Ojalgo #57
Ojalgo #57
Conversation
return to previous ojalgo branch
saved roughly tenth of second
tested and Ojalgo arrays not the best for simply setting and getting array elements, thus used double[]
|
getting there! now down to < 0.5 sec; Elapsed time = 0.487 seconds for 1000 realizations of total = 47000 see what it will take to to serialize the MatrixStore in EnsembleRecord - this is a lead: http://javadox.com/org.ujmp/ujmp-ojalgo/0.3.0/serialized-form.html |
also avoided serialized issue with conversion to array
|
speed is improving well: |
Elapsed time = 0.197 seconds for 1000 realizations of total = 50000 Error function = 2978.482 Change Log Ratio: 152.0 of 152.0 accepted (97.535% total) Change Intensity: 341.0 of 512.0 accepted (66.744% total) Change DF Gain: 56.0 of 74.0 accepted (76.736% total) Change Baseline: 32.0 of 92.0 accepted (34.373% total) Noise: 30.0 of 170.0 accepted (18.913% total) Intervals: in microseconds, each from prev or zero time till new interval Interval1 43 Interval2 53 Interval3 5 Interval4 88 Interval5 6 next fix catching error
Possible fix
probably need to not return a null
|
I am merging another PR from myself first - did some refactoring and tweaking as I found a bug in the logic. Once you fix the conflicts, I am willing to merge this. looking faster! |
|
Currently broken wait for next commit |
Need to clean code of tests I was doing, atm slower then previous baseline. Does blockIntensities in DataModellerOutputRecord need to be a 2 dimensional array or 1? Its treated as both in different parts of the code. Current implementation I changed it too is 1 however I'm guessing when multiple blocks are involved will need to be a 2 dimensional array.
|
@RoyalBoggs - good work! I am merging and then doing a pull request for my additional work including a cleanup of the driverExperiement ... stand by. Then do another pass and make sure we have all the speedups we can get - see my intervals below: %%%%%%%%%%%%%%%%%%%%%%% Tripoli in Java test %%%%%%%%%%%%%%%%%%%%%%% |
Conversion from Jama library to Ojalgo for matrix multiplication