October 2008 - Mpich2-core - mailman3.cels.anl.gov

maint/updatefiles warnings
by David Gingold 04 Nov '08

04 Nov '08

MPICH2 folks -- I'm working on updating our MPI code base this week, and so I'm starting to play with some of the MPICH2 1.1 code. When I check out a clean copy of the 1.1.0a1 code (svn export https:// svn.mcs.anl.gov/repos/mpi/mpich2/tags/release/mpich2-1.1.0a1) and run ./maint/updatefiles, I get various error messages in the output. They fall into these categories: 1. A few of these: Warning: State in FUNC_EXIT(a) does not match STATE_DECL() in src/ include/rlog_macros.h Warning: no FUNCNAME defined for MPID_STATE_MPIR_BCAST in src/mpi/ coll/bcast.c 2. Warning: header file mpidconf.h.in or mpidconf.h.in.in not found in src/mpid/globus/ 3. Shell variable MPID_THREAD_OUTPUT_FILES will not be added to the list of known autoconf files for src/mpid/dcmfd. 4. Warning: source file mpe_thread.c or mpe_thread.c.in not found in src/util/thread/ 5. Many of these, which I believe are NFS lossage: make: Warning: File `mf.tmp' has modification time 0.029 s in the future make: warning: Clock skew detected. Your build may be incomplete. 6. A bunch of these: configure.in:95: warning: AC_CONFIG_SUBDIRS: you should use literals ../../lib/autoconf/status.m4:1103: AC_CONFIG_SUBDIRS is expanded from... configure.in:95: the top level 7. A bunch of errors coming out of slog2sdk, attached below. Are these all normal? In particular, I want to make sure that there is nothing wrong with the autoconf-2.62 that I built to do this. (And if there is a way that I can run this with an older autoconf, like 2.62, that surely would make my life easier....) Thanks in advance for the help. -dg .... Creating configure in /home/gingold/scmpi-update/mpich2-1.1.0a1/src/ mpe2/src/slog2sdk/trace_rlog/ ... configure.in:202: warning: AC_CACHE_VAL (lt_prog_compiler_static_works, ...): suspicious cache-id, must contain _cv_ to be cached ../../lib/autoconf/general.m4:1987: AC_CACHE_VAL is expanded from... ../../lib/autoconf/general.m4:2000: AC_CACHE_CHECK is expanded from... ./libtool.m4:670: AC_LIBTOOL_LINKER_OPTION is expanded from... ./libtool.m4:2654: _LT_AC_LANG_C_CONFIG is expanded from... ./libtool.m4:2550: AC_LIBTOOL_LANG_C_CONFIG is expanded from... ./libtool.m4:219: AC_LIBTOOL_SETUP is expanded from... ./libtool.m4:75: _AC_PROG_LIBTOOL is expanded from... ./libtool.m4:55: AC_PROG_LIBTOOL is expanded from... configure.in:202: the top level configure.in:202: warning: AC_CACHE_VAL (lt_prog_compiler_pic_works, ...): suspicious cache-id, must contain _cv_ to be cached ./libtool.m4:633: AC_LIBTOOL_COMPILER_OPTION is expanded from... ./libtool.m4:5165: AC_LIBTOOL_PROG_COMPILER_PIC is expanded from... configure.in:202: warning: AC_CACHE_VAL (lt_prog_compiler_pic_works_CXX, ...): suspicious cache-id, must contain _cv_ to be cached ./libtool.m4:3612: _LT_AC_LANG_CXX_CONFIG is expanded from... ./libtool.m4:2662: AC_LIBTOOL_LANG_CXX_CONFIG is expanded from... ./libtool.m4:1797: _LT_AC_TAGCONFIG is expanded from... configure.in:202: warning: AC_CACHE_VAL (lt_prog_compiler_pic_works_F77, ...): suspicious cache-id, must contain _cv_ to be cached ./libtool.m4:3853: _LT_AC_LANG_F77_CONFIG is expanded from... ./libtool.m4:3755: AC_LIBTOOL_LANG_F77_CONFIG is expanded from... configure.in:202: warning: AC_CACHE_VAL (lt_prog_compiler_pic_works_GCJ, ...): suspicious cache-id, must contain _cv_ to be cached ./libtool.m4:3915: _LT_AC_LANG_GCJ_CONFIG is expanded from... ./libtool.m4:3861: AC_LIBTOOL_LANG_GCJ_CONFIG is expanded from... configure.in:202: warning: AC_CACHE_VAL (lt_prog_compiler_static_works, ...): suspicious cache-id, must contain _cv_ to be cached ../../lib/autoconf/general.m4:1987: AC_CACHE_VAL is expanded from... ../../lib/autoconf/general.m4:2000: AC_CACHE_CHECK is expanded from... ./libtool.m4:670: AC_LIBTOOL_LINKER_OPTION is expanded from... ./libtool.m4:2654: _LT_AC_LANG_C_CONFIG is expanded from... ./libtool.m4:2550: AC_LIBTOOL_LANG_C_CONFIG is expanded from... ./libtool.m4:219: AC_LIBTOOL_SETUP is expanded from... ./libtool.m4:75: _AC_PROG_LIBTOOL is expanded from... ./libtool.m4:55: AC_PROG_LIBTOOL is expanded from... configure.in:202: the top level configure.in:202: warning: AC_CACHE_VAL (lt_prog_compiler_pic_works, ...): suspicious cache-id, must contain _cv_ to be cached ./libtool.m4:633: AC_LIBTOOL_COMPILER_OPTION is expanded from... ./libtool.m4:5165: AC_LIBTOOL_PROG_COMPILER_PIC is expanded from... configure.in:202: warning: AC_CACHE_VAL (lt_prog_compiler_pic_works_CXX, ...): suspicious cache-id, must contain _cv_ to be cached ./libtool.m4:3612: _LT_AC_LANG_CXX_CONFIG is expanded from... ./libtool.m4:2662: AC_LIBTOOL_LANG_CXX_CONFIG is expanded from... ./libtool.m4:1797: _LT_AC_TAGCONFIG is expanded from... configure.in:202: warning: AC_CACHE_VAL (lt_prog_compiler_pic_works_F77, ...): suspicious cache-id, must contain _cv_ to be cached ./libtool.m4:3853: _LT_AC_LANG_F77_CONFIG is expanded from... ./libtool.m4:3755: AC_LIBTOOL_LANG_F77_CONFIG is expanded from... configure.in:202: warning: AC_CACHE_VAL (lt_prog_compiler_pic_works_GCJ, ...): suspicious cache-id, must contain _cv_ to be cached ./libtool.m4:3915: _LT_AC_LANG_GCJ_CONFIG is expanded from... ./libtool.m4:3861: AC_LIBTOOL_LANG_GCJ_CONFIG is expanded from... Creating configure in /home/gingold/scmpi-update/mpich2-1.1.0a1/src/ mpe2/src/slog2sdk/src/ ... Creating configure in /home/gingold/scmpi-update/mpich2-1.1.0a1/src/ mpe2/src/slog2sdk/ ... configure.in:443: warning: AC_CONFIG_SUBDIRS: you should use literals ../../lib/autoconf/status.m4:1103: AC_CONFIG_SUBDIRS is expanded from... configure.in:443: the top level Creating configure in /home/gingold/scmpi-update/mpich2-1.1.0a1/src/ mpe2/src/slog2sdk/trace_sample/ ... configure.in:202: warning: AC_CACHE_VAL (lt_prog_compiler_static_works, ...): suspicious cache-id, must contain _cv_ to be cached ../../lib/autoconf/general.m4:1987: AC_CACHE_VAL is expanded from... ../../lib/autoconf/general.m4:2000: AC_CACHE_CHECK is expanded from... ./libtool.m4:670: AC_LIBTOOL_LINKER_OPTION is expanded from... ./libtool.m4:2654: _LT_AC_LANG_C_CONFIG is expanded from... ./libtool.m4:2550: AC_LIBTOOL_LANG_C_CONFIG is expanded from... ./libtool.m4:219: AC_LIBTOOL_SETUP is expanded from... ./libtool.m4:75: _AC_PROG_LIBTOOL is expanded from... ./libtool.m4:55: AC_PROG_LIBTOOL is expanded from... configure.in:202: the top level configure.in:202: warning: AC_CACHE_VAL (lt_prog_compiler_pic_works, ...): suspicious cache-id, must contain _cv_ to be cached ./libtool.m4:633: AC_LIBTOOL_COMPILER_OPTION is expanded from... ./libtool.m4:5165: AC_LIBTOOL_PROG_COMPILER_PIC is expanded from... configure.in:202: warning: AC_CACHE_VAL (lt_prog_compiler_pic_works_CXX, ...): suspicious cache-id, must contain _cv_ to be cached ./libtool.m4:3612: _LT_AC_LANG_CXX_CONFIG is expanded from... ./libtool.m4:2662: AC_LIBTOOL_LANG_CXX_CONFIG is expanded from... ./libtool.m4:1797: _LT_AC_TAGCONFIG is expanded from... configure.in:202: warning: AC_CACHE_VAL (lt_prog_compiler_pic_works_F77, ...): suspicious cache-id, must contain _cv_ to be cached ./libtool.m4:3853: _LT_AC_LANG_F77_CONFIG is expanded from... ./libtool.m4:3755: AC_LIBTOOL_LANG_F77_CONFIG is expanded from... configure.in:202: warning: AC_CACHE_VAL (lt_prog_compiler_pic_works_GCJ, ...): suspicious cache-id, must contain _cv_ to be cached ./libtool.m4:3915: _LT_AC_LANG_GCJ_CONFIG is expanded from... ./libtool.m4:3861: AC_LIBTOOL_LANG_GCJ_CONFIG is expanded from... configure.in:202: warning: AC_CACHE_VAL (lt_prog_compiler_static_works, ...): suspicious cache-id, must contain _cv_ to be cached ../../lib/autoconf/general.m4:1987: AC_CACHE_VAL is expanded from... ../../lib/autoconf/general.m4:2000: AC_CACHE_CHECK is expanded from... ./libtool.m4:670: AC_LIBTOOL_LINKER_OPTION is expanded from... ./libtool.m4:2654: _LT_AC_LANG_C_CONFIG is expanded from... ./libtool.m4:2550: AC_LIBTOOL_LANG_C_CONFIG is expanded from... ./libtool.m4:219: AC_LIBTOOL_SETUP is expanded from... ./libtool.m4:75: _AC_PROG_LIBTOOL is expanded from... ./libtool.m4:55: AC_PROG_LIBTOOL is expanded from... configure.in:202: the top level configure.in:202: warning: AC_CACHE_VAL (lt_prog_compiler_pic_works, ...): suspicious cache-id, must contain _cv_ to be cached ./libtool.m4:633: AC_LIBTOOL_COMPILER_OPTION is expanded from... ./libtool.m4:5165: AC_LIBTOOL_PROG_COMPILER_PIC is expanded from... configure.in:202: warning: AC_CACHE_VAL (lt_prog_compiler_pic_works_CXX, ...): suspicious cache-id, must contain _cv_ to be cached ./libtool.m4:3612: _LT_AC_LANG_CXX_CONFIG is expanded from... ./libtool.m4:2662: AC_LIBTOOL_LANG_CXX_CONFIG is expanded from... ./libtool.m4:1797: _LT_AC_TAGCONFIG is expanded from... configure.in:202: warning: AC_CACHE_VAL (lt_prog_compiler_pic_works_F77, ...): suspicious cache-id, must contain _cv_ to be cached ./libtool.m4:3853: _LT_AC_LANG_F77_CONFIG is expanded from... ./libtool.m4:3755: AC_LIBTOOL_LANG_F77_CONFIG is expanded from... configure.in:202: warning: AC_CACHE_VAL (lt_prog_compiler_pic_works_GCJ, ...): suspicious cache-id, must contain _cv_ to be cached ./libtool.m4:3915: _LT_AC_LANG_GCJ_CONFIG is expanded from... ./libtool.m4:3861: AC_LIBTOOL_LANG_GCJ_CONFIG is expanded from...

