Hi again,

so with a fresh update I also now see the recon being x3 slower than this morning. My full recon job takes 140ms/event instead of 50ms/event. 

Can people who have committed code today take a look at this and maybe revert to find the source and if it’s needed? 

Maybe first verify that it’s slow.

/Pelle


On Sep 24, 2015, at 9:13 AM, Alessandra Filippi <[log in to unmask]> wrote:


I’m not sure if I think it runs so much slower. It’s using 100% of one of my cores for sure. Perhaps you can try at a slac node and see if it behaves the same or locally if you are running on a slac machine. Note that if you use a  shared slac machine I’ve seen those sometimes be very slow depending on load.


ok. Since for me it's using just the 22% (while until yesterday it was completely filling the core) I will try to reboot my machine, first thing. Tomorrow, as I want at least one run to be finished to compare results, and it started this morning and still runs. Well, runs... I'd say, walks (lame) ;-)
thanks





/Pelle



On Sep 24, 2015, at 8:42 AM, Alessandra Filippi <[log in to unmask]> wrote:

When you says it “does more” you mean it takes more time? Or do you see it doing something else?

I’m looking into it.

I cannot tell if it's writing more on the slcio file as long as the run isn't finished and I can compare full run sizes. The gbl file dimension looks comparable to the old one (for 1/3 of processed events).
So, actually I don't know it it's doing more. For sure, it's doing it more slowly...
can any of you try on your machines if experiences the same slowdown?
thanks.



/Pelle


On Sep 24, 2015, at 8:19 AM, Alessandra Filippi <[log in to unmask]> wrote:

yes
how do I know which strategies are running??
I just need what was running for pass2, but it looks like the reconstruction it's doing much more now... and it takes forever



On Thu, 24 Sep 2015, Hansson Adrian, Per Ola wrote:


Are you doing evio to lcio?


On Sep 24, 2015, at 8:04 AM, Hansson Adrian, Per Ola <[log in to unmask]> wrote:

Perhaps more tracking strategies are running?


On Sep 24, 2015, at 4:30 AM, Alessandra Filippi <[log in to unmask]> wrote:

Hi all,
after a fresh update yesterday I realized that the reconstruction speed decreased awfully: yesterday I was reconstructing 1000 events in 25 seconds, now it takes more than 3 minutes. Also, while before 100% of the cpu was taken, now not more than 25% is used (and I'm not doing anything else on the pc). The steering file I am using is EngineeringRun2015FullReconGbl2.lcsim with GhostHits removed, the "well-behaving" code (whose version number of course I lost after the last update) was next-to-last updated after the start of pass2 production, no later than one week ago.

What happened in the meanwhile?
thanks,
Alessandra

########################################################################
Use REPLY-ALL to reply to list

To unsubscribe from the HPS-SOFTWARE list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=HPS-SOFTWARE&A=1

########################################################################
Use REPLY-ALL to reply to list

To unsubscribe from the HPS-SOFTWARE list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=HPS-SOFTWARE&A=1



########################################################################
Use REPLY-ALL to reply to list

To unsubscribe from the HPS-SOFTWARE list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=HPS-SOFTWARE&A=1

########################################################################
Use REPLY-ALL to reply to list

To unsubscribe from the HPS-SOFTWARE list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=HPS-SOFTWARE&A=1


Use REPLY-ALL to reply to list

To unsubscribe from the HPS-SOFTWARE list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=HPS-SOFTWARE&A=1