CREATE_STRUCT

2014-10-03 Thread George Bosilca
Diego, I strongly suggest a careful reading of the Datatype chapter in the MPI 3.0 standard. More precisely, the example 4.18 might be of particular interest in your case, as it explains everything that you need to do in order to obtain a portable datatype, one that works in all cases independe

Re: [OMPI users] what order do I get messages coming to MPI Recv from MPI_ANY_SOURCE?

2014-11-11 Thread George Bosilca
Using MPI_ANY_SOURCE will extract one message from the queue of unexpected messages. The fairness is not guaranteed by the MPI standard, thus it is impossible to predict the order between servers. If you need fairness your second choice is the way to go. George. > On Nov 10, 2014, at 20:14

Re: [OMPI users] collective algorithms

2014-11-18 Thread George Bosilca
Daniel, Many papers have been published about the performance modeling of different collective communications algorithms (and fortunately these models are implementation independent). I can pinpoint you to our research in collective modeling which is the underlying infrastructure behind the decisi

Re: [OMPI users] mpi_wtime implementation

2014-11-19 Thread George Bosilca
Dave, You’re right, we screwed up (some #define not correctly set). I have a patch, I’ll push it asap. George. > On Nov 19, 2014, at 05:19 , Dave Love wrote: > > "Daniels, Marcus G" writes: > >> On Mon, 2014-11-17 at 17:31 +, Dave Love wrote: >>> I discovered from looking at the mpiP

Re: [OMPI users] MPI_Neighbor_alltoallw fails with mpi-1.8.3

2014-11-23 Thread George Bosilca
I would argue this is a typical user level bug. The major difference between the dist_create and dist_create_adjacent is that in the later each process provides its neighbors in an order that is expected (and that match the info provided to the MPI_Neighbor_alltoallw call. When the topology is cre

Re: [OMPI users] mpi_wtime implementation

2014-11-24 Thread George Bosilca
https://github.com/open-mpi/ompi/pull/285 is a potential answer. I would like to hear Dave Goodell comment on this before pushing it upstream. George. On Wed, Nov 19, 2014 at 12:56 PM, George Bosilca wrote: > Dave, > > You’re right, we screwed up (some #define not correctly set).

Re: [OMPI users] MPI_Neighbor_alltoallw fails with mpi-1.8.3

2014-11-25 Thread George Bosilca
> On Nov 25, 2014, at 01:12 , Gilles Gouaillardet > wrote: > > Bottom line, though Open MPI implementation of MPI_Dist_graph_create is not > deterministic, it is compliant with the MPI standard. > /* not to mention this is not the right place to argue what the standard > could or should have b

Re: [OMPI users] Fwd: [EXTERNAL] Re: How to find MPI ranks located in remote nodes?

2014-11-28 Thread George Bosilca
The same functionality can be trivially achieved at the user level using Adam's approach. If we provide a shortcut in Open MPI, we should emphasize this is an MPI extension, and offer the opportunity to other MPI to provide a compatible support Thus, I would name all new types MPIX_ instead of OMP

Re: [OMPI users] mpi_wtime implementation

2014-11-28 Thread George Bosilca
; > > > int main (int argc, char *argv[]) { > > int i; > > double t = 0; > > MPI_Init(&argc, &argv); > > for (;;) { > > double _t = MPI_Wtime(); > > if (_t < t) { > > fprintf(stderr, "going back in time %l

Re: [OMPI users] [EXTERNAL] Re: How to find MPI ranks located in remote nodes?

2014-12-02 Thread George Bosilca
> On Dec 2, 2014, at 00:37 , Jeff Squyres (jsquyres) wrote: > > On Nov 28, 2014, at 11:58 AM, George Bosilca wrote: > >> The same functionality can be trivially achieved at the user level using >> Adam's approach. If we provide a shortcut in Open MPI, we should

Re: [OMPI users] [EXTERNAL] Re: How to find MPI ranks located in remote nodes?