3 2

Re: [mpich2-core] Implicit decl warning with "-ansi"+"-Wall" option
by Anthony Chan 31 Oct '08

31 Oct '08

The gcc compiler options swallow the mkstemp declaration. MPICH2 or MPE's configure checks for mkstemp declaration (and other common libc functions) to see if it exists (especially when --enables-strict is used) If not, the configure will insert mkstemp declaration. ----- "Jayesh Krishna" <jayesh(a)mcs.anl.gov> wrote: > Hi, > > ========================================= > #include <stdlib.h> > > /* extern int mkstemp(char *); */ > int main(int argc, char *argv[]){ > int fd = -1; > char template[] = "temp_XXXXXX"; > fd = mkstemp(template); > return 0; > } > ========================================= > > The code above gives me the following warning when compiled with > "-Wall" & > "-ansi" option with gcc, > > ========================================= > shakey:/sandbox/jayesh/Enhancements/NemesisOnWindows/testPgms> gcc > -ansi > -Wall test.c > test.c: In function â?~mainâ?T: > test.c:7: warning: implicit declaration of function â?~mkstempâ?T > ========================================= > > Any ideas ? > > -Jayesh

2 2

Implicit decl warning with "-ansi"+"-Wall" option
by Jayesh Krishna 31 Oct '08

31 Oct '08

Hi, ========================================= #include <stdlib.h> /* extern int mkstemp(char *); */ int main(int argc, char *argv[]){ int fd = -1; char template[] = "temp_XXXXXX"; fd = mkstemp(template); return 0; } ========================================= The code above gives me the following warning when compiled with "-Wall" & "-ansi" option with gcc, ========================================= shakey:/sandbox/jayesh/Enhancements/NemesisOnWindows/testPgms> gcc -ansi -Wall test.c test.c: In function ï¿½?~mainï¿½?T: test.c:7: warning: implicit declaration of function ï¿½?~mkstempï¿½?T ========================================= Any ideas ? -Jayesh

1 0

preliminary results for smp collective performance
by Dave Goodell 22 Oct '08

22 Oct '08

I've been measuring the performance gain we get from SMP collectives on the breadboard cluster and I've put up some graphs here: http:// www.mcs.anl.gov/~goodell/smpcoll.html The gains are substantial, the smp-aware collectives simultaneously reduce the time taken and minimize the performance variation when different root arguments are used. These are only the results for bcast and reduce, and then only for 4- byte and 1024-byte payloads (1 and 256 MPI_INT's, respectively). I still need to measure barrier and allreduce as well as try other payload sizes. The graphs could probably use some tweaking, but this is the best I could whip up in a short time this afternoon. Also, all of this was measured with --with-device=ch3:nemesis:newtcp, so the results probably look a bit better than what you'd see on a legitimate high-speed network. I need to see if the mx module is in good enough shape to take the same measurements. I also want to throw OpenMPI into the mix to see how our smp collectives stack up against theirs :) -Dave

