Since Kyle is very polite, I'll be more direct: in those (limited) cases where head-to-head comparison has been performed, Dierckx is considerably faster than Grid. Grid's performance problems have been known for some time and were already slated to be overhauled with a new package, Interpolations. It's hard for me to imagine that the new design won't close the gap, but at least until that work is done (and quite possibly beyond), Dierckx seems promising. Nice work, Kyle!
Best, --Tim On Monday, October 27, 2014 08:32:01 PM Kyle Barbary wrote: > This is an announcement of Dierckx.jl, a Julia wrapper for the dierckx > Fortran library from netlib. This is the same library underlying the spline > classes in scipy.interpolate. > > http://github.com/kbarbary/Dierckx.jl > > Note: Some of the functionality here overlaps with Grid.jl, Tim Holy's > pure-Julia interpolation package. This package is intended to complement > Grid.jl and to serve as a benchmark in cases of overlapping functionality. > I hope it will be useful to people looking for a direct equivalent of > scipy.interpolate or in any cases where spline interpolation is a major > bottleneck. > > Contributions, or advice on the correct pronunciation of "Dierckx", > gratefully accepted. > > - Kyle