2014-12-02 Thread George Bosilca
On Tue, Dec 2, 2014 at 9:56 AM, Jeff Squyres (jsquyres) wrote: > I like the OMPI_ prefix because it clearly identifies the function as > specific to Open MPI (i.e., you really should enclose it in #if > defined(OPEN_MPI) / #endif). > That's not enough. They will have to check for the right versi

Re: [OMPI users] OMPI users] MPI inside MPI (still)

2014-12-13 Thread George Bosilca
You have to call MPI_Comm_disconnect on both sides of the intercommunicator. On the spawner processes you should call it on the intercom, while on the spawnees you should call it on the MPI_Comm_get_parent. George. > On Dec 12, 2014, at 20:43 , Alex A. Schmidt wrote: > > Gilles, > > MPI_co

Re: [OMPI users] OMPI users] OMPI users] MPI inside MPI (still)

2014-12-13 Thread George Bosilca
d a submit parameter to make the command block until >> the job completes. Or you can write your own wrapper. >> Or you can retrieve the jobid and qstat periodically to get the job state. >> If an api is available, this is also an option. >> >> Cheers, >> >> G

Re: [OMPI users] OMPI users] OMPI users] OMPI users] MPI inside MPI (still)

2014-12-14 Thread George Bosilca
or ? >> >> I also read the man page again, and MPI_Comm_disconnect does not ensure >> the remote processes have finished or called MPI_Comm_disconnect, so that >> might not be the thing you need. >> George, can you please comment on that ? >> >> Cheers, >&g

Re: [OMPI users] OMPI users] OMPI users] OMPI users] MPI inside MPI (still)

2014-12-17 Thread George Bosilca
t;> >>>>> On Mon, Dec 15, 2014 at 9:27 AM, Alex A. Schmidt wrote: >>>>> >>>>>> George, >>>>>> >>>>>> Thanks for the tip. In fact, calling mpi_comm_spawn right away with >>>>>> MPI_COMM_

Re: [OMPI users] [EXTERNAL] Re: How to find MPI ranks located in remote nodes?

2014-12-17 Thread George Bosilca
On Wed, Dec 17, 2014 at 7:29 PM, Jeff Squyres (jsquyres) wrote: > Returning to a super-old thread that was never finished... > > > On Dec 2, 2014, at 6:49 PM, George Bosilca wrote: > > > That's not enough. They will have to check for the right version of Open > MPI

Re: [OMPI users] Deadlock in OpenMPI 1.8.3 and PETSc 3.4.5

2014-12-17 Thread George Bosilca
Ben, I can't find anything in the MPI standard suggesting that a recursive behavior of the attribute deletion is enforced/supported by the MPI standard. Thus, the current behavior of Open MPI (a single lock for all attributes), while maybe a little strict, is standard compliant (and thus correct).

Re: [OMPI users] [EXTERNAL] Re: How to find MPI ranks located in remote nodes?

2014-12-19 Thread George Bosilca
On Thu, Dec 18, 2014 at 2:27 PM, Jeff Squyres (jsquyres) wrote: > On Dec 17, 2014, at 9:52 PM, George Bosilca wrote: > > >> I don't understand how MPIX_ is better. > >> > >> Given that there is *zero* commonality between any MPI extension > implemented

Re: [OMPI users] Deadlock in OpenMPI 1.8.3 and PETSc 3.4.5

2014-12-19 Thread George Bosilca
interest/cycles in taking it over and advancing it with the Forum. > > Two additional points from the PDF listed above: > > - on slide 21, it was decided to no allow the recursive behavior (i.e., > you can ignore the "This is under debate" bullet. > - the "destroy&q

Re: [OMPI users] send and receive vectors + variable length

2015-01-08 Thread George Bosilca
Diego, Non-blocking communications only indicate a communication will happen, it does not force them to happen. They will only complete on the corresponding MPI_Wait, which also marks the moment starting from where the data can be safely altered or accessed (in the case of the MPI_Irecv). Thus dea

Re: [OMPI users] MPI_Type_Create_Struct + MPI_TYPE_CREATE_RESIZED

2015-01-08 Thread George Bosilca
Diego, Please find below the corrected example. There were several issues but the most important one, which is certainly the cause of the segfault, is that "real(dp)" (with dp = selected_real_kind(p=16)) is NOT equal to MPI_DOUBLE_RECISION. For double precision you should use 15 (and not 16). G

