[mvapich-discuss] New vapi-multirail user experiences

Abhinav Vishnu vishnu at cse.ohio-state.edu
Tue Aug 22 11:24:22 EDT 2006


Dr Panda,

I think the "make install" for vapi_multirail may have got broken
after 0.9.5 +, since we did not focus on it much. We did not put the
mpirun patch for vapi_multirail.

Should I just reply to this mail thanking for letting us know about
these problems, and we will fix this in the upcoming release. Please let
me know.

Thanks,

-- Abhinav


-------------------------------
Abhinav Vishnu,
Graduate Research Associate,
Department Of Comp. Sc. & Engg.
The Ohio State University.
-------------------------------

On Tue, 22 Aug 2006, Eric A. Borisch wrote:

> Morning all,
>
> I'm not sure if this is the proper forum for this, but I wanted to pass
> along some problems I ran into while installing mvapich 0.9.8 vapi
> multirail.
>
> I'm installing on ROCKS 4.1, EM64T dual 3.4 GHz machines, and I'm using icc
> instead of gcc, so there are a few extra libs to include for the Fortran
> linker, -O3 -xP optimizations, but on other major changes to the install
> script.
>
> I did not run into any of these problems when installing the non-multirail
> varsion, BTW. I started with a fresh copy of the source for the multirail
> build.
>
>
>    1. tarch had disappeared by the time the install script ran.
>    Re-extraction of the original archive over the build directory with -k (no
>    replacements) fixed this.
>    2. mprun.vapi_multirail and mpirun.vapi_multirail.args were not
>    installed. Copying the .args and mprun.vapi_multirail.in files from
>    mpid/vapiu_multirail seemed to fix this.
>
>
> Any suggestions on if what I've done is appropriate / if I caused any of
> these problems? The multirail system seems to be up and running correctly
> now (my BW's are way up!)
>
> Thanks,
>  Eric Borisch
>  Mayo Clinic
>  Radiology Research
>



More information about the mvapich-discuss mailing list