|
From: Eric G. <da...@fr...> - 2011-05-24 01:43:13
|
Dear all,
I just like to know whether it's possible to disable the MPI feature
when building Valgrind.
I got an error regarding MPI library like
/make[2]: Entering directory
`/home/germaneau/soft/valgrind/valgrind-3.6.1/mpi'
mpicc -g -O -fno-omit-frame-pointer -Wall -fpic -m64 -fpic -shared -o
libmpiwrap-amd64-linux.so libmpiwrap_amd64_linux_so-libmpiwrap.o
/usr/lib64/gcc/x86_64-slackware-linux/4.4.4/../../../../x86_64-slackware-linux/bin/ld:
/usr/lib64/libmpich.a(setbotf.o): relocation R_X86_64_32 against
`mpifcmb2_' can not be used when making a shared object; recompile with
-fPIC
/usr/lib64/libmpich.a: could not read symbols: Bad value
collect2: ld returned 1 exit status
make[2]: *** [libmpiwrap-amd64-linux.so] Error 1
make[2]: Leaving directory
`/home/germaneau/soft/valgrind/valgrind-3.6.1/mpi'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/home/germaneau/soft/valgrind/valgrind-3.6.1'
make: *** [all] Error 2/
And I actually don't need to debug MPI code right now.
Thanks,
Éric.
--
/Be the change you wish to see in the world
/ --- Mahatma Gandhi ---
Dr. Éric Germaneau <mailto:ger...@gu...>
College of Physical Sciences
Graduate University of Chinese Academy of Sciences
Yuquan Road 19A
Beijing 100049
China
/Please, if possible, don't send me MS Word or PowerPoint attachments
Why? See: http://www.gnu.org/philosophy/no-word-attachments.html /
--
/Be the change you wish to see in the world
/ --- Mahatma Gandhi ---
One Human Family <http://onehumanfamily.info/>
/Please, if possible, don't send me MS Word or PowerPoint attachments
Why? See: http://www.gnu.org/philosophy/no-word-attachments.html /
|
|
From: John R. <jr...@bi...> - 2011-05-24 02:36:26
|
> I just like to know whether it's possible to disable the MPI feature when building Valgrind. Here are two things to try: 1) Uninstall MPI (at least 'mpicc') then re-build valgrind, starting with ./autogen.sh . See "Building and installing it" in the README file. or, 2) Remove the "mpi \" line from Makefile.am, then re-build valgrind, starting with ./autogen.sh . -- |
|
From: Eric G. <da...@fr...> - 2011-05-24 03:46:14
|
Thank you,
I removed mpi, build valgrind, and reinstall mpi.
It works.
Best,
Éric.
On 05/23/2011 10:36 PM, John Reiser wrote:
>> I just like to know whether it's possible to disable the MPI feature when building Valgrind.
> Here are two things to try:
>
> 1) Uninstall MPI (at least 'mpicc') then re-build valgrind, starting with
> ./autogen.sh . See "Building and installing it" in the README file.
>
> or,
> 2) Remove the "mpi \" line from Makefile.am, then re-build valgrind,
> starting with ./autogen.sh .
>
--
/Be the change you wish to see in the world
/ --- Mahatma Gandhi ---
One Human Family <http://onehumanfamily.info/>
/Please, if possible, don't send me MS Word or PowerPoint attachments
Why? See: http://www.gnu.org/philosophy/no-word-attachments.html /
|
|
From: Dave G. <go...@mc...> - 2011-05-24 16:04:59
|
On May 23, 2011, at 9:36 PM CDT, John Reiser wrote: >> I just like to know whether it's possible to disable the MPI feature when building Valgrind. > > Here are two things to try: > > 1) Uninstall MPI (at least 'mpicc') then re-build valgrind, starting with > ./autogen.sh . See "Building and installing it" in the README file. > > or, > 2) Remove the "mpi \" line from Makefile.am, then re-build valgrind, > starting with ./autogen.sh . Those are probably the most labor-intensive ways that you can do it. It's usually easier to just pass "--without-mpicc" to configure and use the normal build process in all other respects. -Dave |
|
From: John R. <jr...@bi...> - 2011-05-24 17:01:41
|
> Those are probably the most labor-intensive ways that you can do it. It's usually easier to just pass "--without-mpicc" to configure and use the normal build process in all other respects. 1. No documentation: no mention of MPI in README*, no explicit list of packages that might be eligible for --without-*, etc. 2. The default "-with-mpicc" does not work (it fails to detect the unusable state on the original poster's machine), so the override "--without-mpicc" should be regarded with suspicion. -- |
|
From: Dave G. <go...@mc...> - 2011-05-24 17:31:35
|
On May 24, 2011, at 12:02 PM CDT, John Reiser wrote: >> Those are probably the most labor-intensive ways that you can do it. It's usually easier to just pass "--without-mpicc" to configure and use the normal build process in all other respects. > > 1. No documentation: no mention of MPI in README*, no explicit list of packages > that might be eligible for --without-*, etc. > > 2. The default "-with-mpicc" does not work (it fails to detect the unusable > state on the original poster's machine), so the override "--without-mpicc" > should be regarded with suspicion. I didn't mean it as any kind of personal slight, and I completely agree that this is both under-documented and slightly buggy. I just wanted to get the easiest solution that I know actually works into the mail archives. I've got ~1/2 of a patch to fix the mpicc detection already written that has been shelved for a while, perhaps I should resume work on it. If I can get that working then none of these solutions should be necessary. -Dave |
|
From: Dave G. <go...@mc...> - 2011-05-24 22:53:03
|
On May 24, 2011, at 12:31 PM CDT, Dave Goodell wrote: > I've got ~1/2 of a patch to fix the mpicc detection already written that has been shelved for a while, perhaps I should resume work on it. If I can get that working then none of these solutions should be necessary. FYI: https://bugs.kde.org/show_bug.cgi?id=274078 -Dave |