Re: [OMPI users] send and receive vectors + variable length

2015-01-08 Thread George Bosilca
s -- put them all > in a single array). > > Look at examples 3.8 and 3.9 in the MPI-3.0 document. > > > > On Jan 8, 2015, at 5:15 PM, George Bosilca wrote: > > > Diego, > > > > Non-blocking communications only indicate a communication will happen, > it

Re: [OMPI users] MPI_Type_Create_Struct + MPI_TYPE_CREATE_RESIZED

2015-01-08 Thread George Bosilca
Or use MPI_Type_match_size to find the right type. George. > On Jan 8, 2015, at 19:05 , Gus Correa wrote: > > Hi Diego > > *EITHER* > declare your QQ and PR (?) structure components as DOUBLE PRECISION > *OR* > keep them REAL(dp) but *fix* your "dp&q

Re: [OMPI users] send and receive vectors + variable length

2015-01-09 Thread George Bosilca
I totally agree with Dave here. Moreover, based on the logic exposed by Jeff, there is no right solution because if one choose to first wait on the receive requests this also leads to a deadlock as the send requests might not be progressed. As a side note, posting the receive requests first minim

Re: [OMPI users] MPI_type_create_struct + MPI_Type_vector + MPI_Type_contiguous

2015-01-15 Thread George Bosilca
> On Jan 15, 2015, at 06:02 , Diego Avesani wrote: > > Dear Gus, Dear all, > Thanks a lot. > MPI_Type_Struct works well for the first part of my problem, so I am very > happy to be able to use it. > > Regarding MPI_TYPE_VECTOR. > > I have studied it and for simple case it is clear to me what

Re: [OMPI users] MPI_type_create_struct + MPI_Type_vector + MPI_Type_contiguous

2015-01-16 Thread George Bosilca
orge, dear Gus, dear all, >> Could you please tell me where I can find a good example? >> I am sorry but I can not understand the 3D array. >> >> >> Really Thanks >> >> Diego >> >> >> On 15 January 2015 at 20:13, George Bosilca >

Re: [OMPI users] MPI_type_create_struct + MPI_Type_vector + MPI_Type_contiguous

2015-01-16 Thread George Bosilca
> > I would like to have a sort *MPI_my_TYPE to do that (like * > *MPI_my_STRUCT**) *and not to create every time *MPI_TYPE_VECTOR *because > *DATASEND_REAL *changes size every time. > > I hope to make myself clear. > > So is it correct to use *MPI_TYPE_VECTOR?, *Can I do

Re: [OMPI users] MPI_type_create_struct + MPI_Type_vector + MPI_Type_contiguous

2015-01-16 Thread George Bosilca
emulate Fortran90 and C > structures, as Gus' suggestion. > > Let's me look to that tutorial > What do you think? > > Thanks again > > > > > > > Diego > > > On 16 January 2015 at 16:02, George Bosilca <mailto:bosi...@icl.utk.edu>> wro

Re: [OMPI users] How to handle strides in MPI_Create_type_subarray - Re: MPI_type_create_struct + MPI_Type_vector + MPI_Type_contiguous

2015-01-16 Thread George Bosilca
ng in by then, writing a new book might be worth the effort. George. > > Thank you, > Gus Correa > (Hijacking Diego Avesani's thread, apologies to Diego.) > (Also, I know this question is not about Open MPI, but about MPI in general. > But the lack of examples may

Re: [OMPI users] How to handle strides in MPI_Create_type_subarray - Re: MPI_type_create_struct + MPI_Type_vector + MPI_Type_contiguous

2015-01-17 Thread George Bosilca
t; >> Maybe after the release of MPI 4.0 would be a good target … > > Not any sooner than that? > MPI-2 is already poorly covered in the literature, > MPI-3 only by the standard (yawn ...). > And when MPI 4 comes, would we have to wait for MPI-5 to get > the examples? >

