Secret residues

[Keroberts1]

how many secret tests have been done above 2,000,000? Have any of the residue's come back not matcing? Just wondering if we have any info on the error rate of the tests yet.

[Death]

don't want to create a new topic, so please tell me something.

now I can works with SoB. so I left f-a-d and start 12oB.

username is secret. team Ukraine.

I just want to know. Is there any reason to count secret tests? or it's just fun?
How much secret tests left?
Where's http://www.seventeenorbust.com/supersecret ?
Is supersecrets and secrets merged?
Will my secret jobs count on team scores?

[Mystwalker]

Originally posted by Death
don't want to create a new topic, so please tell me something.

now I can works with SoB. so I left f-a-d and start 12oB.

11oB meanwhile... ;)


username is secret. team Ukraine.
I just want to know. Is there any reason to count secret tests? or it's just fun?

Do you mean 'count' or 'do'?


How much secret tests left?

Well, http://www.seventeenorbust.com/secret/ shows that there are a little less than 30,000 secret tests left. Then, all tests n<1M and 2M<n<3M are complete.


Where's http://www.seventeenorbust.com/supersecret ?

Two more columns ("Supersecret Tests", "Supersecret Upper") and "Checked Twice" --> "Supersecret Lower" should do the trick...
Well, maybe it's possible to change the "Secret Lower" column so that there are no n/a's any more, but the last test done here instead.


Is supersecrets and secrets merged?

AFAIK, a Supersecret test exists when there has been only one residue so far. This can happen when the first searcher doesn't have the residues any more...


Will my secret jobs count on team scores?

Nope. Secret and Supersecret count neither to individual nor to team score. Well, it would be kinda unfair anyway, as "normal" tests take a lot more time.


Hope I could help you answering your questions. :thumbs:

[Keroberts1]

but hav any non-matchng residues come back yet?

[Death]

Originally posted by Mystwalker
11oB meanwhile... ;)

Do you mean 'count' or 'do'?

Well, doing them.
Now I'm finishing 21181. Then I'll switch to 11oB for my account and my team. Want some scores.




Nope. Secret and Supersecret count neither to individual nor to team score. Well, it would be kinda unfair anyway, as "normal" tests take a lot more time.

Hope I could help you answering your questions. :thumbs:

What a pity.

Thank you very much for info. :cheers:

[Mystwalker]

Originally posted by Death
Well, doing them.

It's possible that some calculations were wrong. Double-checking them tests give information
a) which were wrong (either the first test or the double-check, so a third calculation is needed)
b) what is the average error rate.


but hav any non-matchng residues come back yet?

Hm, good question. I guess it's best to PM Louie or David...

[jjjjL]

212 numbers duplicated so far with 1 mismatch.

not enough data to infer anything.

the record for the secret test looks far more suspicious then the original. it hasn't been triple checked yet.

21181•2^2000348+1 is the offending number.

Data should be more interesting when we get closer to an FFT boundry.

Cheers,
Louie

[Keroberts1]

how does a tests record look suspicious? Are there indicateors that a certain test may be incorrect? Perhaps these should be double checked before anyothers are done. Tests don't have to be double checked sequencially.

[jjjjL]

it was suspicious only because it took quite a long time to complete for how small the test is. i've noticed that almost all mismatched residues come from the tests that took an abnormally long time to complete.

perhaps the outlying tests that took over 5x as long as similar tests around them should be retested first.

cheers,
Louie


Originally posted by Keroberts1
how does a tests record look suspicious? Are there indicateors that a certain test may be incorrect? Perhaps these should be double checked before anyothers are done. Tests don't have to be double checked sequencially.

[Death]

Originally posted by jjjjL

21181•2^2000348+1 is the offending number.

Data should be more interesting when we get closer to an FFT boundry.

Cheers,
Louie

well. one block from 21181 takes nearly hour to test on a 1.7GHz Cel.
there is 30 blocks at entire test. so it takes an ~30 hours to complete full test.

btw. is supersecret and secret accounts receive different tests?

[Nuri]

btw. is supersecret and secret accounts receive different tests? Yes.

Secret gets 2,000,000<n<3,000,000

Supersecret currently gets n<800,000 (I guess Louie will continue for numbers for 800,000<n<2,000,000 with supersecret when it reaches n values close to it).

Thus, number of remaining tests for supersecret shows number of tests remaining below 800,000 only, whereas it's 2m<n<3m for secret.

[Keroberts1]

is there any way that the client could flag results returned from a test that encountered several errors during execution? Or perhaps one that was yeilded from an over clocked PC. Some way of flagging results taht are likely to have errors? From what I've always thought a normal PC should almost never return a bad result unless something is physically wrong with it or it is overclocked.