mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   GPU to 72 (https://www.mersenneforum.org/forumdisplay.php?f=95)
-   -   Extremely lucky assignments (https://www.mersenneforum.org/showthread.php?t=20158)

apocalypse 2015-04-06 14:59

Extremely lucky assignments
 
I just checked my results and discovered that I found factors for 10 exponents in the last 24 hours (vs my usual rate of ~2/week) :w00t:

[url]http://www.gpu72.com/reports/worker/8e0582ce97835362f146098b4e529221/[/url]

kladner 2015-04-06 15:06

[QUOTE=apocalypse;399469]I just checked my results and discovered that I found factors for 10 exponents in the last 24 hours (vs my usual rate of ~2/week) :w00t:

[URL]http://www.gpu72.com/reports/worker/8e0582ce97835362f146098b4e529221/[/URL][/QUOTE]

Wow! Seeing even two in a day is pretty remarkable for me!

MatWur-S530113 2015-04-06 18:03

:tu:
yes that looks nice^^
and you raised above the expectet number of factors (100 vs 94.6)
I myself never found more than 2 factors/day with mfaktc

Mark Rose 2015-04-06 18:28

[QUOTE=apocalypse;399469]I just checked my results and discovered that I found factors for 10 exponents in the last 24 hours (vs my usual rate of ~2/week) :w00t:

[url]http://www.gpu72.com/reports/worker/8e0582ce97835362f146098b4e529221/[/url][/QUOTE]

I left those for you :D

That's very unusual. Even though I have roughly triple the throughput, and I'm working at a lower bit depth, I don't believe I've ever found ten in a day.

apocalypse 2015-04-07 01:54

[QUOTE=Mark Rose;399491]I left those for you :D

That's very unusual. Even though I have roughly triple the throughput, and I'm working at a lower bit depth, I don't believe I've ever found ten in a day.[/QUOTE]

I went back to look at the mfakto output logs and, even more surprisingly, I had 10 consecutive assignments with a factor found between 2^71 and 2^74 and the total time from the first reported factor [Sun Apr 5 16:55:30 2015] to the last one [Sun Apr 5 23:34:51 2015] was under 7 hours. :shock:

LaurV 2015-04-07 03:31

No mystery. There was a time when (due to a bug into PrimeNet intestines, i.e. the part that checks the validity of the factors was having indigestion) you could not report "factor found" lines, but only "no factor found". The disease persisted until James/George gave it some medicine. Especially when the server was busy, because the "factor" lines are reported at the end, if the communication crashed on the way, only "no factor" lines were registered. The work which resulted in factors was still registered as not being done, and re-assigned to the next "lucky" guy like you :smile:

There are another 3 or 4 threads here around about these "lucky strikes" and about fixing the problem. The "unlucky" guys who could not report the factors (and did not realize) were in the group that was not using Misfit, because Misfit pops out red windows when he can't register a factor.

What you can do now, if you are curious, you can check the exponent status page, for the range around the exponents you got factors in a row, and see who was the unlucky guy who missed to report the factors.

There was a time when we used to hunt for these errors, like for example, look to "how deep factored" page, and if you see a 71 in the middle of 72's, then there was a mismatched report, and there is a high probability a factor is laying there, which could not be reported at some time in the past, by some unlucky cruncher. In this case, you could do the assignment by yourself and if (most probably) you find a factor, report it and be happy.

I always check if my factors went through, in spite of the fact that Misfit checks it too and pops out red windozes. Just a bit of paranoia...

apocalypse 2015-04-07 04:41

Hmmm...
At first I thought this was unlikely because all 10 of the exponents were assigned to me for the bit levels 71-75. For 7 of the exponents, I completed 1 or 2 bit levels with no factors found, indicating that the ranges weren't partially cleared.

But poking around, it looks like Bill Staffen was reporting no factor ranges like "no factor from 2^71 to 2^74", so presumably an unreported found factor wouldn't clear the lower bit levels in the range.

Sorry, Bill!


All times are UTC. The time now is 03:00.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.