Re: [OctDev] list of primes using D. Bernsteins fast primegen library

[David Bateman]

Dr.-Ing. Torsten Finke wrote:

>
>>Then there is another policy that has been discussed is that we should
>>try hard not to have an overlap between the code in octave and
>>octave-forge to avoid issues of knowing which code is being used and
>>what bug fixes have been applied where. In the octave-forge CVS the
>>overlap with octave 2.9.6 has been completely eliminated. So I'd
>>hesitate to include your code in octave-forge if it isn't accepted in
>>octave core.
>>    
>>
>
>since octave is absolutely high quality code, I fully agree with that policy. 
>
>  
>
>>With a change to use a lookup table for small values of p, I think
>>you've demonstrated the case for the speed-up of your code. The next
>>question is how many people will really use this function and whether
>>the additional support costs of C code is offset by the number of
>>potential users who are impacted by this improvement. The final say is
>>of course up to John.
>>    
>>
>
>I am also not sure how I would decide. There are adavantages (maybe only for
>few users - I don't know) but there several drwabacks as well. 
>
>Should I better never having made my suggestion :-)
>
>  
>
As I said the final say is up to John, but discussion on the list will
help clarify his choice.. He hasn't weighed in on this discussion yet so
who knows.. In any case I'd suggest you resend an updated version of
your code with the lookup table for small values of p to further speed
up your code, and see what he says.

D.


-- 
David Bateman                                address@hidden
Motorola Labs - Paris                        +33 1 69 35 48 04 (Ph) 
Parc Les Algorithmes, Commune de St Aubin    +33 6 72 01 06 33 (Mob) 
91193 Gif-Sur-Yvette FRANCE                  +33 1 69 35 77 01 (Fax) 

The information contained in this communication has been classified as: 

[x] General Business Information 
[ ] Motorola Internal Use Only 
[ ] Motorola Confidential Proprietary