PRPNet released!

[Svenie25]

Done bothe. The email to you with success and switched off the authentication without.

[rogue]

Here are changes from 2.2.0:

• Server: Add ability to expire tests (see prpserver.delay for information).
• Server: Check for pending tests before removing candidates from the server.
• Server: Add support for Twin Prime searches.
• Server: Remove candidates after a successful test if there is no double-ckecking.
• Client: Check for existence of workunit on server before sending rest of message.
• Client: Add support for Twin Prime searches.
• Client: Modularized code for the handling of helper programs.
• Client: Do not return immediately when a prime/prp is found so that no tests are abandoned.
• Client: Use -k feature added in PFGW 1.3.

and from 2.2.1

• Server: More changes to prpserver.delay to remove first test checks. The server will no longer send the same test to two clients unless one of the client's has missed the expiration deadline.
• Client: Fix a segfault if communication with the server is broken while sending workunits back.
• Client: Fixed issue with counting of completed and current work units.
• Client: The client will drop tests that are not known on the server.

You can d/l it from here.

BTW, the 2.2.x client will only work with a 2.2.x server. 2.1.x clients will work a 2.2.x server.

[Svenie25]

Server: Add support for Twin Prime searches.

Is there a special header needed for the candidatesfiles for twins?

[rogue]

Originally Posted by Svenie25

Is there a special header needed for the candidatesfiles for twins?

The twin prime search is only for k*b^2+/-1 numbers. The candidates file should contain that +1 form. The client will test the +1 form. If that is PRP or prime, then the client will also test the -1 form. If the +1 form is composite then a test for the -1 form is not required.

[Mini-Geek]

Originally Posted by rogue

The twin prime search is only for k*b^2+/-1 numbers.

I think you mean n. Or maybe k*2^n+/-1.

Originally Posted by rogue

The candidates file should contain that +1 form. The client will test the +1 form. If that is PRP or prime, then the client will also test the -1 form. If the +1 form is composite then a test for the -1 form is not required.

Isn't it faster (at least with b=2) to check the -1 form? It would save time to check the -1 form first, then only check the +1 form if the -1 is PRP/prime. I know that LLR checks the -1 first, at least for b=2.

[mdettweiler]

Originally Posted by Mini-Geek

Isn't it faster (at least with b=2) to check the -1 form? It would save time to check the -1 form first, then only check the +1 form if the -1 is PRP/prime. I know that LLR checks the -1 first, at least for b=2.

Actually, there is no speed difference between testing the -1 and +1 sides; thus, on a file sieved for a twin search (i.e. with factors removed from both sides), either side can be searched arbitrarily with no speed penalty.

[gd_barnes]

Originally Posted by mdettweiler

Actually, there is no speed difference between testing the -1 and +1 sides; thus, on a file sieved for a twin search (i.e. with factors removed from both sides), either side can be searched arbitrarily with no speed penalty.

Where have you gotten that info. from? Have you tested a Riesel and a Proth prime of the same size?

It has been my experience with LLR that Riesel's test faster than Proth's.

Or are you referring to Phrot in this case?

[Mini-Geek]

Originally Posted by gd_barnes

Where have you gotten that info. from? Have you tested a Riesel and a Proth prime of the same size?

It has been my experience with LLR that Riesel's test faster than Proth's.

Or are you referring to Phrot in this case?

After he posted the info I tested it on the largest known twin prime pair, (ran both in LLR and compared times) and it's correct.

2003663613*2^195000-1 is prime!  Time : 130.578 sec.
2003663613*2^195000+1 is prime!  Time : 130.273 sec.

[rogue]

This is patch release to 2.2.1. The only changes are to the client:

• Fixed issue with restarting client where it would not handle some start options correctly.
• Modified to log messages on the client when the server refuses the workunit. In these cases the client will drop the workunit so that it can continue doing other work for the server.

You can d/l it from here.

[mdettweiler]

Originally Posted by gd_barnes

Where have you gotten that info. from? Have you tested a Riesel and a Proth prime of the same size?

It has been my experience with LLR that Riesel's test faster than Proth's.

Or are you referring to Phrot in this case?

I haven't done any actual testing, though I've heard this in a number of places. And, Mini-Geek's testing seems to bear out this theory.

[rogue]

Here is a list of changes:

• prpserver: Add client ID to log files.
• prpserver: Remove support for 1.x clients.
• prpclient: Pass correct flag to PFGW for Primorial/Factorial primality tests.
• prpclient: Set residue to PRIME when PFGW does primality test and number is prime.
• prpclient: Delete save file if there are no work units.
• prpclient: Use PFGW for non-base 2 numbers as v3.1 is faster than LLR and phrot.

You can d/l it from here.