On Fri, Oct 16, 2009 at 8:16 AM, Robert Bradshaw <rober...@math.washington.edu> wrote: > > Sounds interesting. I'm curious--what kind of savings are you getting > over the Python and Cython versions (for, say, the 100-1000 bit range)? > > - Robert
It's 6-7 times faster than my prototype Cython code in that range. However, the Cython code was slightly less optimized, did more memory allocation, and also created Python objects and had some other overheads. There's probably no difference if the current code is translated to Cython line by line. I should clarify that my first benchmarks were a bit off, and I updated the page to reflect this. The code is 3 times faster than MPFR up to less than 300 bits and 2 times faster up to less than 900 bits. MPFR should also be able to regain a fair amount of that difference by switching to the same algorithm for exp, though I think my code can also be optimized further. In any case, exp is the kind of function where I expect minimal difference (because it's so fast already) and there is more potential elsewhere. Fredrik --~--~---------~--~----~------------~-------~--~----~ To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel-unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org -~----------~----~----~----~------~----~------~--~---
