PyPspline

PyPpspline is Python module for interpolating and computing derivatives of fields in up to three dimensions. Boundary conditions can be of type not-a-knot, periodic, or be expressed in terms of first or second order derivatives. Function value and derivatives can be computed on isolated points (point interpolation), on a list of scattered points (cloud interpolation) or on a structured grid (array interpolation).

PyPspline uses f2py generated code.

• Prerequisites:
  • a C compiler (for instance gcc)
  • a Fortran 90 compiler (to build PSPLINE).
  • NetCDF
  • Python
  • NumPy
  • ScientificPython (to read and write NetCDF files)
• Download
• Documentation:
  • 1-D spline (Float32)
  • 1-D spline (Float64)
  • 2-D spline (Float32)
  • 2-D spline (Float64)
  • 3-D spline (Float32)
  • 3-D spline (Float64)
  • Interface to OpenDX Visualization
• Examples
• Links:
  • Matlab(R) users can download ezsplinetoolbox. Also based on PSPLINE, ezsplinetoolbox is a MEX-based toolbox developed by Remik Ziemlinski with comparable functionality to PyPspline.
• Performance:
  In general, most of the number crunching is done in PSPLINE so the overall pypspline performance will closely reflect the quality of your Fortran compiler. For the test program mtest_pypspline.py comprising of 10+ million interpolations on a non-uniform mesh, the following timings apply on a 2.4Ghz Intel Pentium 4 box:
  • Intel ifort Version 8.0 (-O1): 7.4 secs
  • Lahey/Fujitsu Fortran 95 Express Release L6.10b (-O --trap --trace): 8.5 secs
  • The Portland Group, Inc. pgf90 3.2-4 (-O -Mdalign -Mdefaultunit -Ktrap=fp): 8.9 secs
• Mailing list