A suggestion for the initial estimate of tasks 2 through N of NCI tasks for any 
particular application is to use the mean of the elapsed wall time for all 
tasks for that application.  Note that this would only apply while the reported 
% complete is 0.

-----Original Message-----
From: boinc_dev [mailto:[email protected]] On Behalf Of David 
Anderson
Sent: Thursday, April 25, 2013 4:09 AM
To: [email protected]
Subject: Re: [boinc_dev] [boinc_projects] BOINC 7.0.64/7.0.65 released for the 
public

Actually, since NCI jobs don't use much CPU,
using FLOPS info to compute time remaining is the wrong idea.

I changed things so that for NCI apps, it will use an estimate based on
elapsed time / fraction done
If fraction_done is zero, it will show ---

-- David

On 23-Apr-2013 1:32 PM, Carl Christensen wrote:
> That's cool.  One thing I should report, not particularly for this version
> but it's been in BOINC clients for quite awhile - is that it seems the
> workunit remaining calculation for non-CPU intensive apps is really off.  It
> used to be that it would report based on elapsed run-time and fraction done,
> which is sensible, i.e. QCN workunits would report (soon after starting) that
> 24 hours remain.  For awhile it reports 800 hours remain (over a month),
> which I guess may be a flop estimate I copied over from my old CPDN work
> units! _______________________________________________ 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.
_______________________________________________
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.

Reply via email to