Re: certification of a prime gap
- Jose Luis Gomez Pardo wrote
> I realize now that the certification of the top andThe snappy pace of 8342 bits in 66 hours (1GHz Athlon) in
> bottom requires much more work than what I had assumed
> from earlier David comments.
is about as fast as ECPP gets, to date.
I reckon that the median Primo time (same machine) is
about half a week at 8000 bits. One week, or even two,
might be required if Run 3 appears, and/or there is need of
significant backtracking after discriminant exhaustion.
I wonder what Hans and Giovanni are doing right now?
5k decimal digits by someone, in time for Christmas?
In my bones I feel that, *within operational limits*,
Primo scales more like digts^4 than the asymptotic
estimate digits^6. If 2.5k digits takes half a week,
5k digits might take only only 2 or 3 months,
on a wing and a prayer...
- David Broadhurst wrote:
> I wonder what Hans and Giovanni are doing right now?Some less digits from me and some more digits from Giovanni
> 5k decimal digits by someone, in time for Christmas?
(and from someone else) seem to be likely before Christmas :)
> In my bones I feel that, *within operational limits*,In my bones I'm feeling the same, but too much depends on the
> Primo scales more like digts^4 than the asymptotic
> estimate digits^6. If 2.5k digits takes half a week,
> 5k digits might take only only 2 or 3 months,
> on a wing and a prayer...
*individual* number being tested. You'll know what a certain
5k-+ number takes to be proven prime with Tx/Po on a certain
PC using certain methods this year, that's for sure.
PS: I indeed don't know what Giovanni is testing at present,
I only know what I am working on with Primo.
- On Thu, 30 Aug 2001 21:30:16 -0400, Nathan Russell
>On Thu, 30 Aug 2001 18:35:11 -0000, d.broadhurst@... wrote:Now finished.
>>Jose Luis Gomez Pardo asked about the certification
>>of the top and bottom PrPs for his gap.
>>My experience is that a 8000 bit PrP takes
>>between 3 and 4 days to certify, running
>>Primo on a 1GHz Athlon, if the chain of elliptic
>>curves never hits "Run 3". So with a bit of luck
>>Jose Luis' gap could be certified in a week.
>>I have no spare machine, at present, but I think
>>that someone here surely has! (Greg?)
>>The top and bottom would then be archivable as
>>ECPP primes, for a while.
>I will do the bottom prime of this gap, devoting my P3-600 with Win98
>and 384 MB of 133 MHz memory completely to it until it is finished.
cert_val \quick-check passed, though i didn't clean out all the
headers and didn't change any test numbers. I don't know if that's
important to anyone...
I'm running a full cert_val now, and I'll send out the certificate (in
compressed form, of course) to Jose and anyone else who expresses an
interest after that passes. It's about 2/3 of a megabyte.