What is the command you are using with mpirun (ie. C, N, etc.)? I am
not very familiar with the "no-schedule" option in lamboot, but the
user's guide uses "schedule=no". If the boot schema worked before, then
that is probably not the problem.
However, the user's guide states that even a "non-schedulable" node can
be used in the mpirun command if it is explicitly stated (ie. nX, cX).
If you only want to perform a computation using 4 CPU's, then I would
explicitly select the machines to use, and put them in as arguments to
mpirun.
Hope this helps.
------------------------------
Joshua Stewart
Software Engineer
Applied Optimization
joshua.stewart_at_[hidden]
-----Original Message-----
From: lam-bounces_at_[hidden] [mailto:lam-bounces_at_[hidden]] On Behalf
Of slaton
Sent: Thursday, November 11, 2004 11:25 PM
To: lam_at_[hidden]
Subject: LAM: no-schedule=1 broken in 7.11?
Hi,
My lam-bhost.def file looks like:
qcn00.domain.net no-schedule=1
qcn01 cpu=2
qcn02 cpu=2
qcn03 cpu=2
qcn04 cpu=2
(etc.)
The first host listed is the localhost where lamboot is started from.
If i start a job requesting 4 cpus, i get one thread running on qnc00,
two
on qcn01, and one on qcn02.
If i understand the no-schedule option correctly, what i should get is
two
threads on qcn01 and two on qcn02.
What have i done wrong?
Each machine has two (or 3) network interfaces. qcn00.domain.net is the
actual hostname, but also an interface name. There is no "localhost"
line
in lam-bhost.def.
Platform = AMD Opteron SMP.
thanks
slaton
_______________________________________________
This list is archived at http://www.lam-mpi.org/MailArchives/lam/
|