Other

drummers-lowrise

Message boards : General discussion : A few questions / statements

 Subscribe SortOldest firstNewest firstHighest rated posts first
Author Message
KEP

Joined: 10 Aug 05
Posts: 280
ID: 110
Credit: 8,528,983
RAC: 1,013

Message 3225 - Posted: 5 May 2006 | 19:30:46 UTC

Everyone with the knowledge to tell me, please feel free to submit your answer. I was primarily wondering why the primegen5.11 seems to slow down as i gratualy gets throgh the prime searc. Given example is that the first 5% takes 20 sec. on my machine and the last 5% appears to take several minutes. So howcome is the primegen slowing down as i gradually moves along within its search string?

Also I'm puzzled to the fact, that the WU I'm handeling at the moment is above 6 billion as a start, but we have only found a few million primes how come is that? I know it can't be, since the amount of primes is rising exponentialy with a frequenzy there roughly can be found this way: example given "(decimal digits)*2,3=Frequenzy of a prime appering" which for 6 billion plus alone should give an answer, that 1 in every 23 numbers is a prime.

And Rytis, will there come a public database where people can lookup those primes that we have found yet? If the prime database should ever come, will you then link to it or should the folks that primegrids have the trouble finding the prime database them self?

Regards!
____________

Rytis
Volunteer moderator

Joined: 22 Jun 05
Posts: 2648
ID: 1
Credit: 23,686,625
RAC: 40,284

Message 3246 - Posted: 8 May 2006 | 17:28:57 UTC - in response to Message 3225.

Everyone with the knowledge to tell me, please feel free to submit your answer. I was primarily wondering why the primegen5.11 seems to slow down as i gratualy gets throgh the prime searc. Given example is that the first 5% takes 20 sec. on my machine and the last 5% appears to take several minutes. So howcome is the primegen slowing down as i gradually moves along within its search string?

This is due to the fact that Yacas (the algebra system that we are using) stores some internal values (factors of the numbers previously crunched) to speed up (theoretically) factorization of the given numbers. If they had the same factors as the previous ones, they would be discarded more quickly. I am looking for a way to discard that data, since it does not benefit us, but as of yet I have not found it. Oh well, this is alpha :)

Also I'm puzzled to the fact, that the WU I'm handeling at the moment is above 6 billion as a start, but we have only found a few million primes how come is that?

Not all results before the workunit you are crunching have been returned yet; not every number in the range is a prime.

I know it can't be, since the amount of primes is rising exponentialy with a frequenzy there roughly can be found this way: example given "(decimal digits)*2,3=Frequenzy of a prime appering" which for 6 billion plus alone should give an answer, that 1 in every 23 numbers is a prime.

Not sure what you are stating there, but the number of primes in the range is not increasing when going further, but decreasing.

And Rytis, will there come a public database where people can lookup those primes that we have found yet? If the prime database should ever come, will you then link to it or should the folks that primegrids have the trouble finding the prime database them self?

Yes, the database infrastructure is ready, I'm just waiting for full ranges of workunits (every 1000 workunits, 100 000 000 numeric range) to be returned so I can dump a sequential file. The downloads will be done via bittorrent, because the server would not be able to handle the bandwidth.
____________

Rytis
Volunteer moderator

Joined: 22 Jun 05
Posts: 2648
ID: 1
Credit: 23,686,625
RAC: 40,284

Message 3263 - Posted: 9 May 2006 | 17:07:13 UTC

Update :) Actually there was an error importing prime numbers from output files into the database. I fixed the bug and restarted the process. Will a lot of workunits still not imported we are already above the previous count.
____________

KEP

Joined: 10 Aug 05
Posts: 280
ID: 110
Credit: 8,528,983
RAC: 1,013

Message 3276 - Posted: 11 May 2006 | 15:38:13 UTC

Thanks Rytis! It now looks more natural, since we are now closing in to the 25 million primes. Did you by the way reach the goal of 10 billion numeric range, within the 10 days you mentioned in an earlier thread?

Looking forward to be able to help spread the primes by the bittorrent system. What size are these files going to be, and are they based on the amount of primes rather than a byte size?

Regards!
____________

Rytis
Volunteer moderator

Joined: 22 Jun 05
Posts: 2648
ID: 1
Credit: 23,686,625
RAC: 40,284

Message 3277 - Posted: 11 May 2006 | 15:49:57 UTC - in response to Message 3276.

Thanks Rytis! It now looks more natural, since we are now closing in to the 25 million primes. Did you by the way reach the goal of 10 billion numeric range, within the 10 days you mentioned in an earlier thread?

Started sending workunits above 10 billion mark today.

As for the assimilated number of primes, we're now held back by the size of the prime database (nearing 1GB), it is very slow to do lookups in it. Will dump first prime numbers the next week, and will enable bittorent tracker there.

Looking forward to be able to help spread the primes by the bittorrent system. What size are these files going to be, and are they based on the amount of primes rather than a byte size?

Each file will hold primes in the range of 100 million numbers (1000 workunits) and will be compressed with gzip. I'm not sure about the size yet.
____________

Pentium100

Joined: 11 Jul 05
Posts: 40
ID: 3
Credit: 2,073,875
RAC: 0

Message 3354 - Posted: 17 May 2006 | 17:54:15 UTC - in response to Message 3277.

Each file will hold primes in the range of 100 million numbers (1000 workunits) and will be compressed with gzip. I'm not sure about the size yet.

Uncompressed it will be about 400MB in plaintext.
____________

Message boards : General discussion : A few questions / statements