On Mon, Mar 16, 2009 at 1:21 PM, Ahmed Fasih <wuzzyv...@gmail.com> wrote: > > Greetings. This might not qualify as interesting or relevant, but for > my own research, I wrote a Sage app to calculate via Monte Carlo > approximation the Cramer-Rao bounds for estimating certain target > parameters from synthetic aperture radar data. MPMath's arbitrary- > precision float functionality and quadrature integration routines were > essential (I had integrands needing 30-200 decimal digits of > precision). The @parallel decorator worked fantastically on our 4- and > 8-core boxes. The app used Numpy as well. We haven't published results > from this yet but will in the coming few months.
Hi, I'm the main author of mpmath. This is awesome, and I'd love to hear more about the details. Were you actually able to use @parallel to speed up integrations in mpmath? If so, I'm curious how you did it because mpmath is not currently designed to take advantage of parallel evaluation. Vinzent Steinberg, a contributor to mpmath, has looked at implementing parallel quadrature, but he ran into some obstacles. > MPMath (and it being nicely packaged in Sage via SymPy) made this > application possible---I couldn't find support for high-precision math > in Matlab so I can't really benchmark its speed. In this case benchmarking should probably be done against Mathematica or Maple. Without having done any systematic comparisons, I think quadrature of smooth functions in the 30-200 digit range is more or less the same speed (within an order of magnitude either way) in mpmath and Mathematica. > it might not be of interest to SIAM but it > has helped us answer some of our research questions. Well, it made my day at least :-) Fredrik --~--~---------~--~----~------------~-------~--~----~ To post to this group, send email to sage-devel@googlegroups.com To unsubscribe from this group, send email to sage-devel-unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://www.sagemath.org -~----------~----~----~----~------~----~------~--~---
