This issue tracker has been migrated to GitHub, and is currently read-only.
For more information, see the GitHub FAQs in the Python's Developer Guide.

Author mark.dickinson
Recipients fredrikj, mark.dickinson
Date 2008-08-11.10:40:26
SpamBayes Score 2.7804197e-07
Marked as misclassified No
Message-id <>
> Indeed, that seems to be much faster. In the mean time, do you mind if I
> steal the code? :-)

Not at all!

I guess constant factors could well appear and/or disappear when
recoding in C; it might well be worth trying to implement both
methods to see which is faster.

There are some semi-obvious tricks available that might speed up the
recursive version.  For one thing, all shifts should be in multiples
of 15 bits (because longs are stored in base 2**15).  For another, it
ought to be possible to avoid the single-bit normalization shifts
every time the number of bits ('n') is odd---one can do a single
shift at the beginning of the calculation instead.
Date User Action Args
2008-08-11 10:40:28mark.dickinsonsetrecipients: + mark.dickinson, fredrikj
2008-08-11 10:40:28mark.dickinsonsetmessageid: <>
2008-08-11 10:40:27mark.dickinsonlinkissue3451 messages
2008-08-11 10:40:26mark.dickinsoncreate