7 10

Fwd: New MPI text
by Rusty Lusk 21 Oct '08

21 Oct '08

I haven't read the whole thing, but the table of contents looks OK. Begin forwarded message: > From: "W Levy" <wlevy.is(a)gmail.com> > Date: October 17, 2008 5:11:49 PM CDT > To: lusk(a)mcs.anl.gov > Subject: New MPI text > > Dear Dr. Lusk, > > > We are hoping that you would add our MPI teaching material to your > list of "Books on and about MPI" (http://www-unix.mcs.anl.gov/mpi/learning.html > ). Our teaching material is aimed at non-computer scientists and > includes both a booklet and live CD entitled, SuperComputing > Simplified: The Bare Necessities for Parallel C Programming with > MPI. The live CD allows the student to compile and run the MPI > programming lessons on a conventional PC. > > > Our contribution is many levels simpler and infinitely less complete > than the book you co-authored with Dr. Gropp (Using MPI, 2nd > edition). However, from our perspective, and I hope yours too, our > booklet and live CD is an appropriate stepping stone to your > advanced and indispensable treatment of MPI. You can find out more > about our booklet and live CD at:www.supercomputingsimplified.com. > > > We will be happy to send a complimentary copy of the booklet and > live-CD (the live-CD turns a standard PC into an MPI emulation > system running under Linux). In order for the live-CD to work, we > will need to know the MAC address of the computer that it will be > run on. > > > In addition, we have begun working on the second edition. Could you > suggest any people who have or who are currently teaching MPI to non- > computer scientists? We are looking for suggestions that would > improve the utility and teachability of our presentation. If you or > any of your colleagues have time to look over the booklet, we would > like to know how to enhance the transition between our elementary > treatment and your advanced work. > > > Thank you for your consideration and any assistance that you may > offer. > > > William B. Levy, Ph.D. > > Owner, Informed Simplifications, LLC > > Professor, University of Virginia >

3 7

Fwd: mpich2-1.0.8rc1 tarball available
by Dave Goodell 17 Oct '08

17 Oct '08

FYI, below is a link to the release candidate. I'm doing some basic sanity checking right now, but now is the time for anyone else who wants to test the RC before the release. https://svn.mcs.anl.gov/repos/mpi/www/downloads/tarballs/1.0.8rc1/ mpich2-1.0.8rc1.tar.gz -Dave Begin forwarded message: > From: "Jayesh Krishna" <jayesh(a)mcs.anl.gov> > Date: October 17, 2008 Oct 17 1:29:28 PM CDT > To: "'Dave Goodell'" <goodell(a)mcs.anl.gov> > Subject: RE: mpich2-1.0.8rc1 tarball available > > You might also want to check with Anthony -- he usually finds > pblms (autoconf, man pages) etc > > -Jayesh > > -----Original Message----- > From: Dave Goodell [mailto:[email protected]] > Sent: Friday, October 17, 2008 1:21 PM > To: Jayesh Krishna > Cc: Dave Goodell > Subject: mpich2-1.0.8rc1 tarball available > > You can get it here: https://svn.mcs.anl.gov/repos/mpi/www/ > downloads/tarballs/1.0.8rc1/mpich2-1.0.8rc1.tar.gz > > I still have to do some sanity testing with it, but that's no > reason to delay your packaging and testing. > > -Dave >

2 1

Re: [mpich2-core] Fwd: mpich2-1.0.8rc1 tarball available
by Anthony Chan 17 Oct '08

17 Oct '08

the rc1 tarball is fine. I just added a statement in CHANGES regarding to autoconf-2.62. A.Chan ----- "Dave Goodell" <goodell(a)mcs.anl.gov> wrote: > FYI, below is a link to the release candidate. I'm doing some basic > > sanity checking right now, but now is the time for anyone else who > wants to test the RC before the release. > > https://svn.mcs.anl.gov/repos/mpi/www/downloads/tarballs/1.0.8rc1/ > mpich2-1.0.8rc1.tar.gz > > -Dave > > Begin forwarded message: > > > From: "Jayesh Krishna" <jayesh(a)mcs.anl.gov> > > Date: October 17, 2008 Oct 17 1:29:28 PM CDT > > To: "'Dave Goodell'" <goodell(a)mcs.anl.gov> > > Subject: RE: mpich2-1.0.8rc1 tarball available > > > > You might also want to check with Anthony -- he usually finds > > pblms (autoconf, man pages) etc > > > > -Jayesh > > > > -----Original Message----- > > From: Dave Goodell [mailto:[email protected]] > > Sent: Friday, October 17, 2008 1:21 PM > > To: Jayesh Krishna > > Cc: Dave Goodell > > Subject: mpich2-1.0.8rc1 tarball available > > > > You can get it here: https://svn.mcs.anl.gov/repos/mpi/www/ > > downloads/tarballs/1.0.8rc1/mpich2-1.0.8rc1.tar.gz > > > > I still have to do some sanity testing with it, but that's no > > reason to delay your packaging and testing. > > > > -Dave > >

