On Dec 18, 2009, at 2:42 PM, John Cremona wrote:

> 2009/12/18 Robert Bradshaw <rober...@math.washington.edu>:
>> On Dec 18, 2009, at 2:20 PM, John Cremona wrote:
>>
>>> That looks brilliant.  I think it would be fine to have the faster
>>> version in unreadable mpfr provided that the "real" formula was
>>> included as comments.
>>>
>>> Robert, can you post some version of what you have so far so I can  
>>> see
>>> how you have made it faster?
>>
>> Essentially, I used double complex values--the code otherwise looks
>> pretty much the same. I'll be putting it up on trac shortly.
>
> Hmmm -- it's not so great to have the speedup if it's not
> multi-precision.  The AGM algorithm is doubly exponential (i.e. the
> number of correct digits doubles with each iteration) so in
> single/double precision you hardly even need to have a loop.
> Obviously that's still useful, as long as we have a multi version as
> well.

Yeah, sorry if I wasn't clear about this. For such low precision, the  
overhead kills you and so I thought passing to CDF would be good. For  
higher precision, at least some of the time in the code will be spent  
doing "real work."

- Robert

-- 
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

Reply via email to