Seite 1 von 1

Report on the new stage of Odd Weird Search

Verfasst: 24.02.2014 21:42
von fwjmath
Hi everyone,

First of all, thank you a lot for contributing to Odd Weird Search! With your help, we have finished the search for numbers under 10^21, two orders of magnitude greater than the previous record. We can never do that without your help!

Here are some numbers of your contribution:
About 530000 workunits are released in total
24298561219965 = 2.43e13 abundant numbers are checked thorough the search, averaging to about 4.60e7 numbers per workunit
It will take roughly 100 years to finish the whole work on a single core

Since we haven't found our cunning little target, we will now proceed to a new stage of our search. This time, we will search for numbers up to 10^28, much higher than the range before, thus increasing (I hope) the chance of finding an odd weird number.

But there is a catch. If we do just as before, even with BOINC we cannot handle the work in manageable time. There are some compromise to take. We cannot really search for every number in the range, that would take too long. We can just look for some of them. So we choose to look only at numbers with lower abundance, i.e. a number with the sum of its factors not much larger than twice of it. With a heuristic argument, we are more or less sure that this choice will not lower the probability of finding one odd weird number by much.

In the case of this stage, we will only check for numbers with abundance less than 10^14. They will be bundled into about 48000 workunits, and as always the workunits are recycled for a better crunching experience. Since we only check the part of numbers where an odd weird one would most hide, even though we search for a much higher range, the overall workload may not increase much. Considering the time taken for the current stage, we might even have a third stage of searching.

Now currently we only have linux application, but we are working on a Windows one. So Windows users, stay tuned!

All in all, enjoy the new hunt for numbers and for badges!

fwjmath.

Re: Report on the new stage of Odd Weird Search

Verfasst: 24.02.2014 23:07
von Ananas
fwjmath hat geschrieben:...
Now currently we only have linux application, but we are working on a Windows one. So Windows users, stay tuned! ...
This explains why my XP x64 box produces nothing but "division by zero" errors - but then ... why do I receive work at all on that box?

oops ... found the answer myself, an old app_info.xml from the time when we tested the fixed windows application (truncated output values on XP) - sorry, my bad :-/

Re: Report on the new stage of Odd Weird Search

Verfasst: 26.02.2014 06:59
von DoctorNow
fwjmath hat geschrieben:All in all, enjoy the new hunt for numbers and for badges!
If the WUs were counted...
Currently it looks like they don't. I have already returned several valid ones since yesterday but the subproject count for OWS did not increase yet!
Please fix this.

Re: Report on the new stage of Odd Weird Search

Verfasst: 26.02.2014 19:55
von yoyo
The backend which collects the results and handles the batch counter was not running. Now it should work and all remaining credits are added to the batch counter.

Re: Report on the new stage of Odd Weird Search

Verfasst: 01.03.2014 00:35
von zombie67
Out of work?

Re: Report on the new stage of Odd Weird Search

Verfasst: 01.03.2014 07:33
von yoyo
No, we had a bug in the App and restarted the search and cancelled the old workunits. Currently we have only a Windows App, but the Linux version will follow in the next days.
If anybody has still version 1.01 of the App running, please abort the workunits.

yoyo

Re: Report on the new stage of Odd Weird Search

Verfasst: 01.03.2014 15:27
von zombie67
Ah. That explains the 24 hour back off my linux machines. Got it.

Re: Report on the new stage of Odd Weird Search

Verfasst: 05.03.2014 07:11
von DoctorNow
yoyo hat geschrieben:No, we had a bug in the App and restarted the search and cancelled the old workunits.
If anybody has still version 1.01 of the App running, please abort the workunits.
Nice...
I guess that explains why there are a lot of completed WU-packs which still are pending (like this, this or this one).
Do we get at least some credit for it?

Re: Report on the new stage of Odd Weird Search

Verfasst: 05.03.2014 08:34
von yoyo
These results didn't match, they are not valid. Means 2 people were running the same computation and at the end the results are different. Would be not OK to give credits for inconclusive results.

Re: Report on the new stage of Odd Weird Search

Verfasst: 05.03.2014 12:12
von DoctorNow
Hm, ok, when there were miscalculations it's too bad but understandable - crunched a lot of WUs for nothing then. ;-)
First I thought it could be the validator because all WUs were pending.
You should get rid of these from the database as soon as possible so that there is a better overview. I have so much pending it isn't viewable which are wrong and which are normal now.

Re: Report on the new stage of Odd Weird Search

Verfasst: 05.03.2014 14:06
von fwjmath
I am very sorry for the bug. It is due to a new mechanism I introduced to shorten periods between checkpoints. But it seems that my tests are not so complete. The application now has the buggy mechanism removed, the checkpoint period is a bit longer (3 minutes or more), but should work well as in the previous run. However I am not sure how the situation will be in a foreseeable future run. We will then see if it is necessary to work more on it.

Re: Report on the new stage of Odd Weird Search

Verfasst: 25.10.2014 17:46
von Werinbert
Now that 10^28 is nearing conclusion are there any plans to continue looking at numbers beyond 10^28 or filling in the gaps of the 10^28 search?