- devel - mailman3.cels.anl.gov

Re: [mpich-devel] small MPI_Win_allocate
by Balaji, Pavan 17 May '16

17 May '16

It does allocate from shared memory irrespective of the size. It's not page-aligned unless you ask for noncontig allocations. With respect to wasted VM space, you'll run out of window contexts much before you run out of VM space, so I'm not sure this is a worthwhile optimization. -- Pavan > On May 12, 2016, at 4:54 PM, Jeff Hammond <jeff.science(a)gmail.com> wrote: > > If I call MPI_Win_allocate for e.g. 8 bytes, does it allocate shared memory every time, and thus potentially allocate with page alignment, which could cause a lot of empty VM space? > > Perhaps I do not understand POSIX shared memory well enough, but it strikes me that it might be a worthwhile optimization to have a memory pool for small-size MPI_Win_allocate(_shared) calls. > > Thanks, > > Jeff > > -- > Jeff Hammond > jeff.science(a)gmail.com > http://jeffhammond.github.io/ > _______________________________________________ > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/devel

1 0

romio build failure on OpenBSD >= 5.7
by Gilles Gouaillardet 15 May '16

15 May '16

Folks, currently, mpich (master and previous releases) cannot be built on OpenBSD 5.7 and later. the root cause is 'struct statfs' has no f_type field. here is attached a patch that fixes that, and add some extra #ifdef protections this issue was initially reported at https://www.open-mpi.org/community/lists/devel/2015/05/17449.php and more recently at https://github.com/open-mpi/ompi/issues/1635 Cheers, Gilles

1 0

small MPI_Win_allocate
by Jeff Hammond 12 May '16

12 May '16

If I call MPI_Win_allocate for e.g. 8 bytes, does it allocate shared memory every time, and thus potentially allocate with page alignment, which could cause a lot of empty VM space? Perhaps I do not understand POSIX shared memory well enough, but it strikes me that it might be a worthwhile optimization to have a memory pool for small-size MPI_Win_allocate(_shared) calls. Thanks, Jeff -- Jeff Hammond jeff.science(a)gmail.com http://jeffhammond.github.io/

1 0

Re: [mpich-devel] ./configure --enable-timer-type help output missing
by Halim Amer 07 Apr '16

07 Apr '16

> At the very least, if I know to set the --enable-timer-type option, will > it be passed on to the child configure scripts if I pass it to the > top-level configure invocation? Yes, configure options get propagated to the child scripts. That's part of the difficulty I think, because printing help information and doing option checking at the parent is difficult because the option might only be relevant for some descendant configure script. --Halim On 4/7/16 5:24 PM, Jeff Hammond wrote: > Thanks. I don't know the answer, but I thought there was a standard > method for implementing this. If the MPICH team doesn't know if, then I > am certainly no help, since essentially all of my build systems and > knowledge thereof are derived from MPICH :-) > > At the very least, if I know to set the --enable-timer-type option, will > it be passed on to the child configure scripts if I pass it to the > top-level configure invocation? > > Jeff > > On Thu, Apr 7, 2016 at 2:28 PM, Balaji, Pavan <balaji(a)anl.gov > <mailto:[email protected]>> wrote: > > > It is convenient to have all the help output be present in the > top-level configure. We need to figure out how to make that happen > while still allowing multiple independent packages inside MPICH. We > don't have a good solution for that yet, but we are open to suggestions. > > -- Pavan > > > On Apr 7, 2016, at 3:24 PM, Jeff Hammond <jeff.science(a)gmail.com > <mailto:[email protected]>> wrote: > > > > I recall that the base configure script used to emit the help > associated with --enable-timer-type. It no longer does so, although > the option is still present in src/mpl/configure. > > > > Do I have a reasonable expectation of seeing --help output for > all options in the base configure, or am I expected to run --help > against every configure in the tree? > > > > Thanks, > > > > Jeff > > > > jrhammon-mac01:git jrhammon$ ./configure --help | grep time > > runtime - error checking controllable at runtime > through environment > > runtime - Runtime control of data collected > > runtime - Alias to "multiple" > > --enable-mutex-timing calculate the time spent waiting on mutexes > > speeds up one-time build > > sometimes confusing) to the casual > installer > > User-defined run-time library search path. > > jrhammon-mac01:git jrhammon$ ./src/mpl/configure --help | grep time > > speeds up one-time build > > --enable-timer-type=name - Select the timer to use for > MPI_Wtime and > > internal timestamps. > > gethrtime - Solaris timer (Solaris systems only) > > clock_gettime - Posix timer (where available) > > gettimeofday - Most Unix systems > > linux86_cycle - Linux x86; returns cycle counts, not > time in seconds* > > gcc_ia64_cycle - IPF ar.itc timer* > > mach_absolute_time - Mach absolute time (alternative to clock_gettime > > device - The timer is provided by the device > > *Note that the cycle timers are intended to be used by > > User-defined run-time library search path. > > > > > > -- > > Jeff Hammond > > jeff.science(a)gmail.com <mailto:[email protected]> > > http://jeffhammond.github.io/ > > _______________________________________________ > > To manage subscription options or unsubscribe: > > https://lists.mpich.org/mailman/listinfo/devel > > _______________________________________________ > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/devel > > > > > -- > Jeff Hammond > jeff.science(a)gmail.com <mailto:[email protected]> > http://jeffhammond.github.io/ > > > _______________________________________________ > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/devel >