1 0

Re: [mpich2-core] Fwd: mpich2-1.0.8rc1 tarball available
by Anthony Chan 17 Oct '08

17 Oct '08

What is the best way to get the file without a web-brower ? A.Chan /sandbox/chan/mpich2_work> wget https://svn.mcs.anl.gov/repos/mpi/www/downloads/tarballs/1.0.8rc1/mpich2-1.… . --15:00:27-- https://svn.mcs.anl.gov/repos/mpi/www/downloads/tarballs/1.0.8rc1/mpich2-1.… => `mpich2-1.0.8rc1.tar.gz' Resolving svn.mcs.anl.gov... 140.221.9.28 Connecting to svn.mcs.anl.gov|140.221.9.28|:443... connected. HTTP request sent, awaiting response... 401 Authorization Required Authorization failed. --15:00:27-- http://./ => `index.html' Resolving .... failed: Name or service not known. FINISHED --15:00:27-- Downloaded: 0 bytes in 0 files ----- "Dave Goodell" <goodell(a)mcs.anl.gov> wrote: > FYI, below is a link to the release candidate. I'm doing some basic > > sanity checking right now, but now is the time for anyone else who > wants to test the RC before the release. > > https://svn.mcs.anl.gov/repos/mpi/www/downloads/tarballs/1.0.8rc1/mpich2-1.… > > -Dave > > Begin forwarded message: > > > From: "Jayesh Krishna" <jayesh(a)mcs.anl.gov> > > Date: October 17, 2008 Oct 17 1:29:28 PM CDT > > To: "'Dave Goodell'" <goodell(a)mcs.anl.gov> > > Subject: RE: mpich2-1.0.8rc1 tarball available > > > > You might also want to check with Anthony -- he usually finds > > pblms (autoconf, man pages) etc > > > > -Jayesh > > > > -----Original Message----- > > From: Dave Goodell [mailto:[email protected]] > > Sent: Friday, October 17, 2008 1:21 PM > > To: Jayesh Krishna > > Cc: Dave Goodell > > Subject: mpich2-1.0.8rc1 tarball available > > > > You can get it here: https://svn.mcs.anl.gov/repos/mpi/www/ > > downloads/tarballs/1.0.8rc1/mpich2-1.0.8rc1.tar.gz > > > > I still have to do some sanity testing with it, but that's no > > reason to delay your packaging and testing. > > > > -Dave > >

1 1

RE: 1.0.8 rc
by Rajeev Thakur 16 Oct '08

16 Oct '08

Pavan, If you get a chance before Tuesday, can you cut a release candidate tar ball for Unix. We can start testing that before the Windows version is ready. Rajeev

3 4

Brice Goglin Visit
by Darius Buntinas 16 Oct '08

16 Oct '08

Brice Goglin is planning on visiting for a week in December. He was a post-doc at Myricom where he was working on and is still working on Open-MX. He wanted to meet with members of the MPICH2 and PVFS teams. Most of the MPICH2 team will be out the third week of December at the MPI Forum, so that just leaves the first two weeks. Rob R, L and/or Sam, will you be available to meet with Brice during this time? Rajeev, Pavan and Dave, will you guys be around? Thanks, -d

5 4