Re: [OMPI users] TCP BTL and virtual network interfaces, bug #3339

2015-01-26 Thread George Bosilca
Use mpirun --mca btl_tcp_if_exclude eth0 should fix your problem. Otherwise you can add it to your configuration file. Everything is extensively described in the FAQ. George. On Jan 26, 2015 12:11 PM, "Kris Kersten" wrote: > I'm working on an ethernet cluster that uses virtual eth0:* interfaces

Re: [OMPI users] Several Bcast calls in a loop causing the code to hang

2015-02-22 Thread George Bosilca
Sachin, I cant replicate your issue neither with the latest 1.8 nor with the trunk. I tried using a single host, while forcing SM and then TP to no avail. Can you try restricting the collective modules in use (adding --mca coll tuned,basic) to your mpirun command? George. On Fri, Feb 20, 201

Re: [OMPI users] Several Bcast calls in a loop causing the code to hang

2015-02-23 Thread George Bosilca
m = 12000 >> >rank 0, m = 12000 >> >rank 1, m = 13000 >> >rank 0, m = 13000 >> >rank 2, m = 13000 >> >rank 1, m = 14000 >> >rank 2, m = 14000 >> >rank 0, m = 14000 >> >rank 1, m = 15000 >> >

Re: [OMPI users] Questions regarding MPI intercommunicators & collectives

2015-02-24 Thread George Bosilca
> On Feb 23, 2015, at 10:20 , Harald Servat wrote: > > Hello list, > > we have several questions regarding calls to collectives using > intercommunicators. In man for MPI_Bcast, there is a notice for the > inter-communicator case that reads the text below our questions. > > If an I is an i

Re: [OMPI users] Segmentation fault with MPI_Type_indexed

2015-03-05 Thread George Bosilca
Bogdan, As far as I can tell your code is correct, and the problem is coming from Open MPI. More specifically, I used alloca in the optimization stage in MPI_Type_commit, and as your arrays of length were too large, alloca failed and lead to a segfault. I fixed in the trunk (3c489ea), and this wil

Re: [OMPI users] Segmentation fault with MPI_Type_indexed

2015-03-05 Thread George Bosilca
f Technical Sciences, Novi Sad, Serbia > > On Thu, Mar 5, 2015 at 6:31 PM, George Bosilca > wrote: > >> Bogdan, >> >> As far as I can tell your code is correct, and the problem is coming from >> Open MPI. More specifically, I used alloca in the optimization s

Re: [OMPI users] Segmentation fault with MPI_Type_indexed

2015-03-07 Thread George Bosilca
> > Bogdan Sataric > > email: bogdan.sata...@gmail.com > phone: +381 21-485-2441 > > Teaching & Research Assistant > Chair for Applied Computer Science > Faculty of Technical Sciences, Novi Sad, Serbia > > On Fri, Mar 6, 2015 at 12:52 AM, George Bosilca > wrote: &g

Re: [OMPI users] Open MPI collectives algorithm selection

2015-03-10 Thread George Bosilca
Khalid, The decision is rechecked every time we create a new communicator. So, you might create a solution that force the algorithm to whatever you think it is best (using the environment variables you mentioned), then create a communicator, and free it once you’re done. I have no idea what yo

Re: [OMPI users] Custom MPI_DATA_TYPE with allocatable arrays in fortran derived data type

2015-03-15 Thread George Bosilca
Thomas, IWhat exactly is 'local_tlr_lookup(1)%wtlr'? I think the problem is that your MPI derived datatype use the pointer to the allocatable arrays instead of using the pointer to the first element of these arrays. As an example instead of doing call mpi_get_address(local_tlr_lookup(1)%wtlr,

Re: [OMPI users] Custom MPI_DATA_TYPE with allocatable arrays in fortran derived data type

2015-03-15 Thread George Bosilca
s issue of derived data types, > though. Perhaps that'll help. > > Best, > Thomas > > On Sun, Mar 15, 2015 at 9:00 PM George Bosilca <mailto:bosi...@icl.utk.edu>> wrote: > Thomas, > > IWhat exactly is 'local_tlr_lookup(1)%wtlr'? > > I thi

