richard, please keep the discussion on list as some of the issuesbeing discussed might be of general interest
On 28 May 2012, at 18:13, richard wrote: > I see a speed up of close to six times on four cores. > Your method of 'vectorisation' is of interest. I would like > to be able to standardise this as much as I can. My expectation is > that the method could be adapted to support improvements in the Model. > If you don't mind, a couple of questions ... > Some code in postprocess.m seems similar to the code in increment.m > My instincts push me toward asking postprocess.m to provide a matrix, > call it A(16,16), as part of the Result, and to then pass A to > increment.m I am not sure whether that is possible, postprocess must be run AFTER the whole dynamic simulation using "increment" as the increment function has been run. So I am not sure you can use results from "postprocess" in "increment". You can for sure do the opposite, but the gain in efficiency would be negligible so I didn't bother ... It does reduce code duplication, though, so I do recommend making this change. > Is there something in pararrayfun or parcellfun I should worry about > before doing this? vectorization and parallelization were applied to your example completely independently: you are solving ten times the same system of ODEs with different model parameters. each run is completely idependent from the others so all the ODEs may be solved in parallel. what you do in each run will not affect others. so you can safely change how "increment" and "postprocess" interact. > TIA c. ------------------------------------------------------------------------------ Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ _______________________________________________ Octave-dev mailing list Octave-dev@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/octave-dev
