Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
[january-dev] Some problems with larger dataset maths


I have been having problems again with maths when things get larger.

I wrote the attached test to show the issues. Will make a merge request with the @Ignore set at the top of the file this afternoon so that if we do nothing, at least it is there for posterity.

The compiler is able to optimize certain operations, which explains the behaviour. Once things get large, the effects show. I am willing to contribute a fix or workaround to January as a merge request. Is there a suggested fix that you prefer? From what I can see it would mean editing the auto-generated code to add a test to see if the naive operation can be performed, giving the compiler a clue to do the optimization. Would such a change be acceptable? This is similar to what numpy does to optimize some cases at the hardware level I think. It is not hard in concept to do some things with basic loops if they can be but I am not sure what the test for can be is. (Perhaps non-compound is enough.)


Description: Binary data

Back to the top