|
2022-08-09, 12:36
|
#1 |
|
Apr 2004
22×7 Posts |
similar exponents - different backup file sizes
Hi,
just started to PRP test 4 similar exponents: 1xxxxx327, 1xxxxx329 and 1xxxxx681, 1xxxxx683 ... two pairs of twin primes. Looking at the file sizes of .bu, .bu2 etc. I found for first exponent 43MB, for second: 50, third: 55 and fourth: 52, residues files are all of same size, 5.1 GB. Shouldn't the save files of similar exponents be of comparable sizes? Restarted completely from the beginning, same observation. Is something going wrong here? PC is using a 4-core cpu, ht not used in calculations (4 workers, each with 1 thread). Greetings Last fiddled with by stippix on 2022-08-09 at 12:38 Reason: typo |
|
|
|
2022-08-09, 15:29
|
#2 |
|
P90 years forever!
Aug 2002
Yeehaw, FL
17·487 Posts |
Yes, they should be similar in size. I cannot explain the symptoms you are seeing.
If you stop and resume PRP testing does it work as expected? |
|
|
|
|
2022-08-09, 16:28
|
#3 |
Dec 2011
After 1.58M nines:)
1,699 Posts |
Maybe it is connected to "old know bug": if you run PRP with same exponent, then sometimes it produced error, since Prime95 use one backup file (since they have same backup name) for different K and problem occur.
Could you change algo for backup naming? Or at least add some config settings so backup name can have different format or something like that? Last fiddled with by pepi37 on 2022-08-09 at 16:29 |
|
|
|
2022-08-09, 17:46
|
#4 |
|
P90 years forever!
Aug 2002
Yeehaw, FL
100000010101112 Posts |
OP said these were different exponents, so naming conflicts are not the issue.
|
|
|
|
|
2022-08-09, 20:55
|
#5 |
|
Apr 2004
22·7 Posts |
stopped calculation, shutdown P95, restart P95 and calculation -> no changes
complete new setup in new folder, starting from scatch: no changes in general behavior, but now the file sizes are: 47, 55, 51 and 54MB ... even stranger behavior ... e.g. test with 150.007.157 and 150.007.159 results in (bu, bu2, 00x, etc.) file sizes of 54135 resp. 47499 kB, after complete new setup in new folder, starting from scratch: 51413 resp. 37497 kB ...  residues in the different runs are matching, so mathematically everything seems to work correctly ... btw. using version 30.7b9.win64 greetings |
|
|
|
|
2022-08-10, 07:41
|
#6 |
|
Apr 2004
22·7 Posts |
happens for both versions 30.3b6.win64 and 30.7b9.win64, also with smaller exponents 20996009 and 20996011, please see attached pics ...
Last fiddled with by stippix on 2022-08-10 at 07:56 Reason: typo |
|
|
|
|
2022-08-10, 09:19
|
#7 |
|
P90 years forever!
Aug 2002
Yeehaw, FL
201278 Posts |
Further investigation finds that you are correct! Until the first Gerbicz error check is passed the save file size is indeed somewhat variable in size. By default the first Gerbicz error check is at iteration 1 million.
|
|
|
|
|
2022-08-10, 09:33
|
#8 |
|
Apr 2004
22×7 Posts |
ok, thank you, valueable information, indeed I never tested until first GEC at 1 Mio ...
so I can go on with my tests looking forward the 1 Mio iteration, or probably using undoc infos to lower that limit: PRPGerbiczCompareInterval=n (default is 1000000) 
|
|
|
|
|
2022-08-10, 12:35
|
#9 |
|
Apr 2004
22·7 Posts |
hi George,
indeed, everything works as expected and you mentioned after first GEC  now looking forward to go on with the running prp tests  greetings Last fiddled with by stippix on 2022-08-10 at 12:38 |
|
|
|
 |
| Thread Tools | |
Similar Threads
|
||||
| Thread | Thread Starter | Forum | Replies | Last Post |
| Variable FFT sizes | TimSorbet | Software | 7 | 2008-01-14 17:33 |
| FFT sizes | Cruelty | Riesel Prime Search | 3 | 2006-07-12 15:15 |
| How Much Memory at Various Sizes? | wblipp | GMP-ECM | 5 | 2005-04-24 20:04 |
| Different word sizes/accuracy for forward/inverse transform? | Dresdenboy | Math | 2 | 2003-12-29 22:55 |
| Cache Sizes | Unregistered | Hardware | 4 | 2003-12-06 13:43 |
