Hi Chi and others,
when lrd crunches from a dst after introducing v3.3, it takes about ten
times longer (!) than in a case where the dst had been generated with the
old online BlastLib version. You see it in the logfiles from the CPUtime
required, around run 10004 when you introduced the new BlastLib version:
Run CPUtime dst generated
9989 331.78 August 10
9990 338.6 August 10
9991 344.05 August 10
9992 353.01 August 10
9993 325.46 August 10
9994 337.8 August 10
9995 321.58 August 10
9997 243.95 August 10
10003 1077.34 August 14
10004 1086.2 August 14
10005 11485.2 August 14
10007 11367.7 August 14
10008 11567.8 August 14
The increased file size effect (500k events instead of 250k) looks
more like a factor 3 in cputime though, introduced on September 3 (run
10826):
Run CPUtime
10794 10648.3
10795 10907.8
10796 11021
10797 10847.7
10798 10646.7
10799 10567.9
10801 9200.24
10881 2575.81
10882 48873.1
10883 45581.8
10894 25971.6
10901 36234.4
10902 33244.1
10903 33609.9
To prove that it is really the BlastLib version of the generated dst, see
e.g.:
Run CPUtime dst generated on
7239 1205.71 May 7
7240 1194.63 May 7
7241 13215.4 August 11
7242 1077.61 May7
There are still about 900 runs (> run 10000) to be recrunched. Assuming in
average 20000 cpusec per run and 24 buds, the remainder will take 750000
cpusec or 200h (=2 weeks).
There must be a reason why lrd takes so much longer now. Let's figure it
out quickly!
Regards,
Michael
--+-------------------------------------+--------------------------+ | Office: | Home: | |-------------------------------------|--------------------------| | Dr. Michael Kohl | Michael Kohl | | Laboratory for Nuclear Science | 5 Ibbetson Street | | MIT-Bates Linear Accelerator Center | Somerville, MA 02143 | | Middleton, MA 01949 | U.S.A. | | U.S.A. | | | - - - - - - - - - - - - | - - - - - - - - -| | Email: kohlm@mit.edu | K.Michael.Kohl@gmx.de | | Work: +1-617-253-9207 | Home: +1-617-629-3147 | | Fax: +1-617-253-9599 | Mobile: +1-978-580-4190 | | http://blast.lns.mit.edu | | +-------------------------------------+--------------------------+
This archive was generated by hypermail 2.1.2 : Mon Feb 24 2014 - 14:07:31 EST