1 0

Re: [mpich-devel] ./configure --enable-timer-type help output missing
by Jeff Hammond 07 Apr '16

07 Apr '16

Thanks. I don't know the answer, but I thought there was a standard method for implementing this. If the MPICH team doesn't know if, then I am certainly no help, since essentially all of my build systems and knowledge thereof are derived from MPICH :-) At the very least, if I know to set the --enable-timer-type option, will it be passed on to the child configure scripts if I pass it to the top-level configure invocation? Jeff On Thu, Apr 7, 2016 at 2:28 PM, Balaji, Pavan <balaji(a)anl.gov> wrote: > > It is convenient to have all the help output be present in the top-level > configure. We need to figure out how to make that happen while still > allowing multiple independent packages inside MPICH. We don't have a good > solution for that yet, but we are open to suggestions. > > -- Pavan > > > On Apr 7, 2016, at 3:24 PM, Jeff Hammond <jeff.science(a)gmail.com> wrote: > > > > I recall that the base configure script used to emit the help associated > with --enable-timer-type. It no longer does so, although the option is > still present in src/mpl/configure. > > > > Do I have a reasonable expectation of seeing --help output for all > options in the base configure, or am I expected to run --help against every > configure in the tree? > > > > Thanks, > > > > Jeff > > > > jrhammon-mac01:git jrhammon$ ./configure --help | grep time > > runtime - error checking controllable at runtime through > environment > > runtime - Runtime control of data collected > > runtime - Alias to "multiple" > > --enable-mutex-timing calculate the time spent waiting on mutexes > > speeds up one-time build > > sometimes confusing) to the casual installer > > User-defined run-time library search path. > > jrhammon-mac01:git jrhammon$ ./src/mpl/configure --help | grep time > > speeds up one-time build > > --enable-timer-type=name - Select the timer to use for MPI_Wtime and > > internal timestamps. > > gethrtime - Solaris timer (Solaris systems only) > > clock_gettime - Posix timer (where available) > > gettimeofday - Most Unix systems > > linux86_cycle - Linux x86; returns cycle counts, not time > in seconds* > > gcc_ia64_cycle - IPF ar.itc timer* > > mach_absolute_time - Mach absolute time (alternative to clock_gettime > > device - The timer is provided by the device > > *Note that the cycle timers are intended to be used by > > User-defined run-time library search path. > > > > > > -- > > Jeff Hammond > > jeff.science(a)gmail.com > > http://jeffhammond.github.io/ > > _______________________________________________ > > To manage subscription options or unsubscribe: > > https://lists.mpich.org/mailman/listinfo/devel > > _______________________________________________ > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/devel > -- Jeff Hammond jeff.science(a)gmail.com http://jeffhammond.github.io/

1 0

Re: [mpich-devel] ./configure --enable-timer-type help output missing
by Balaji, Pavan 07 Apr '16

07 Apr '16

It is convenient to have all the help output be present in the top-level configure. We need to figure out how to make that happen while still allowing multiple independent packages inside MPICH. We don't have a good solution for that yet, but we are open to suggestions. -- Pavan > On Apr 7, 2016, at 3:24 PM, Jeff Hammond <jeff.science(a)gmail.com> wrote: > > I recall that the base configure script used to emit the help associated with --enable-timer-type. It no longer does so, although the option is still present in src/mpl/configure. > > Do I have a reasonable expectation of seeing --help output for all options in the base configure, or am I expected to run --help against every configure in the tree? > > Thanks, > > Jeff > > jrhammon-mac01:git jrhammon$ ./configure --help | grep time > runtime - error checking controllable at runtime through environment > runtime - Runtime control of data collected > runtime - Alias to "multiple" > --enable-mutex-timing calculate the time spent waiting on mutexes > speeds up one-time build > sometimes confusing) to the casual installer > User-defined run-time library search path. > jrhammon-mac01:git jrhammon$ ./src/mpl/configure --help | grep time > speeds up one-time build > --enable-timer-type=name - Select the timer to use for MPI_Wtime and > internal timestamps. > gethrtime - Solaris timer (Solaris systems only) > clock_gettime - Posix timer (where available) > gettimeofday - Most Unix systems > linux86_cycle - Linux x86; returns cycle counts, not time in seconds* > gcc_ia64_cycle - IPF ar.itc timer* > mach_absolute_time - Mach absolute time (alternative to clock_gettime > device - The timer is provided by the device > *Note that the cycle timers are intended to be used by > User-defined run-time library search path. > > > -- > Jeff Hammond > jeff.science(a)gmail.com > http://jeffhammond.github.io/ > _______________________________________________ > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/devel

