URL: <http://gna.org/task/?7793> Summary: Speed-up of dispersion models Project: relax Submitted by: tlinnet Submitted on: Sat 10 May 2014 04:49:54 PM UTC Should Start On: Sat 10 May 2014 12:00:00 AM UTC Should be Finished on: Sat 10 May 2014 12:00:00 AM UTC Category: relax's source code Priority: 5 - Normal Status: None Percent Complete: 0% Assigned to: tlinnet Open/Closed: Open Discussion Lock: Any Effort: 0.00 _______________________________________________________ Details: This was discussed in: http://thread.gmane.org/gmane.science.nmr.relax.devel/5718 It will consist of two things. Instead of looping over each point and check math domain errors, and then return each value to backcalc list, the whole numpy array is calculated in one go. Thereafter there will be a check for the array. If any array holds nan/Nan, the whole array is filled with 10**20 and returned. But still keep the looping over each point when returning values. This testing can be done in the trunk. The next step is to pass back the whole array of values, to the target function, which then updates the class list in one go. This should be done in a separate branch, to perform speed testing. _______________________________________________________ Reply to this item at: <http://gna.org/task/?7793> _______________________________________________ Message sent via/by Gna! http://gna.org/