Dear PDL folks, I have just uploaded PDL 2.027. Changes from 2.026:
- native support for complex numbers - thanks Ingo Schmid - define and use C macros in PP for shorter, more comprehensible XS Note that the native complex numbers are as defined in C99, and no attempt has (yet) been made to integrate this with PDL::Complex. Additionally, it’s not yet clear to me whether PDL performs better on complex numbers via PDL::Complex, or natively. Could someone more knowledgeable than me with PDL complex numbers (Luis?) come up with a benchmark, or at least a plausible, representative set of calculations to do with complex numbers so I can make a comparison? Adding this in case it’s useful, because I can’t figure out a sensible place to document it that anyone interested would be likely to find it: as discovered in fixing the temporary regression in asin(2) not returning NaN, it is highly likely that PDL functions given longlong data losing precision by converting the numbers to double is caused by this: when a function is given data not in its “GenericTypes” definition, PP defaults to the last entry in that definition. Therefore, if the function hasn’t specified it handles longlong, it will be treated as the last entry (often “D”, a double). Therefore, to fix this, all that would be needed is to add a “Q” entry to the relevant function to explicitly handle longlong. As usual, please report any problems with a pull-request fix (best), a GitHub issue with enough info to reproduce (still great), or at least a report on here (also fine!). Next steps: split out PDL::Core, then other large chunks, into their own distributions, for easier maintenance, and easier use by packagers. Best regards, Ed
_______________________________________________ pdl-general mailing list pdl-general@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/pdl-general