pityka

Pushed a commit with updated jmh runs. @tnielens You asked a few weeks ago about those macro based inlined ops. It turns out there are some jmh benchmarks in the...

Very interesting. Thank you! Unfortunately the Op is very likely to be strongly polymorphic in real world client code ie that virtual call ordeoptimization will occur unless inlined by the...

Thank you for the investigation. If I understand that means that if we sprinkle enough annotations and the very end user of the final application adds -opt:inline:** (and potentially -opt:local...

In this latest you compile saddle with -opt:inline:org.saddle.** ?

Out of the three alternatives (macro, inline later, inline within saddle) which one do you prefer? If we can solve this problem with inlinin then I would remove the macro....

Thanks! I agree this is a good way forward. I can probably find some time for this during summer or let me know if you will contribute this refactor to...

Another possibility is to do nothing and hope that scala3 will have a tasty based whole program optimizer one day. Hard to believe that ad hoc template polymorphism (which the...

I found some rumors that the JVM will have automatic specialisation in the mid future. This should be weighed in when considering the code generation route. https://github.com/lampepfl/dotty/issues/11114#issuecomment-1019956981

Thanks for finding this. What do you think of Panel? In the last 10 years I never used it.

No, in this fork Frame is Frame[RowIndextype, ColIndexType, ValueType] thus the values in the frame must be of a single type. For the categorical variables you can overcome this by...