1 0

./configure --enable-timer-type help output missing
by Jeff Hammond 07 Apr '16

07 Apr '16

I recall that the base configure script used to emit the help associated with --enable-timer-type. It no longer does so, although the option is still present in src/mpl/configure. Do I have a reasonable expectation of seeing --help output for all options in the base configure, or am I expected to run --help against every configure in the tree? Thanks, Jeff jrhammon-mac01:git jrhammon$ ./configure --help | grep time runtime - error checking controllable at runtime through environment runtime - Runtime control of data collected runtime - Alias to "multiple" --enable-mutex-timing calculate the time spent waiting on mutexes speeds up one-time build sometimes confusing) to the casual installer User-defined run-time library search path. jrhammon-mac01:git jrhammon$ ./src/mpl/configure --help | grep time speeds up one-time build --enable-timer-type=name - Select the timer to use for MPI_Wtime and internal timestamps. gethrtime - Solaris timer (Solaris systems only) clock_gettime - Posix timer (where available) gettimeofday - Most Unix systems linux86_cycle - Linux x86; returns cycle counts, not time in seconds* gcc_ia64_cycle - IPF ar.itc timer* mach_absolute_time - Mach absolute time (alternative to clock_gettime device - The timer is provided by the device *Note that the cycle timers are intended to be used by User-defined run-time library search path. -- Jeff Hammond jeff.science(a)gmail.com http://jeffhammond.github.io/

1 0

Re: [mpich-devel] MPICH on Windows via WSL?
by Jed Brown 07 Apr '16

07 Apr '16

Jeff Hammond <jeff.science(a)gmail.com> writes: > On Wed, Apr 6, 2016 at 12:02 PM, Rob Latham <robl(a)mcs.anl.gov> wrote: > >> >> >> On 04/06/2016 01:52 PM, Jeff Hammond wrote: >> >>> I am still figuring out how to get it, >>> >> >> it's not available yet. >> >> > Good to know I am not stupid for not being able to find it. Looks like it's up now. https://blogs.msdn.microsoft.com/commandline/2016/04/06/bash-on-ubuntu-on-w…

1 0

Re: [mpich-devel] MPICH on Windows via WSL?
by Rob Latham 06 Apr '16

06 Apr '16

On 04/06/2016 02:48 PM, Jeff Hammond wrote: > > > On Wed, Apr 6, 2016 at 12:02 PM, Rob Latham <robl(a)mcs.anl.gov > <mailto:[email protected]>> wrote: > > > > On 04/06/2016 01:52 PM, Jeff Hammond wrote: > > I am still figuring out how to get it, > > > it's not available yet. > > > Good to know I am not stupid for not being able to find it. > > when it /is/ available, I'm betting Pavan is going to fire the first > person who wastes any time on building/testing it. > > > I don't have to worry about that problem, at least until Min masters > diagrammatic perturbation theory :-) Glad you understood my joke. Worded better: "we welcome members of the MPICH community to try it out and share their experiences". ==rob

1 0

Re: [mpich-devel] MPICH on Windows via WSL?
by Kenneth Raffenetti 06 Apr '16

06 Apr '16

I have to admit I am curious about this as a dev environment. Would like to hear about experiences when folks get their hands on a working version. Ken On 04/06/2016 12:48 PM, Jeff Hammond wrote: > > > On Wed, Apr 6, 2016 at 12:02 PM, Rob Latham <robl(a)mcs.anl.gov > <mailto:[email protected]>> wrote: > > > > On 04/06/2016 01:52 PM, Jeff Hammond wrote: > > I am still figuring out how to get it, > > > it's not available yet. > > > Good to know I am not stupid for not being able to find it. > > when it /is/ available, I'm betting Pavan is going to fire the first > person who wastes any time on building/testing it. > > > I don't have to worry about that problem, at least until Min masters > diagrammatic perturbation theory :-) > > Windows Services for UNIX and the Microsoft POSIX Subsystem have > been options for a long time. They did not make "port MPICH to > windows" trivial. I suspect neither will WSL, though it takes a > distinctly different approach. > > > I am more optimistic. Maybe there will be some issues with hostname > resolution as there are today with VPNs and such, but if they make good > on the publicity, _compiling_ MPICH should not be an issue. > > My motivation for getting this to work is that I have to shut my laptop > for the first and last 20 minutes of every flight I take, whereas the > tablet users get to keep theirs out end-to-end. If I can get a *nix > terminal on a tablet, I'll be able to write that many more hours of > MPI-3 code every year. Plus, even if I could get a BSD terminal on iPad > that way I can on a Mac Air, it supports the wrong ISA. Fortunately, > Microsoft Surface already supports the ISA of champions, so I just need > to solve the software problem. > > Jeff > > ==rob > > _______________________________________________ > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/devel > > > > > -- > Jeff Hammond > jeff.science(a)gmail.com <mailto:[email protected]> > http://jeffhammond.github.io/ > > > _______________________________________________ > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/devel >

1 0