mersenneforum.org PSP goes prpnet
 Register FAQ Search Today's Posts Mark Forums Read

2009-10-05, 16:48   #34
rogue

"Mark"
Apr 2003
Between here and the

22×3×487 Posts

Quote:
 Originally Posted by opyrt According to Louie of Seventeen or Bust, Jean Penné is working on a new version of llr based on a newer gwnum library: http://www.free-dc.org/forum/showthr...762#post139762 I'm so glad we're running PRPNet now so we can start using the new llr as soon as it's ready! :-)
You can use the latest PFGW, which already has those improvements, but as I understand it, Jean will be adding some additional improvements to LLR. I believe that he was going to support primality proofs for non-base 2 numbers.

 2009-10-05, 18:42 #35 ltd     Apr 2003 22·193 Posts Does pfgw create compatible residues?
2009-10-05, 19:15   #36
rogue

"Mark"
Apr 2003
Between here and the

22×3×487 Posts

Quote:
 Originally Posted by ltd Does pfgw create compatible residues?
I forgot that PSP is base 2, so I guess PFGW is out of the question right now. The choice of PFGW is best when the base is not a power of 2.

I could modify PRPNet to make PFGW do a primality test for +1 base 2/4/8/... numbers instead of a PRP test. But I would also need to modify PFGW to output a residue for these bases when it is performing a Proth primality test. That shouldn't be too difficult.

I do not expect LLR to gain much speed for base 2 tests when it is updated to 25.12.

Last fiddled with by rogue on 2009-10-05 at 19:21

2009-10-06, 04:23   #37
mdettweiler
A Sunny Moo

Aug 2007
USA (GMT-5)

624910 Posts

Quote:
 Originally Posted by rogue I forgot that PSP is base 2, so I guess PFGW is out of the question right now. The choice of PFGW is best when the base is not a power of 2. I could modify PRPNet to make PFGW do a primality test for +1 base 2/4/8/... numbers instead of a PRP test. But I would also need to modify PFGW to output a residue for these bases when it is performing a Proth primality test. That shouldn't be too difficult. I do not expect LLR to gain much speed for base 2 tests when it is updated to 25.12.
What would be really handy, IMO, would be to have PFGW be able to do LLR tests. I believe you've stated before that this would be rather hard to implement, but, would it be possible to simply build Jean's code into PFGW? If that could be done, along with a tweak to make it do Proth tests and output such residues for base 2 +1, then PFGW could truly be a "universal prime search application". Of course LLR would still be useful for some things, but nonetheless it might simplify coding PRPnet a bit.

2009-10-06, 12:49   #38
rogue

"Mark"
Apr 2003
Between here and the

584410 Posts

Quote:
 Originally Posted by mdettweiler What would be really handy, IMO, would be to have PFGW be able to do LLR tests. I believe you've stated before that this would be rather hard to implement, but, would it be possible to simply build Jean's code into PFGW? If that could be done, along with a tweak to make it do Proth tests and output such residues for base 2 +1, then PFGW could truly be a "universal prime search application". Of course LLR would still be useful for some things, but nonetheless it might simplify coding PRPnet a bit.
I don't want to take away from Jean's work, so I hesitate to have PFGW compete directly with LLR on certain things. If he decides to add support for generic forms to LLR, then all gloves would be off, so to speak. If he wants to contribute to PFGW, he is welcome to do so. My other concern is that I wouldn't want to significantly alter default behavior of PFGW as that might confuse other users.

Last fiddled with by rogue on 2009-10-06 at 12:50

2009-10-06, 14:50   #39
mdettweiler
A Sunny Moo

Aug 2007
USA (GMT-5)

3×2,083 Posts

Quote:
 Originally Posted by rogue I don't want to take away from Jean's work, so I hesitate to have PFGW compete directly with LLR on certain things. If he decides to add support for generic forms to LLR, then all gloves would be off, so to speak. If he wants to contribute to PFGW, he is welcome to do so. My other concern is that I wouldn't want to significantly alter default behavior of PFGW as that might confuse other users.
Ah, right--good points. Okay, that makes sense.

