Differences between revisions 6 and 10 (spanning 4 versions)
⇤ ← Revision 6 as of 2008-11-09 11:05:16 →
Size: 1208
Editor: was
Comment:
← Revision 10 as of 2008-11-14 13:42:12 → ⇥
Size: 1738
Editor: anonymous
Comment: converted to 1.6 markup
Deletions are marked like this. Additions are marked like this.
Line 6: Line 6:
    1. Stuff still computed using gp:     1. Modularize the code (break into the following classes and sort dependencies)
        * LSeries (Is already in there. Have to free it from the other 3 modules)
        * GammaFactor (Done and documented)
        * IncompleteGeneralizedGammaFactor (To be done)
        * InverseMellinGammaFactor (To be done)

    2. Stuff still computed using gp:
Line 12: Line 18:
    2. Doctest.     3. Doctest everything, making sure it all works 100% and fix issues with coercion, complex inputs, etc., as they are systematically uncovered.
    4. Possibly maybe change from digits to bits prec.
    5. Optimize.
Line 34: Line 42:
Dokchiter's Paper: attachment:dokchitser.pdf Dokchiter's Paper: [[attachment:dokchitser.pdf]]

Dokchitser Project for Sage Days 11

TODO:

  1. Modularize the code (break into the following classes and sort dependencies)
  2. Stuff still computed using gp:
    • Delta polynomials in _recursions_at_infinity (search for comment below)
    • _without_gp (gamma_series) has this line
      • sinser = sage_eval(rs(gp_eval('Vec(sin(Pi*(%s)))'%(z0+x))))
    • init_Ginf: still uses pari (see below)
    • Ginf: still uses pari to evaluate continued fraction
  3. Doctest everything, making sure it all works 100% and fix issues with coercion, complex inputs, etc., as they are systematically uncovered.
  4. Possibly maybe change from digits to bits prec.
  5. Optimize.

From Jen:

Here's the version (closest to Dokchitser's original pari code) that still uses continued fraction approximation:

http://sage.math.washington.edu/home/jen/sage-3.0.5-x86_64-Linux/l4.py

(needs gamma_series.py to run:

http://sage.math.washington.edu/home/jen/sage-3.0.5-x86_64-Linux/gamma_series.py)

The version with Pade approximation (l5.py) has a negligible speedup but only really works for low precision. I'm not sure if Pade gives us a means of computing bounds (I think Mike Rubinstein said that continued fractions won't). Also, l4.py doesn't work for imaginary inputs yet - some coercion with SymbolicRing that I didn't try.

Dokchiter's Paper: dokchitser.pdf

days11/projects/dokchitser (last edited 2008-11-14 13:42:12 by anonymous)