Re: [OMPI devel] trace the openmpi internal function calls in MPI user program

Thanks, Jeff. I am trying to understand in detail how Open MPI works in the run time. What main functions does it call to select and initialize the coll components? Using the "helloworld" as an example,  how does it select and initialize the MPI_Barrier algorithm?  which C functions are involved

Re: [OMPI devel] v1.10.1rc1 released

Thanks Paul - always appreciate your help! > On Oct 3, 2015, at 5:07 PM, Paul Hargrove wrote: > > > On Sat, Oct 3, 2015 at 4:47 PM, Paul Hargrove > wrote: > So, I am currently willing to bet that I've got a broken MOFED/OFED hybrid > installation. > If that is the c

Re: [OMPI devel] v1.10.1rc1 released

On Sat, Oct 3, 2015 at 4:47 PM, Paul Hargrove wrote: > So, I am currently willing to bet that I've got a broken MOFED/OFED hybrid > installation. > If that is the case (and I am looking into it now) then this is NOT an > Open MPI problem. > I will try to let you know what I discover. > OK, mea c

Re: [OMPI devel] v1.10.1rc1 released

Ralph, It doesn't look to me like there is a lack of a dl component. It looks to me like libinfinipath is doing something bad when dl_open()ed by mtl_ofi. Regardless, the requested verbose output is attached. As for cores, I cannot get any because somebody is catching SIGABRT. What I do see is fi

Re: [OMPI devel] v1.10.1rc1 released

Paul At a glance, it looks to me like no opal_dl component was selected. Any chance you could add -mca dl_base_verbose 100 and get a line number for us from one of those core files? > On Oct 3, 2015, at 10:41 AM, Paul Hargrove wrote: > > > On Sat, Oct 3, 2015 at 3:13 AM, Jeff Squyres (jsquy

Re: [OMPI devel] v1.10.1rc1 released

No it doesn’t - those haven’t been accepted yet. As was discussed there, the patches need to go into master and then can be worked back into the release cycle. > On Oct 3, 2015, at 2:33 PM, Joshua Ladd wrote: > > This doesn't contain the three patches that we discussed on PR: > https://github

Re: [OMPI devel] v1.10.1rc1 released

This doesn't contain the three patches that we discussed on PR: https://github.com/open-mpi/ompi-release/pull/621 Josh On Sat, Oct 3, 2015 at 6:13 AM, Jeff Squyres (jsquyres) wrote: > v1.10.1 is primarily a bug-fix release. rc1 has been released; it's in > the usual place: > > http://www.o

Re: [OMPI devel] v1.10.1rc1 released

On Sat, Oct 3, 2015 at 3:13 AM, Jeff Squyres (jsquyres) wrote: > v1.10.1 is primarily a bug-fix release. rc1 has been released; it's in > the usual place: > > http://www.open-mpi.org/software/ompi/v1.10/ > > Please test! > My tests experienced a failure (SEGV) on a fairly vanilla Linux VM

[OMPI devel] v1.10.1rc1 released

v1.10.1 is primarily a bug-fix release. rc1 has been released; it's in the usual place: http://www.open-mpi.org/software/ompi/v1.10/ Please test! Changes since v1.10.0: - Fix segv when invoking non-blocking reductions with a user-defined operation. Thanks to Rupert Nash and Georg Geise

Re: [OMPI devel] Access to old users@ and devel@ Open MPI mails

Probably so, sure. > On Oct 3, 2015, at 1:34 AM, George Bosilca wrote: > > Excellent remark. I concur with Gilles, there is no point in increasing the > amount of spam we already receive. > > Can we sanitize the mbox in order to remove all email addresses before we > entrust them with it? >

Re: [OMPI devel] Access to old users@ and devel@ Open MPI mails

Excellent remark. I concur with Gilles, there is no point in increasing the amount of spam we already receive. Can we sanitize the mbox in order to remove all email addresses before we entrust them with it? George. On Sat, Oct 3, 2015 at 12:27 AM, Gilles Gouaillardet < gilles.gouaillar...@gmail.

Re: [OMPI devel] Access to old users@ and devel@ Open MPI mails

Jeff, the minor distinction includes the fact that the web archive does not include email addresses, but the mbox does. I am fine handing them the mbox, with a note asking them not to redistribute it, and keeping it in a secure place because no one like being spammed. Cheers, Gilles On Saturday