mersenneforum.org Prime95 version 26.2
 Register FAQ Search Today's Posts Mark Forums Read

 2010-10-01, 18:22 #111 ewmayer ∂2ω=0     Sep 2002 República de California 3·3,877 Posts Just got a new Win7-running quad-core Intel laptop at work, downloaded v26.2 and started it off on 2 cores (see how the fan takes that for a few days). I started it off on a partially-completed DC assignment ~25M, most interesting to me was this slightly cryptic note in the client diagnostic-echo display: "using Core2 type-3 FFT length 1344K, Pass1=896, Pass2=1536" So it looks like some kind of hybrid of previously-supported FFT lengths [896K = 7*2^17] and [1536K = 3*2^19] is being used to effect the equivalent of a single length [1344K = 3*7*2^15] transform. George, care to elaborate?
2010-10-01, 19:31   #112
Prime95
P90 years forever!

Aug 2002
Yeehaw, FL

22×1,873 Posts

Quote:
 Originally Posted by ewmayer "using Core2 type-3 FFT length 1344K, Pass1=896, Pass2=1536" So it looks like some kind of hybrid of previously-supported FFT lengths [896K = 7*2^17] and [1536K = 3*2^19] is being used to effect the equivalent of a single length [1344K = 3*7*2^15] transform. George, care to elaborate?
Um, 896 = 3*2^7, 1536=3*2^9 to effect a 1536K=7*3*2^16.

Oversimplifying:

25.11 supported a 5,6,7,or 8 reals first step followed by any number of 2 or 4 complex steps. This gave us 5,6,7,8 * any power of 2 FFT lengths.

26.2 has new building blocks for 3 and 5 complex values. These can appear (with some limitations for easy looping constructs and cache line padding) at any point in pass 1 or 2. The pass 1 sizes supported are 5,6,7,8 * 3,4,5 * a power of two. The pass 2 sizes supported are 3,4,5 * 3,4,5 * a power of two. Obviously this leads to many combinations and many new FFT lengths.

2010-10-01, 21:47   #113
chessmc

Sep 2010

3 Posts

Quote:
 Originally Posted by Prime95 This bug has been around for ages. I've never been able to find it. By any chance did you stop and restart during stage 2? Did you change memory settings (or did you get a restart due to a day/night memory changeover)?
I have noticed this bug as well. It appears to occur when stage 2 restarts with lower memory settings, either from the day/night memory changeover or another worker entering stage 2.

2010-10-01, 23:28   #114
ewmayer
2ω=0

Sep 2002
República de California

3×3,877 Posts

Quote:
 Originally Posted by Prime95 Um, 896 = 3*2^7, 1536=3*2^9 to effect a 1536K=7*3*2^16.
I think you mean "896 = 7*2^7, 1536=3*2^9 to effect a 1344K=7*3*2^16". ... and now I understand my misunderstanding: I thought there was an implied "K" after those FFT lengths, and that you had found some funky way to combine 2 different-length FFTs bracketing the normal 1280K length needed for the exponent in question.

But you're referring to your usual 2-pass FFT, using length 896 * 1536. Got it.

2010-10-02, 16:14   #115
Prime95
P90 years forever!

Aug 2002
Yeehaw, FL

1D4416 Posts

Quote:
 Originally Posted by Mini-Geek On further investigation, I've found that it is only on the calculation of optimal bounds that it incorrectly assumes it can have the full amount. When you actually get to the part where it uses the memory, it splits it using the worker-specific settings that are in local.txt.
That makes sense. I'll fix the optimal bounds calculations.

2010-10-02, 17:04   #116
Xyzzy

"Mike"
Aug 2002

8,167 Posts

Quote:
 That makes sense. I'll fix the optimal bounds calculations.

 2010-10-04, 07:59 #117 lpmurray     Sep 2002 8910 Posts mersenne M45153469 I upgraded to 26.2 build 1 about halfway through M45153469. When it finished it downloaded a new number but never uploaded the results how do I get credit for it nowadays here is what was posted in my results file. [Mon Sep 27 22:20:59 2010] M45153469 is not prime. Res64: XXXXXXXXXXXXXXXX. We1: 7F42685C,27479492,00000000, AID: 00000000000000000000000002B0FCBD Last fiddled with by garo on 2010-10-04 at 08:41
 2010-10-04, 08:05 #118 lycorn     Sep 2002 Oeiras, Portugal 3×487 Posts Will some mod mask the residue, please?
 2010-10-04, 08:42 #119 garo     Aug 2002 Termonfeckin, IE 22×691 Posts You can go to the manual results submission page after logging into your account and submit the results there. http://www.mersenne.org/manual_result/ Have you tried forcing Manual Communication? Advanced->Manual Communication->Contact Primenet now. Last fiddled with by garo on 2010-10-04 at 08:44
2010-10-04, 12:25   #120
Mini-Geek
Account Deleted

"Tim Sorbera"
Aug 2006
San Antonio, TX USA

426710 Posts

Quote:
I see that garo edited it to mask it, but the last two (or more, or all) hexdigits should be masked, not the first twelve. That's how GIMPS hides them in the reports so that people can't fake it. If the last four are visible here, once it goes into the report, people can put it together to make the whole residue and submit it without doing any work.

Last fiddled with by Mini-Geek on 2010-10-04 at 12:26

 2010-10-04, 21:00 #121 lpmurray     Sep 2002 89 Posts M45153469 thanks for the help i went to the manual results page and sent it up sorry for being so dumb, I can't believe people are so petty about increasing the score that the would steal other peoples work. I have been doing this for over 10 years and like to see my credit go up but only as a personal accomplishment of how I've given to this great community. I figure through the years its cost me several thousand dollars and before my injury 8 years ago i bought 4 duel 2.8 processor boxes, its funny that now my duel core laptop is twice as fast as running a number on both processors of the old boxes. I can't wait for the time when the amd 12cores come down in price i would love to put together a 48core box it would clear 16-24 numbers a month. Nice to dream.

 Similar Threads Thread Thread Starter Forum Replies Last Post Prime95 Software 148 2012-03-18 19:24 Prime95 Software 76 2010-12-11 00:11 Prime95 PrimeNet 369 2008-02-26 05:21 Prime95 PrimeNet 143 2007-09-24 21:01 pacionet Software 74 2006-12-07 20:30

All times are UTC. The time now is 13:01.

Sun May 16 13:01:14 UTC 2021 up 38 days, 7:42, 0 users, load averages: 2.43, 2.33, 2.35