We'll take a look at the prefix behavior. As to the btl, you can always just force it: for example, -mca btl sm,self,openib would restrict it to shared memory and IB.
On Mar 2, 2012, at 7:48 AM, Yiguang Yan wrote:
> Hi Jeff, Ralph--
> Please let me follow the thread, here are what I observed:
> (All with the same test script test.bash I post in my previous emails, so run with app file fed to mpirun command.)
> (1) If I put the --prefix in the app file, on each line of it, it works fine as Jeff said.
> (2) Since in the manual, it is said that the full path of mpirun is the same as setting "--prefix". However, with app file,
> this is not the case. Without "--prefix" on each line of the app file, the full path of mpirun does not work.
> (3) With "--prefix $adinahome" set on each line of the app file, it is exclusively put, on each node, the
> $adinahome/bin into the PATH, and $adinahome/lib into the LD_LIBRARY_PATH(not the $adinahome/lib64 as said
> in mpirun manual(v1.4.x)). The envars $PATH and $LD_LIBARARY_PATH set in test.bash script only affect the
> envars on the submitting node(gulftown in my case). No $PATH or $LD_LIBRARY_PATH is passed to slave nodes
> even if I use "-x PATH -x LD_LIBRARY_PATH", either fed to mpirun or put on each line of the app file. I am not sure
> if this is intended, since "--prefix" overwrite the effect of "-x" option, this is different from what I see from the mpirun
> man page.
> I have another question about the btl used for communication. I noticed that rsh is using the tcp for connection, I
> understand that tcp may be used for initial connection, but how can I know that openib(infiniband) btl is used for my
> data communication? Any explicit way?
> users mailing list