Menu
▾
▴
Re: [Swiftng-users] long run times
|
From: Tom S. <ts...@sa...> - 2009-04-27 15:23:34
|
It got hung up in cross-talk correction, by the look of it. Odd, that's usually quite quick, IIRC. There are no Memstats lines in the output. What's become of them? --TS Nava Whiteford wrote: > Though it says it's not using the reference, it maybe processing a large > number of smaller contigs. I've corrected this in the latest checkin of > swift but it would be worth trying this tile with no reference > specified. > > Other than that, it does look like the tile is generating an abnormally > large number of clusters. If possible can you send me a sample tile? > > On Mon, Apr 27, 2009 at 04:41:51PM +0200, Bernd Jagla wrote: > >> Here are the associated output-files... >> The one with the worst times unfortunately hasn't gotten any stdout/err for >> some odd reason. Here is second "best"... >> >> >> -----Original Message----- >> From: Nava Whiteford [mailto:ne...@sg...] >> Sent: Monday, April 27, 2009 4:30 PM >> To: Bernd Jagla >> Cc: 'Tom Skelly'; Swi...@li... >> Subject: Re: long run times >> >> How big is dm3_ucsc.fa? Swifts aligner is very slow and only designed to >> deal with 5000bp sequences. >> >> ->Dm3 is not being used (too big)... >> >> Are you saying it took 39 hours using the parameters settings in the >> svn? That would be a bit odd, if you can send me the stdout/err that >> would be useful. >> >> ->See attached >> >> Sorry, for the delay in getting back to you on your xml questions. I've >> not forgotten! Just snowed under. >> >> ->no worries... it is cold here in Paris too today, but not that cold ;)... >> Good luck !! >> >> Nav >> >> On Mon, Apr 27, 2009 at 04:07:06PM +0200, Bernd Jagla wrote: >> >>> Nava, >>> >>> >>> >>> Here is a list of parameters I tested and associated run-times (143,343.85 >>> sec = 39hrs!!!!) for one tile. >>> >>> Could you please comment on where all the time might be spent? >>> >>> >>> >>> I would like to understand the problem to be able to optimize our >>> >> protocols. >> >>> >>> >>> It doesn't really seem to be possible to reduce the time without loosing >>> >> too >> >>> many clusters. >>> >>> >>> >>> Thanks, >>> >>> >>> >>> Bernd >>> >>> >>> >>> Time = [sec] user time from time command >>> >>> Wc pf = number of lines in *.pf files >>> >>> Wc nonpf = numbe of lines in *.nonpf files >>> >>> There is only one parameter changed compared to the standard (normal) call >>> of the program. >>> >>> >>> >>> idx parameter time wc pf wc nonpf >>> >>> 1 normal 143,343.85 182,256 1,720,580 >>> >>> 2 --ref dm3_ucsc.fa 27,681.68 110,988 1,809,448 >>> >>> 3 --correlation_cc_subimage_multiplier 20 114,063.20 179,724 >>> 1,695,836 >>> >>> 4 --correlation_cc_subimage_multiplier 10 34,372.92 154,840 >>> 1,751,684 >>> >>> 5 --correlation_cc_subimage_multiplier 40 4,732.95 116,920 >>> 1,801,616 >>> >>> 6 --threshold_window 3 2,311.50 131,260 664,312 >>> >>> 7 --threshold_window 4 2,226.00 106,416 261,468 >>> >>> 8 --segment_cycles 2 3,343.96 143,712 1,301,216 >>> >>> 9 --segment_cycles 1 2,266.62 142,016 772,496 >>> >>> >> -- >> Nav >> >> Work: 01865 854873 >> Mob : 07518-358405 >> > > > > -- The Wellcome Trust Sanger Institute is operated by Genome Research Limited, a charity registered in England with number 1021457 and a company registered in England with number 2742969, whose registered office is 215 Euston Road, London, NW1 2BE. |
