Re: [PrimeNumbers] Sieving Prime Tools
- PrimeGrid is currently running a GFN search. They are covering some
of the same ground as Yves's search, but only because they want to be
thorough as some ranges from the original search were never returned.
Yves's genefer also failed to report roundoff errors correctly, so I
suspect that results for higher b might be correct.
I would recommend srsieve. It is much faster than any other choices
for fixed-k sieving. There is also gcwsieve (from Geoff Reynolds as
well) for Cullens and Woodalls. There is also fpsieve (psieve and
fsieve as well) developed by me being used by PrimeGrid.
On Oct 24, 2010, at 8:12 PM, Michael wrote:
> Since I was last active in 2004 and there appear to be a lot of new
> programs and I would appreciate your wisdom learning about the tools.
> The venerable Latest version of NewPGen v2.82 appears to be from
> 2004 although in a 2010 zip file http://primes.utm.edu/programs/NewPGen/
> The Linux program reports 2007, so I wonder if it has new features.
> I vaguely recall 64-bit needed a massive rewrite. TwinGenX has 64-
> bit although supports less forms and appears slower in some cases.
> Multi-Math also looks a bit tricky to use well with very large files.
> What are the best tools now available? When do you use TwinGenX or
> NewPGen or other tools?
> Michael Angel
> Btw, Is the GFN search still active? Yves's website seams out of date.
> Many of these program did not give much of any search results in the
> forums so I list some below for discussion. If their are other good
> programs please share.
> Then their are an array new specialized choices.
> SRSieve, SRSieve1/2/5.
> sgsieve, not publicly released that I found
> TPSieve: originally developed by Geoff Reynolds for the Twin Prime
> Search, was meant for use in a sieve with one or a few n's. It was
> then modified by Ken Brazier, in collaboration with Geoff Reynolds,
> to make many-n searching efficient, within the fixed-n format.
> Has NVidia CUDA GPU version.
> PPSieve: developed by Ken Brazier, is a modified version of TPSieve
> that sieves for single primes of the form k*2^n+1. Its strength is
> the many-n optimization.
> Trial Division for CUDA, 1 bit at a time for mersenne candidates
[Non-text portions of this message have been removed]