Join PrimeGrid
Returning Participants
Community
Leader Boards
Results
Other
drummers-lowrise
|
Message boards :
Sieving :
PPS Sieve New WUs just released?
| Author |
Message |
|---|
|
|
|
Hi. I was just wondering if anyone knew if there was a new batch of WUs with different coding released for PPS Sieve since the WU times just increased across the board by around 16-17% for completion times without a change in version number. | |
|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
                 
|
Hi. I was just wondering if anyone knew if there was a new batch of WUs...
We are reprocessing some ranges below 1P with the new tpsieve 1.38 to explore abnormal factor gaps. It appears that under certain unknown circumstances, incomplete factor files were being returned. v1.38 has a "work around" check for this.
The normal missed factor rate is historically under 0.1%. However, we discovered from 100T-210T that rate was 0.27%. Although this number is still quite small, we wish to recover as many factors as possible. We also hope to confirm that the problem is resolved.
We will explore all gaps under 1P and possibly up to 2P. Up to 1P should only take a day or two and up to 2P would double that.
BTW, the WU's are longer only because of the lower p and not because of any processing changes.
____________
| |
|
|
|
|
OK. Thanks. So I guess the higher "P" values are more efficient, correct? What is "P" anyways? I was trying to figure that out here: http://primegrid.com/stats_pps_sieve.php | |
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
OK. Thanks. So I guess the higher "P" values are more efficient, correct? What is "P" anyways? I was trying to figure that out here: http://primegrid.com/stats_pps_sieve.php
[EDIT]Yes, the higher p values run faster.
Sieving depth is usually identified as p=factor size. And factor size is usually referenced in terms of M, G, T, P, and for the GFN search, E. These are just SI prefixes (mega, giga, tera, peta, exa). For example:
p=100T is a factor size of 100*10^12 or 100,000,000,000,000
The current depth of the PPS sieve is about p=18P or 18*10^15 or 18,000,000,000,000,000. An example of some recent factors found this past week are:
16032266992594319 | 5471*2^355177+1
16032267065197489 | 1589*2^2487823+1
16032267313245223 | 6825*2^717270+1
16032267498323101 | 3559*2^2141545-1
16032267608905127 | 2925*2^2686855+1
This means that 16032266992594319 is a factor of 5471*2^355177+1.
____________
| |
|
|
|
|
Got it. Thanks again. | |
|
|
|
We are reprocessing some ranges below 1P with the new tpsieve 1.38 to explore abnormal factor gaps. It appears that under certain unknown circumstances, incomplete factor files were being returned. v1.38 has a "work around" check for this.
The normal missed factor rate is historically under 0.1%. However, we discovered from 100T-210T that rate was 0.27%. Although this number is still quite small, we wish to recover as many factors as possible. We also hope to confirm that the problem is resolved.
John, just as a side issue:
Since I began cruncing PPS Sieve with my Fermi (instead of CPU) back in late November, I was curious about why my Average factors per workunit stat had dropped steadily from ~110 to less than half that. Just since the new batch of WUs came out, it's already back up to ~71.
I assume this change in the missed factor rate answers that question? Also, is that statistic an overall average or a moving average (like RAC for example)?
____________
  | |
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
Since I began cruncing PPS Sieve with my Fermi (instead of CPU) back in late November, I was curious about why my Average factors per workunit stat had dropped steadily from ~110 to less than half that. Just since the new batch of WUs came out, it's already back up to ~71.
The factor density drops the deeper we sieve. There are more factors at 1P than there are at 18P. Therefore, the Average factors per workunit will return to declining after we finish reprocessing these lower gaps.
____________
| |
|
|
|
The factor density drops the deeper we sieve. There are more factors at 1P than there are at 18P. Therefore, the Average factors per workunit will return to declining after we finish reprocessing these lower gaps.
Got it, thanks. | |
|
|
|
|
may i ask when this adjustment will finished,and the runtime will return to the normal? | |
|
|
|
may i ask when this adjustment will finished,and the runtime will return to the normal?
on it's way:
24 Jan 2011 13:48:32 UTC Completed and validated 30,589.84
23 Jan 2011 23:49:01 UTC Completed and validated 32,263.73
22 Jan 2011 8:50:17 UTC Completed and validated 35,675.31
that host was below 25k @ 18p, but now it's hitting around 300 instead of less than 20 per WU.
| |
|
|
|
|
thanks for the detail! | |
|
Message boards :
Sieving :
PPS Sieve New WUs just released? |
