We may have to set a time limit on how long to wait for a checkpoint. Some
applications still never checkpoint. At least one application does not
checkpoint, and loses hours of work when it is left in memory when
suspended.
jm7
David Anderson
<[email protected]
ey.edu> To
Sent by: [email protected]
<boinc_dev-bounce cc
[email protected] BOINC Developers Mailing List
u> <[email protected]>,
[email protected]
Subject
10/26/2009 02:04 Re: [boinc_dev] [boinc_alpha] CPUs
PM Idle BM v6.10.16 Vista64
[email protected] wrote:
> The application DOES specify how many threads it is going to use.
> Unfortunately, it is fixed at the time the task is created on the server.
Actually it's fixed when the job is sent to a host;
i.e. it can reflect the number of cores on the host.
In response to Raistmer's idea of starting MT jobs
overcommitted and dropping out the 1-CPU jobs as they checkpoint:
that's a good idea and I'll put it on my list.
It should be a small code change in case anyone wants to
tackle it first (watch out for unintended consequences, though).
-- David
_______________________________________________
boinc_dev mailing list
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.
_______________________________________________
boinc_dev mailing list
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.
