GPU speed up?

[PeterPirog]

Is possible to speed up optapy calculations by usage of CUDA GPU?

[Christopher-Chianelli]

You probably could use CUDA in an @easy_score_calculator (https://www.optapy.org/docs/latest/score-calculation/score-calculation.html#easyPythonScoreCalculation), but then there will be frequent transfers between CPython and Java, which will probably eat all possible performance gains (and then some). With a constraint provider, incremental score calculation is used (https://www.optapy.org/docs/latest/score-calculation/score-calculation.html#incrementalScoreCalculation) and functions are converted into Java bytecode (if possible). CUDA would probably be suitable if you have a constraint that cannot be calculated incrementally, but it would need to be benchmarked (I strongly suspect a constraint provider will still beat it, since transfer between CPython and Java have a lot of overhead).

[PeterPirog]

@Christopher-Chianelli Thank You for the answer. I asked because I wonder if is possible to use JCuda http://javagl.de/jcuda.org/tutorial/TutorialIndex.html#Introduction , typically I use CUDA when I work with neural nets Tensorflow framework.