Sieving discussion thread

[MooMoo2]

Here's the link to the sieve file:

http://www.sendspace.com/file/dgnywn

I think the links last for about a week or two before they get deleted.

[mdettweiler]

Originally Posted by MooooMoo

Here's the link to the sieve file:

http://www.sendspace.com/file/dgnywn

I think the links last for about a week or two before they get deleted.

Okay, thanks!

[Joshua2]

how is seiving for next n coming? Are you going to upload it and make it public soon?

[MooMoo2]

Originally Posted by Joshua2

how is seiving for next n coming? Are you going to upload it and make it public soon?

I haven't made any final decisions yet, but the next value after n=333,333 may be a range of n, not a single n-value.

This means we would be sieving something like n=460K-525K from k=3 to 3M.

See the discussion in this thread for more info:

http://www.mersenneforum.org/showthread.php?t=8479

[Kosmaj]

IMO, It's enough sieving. Extra sieving at this level produces a miniscule improvement of the probability to find a twin. What's badly needed are more LLR tests. If you have available sieving resources it's better to start sieving a new exponent.

[biwema]

for the whole process, we have:

Optimal sieve depth for n=333,333: ~25000T (25P).

the current Optimal sieve depth is: d_Opt * Range_Remaining * (1+DoubleCheck%)

= 25T * 0.5 * 1.3 = 16.25T

At the moment, sieving is removing candidates 2.5 times as fast as LLR but depends on the Architecture.
These factors don't need to be doublechecked, either.

Nevertheless:
- You won't find a Twin when sieving.
- I recommend LLR, when using a P4 or Core2 and sieving when using a Centrino or AMD CPU (see benchmark for details)
- Sieving the next Exponent is only recommended, if having lots of memory.

maybe popandbob is also interested to come back

[cipher]

Guys as soon as PG finishes 20P are we stopping sieving? Is 20P optimal sieve depth?

on My PC to LLR 1 candidate it takes 3:10seconds and it takes 4:57seconds to Sieve one candidate out. so switching to LLR finally makes sense.

thanks
cipher

[MooMoo2]

Originally Posted by cipher

Is 20P optimal sieve depth?

on My PC to LLR 1 candidate it takes 3:10seconds and it takes 4:57seconds to Sieve one candidate out. so switching to LLR finally makes sense.

Is 4:57 the time needed to find a factor at your range of 16P, or is it the time needed at 20P?

[Lennart]

Originally Posted by cipher

Guys as soon as PG finishes 20P are we stopping sieving? Is 20P optimal sieve depth?

on My PC to LLR 1 candidate it takes 3:10seconds and it takes 4:57seconds to Sieve one candidate out. so switching to LLR finally makes sense.

thanks
cipher

If that is correct time, we will stop at 20P. We have stoped reservasion and the guy doing the highest range do some timings that we will get today i think.

/Lennart

[cipher]

Originally Posted by MooooMoo

Is 4:57 the time needed to find a factor at your range of 16P, or is it the time needed at 20P?

That time is at my range of 16P moooomoo and not @ 20P.

[Lennart]

Originally Posted by cipher

That time is at my range of 16P moooomoo and not @ 20P.

I have tested a small range on 20100T

224sec/factor

LLR on the same computer 300 sec


231sec *1,3=300 sec

If thats corect we can sieve some more.

This is on a q6420@2,2Ghz
sieved with tpsieve

Lennart