Dear Xavier,
You were right ! The problem was with shared memory.
The original mpich was installed in that way.
I installed my own mpich (mpich2) and everything is fine now.
I can run siesta in parallel mode !
Thanks a lot !
Regards,
Mariusz.
---
Mariusz,
.bash_profile is only called for login shells, not for rsh or ssh non-login
commands. Try putting it in .bashrc.
Lucas
On 5/31/07, Mariusz Krawiec <[EMAIL PROTECTED]> wrote:
Dear Heribert,
Of course. I put it into ma .bash_profile file to have it on all nodes.
It didn't help.
Do yo
Additional comment:
on some shells "setenv" command should be used instead of "export":
setenv P4_GLOBMEMSIZE 5000
to be sure that the variable is really set type "export" or "setenv"
without parameters and find
P4_GLOBMEMSIZE in the list.
On 31/05/07, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wr
Mariusz,
I don't know how your machine is set up, but it seems at least from the
message
you get, it didn't work with .bash_profile.
Maybe you try it as Yurko suggests:
> you're quite right. I also have had this issue for big cells. just try
>export P4_GLOBMEMSIZE=5000
>or smth like this, b
Dear Heribert,
Of course. I put it into ma .bash_profile file to have it on all nodes.
It didn't help.
Do you think I should do this before compilation of siesta or just before
running?
All the best,
Mariusz.
--
Dr Mariusz K
you're quite right. I also have had this issue for big cells. just try
export P4_GLOBMEMSIZE=5000
or smth like this, before running siesta - it should work.
On 31/05/07, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote:
Dear Mariusz,
Did you try what the error message suggests?
InitMesh: Mesh c
Dear Mariusz,
Did you try what the error message suggests?
InitMesh: Mesh cutoff (required, used) = 225.000 244.755 Ry
p0_5146: (60.036989) xx_shmalloc: returning NULL; requested 4976688 bytes
p0_5146: (60.036989) p4_shmalloc returning NULL; request = 4976688 bytes
=> You can increase the amount
Dear Mariusz,
It seems to me from the error message that you have compiled mpi with
the shared memory device (ch_shmem ?). I seem to remember that the
default limits on the memory that could be shared between processors was
too low.
In any case, recompiling mpi without the use of shared mem
Dear Marcos,
Thanks for your suggestions.
Changing optimization from O3 to O2 didn't help but the problem, as you
suggested, is probably in mpi or libraries. At the momemnt I can run
parallel siesta to the point where the memory is checked. Then it crashes.
Here is a part of out file:
InitM
9 matches
Mail list logo