Both values are same!
So llr-ing is continued :)

Guys, sorry for the confussion.

Pepi,
Below n=3M the files 2930-3000 you got (from our server) are the latest ones, and you can keep on testing them.

Only for n>3M we need new files, and I'll post them soon.
Carlos, thanks for the link.

Attaching k=17 from n=3M to n=4M sieved up to 277P by PrimeGrid.

Source: [URL]http://home.comcast.net/~jimb_primegrid/rsp/[/URL]

EDIT: Didn't see your post Kosmaj. You can delete this post of mine if you wish.

Carlos

Thanks :)
Kosmaj: I follow link for the riesel sieve, and noticed it is updated around every 30 days.
So wait a few days more to get latest sieve ( if you dont get latest data) difference in small range we tested will only be only few candidates ( if any)...
So latest sieve is named 20130424.7z ( 24.04.2013) so new one will come in ten days :)

Pepi, I cannot wait a few more days, because next week I'll be on a biz trip, with unpredictable Internet access :smile:

Carlos, thanks for the file, I just uploaded new test files from n=3M.

Happy crunching!

And, yes, small little digression :)
In last set of candidates I take, you say that lenght is 192K, but my LLR is show 168 K ( on highest candidate) .
So I dont know why is difference but it is :)

Will have to consult Thomas about the FFT length.

Maybe on i5/i7 cpu's the FFT jumps happen at larger exponents in comparison to "old" ones?

[QUOTE=pepi37;340666]
In last set of candidates I take, you say that lenght is 192K, but my LLR is show 168 K ( on highest candidate).[/QUOTE]

The given FFT lengths are for non-AVX but SSE2 capable cpus (e.g. P4, Core2 and early i3, i5, i7). On the second and third generation i3, i5, and i7's LLR switches to faster AVX code which uses slightly different turning points for the FFT lengths. As far as I could see, these are typically shifted to higher n, e.g. LLR uses the lower FFT length a bit longer than on non-AVX machines.

Thanks for reply and explanation!.

33 % done
no primes

66 % done
17*2^2946584-1 is prime :)