[mira_talk] Re: memory seetings
- From: David Lambert <dlamber2@xxxxxxxxxxxxxxxxxx>
- To: mira_talk@xxxxxxxxxxxxx
- Date: Mon, 21 Dec 2009 16:44:25 -0500
Hi all,
I am having similar memory management problems running the current
version 2.8.3 on a cluster, and I would like to try these parameters
((-AS:amm:kpmf:mps), but I can't find them in the manual or usage
docs, can someone point me to where they are discribed? Or are they
only for an earlier version?
Thanks,
Dave
At 01:31 PM 12/7/2009, you wrote:
On Montag 07 Dezember 2009 Charles Imbusch wrote:
> recently I was running an EST assembly with MIRA and the
> computer crashed. MIRA's last words ...
> [...]
> Mem avail: -4084002406
Hi Charles,
which version is that? Seems like a bug which was fixed lately. Can
you please
try out the latest one? That would be
http://www.chevreux.org/tmp/mira-3rc4d.tar.bz2
or compiled in
http://www.chevreux.org/tmp/mira_3rc4d_dev_linux-gnu_x86_64_static.tar.bz2
> To me it seems that MIRA ran out of memory. Two questions I have in mind:
> 1) is it possible to tell MIRA to terminate if it recognizes, that it's
> running out of memory? So it would not disturb any other running processes
> on that machine.
The definition of "running out of memory" is a bit fuzzy: MIRA has to rely on
what the operating system is willing (or able) to give. If MIRA asks for a
chunk of memory and the OS says "there you have it", the MIRA will take it
even if the OS is swapping out tons of other stuff for it.
If the OS says "nope, not possible" then MIRA will say good-bye and bail out.
> 2) it it possible to tell MIRA to work with a certain amount of memory?
As a matter of fact, yes. Have a look at the parameters for automatic memory
management (-AS:amm:kpmf:mps) which have a mild barrier turned by default.
Please keep in mind that these barriers are only valid for parts of the
algorithms (exactly those in your mail) and also MIRA won't be able to do
magic: if the reads alone eat away 10 GB RAM and you ask MIRA to work in 12
GB, be prepared for some unhappy MIRA running at minimum mem but maximum time.
And then there's always the possibility to limit processes via the shell
'limits' command which have no mercy if a program wants even one byte more
than allowed :-)
> ==============
> Skim summary:
> accepted: 593845227
> possible: 846197766
> permbans: 46674712
> Hits chosen: 91677407
Uh, what's that for a project? 593 million accepted SKIMs, still 91 million
chosen? Somewhat bigger, isn't it?
Regards,
Bastien
--
You have received this mail because you are subscribed to the
mira_talk mailing list. For information on how to subscribe or
unsubscribe, please visit http://www.chevreux.org/mira_mailinglists.html
-------------------------------------------------------------------------
David Lambert
Assistant Professor
Department of Biology
University of Rochester
http://www.rochester.edu/College/BIO/labs/LambertLab/index.html
585 273-2482 (office)
585 275-7521 (lab)
585 275-2070 (fax)
Shipping address:
University of Rochester
Department of Biology
213 Hutchison Hall
River Campus
Rochester, NY 14627-0211
Mailing address:
University of Rochester
Hutchison Hall
RC Box # 270211
Rochester NY 14627-0211
--
You have received this mail because you are subscribed to the mira_talk mailing
list. For information on how to subscribe or unsubscribe, please visit
http://www.chevreux.org/mira_mailinglists.html
Other related posts:
