[Users] MPI error in tutorial for new users
max.praveer at gmail.com
Sat Mar 28 15:14:02 CDT 2020
Dear Dr. Roland,
Thank you for your detailed response. I really appreciate the time and
thought you must've put into it.
I have followed your suggestion of removing the current Cactus build and
compiling it again but received yet another error. The problem seems to be
a mixup of multiple MPI implementations, just as you had suggested. I
couldn't finesse my way out of it, so I re-installed Ubuntu
instead(deleting everything in the process) and began from scratch. I
installed only those packages mentioned on the Jupyter tutorial and the
code compiled successfully this time.
Thank you once again for your helpful advice.
On Fri, Mar 27, 2020 at 8:42 PM Roland Haas <rhaas at illinois.edu> wrote:
> Hello Praveer Krishna,
> since mpirun was not found initially it would seem that you had not MPI
> stack installed when compiling Cactus.
> Cactus is best used with an MPI stack installed beforehand. The new
> user tutorial
> contains instructions on how to install a set up packages that simplify
> building Cactus, including MPI. Did you follow those instructions or
> started from a vanilla, freshly installed OS? Also, what OS are you
> using (OSX, Linux, Linux-Subsystem-on-Windows)?
> The easiest way to get a working copy of Cactus, if you did not install
> an MPI stack before compiling Cactus, is:
> * install an MPI stack (OpenMPI is fine)
> * remove the current Cactus build using "rm -rf configs/sim"
> * build once more
> The solution above is the suggested solution.
> If you would like to avoid recompiling (and risk triggering other
> issues instead), here's an alternative: When it detected that no MPI
> stack was installed then Cactus built its own copy of OpenMPI. It will
> have copied "mpirun" to "exe/sim/mpirun" and you can use that by making
> sure that the full path to the directory "exe/sim" appears in your PATH
> before you start the simulation. Using an mpirun from a different MPI
> stack than the one used to compile Cactus is what gives the error that
> Erik descibes in the email you found.
> > Dear ET Users,
> > I'm a beginner attempting to follow the Jupyter tutorial for new users. I
> > have been able to successfully run the helloworld.par and tov_ET.par
> > with procs set to 1 but I'm having issues when I try to use multiple
> > processors on my laptop.
> > I initially got this error <https://imgur.com/a/cuLLp79> saying "mpirun:
> > not found", following which I installed Open MPI on my system. The error
> > persisted, so I removed Open MPI and installed mpich instead. This
> > seemingly fixed the issue, but I received another error
> > <https://imgur.com/gdXSqCj> saying "CACTUS_NUM_PROCS is set to 4 but
> > are 1 MPI processes"
> > I found one
> > <
> > or two
> > <https://email@example.com/msg01591.html>
> > previous queries regarding this exact error, but the suggestions have not
> > worked for me so far. Having never worked with MPI before, I am not sure
> > what else I should try.
> > I'd be grateful for a few words of advice on this issue from anyone in
> > community.
> > Thank you in advance for your time.
> > Regards,
> > Praveer Krishna
> My email is as private as my paper mail. I therefore support encrypting
> and signing email messages. Get my PGP key from http://pgp.mit.edu .
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Users