Alternative Sieving for 10M digit prime search

[Citrix]

Just to clearify everything

1) We need our own thread, it is getting kind of confusing in here. I am thinking of starting a thread in the PSP forum, where I have mod rights.
2) Cerdric, if you want to start a webpage you can. I don't have much time to generate stats etc.
3) Mhf, regarding your questions

a) You really do not need the new dat file. If you update it or not, it makes no difference to the speed of the sieve. So just to save time I prefer not to update the dat file everytime some one has factors. (I will update the dat file, once we have enough factors, to reduce the size of the file.)

b) Regarding the factors, you can post them on the forum or PM me for my email address and then email me.

c) Regarding the new dat file and the website (sharebig file) not working. You don't really need a new file. You can use the old one, the speed will not change. For the website, I don't know of any good reliable sites where I can host the .dat file. Anyone know of any?


I hope this answers your questions!

[m_f_h]

Originally Posted by Citrix

Just to clearify everything

1) We need our own thread, it is getting kind of confusing in here. I am thinking of starting a thread in the PSP forum, where I have mod rights.
(...)
a) You really do not need the new dat file. If you update it or not, it makes no difference to the speed of the sieve. So just to save time I prefer not to (...)
not change. For the website, I don't know of any good reliable sites where I can host the .dat file. Anyone know of any?

thanks for many answers.

I have a proposal: What about creating a group on google ?
I think one can share files, see e.g.
http://groups.google.com/support/bin...y?answer=46489

and to my experience, google is quite "stable" and generous regarding big files.

as to your answers :
I understood before I don't need the up-to-date data file, but I wanted to avoid you duplicate factors.
All factors I found are in the last attached factor.zip (post from this morning).

[Citrix]

Originally Posted by m_f_h

as to your answers :
I understood before I don't need the up-to-date data file, but I wanted to avoid you duplicate factors.
All factors I found are in the last attached factor.zip (post from this morning).

Duplicate factors are not a problem. The program removes them automatically.

Also I am not so interested in starting a group right now, since there are very few users. Once we have 5+ users, I will request to start a sub-forum.

If you want the latest sieve file I can email it to you, but things will not get faster.

[ValerieVonck]

5000 - 6000 CedricVonck - Reserved

[ValerieVonck]

New version of sr1sieve is out... I am sieving with a speed of 6297052 or 6M /sec on a P4 -3Ghz.

[em99010pepe]

Just testing the client on an AMD 64 3000+.

sr1sieve 1.0.20 -- A sieve for one sequence k*b^n+/-1.
Using SSE2 code path, L1 data cache 16Kb (default), L2 cache 256Kb (default).
WARNING: --pmin=6000000000000 from command line overrides pmin=40100000000 from
`3_16data.txt'
Read 658088 terms for 43046721*2^n+1 from NewPGen file `3_16data.txt'.
Split 1 base 2 sequence into 15 base 2^240 subsequences.
Recognised Generalised Fermat sequence A^16+1
Using 16 Kb for the baby-steps giant-steps hashtable, maximum density 0.21.
Using 128Kb for the Sieve of Eratosthenes bitmap.
Expecting to find factors for about 3429.86 terms.
sr1sieve started: 1000000 <= n <= 49999888, 6000000000000 <= p <= 7000000000000
6000031008577 | 43046721*2^42038368+1
p=6001621099169, 8971123 p/sec, 1 factor, 0.16% done, ETA 17 Apr 02:40

[m_f_h]

Originally Posted by em99010pepe

Just testing the client on an AMD 64 3000+.

sr1sieve 1.0.20 -- ...
p=6001621099169, 8971123 p/sec, 1 factor, 0.16% done, ETA 17 Apr 02:40

Interesting - and shows the importance of CPU cache (if I interpret well):
yours runs almost at same p/sec than my old 1.0.15 on a 2000 AMD64 (with an additional mprime running on the 2nd core!), but in my case it used 32K L1, 512K L2 (also "default" !?) (see my post from 23-Mar-07)
Also astonishing (for me) that it's "so much" slower on Cedric's 3Ghz P4...(?)

[ValerieVonck]

m_f_h,

Pentium 4 will always be slower in sieving, no matter what cpu speed you throw against it. E.g. My (deceasesed) amd 1 ghz sieved for rieselsieve @ 70k/sec, my p4 3ghz had the same speed.

P4 = llr
Amd = sieve.

[Cruelty]

Actually, right now Core2 is better than anything else both with sieving/factoring and LL testing

[Citrix]

Range overview:

0 - 200G - unknown
200G - 4000G - CedricVonck - Done
4000G - 4005G - Citrix
4005G - 4010G - Citrix
4010G - 4040G - M_f_h
4040G - 4550G - CedricVonck
4550 - 5000G - M_f_h
Done till here...

5000 - 6000G - CedricVonck
6000 - 7000G - Carlos

Updated file contains 550046 candidates.

[em99010pepe]

I'm not sieving that range, I only tested the client.

Carlos