New Sieve Thread Discussion

[Citrix]

Lars,

I was wondering that has anyone found any factors that proth_Sieve previously missed? Could you post them below!


Thanks,
Citrix

[em99010pepe]

Citrix,

Lars told me one week ago (16 Jul) that my ranges found already 19 missed factors where we have no PRP test yet. In Lars ranges he found one. (All missed factor where n>2Mil)

One more thing. I started the 98700-99000 range with the old dat file and found the following factors:

98702674984699 | 90527*2^14752511+1
98705626374527 | 79817*2^3437591+1
98725913745839 | 237019*2^15283258+1
98728617700901 | 258317*2^13774959+1

Then I restarted it with the new dat file and

98701195679903 | 90527*2^22380191+1
98702674984699 | 90527*2^14752511+1
98703968070791 | 79817*2^35840703+1
98705626374527 | 79817*2^3437591+1
98707048330861 | 222361*2^35969432+1
98709139652681 | 214519*2^24565634+1
98710294034071 | 168451*2^47186640+1
98713092107897 | 222361*2^38408320+1
98719327707491 | 90527*2^39919751+1
98720039885147 | 90527*2^30822647+1
98720263609273 | 168451*2^44530128+1
98720326492903 | 90527*2^45747167+1
98723758825841 | 222113*2^30695621+1
98724565964807 | 222113*2^47054597+1
98725913745839 | 237019*2^15283258+1
98726907126721 | 222113*2^35183621+1
98727187829681 | 237019*2^36555406+1
98727258317549 | 222113*2^35498597+1
98728617700901 | 258317*2^13774959+1

Cheers,

Carlos

[Citrix]

If in 3T we just found 19 missed factors then I do not think it is worth double checking the dat ranges. May be we should just work on 20-50M and then join the dats around 85T. We are also likely to find duplicate factors for these numbers, so there is no need to find the smallest factor.

If we agree to this it would double are sieving speed.

Citrix

[hhh]

We had about the same discussion at SoB, and we agreed not to give up doublecheck.
Did you think about the amount of time needed to run PRP for 20 numbers? It takes quite a while, I think more then the time spent for re sieve. IMHO, we should do this well, once for all, and not running the risk to regret something later. We will have sieved 4 G in the first month, so, if SoB finishes Firstpass sieving in 8 months, we will be at 40G, at that time. Thats fine, because it is about the level at which PSPsieve alone and SoB+PSP sieve at 60G will be the same efficient (for factor throughput). So, there is no hurry. The extend of the sieving range is something for the far future, anyway. No need to bother with short term speed increases.
For the later ranges which will be done with SoB, I doubt if there is a necessity to do DC, because the SoB DC can show if there are missed factors (and eventual problems with hardware or submission).
Yours H.

[em99010pepe]

Originally Posted by Citrix

If in 3T we just found 19 missed factors then I do not think it is worth double checking the dat ranges. May be we should just work on 20-50M and then join the dats around 85T. We are also likely to find duplicate factors for these numbers, so there is no need to find the smallest factor.

If we agree to this it would double are sieving speed.

Citrix

Correction. In 1T we found 19 missed factors.

Carlos

3100-3300 reserved by shoelace

do you have a prefered range size or range completetion time?

[VJS]

Didn't ready through everything but...

- finding 19 factors in 1T doublecheck
- Speed decrease by using 1k<n<50M compared to 20M<n<50M (33%??)

Basically those 19 factors per T = 57 factors per T (time wise)

That's pretty dense. Especially considering your planning on finding primes for 15k's.

Also the chances of finding the second smallest factor when you missed the first one is not great. I think we only found about 3 or 4 of these with SoB.

[ltd]

Newest numbers:

Within the first 2.5 T we have found 33 lost factors.
All these factors are above n=2Mil.
But that was to be expected as we have already increased the sieving range from 2Mil to 20Mil and had a second pass for all values below 2Mil for the first 5-10T already.


Lars

[ltd]

@shoelace:

I have no prefered range size. I am happy about each calculated result that i will not restrict the range reservation.

For the completion time i like to see results latest after a month. If you want to reserve a range that needs longer then a month either send in intermediate results or send me a note that the range will take longer.

Lars

[OmbooHankvald]

Originally Posted by VJS

...considering your planning on finding primes for 15k's.

I haven't heard about this. Are you merging with 15k too?

[Citrix]

Originally Posted by ltd

Newest numbers:

Within the first 2.5 T we have found 33 lost factors.
All these factors are above n=2Mil.
But that was to be expected as we have already increased the sieving range from 2Mil to 20Mil and had a second pass for all values below 2Mil for the first 5-10T already.


Lars

I suggest that it is not worth to continue with 0-20M dat because as the density of factors decreases with increase in P, so will the density of missed factors decrease. We should get match up with SOBsieve as soon as possible so few resources are wasted.

Citrix