LLRnet server rally March 8th-9th

[mdettweiler]

Not meaning to change the subject, but which server does everybody think we should have our next rally on? Should we do port 5000 again to push 300<k<400 closer to 600K? Or should we do Carlos/AES's port 300 servers for 400<k<1001? We could, alternately, do ports 500 and 100 to help clean up "somebody's mess".

Opinions, anyone?

[IronBits]

Originally Posted by em99010pepe

His garage:

That's only TWO of the FIVE Towers (500GHz total).
He's still building them

And no, there is no llrnet server software that can handle that load.

Beyond or Mr. Sheep can bring down the server, any llrnet server, anytime, with both hands tied behind their backs.

[mdettweiler]

Originally Posted by IronBits

That's only TWO of the FIVE Towers (500GHz total).
He's still building them

And no, there is no llrnet server software that can handle that load.

Beyond or Mr. Sheep can bring down the server, any llrnet server, anytime, with both hands tied behind their backs.

Hmm...maybe he can put at least a fraction of his power on LLRnet, then?

[IronBits]

Originally Posted by Anonymous

80+ cores--hey IronBits, I don't see 'em here!
Is NPLB not listed on that page for whatever reason?

I have 10 PS3s, with 7 3.2GHz cores each and 3 quads, one of them is the linux server.
Get someone to port llrnet to the PS3 platform, then we will see what happens :p

When I get a chance, I'll add NPLB to the project listing.
We don't add projects unless someone requests it as there are just too many and, some day it will get re-written.

[IronBits]

Originally Posted by Anonymous

Hmm...maybe he can put at least a fraction of his power on LLRnet, then?

That remains to be seen. Seeing that I help manage it remotely from time to time ;)

[mdettweiler]

Originally Posted by IronBits

I have 10 PS3s, with 7 3.2GHz cores each and 3 quads, one of them is the linux server.
Get someone to port llrnet to the PS3 platform, then we will see what happens :p

When I get a chance, I'll add NPLB to the project listing.
We don't add projects unless someone requests it as there are just too many and, some day it will get re-written.

You mean that the LLRnet port 5000 server is actually a PS3?

As for porting LLRnet to the PS3 platform--well, first we'd need to port LLR to the PS3, which is actually kinda hard because the SPE's (where most of the PS3's power comes from) do only single-precision floating-point operations (I think). Not to mention that LLR contains a lot of x86-only assembly code that would have to be written from scratch for any other architecture.

Still, I'm sure it would be possible--maybe it could be based off of the Mlucas PowerPC software (Mlucas does LL testing, which is like LLR but for Mersenne numbers).

Originally Posted by IronBits

That remains to be seen. Seeing that I help manage it remotely from time to time ;)

Hmm, I see.

[kar_bon]

another point related to the race!

i read about the prime found by BlisteringSheep.
he mentioned himself this prime was not found during the race!

i counted that prime because of the submission date/time!

to Blistering:
so how do i handle this? let the prime in or not?

[gd_barnes]

Originally Posted by kar_bon

another point related to the race!

i read about the prime found by BlisteringSheep.
he mentioned himself this prime was not found during the race!

i counted that prime because of the submission date/time!

to Blistering:
so how do i handle this? let the prime in or not?

If it was submitted during the rally, it should count, regardless if the person who found it realized it or whether they had double-checked it.

But the time stamp was 17:05 Eastern daylight time U.S. on Sunday. That would equate to 21:05 GMT, more than 2 hours after the race. I guess I'm confused here.

Perhaps the time stamp was for the double check and the prime had already been submitted to the server earlier. Or maybe there's some techie stuff that I don't understand related to his proxy server.

I guess the question would be...were the k/n pair tests directly before and after the prime counted in the rally because they were submitted during the rally? If so, the prime should certainly count in the rally.


Gary

[mdettweiler]

Originally Posted by gd_barnes

If it was submitted during the rally, it should count, regardless if the person who found it realized it or whether they had double-checked it.

But the time stamp was 17:05 Eastern daylight time U.S. on Sunday. That would equate to 21:05 GMT, more than 2 hours after the race. I guess I'm confused here.

Perhaps the time stamp was for the double check and the prime had already been submitted to the server earlier. Or maybe there's some techie stuff that I don't understand related to his proxy server.

I guess the question would be...were the k/n pair tests directly before and after the prime counted in the rally because they were submitted during the rally? If so, the prime should certainly count in the rally.


Gary

I think you misunderstood what Sheep meant by a doublecheck. He found the prime with LLRnet, and then the prime was submitted to the server--later, he doublechecked it with LLR 3.7.1c, which of course wouldn't have been submitted to the server since it was done with standalone LLR.

Thus, the question simply is: is the date on the prime in the LLRnet results file during the rally, or not? That alone will determine whether the prime is to be counted with the rally.

[kar_bon]

so this is what i got from the result-files:

user=kar_bon
[03/09/08 14:03:47]
363*2^387674-1 is not prime.  Res64: 91428AF436339CB7  Time : 165.0 sec.
user=Flatlander
[03/09/08 14:04:06]
357*2^387598-1 is not prime.  Res64: D574F5D9A6B81087  Time : 1285.0 sec.
user=ArsTechnica_BlisteringSheep
[03/09/08 14:04:18]
309*2^387221-1 is prime!  Time : 5419.0 sec.
user=gd_barnes
[03/09/08 14:04:35]
319*2^387557-1 is not prime.  Res64: 0A0466CD84EE5246  Time : 1925.0 sec.
user=Free-DC_IronBits
[03/09/08 14:04:42]
315*2^387677-1 is not prime.  Res64: 99ECF218B40FCBE7  Time : 162.0 sec.

in my spreadsheet Blisterings prime-pair is at position 21486 from 25351 totally.

post #328

user=ArsTechnica_BlisteringSheep
[Sun Mar 9 17:05:48 2008]
309*2^387221-1 is prime!

i don't know where the time "17:05" come from? local time?

The time of 5419s says, that from handing out this pair from the server to the client to submitting back to the server there were 90min between.

so this prime will count due to the date/time and position in IronBits result file!

karsten

[mdettweiler]

Originally Posted by kar_bon

so this is what i got from the result-files:

user=kar_bon
[03/09/08 14:03:47]
363*2^387674-1 is not prime.  Res64: 91428AF436339CB7  Time : 165.0 sec.
user=Flatlander
[03/09/08 14:04:06]
357*2^387598-1 is not prime.  Res64: D574F5D9A6B81087  Time : 1285.0 sec.
user=ArsTechnica_BlisteringSheep
[03/09/08 14:04:18]
309*2^387221-1 is prime!  Time : 5419.0 sec.
user=gd_barnes
[03/09/08 14:04:35]
319*2^387557-1 is not prime.  Res64: 0A0466CD84EE5246  Time : 1925.0 sec.
user=Free-DC_IronBits
[03/09/08 14:04:42]
315*2^387677-1 is not prime.  Res64: 99ECF218B40FCBE7  Time : 162.0 sec.

in my spreadsheet Blisterings prime-pair is at position 21486 from 25351 totally.

post #328

user=ArsTechnica_BlisteringSheep
[Sun Mar 9 17:05:48 2008]
309*2^387221-1 is prime!

i don't know where the time "17:05" come from? local time?

The time of 5419s says, that from handing out this pair from the server to the client to submitting back to the server there were 90min between.

so this prime will count due to the date/time and position in IronBits result file!

karsten

Oh, I get it now what that 17:05 came from: Sheep's proxy! That must have been posted from his proxy's results file.

In that case, since the result in IronBits' results file was definitely during the rally, the prime should be counted with the rally.