On Oct 30 02:06, Bogdan Costescu wrote: > On Mon, 29 Oct 2007, Tom Mitchell wrote: > > > Debian is not supported by the PathScale compiler. > > Thanks for your answer. In the meantime I have found a workaround for > my situation, but I'll answer for the sake of the e-mail archives. > > > which one is is being used? The new Debian etch 64bit has minimal > > support for 32 bit user space programs by default. It may be > > necessary to install 32 bit libs and or compatability libs to get it > > to run correctly. > > It is indeed the 64bit version. However I was able to use the > PathScale compiler to compile other applications without problems, so > it was a specific interaction, probably with the threading support, > that was triggered by Open MPI which showed up. I guess that if the > necessary libs would be missing, the compiler wouldn't work at all or > ldd would show missing libs for the other executables that the > compiler produced. > > > I do not know which gcc compiler suite Debian GCC is based on. > > gcc version 4.1.2 20061115 (prerelease) (Debian 4.1.1-21) > > > In RHEL4 compare and contrast gcc4-4.1.1-53 .vs. gcc-3.4.6-8. The > > Pathscale compiler is recently rebased to also work with gcc4 based > > systems. > > Should I understand from this statement that the PathScale compilers > version 3.0 are only compatible with gcc 3.4 and not with gcc 4.1 > based systems ?
No but there are choices in how pathcc can be invoked and thus what frontend and other chunks are active. Example on CentOS release 4.5 (RHEL4.5). $ pathcc -v QLogic PathScale(TM) Compiler Suite: Version 3.0 Built on: 2007-02-22 13:05:15 -0800 Thread model: posix GNU gcc version 3.3.1 (PathScale 3.0 driver) $ pathcc -v -gnu4 QLogic PathScale(TM) Compiler Suite: Version 3.0 Built on: 2007-02-22 13:05:15 -0800 Thread model: posix GNU gcc version 4.0.2 (PathScale 3.0 driver) Try adding -gnu4 to the "C" flags and see if the segmentation faults exit. If .../bin/orterun segfaults > > > Do foundation Open MPI tools like .../bin/orterun compile and run > > well enough to return errors? > > They compile fine, but they give segmentation faults when run. > > > Scan the eko man page for PathScale compiler specific options. > > I don't quite understand what you mean here... especially what > 'specific options' refer to. Since major parts of the PathScale are gcc or gcc compatable the most common flags are in the GCC suite and man pages. Flags like -gnu4 are covered in the eko man page and are specific to the pathscale compiler. > > > ompi_info does not capture compiler flags so report your configure > > line, make line and environment variables. > > One of my earlier messages already included the configure line; the > environment is only modified to set CC=pathcc, etc. and there are no > options on the make line; just like the documentation says that I > should do... Thanks. Since you are on Debian it may be interesting to see what 'configure' found and what the generated compile line for orterun and it's library functions ended up being. This can include the "libtool" lines. > > -- > Bogdan Costescu > > IWR - Interdisziplinaeres Zentrum fuer Wissenschaftliches Rechnen > Universitaet Heidelberg, INF 368, D-69120 Heidelberg, GERMANY > Telephone: +49 6221 54 8869, Telefax: +49 6221 54 8868 > E-mail: bogdan.coste...@iwr.uni-heidelberg.de > _______________________________________________ > devel mailing list > de...@open-mpi.org > http://www.open-mpi.org/mailman/listinfo.cgi/devel -- T o m M i t c h e l l Host Solutions Group, QLogic Corp. http://www.qlogic.com http://support.qlogic.com
