http://wiki.seventeenorbust.com/ defaced

[jamroga]

I was trying to suggest some featues for v3...went to http://wiki.seventeenorbust.com/ only to find it has
been defaced with random links.

Here is my request for a version 3.0 (Diagnostics mode SOB)

After finding at one of my clients, some malfunctioning hardware ( that could have been bad for months or years). He asked if I could run diagnostics using idle time on the cpu.

I installed SOB with a secondpass designation, on some machines just to see if they would catch any errors. The idea would of course be to see if any secondpass calculations/residuals did not match the first pass.

I think that if the SOB client could be tweaked a little, its use as a diagnostics program could be greatly enhanced. Specifically, I could get its installation approved by almost every client on almost any machine.

Here is some tweaking that I think would make it a better diagnositic:
(Some of these have already been requested for version 3)

1) A client with a second pass only option.
2) A different machine ID/name for each install.
3) A group ID/name to associate groups of machines.
3) A status screen with statistics for each machine.
(This would help track any errors, performance issues, or downtime)
4) An E-mail notification system, in case a machine calculates
incorrectly...suddenly stops or for other problems.

Additionally, the client should probably have a more diagnostic like interface screen perhaps listing uptime, valid tests, invalid tests etc.

I am unaware currently how the client picks memory areas for calculations, but perhaps this should systematically move around so all of memory could be evenly exercised (tested).

I would also like to have the option of runing a simple CPU test/diagnostic between blocks (or otherwise periodically) just to catch any other problems. (This block of code should be available as open source (at least for intel x86)

In summary, I think a diagnostic mode for SOB could harness a significant contribution of CPU machines and time. This would be from people less interested in the prime finding purpose of the project; these people would instead view the diagnostic benefits (derived from doing second passes).

I welcome any and all feedback.

Thank You,

[Ken_g6[TA]]

Great, Wiki spam! :gangpunch

Would it be possible to have the Wiki password-protected with the same passwords used on the seventeenorbust.com site?

As far as diagnostic programs, there's already a prime-searching program that's far better at this task than SoB. GIMPS (http://www.mersenne.org/), which is sort of the parent program to SoB, includes several diagnostic modes (http://mersenneforum.org/showthread.php?t=3813) (sorry, can't find a better link) you may find useful.

[jamroga]

I was aware of the use of the GIMPS program as a diagnostic/stability tester especially for OCed (overclocked) computers. From my reading up on it, the diagnostic modes test your computer against a fixed set of known good results.
Although this works as a good or great diagnostic, it does nothing for the greater good. Our second pass retesting, used as a diagnostic would both benefit the computer user and SOB (and the world in general). I also think that GIMPS serves many masters and our smaller group of project administers/programmers would be able to a better job.

Regards,

[Matt]

The whole point of the diagnostic is that it checks against a known result so that it if you come to a different answer to the known correct result there is something wrong. Do we really want people running SoB on machines that are likely to be unstable (hence the reason to run the test) and just hand back incorrect data?

[Matt]

A wiki is user editable, this means you can fix the defacement as well , as I have just done.

[jamroga]

The plan is not to run SOB on unstable computers; this was never mentioned or implied. (In any case, the GIMPS program in diagnostic mode is already a very popular tool for overclockers to do testing)

The goal is to increase participation in this project by making our secondpass doublecheck a usefull diagnostic/integrity test. The adding in of some simple managment functions as outlined above would go a long way.

If the features were added and worked well, and were then well marketed,
we could probably increase participation by multiples. Would having
10x as many cpu's on the problem help, would 100x be helpful?

How much effort would these changes (mostly already planned for v3) take?
Would having the features make the project more appealing to a significantly larger audience?
How much would having a much larger user base help this and future projects?

Things to think about...and discuss.