Re: [OMPI users] MPI_Put/Get with many nested derived type

2016-05-21 Thread George Bosilca
16d9f71d01cc should provide a fix for this issue. George. On Sat, May 21, 2016 at 12:08 PM, Akihiro Tabuchi < tabu...@hpcs.cs.tsukuba.ac.jp> wrote: > Hi Gilles, > > Thanks for your quick response and patch. > > After applying the patch to 1.10.2, the test code and our program which > uses nes

Re: [OMPI users] openmpi-dev-4221-gb707d13: referenced symbol

2016-06-08 Thread George Bosilca
Apparently Solaris 10 lacks support for strnlen. We should add it to our configure and provide a replacement where needed. George. On Wed, Jun 8, 2016 at 4:30 PM, Siegmar Gross < siegmar.gr...@informatik.hs-fulda.de> wrote: > Hi, > > I have built openmpi-dev-4221-gb707d13 on my machines (Solar

Re: [OMPI users] mpi_reduce implementation

2016-07-08 Thread George Bosilca
On Jul 8, 2016 3:16 PM, "Juan Francisco Martínez" < juan.francisco.marti...@est.fib.upc.edu> wrote: > > Hi everybody! > > First of all I want to congratulate all of you because the quality of > the community, I have solved a lot of doubts just reading the mailing > list. > > However I have a questi

Re: [OMPI users] mpi_reduce implementation

2016-07-08 Thread George Bosilca
Not that I know of. Ompi_info should be good enough no? George > > - Fran > > On Fri, 2016-07-08 at 15:40 +0200, George Bosilca wrote: > > > > On Jul 8, 2016 3:16 PM, "Juan Francisco Martínez" < > > juan.francisco.marti...@est.fib.upc.edu> wrote: > &

Re: [OMPI users] Multi-Threaded Performance Question

2016-09-05 Thread George Bosilca
Steve, Some of these interferences are due to the design choices in Open MPI, but others are due to the constraints imposed by MPI. As an example, MPI requires FIFO ordering on message delivery on each communication channel (communicator/peer). If you inject messages from multiple threads for the

Re: [OMPI users] OS X + Xcode 8 : dyld: Symbol not found: _clock_gettime

2016-10-03 Thread George Bosilca
This function is not available on OS X, and the corresponding OMPI module shouldn't have been compiled. How did you configure you OMPI install ? George. On Mon, Oct 3, 2016 at 10:09 AM, Christophe Peyret < christophe.pey...@onera.fr> wrote: > Hello, > > since Xcode8 update, I have problem pro

Re: [OMPI users] OS X + Xcode 8 : dyld: Symbol not found: _clock_gettime

2016-10-03 Thread George Bosilca
nfirms that. > > ompi-release git:(v1.10) ✗ find . -name '*.[ch]' | xargs grep > clock_gettime > > ompi-release git:(v1.10) ✗ > > -Nathan > > On Oct 03, 2016, at 10:50 AM, George Bosilca wrote: > > This function is not available on OS X, and the corresponding O

Re: [OMPI users] Crash during MPI_Finalize

2016-10-10 Thread George Bosilca
George, There is too much information missing from your example. If I try to run the code on the top assuming the process is is_host(NC.node), I have on NC.commd 3 communications (ignore the others): rc = MPI_Send(&ival, 1, MPI_INT, NC.dmsgid, SHUTDOWN_ANDMSG, NC.commd); MPI_Recv(&ival, 1, MPI_IN

Re: [OMPI users] communications groups

2016-10-14 Thread George Bosilca
Rick, Let's assume that you have started 2 processes, and that your sensorList is {1}. The worldgroup will then be {P0, P1}, which trimmed via the sensorList will give the sensorgroup {MPI_GROUP_EMPTY} on P0 and the sensorgroup {P1} on P1. As a result on P0 you will create a MPI_COMM_NULL communic

Re: [OMPI users] Performing partial calculation on a single node in an MPI job

2016-10-16 Thread George Bosilca
Vahid, You cannot use Fortan's vector subscript with MPI. Are you certain that the arrays used in your bcast are contiguous ? If not you would either need to move the data first into a single dimension array (which will then have the elements contiguously in memory), or define specialized datatyp

  1   2   3   4   5   6   7   8   >