Hi Troels,
This is a truly impressive speed up. However I think you are capable
of pushing this much, much further. The key will be the R20 and dw
data structure creation:
- http://thread.gmane.org/gmane.science.nmr.relax.devel/6135
- http://thread.gmane.org/gmane.science.nmr.relax.devel/6135/focus=6136
The elimination of the pA_arr and kex_arr structures:
- http://thread.gmane.org/gmane.science.nmr.relax.scm/21481/focus=6133
And the fast handling of missing data:
- http://article.gmane.org/gmane.science.nmr.relax.devel/5726
I also have ideas for avoiding temporary data structure creation and
garbage collection in the lib.dispersion modules which will also give
you speed boosts. I think you are close to being able to drop not
from 43 seconds to 2, but from 42 seconds to a few hundreds of
milliseconds!
Regards,
Edward
On 8 June 2014 20:26, Troels E. Linnet <NO-REPLY.INVALID-ADDRESS@xxxxxxx>
wrote:
Follow-up Comment #23, task #7807 (project relax):
The time spent in cr72.py goes from:
38.743 0.000 43.050 0.000 cr72.py:100(r2eff_CR72)
to
1.403 0.001 2.184 0.002 cr72.py:101(r2eff_CR72).
This is awesome!
The time in:
5.241 0.005 54.380 0.054 relax_disp.py:456(calc_CR72_chi2)
15.091 0.015 44.520 0.045 relax_disp.py:496(calc_CR72_chi2)
This is triple time as much, which is probably related to the increased
looping, to get the disp point structures back.
The NEW time consumer is:
1510108 20.473 0.000 20.473 0.000 {numpy.core.multiarray.array}.
_______________________________________________________
Reply to this item at:
<http://gna.org/task/?7807>
_______________________________________________
Message sent via/by Gna!
http://gna.org/
_______________________________________________
relax (http://www.nmr-relax.com)
This is the relax-devel mailing list
relax-devel@xxxxxxx
To unsubscribe from this list, get a password
reminder, or change your subscription options,
visit the list information page at
https://mail.gna.org/listinfo/relax-devel
Re: [task #7807] Speed-up of dispersion models for Clustered analysis