Re: [O-MPI users] pathscale 2.1/2.3 build problem
A quick message just to confirm that this problem really had nothing to do with OpenMPI. It was caused by an incorrect wrapper script for gcc. Dries Greg Lindahl wrote: > Dries, > > We have built OpenMPI-1.0 with our compilers, but don't build SVN. > > In your case you seem to have some kind of SUSE weirdness going on. Is > it possible you have a gcc-3.3 compatibility package of some kind > installed? We should probably take this conversation over to > supp...@pathscale.com, but the next step is to try "pathcc -v > hello.c", and also > > /usr/bin/gcc -E "foo bar" > > -- greg > ___ > users mailing list > us...@open-mpi.org > http://www.open-mpi.org/mailman/listinfo.cgi/users Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm
Re: [O-MPI users] Compiling RC4 on Alpha56 and SGI Origin
Hi Brain If the problems with Alphas is still present and you can fix this if you have access to such machines, I can give you an account one one of my alpha stations (they are 24h online ;). Are you interested? cu Eugen On Fri, 2005-10-28 at 09:10 -0500, Brian Barrett wrote: > On Oct 27, 2005, at 8:22 AM, Eugen Wintersberger wrote: > > > Second we have an SGI Origin at our university. After unpacking, the > > configure command runs without any problems. But a subsequent make > > gives > > > > /bin/sh ../../libtool --mode=compile gcc -O3 -DNDEBUG - > > fno-strict-aliasing -c -o atomic-asm.lo atomic-asm.s > > gcc -O3 -DNDEBUG -fno-strict-aliasing -c atomic-asm.s -DPIC - > > o .libs/atomic-asm.o > > as: Error: /atomic-asm.s, line 146: Not all branch-label symbols > > were defined > > as: Error: /atomic-asm.s, line 41: Label referenced but not > > defined: Lretry1 > > *** Error code 1 (bu21) > > *** Error code 1 (bu21) > > *** Error code 1 (bu21) > > > > Any idea why make breaks? > > > This has been fixed in the subversion trunk. It should be in the > nightly snapshots for both the trunk and the 1.0 branch tonight. It > will be part of RC5 when that is rolled. > > Thanks for the bug report! > > Brian > >
[O-MPI users] Compiler problems on SGI Origin
Hi there I have a problem on compiling openmpi (1.0.1rc4) on an SGI origin. My configure command looks like this ./configure --prefix=/scratch/hlphysik/k353443 --disable-mpi-f77 --disable-mpi-f90 --enable-io-romio --enable-cxx-exceptions a subsequent make funs for a while and breaks with Making all in gpr Making all in base depbase=`echo gpr_base_open.lo | sed 's|[^/]*$|.deps/&|;s|\.lo$||'`; \ if /bin/sh ../../../../libtool --tag=CC --mode=compile gcc -DHAVE_CONFIG_H -I. -I. -I../../../../include -I../../../../include -I../../../../include -I../../../.. -I../../../.. -I../../../../include -I../../../../opal -I../../../../orte -I../../../../ompi -D_REENTRANT -O3 -DNDEBUG -fno-strict-aliasing -MT gpr_base_open.lo -MD -MP -MF "$depbase.Tpo" -c -o gpr_base_open.lo gpr_base_open.c; \ then mv -f "$depbase.Tpo" "$depbase.Plo"; else rm -f "$depbase.Tpo"; exit 1; fi mkdir .libs gcc -DHAVE_CONFIG_H -I. -I. -I../../../../include -I../../../../include -I../../../../include -I../../../.. -I../../../.. -I../../../../include -I../../../../opal -I../../../../orte -I../../../../ompi -D_REENTRANT -O3 -DNDEBUG -fno-strict-aliasing -MT gpr_base_open.lo -MD -MP -MF .deps/gpr_base_open.Tpo -c gpr_base_open.c -DPIC -o .libs/gpr_base_open.o gpr_base_open.c: In function `orte_gpr_notify_data_construct': gpr_base_open.c:143: `SIZE_MAX' undeclared (first use in this function) gpr_base_open.c:143: (Each undeclared identifier is reported only once gpr_base_open.c:143: for each function it appears in.) gpr_base_open.c: In function `orte_gpr_subscription_construct': gpr_base_open.c:186: `SIZE_MAX' undeclared (first use in this function) gpr_base_open.c: In function `orte_gpr_trigger_construct': gpr_base_open.c:222: `SIZE_MAX' undeclared (first use in this function) gpr_base_open.c: In function `orte_gpr_notify_message_construct': gpr_base_open.c:258: `SIZE_MAX' undeclared (first use in this function) *** Error code 1 (bu21) *** Error code 1 (bu21) *** Error code 1 (bu21) *** Error code 1 (bu21) *** Error code 1 (bu21) Has anyone an idea whats going on here? thanks Eugen -- | | | Dipl.- Ing. Eugen Wintersberger | | | | Department of semiconductor physics | | Johannes Kepler University | | Altenbergerstr. 69 | | A-4040 Linz, Austria | | | | Tel.: +43 732 2468 9605 | | Mobil: +43 664 311 286 1 | | | | mail: eugen.wintersber...@jku.at | | eugen.wintersber...@gmx.net| | |
Re: [O-MPI users] Compiler problems on SGI Origin
On Dec 1, 2005, at 7:31 AM, Eugen Wintersberger wrote: Hi there I have a problem on compiling openmpi (1.0.1rc4) on an SGI origin. My configure command looks like this ./configure --prefix=/scratch/hlphysik/k353443 --disable-mpi-f77 -- disable-mpi-f90 --enable-io-romio --enable-cxx-exceptions Thanks for the bug report. I was able to replicate your problem - it looks like a couple of things slipped in since I last tested on IRIX (IRIX isn't officially supported by Open MPI, and gets only light testing - usually when someone complains that it is broken). I've attached a patch that will solve most of the compile issues. This patch should apply cleanly to the 1.0 release and will be part of the 1.0.1 release. There will still be one problem (at least, on our IRIX box) with a undefined symbol dirname. The solution for now is to configure with: ./configure LIBS="-lgen" --with-wrapper-libs="-lgen" On all the platforms we officially support, dirname is in libc. On IRIX, it appears to be in libgen, causing the linking issues. I'm going to commit a fix to the trunk tonight that will solve the problem in configure, but due to time constraints it will likely not be in the 1.0.1 release. It should be in the 1.0.2 release, although I can't give you a time table as to when we will have a 1.0.2 release. Thanks for the report! Brian -- Brian Barrett Open MPI developer http://www.open-mpi.org/ irix_header_fixes.diff Description: Binary data
Re: [O-MPI users] problems with bproc cluster
John, Any progress on this? John Ouellette wrote: Hi Tim, H, nope. I recompiled OpenMPI to produce the static libs, and even recompiled my app statically, and received the same error messages. If orted isn't starting on the compute nodes, is there any way I can debug this to find out where it is failing? You could try running orted standalone to verify that it starts up correctly: bpsh orted --debug Actually, I just tried your suggestion of running ldd on one of the compute nodes (should've tried this before recompiling, I guess...). I get a strange error, which probably indicates a problem with bproc: ldd /usr/local/intel-libs/9.0/bin/orted /proc/self/fd/3: line 125: cat: command not found I know I've run ldd on a node before without problems before I don't know if this is related to my OpenMPI problems, but I'm going to have to look into it. Tim