 2009-10-07, 17:32 #40 opyrt     Apr 2008 Oslo, Norway 7×31 Posts I wonder if this is expected behaviour, or a bug? Got this config on the client: Code:  server=PSPtest:100:0:www.psp-project.de:7100 server=PSPtestdc:0:1:www.psp-project.de:7101 Note that PSPtest is set to 100:0. I thought that was the correct way to say "finish this wu, and don't start a new one". This happens: Code: [2009-10-07 17:55:29 GMT] PSPtest: 156511*2^7955400+1 is not prime. Residue BBBBBBBBBBBBBBB [2009-10-07 17:55:29 GMT] Total Time: 1:03:18 Total Tests: 1 Total PRPs Found: 0 [2009-10-07 17:55:29 GMT] PSPtest: Returning work to server www.psp-project.de at port 7100 [2009-10-07 17:55:34 GMT] PSPtest: ERROR: Workunit 156511*2^7955400+1 not found on server [2009-10-07 17:55:34 GMT] PSPtest: The client will delete this workunit [2009-10-07 17:55:34 GMT] PSPtest: INFO: 0 of 1 test results were accepted [2009-10-07 17:55:34 GMT] PSPtest: Getting work from server www.psp-project.de at port 7100 [2009-10-07 17:55:39 GMT] PSPtest: INFO: No available candidates are left on this server. [2009-10-07 17:55:39 GMT] PSPtest: Getting work from server www.psp-project.de at port 7100 [2009-10-07 17:55:44 GMT] PSPtest: INFO: No available candidates are left on this server. [2009-10-07 17:55:44 GMT] PSPtest: Getting work from server www.psp-project.de at port 7100 [2009-10-07 17:56:05 GMT] PSPtest: INFO: No available candidates are left on this server. [2009-10-07 17:56:05 GMT] PSPtest: Getting work from server www.psp-project.de at port 7100 [2009-10-07 17:56:10 GMT] PSPtest: INFO: No available candidates are left on this server. [2009-10-07 17:56:10 GMT] PSPtest: Getting work from server www.psp-project.de at port 7100 [2009-10-07 17:56:45 GMT] PSPtest: INFO: No available candidates are left on this server. [2009-10-07 17:56:45 GMT] PSPtestdc: Getting work from server www.psp-project.de at port 7101 Is it supposed to behave like that? If yes, what is the correct way to tell the client to finish the WU it's running and don't download a new one?
 2009-10-07, 19:59 #41 Sloth     Mar 2006 2×47 Posts Code:  server=PSPtest:100:0:www.psp-project.de:7100 server=PSPtestdc:0:1:www.psp-project.de:7101 I think you are close but what you have is PSPtest:100:0 - 100% of your work is to be from this server. But you are getting 0 work units PSPtestdc:0:1 - 0 showing it is a fallback if the other server does not give you anything and you are getting 1 work unit. Not sure what the command is if you want to finish your existing work but not get anything else. Rogue will be around and can correct me if I am wrong. S.
 2009-10-07, 20:08 #42 opyrt     Apr 2008 Oslo, Norway 7×31 Posts Thanks for the reply, Sloth. This seems to give the same result: Code: server=PSPtest:100:0:www.psp-project.de:7100 server=PSPtestdc:0:0:www.psp-project.de:7101 Except it will get work from the server on port 7100. BTW: The "workunit not found on server" message seems to be unrelated.
 2009-10-07, 23:28 #43 rogue     "Mark" Apr 2003 Between here and the 22×3×487 Posts To have PRPNet complete a test for PSPTest, set PSPTest to 0:1. This will complete any assigned workunits and return them to the server. The "1" will have it get a workunit from that server only if PSPTestdc is unavailable. PSPTestdc should be 100:x, with x being the number of workunits you want to get from that server.
 2009-10-08, 06:59 #44 opyrt     Apr 2008 Oslo, Norway 7×31 Posts Hi rogue. I think you misunderstood... I want the prpclient to finish the test it's running and don't get a new one. Kind of like "no new tasks" or what it's called in boinc.

 Similar Threads Thread Thread Starter Forum Replies Last Post Mattyp101 Conjectures 'R Us 2 2011-02-07 13:53 Joe O Sierpinski/Riesel Base 5 1 2010-10-22 20:11 rogue Conjectures 'R Us 220 2010-10-12 20:48 mdettweiler No Prime Left Behind 80 2010-02-09 21:31 rogue Conjectures 'R Us 250 2009-12-27 21:29

All times are UTC. The time now is 15:00.

Tue Aug 4 15:00:08 UTC 2020 up 18 days, 10:46, 0 users, load averages: 1.39, 1.73, 1.61