Prime95 version 27.6

[Prime95]

Originally Posted by Dubslow

Edit2: Notice that in both my case and Brian's, the roundoff was not even less than 1, rather it was E3, E4, E75. How could that possibly be a valid roundoff?

Problem found!! Thanks for uncovering it. I'll get a new build uploaded as soon as possible.

The bug was improper initialization of the maximum round off error. This code changed in 27.6 with the upgraded multithreading code.

Note: I'm somewhat surprised the problem didn't occur more often.

[BrianOC]

Now this is hardware or software failure?


Edit:

Originally Posted by Prime95

Problem found!! Thanks for uncovering it. I'll get a new build uploaded as soon as possible.

The bug was improper initialization of the maximum round off error. This code changed in 27.6 with the upgraded multithreading code.

Note: I'm somewhat surprised the problem didn't occur more often.

[Dubslow]

Originally Posted by Prime95

Problem found!! Thanks for uncovering it. I'll get a new build uploaded as soon as possible.

The bug was improper initialization of the maximum round off error. This code changed in 27.6 with the upgraded multithreading code.

Note: I'm somewhat surprised the problem didn't occur more often.

FWIW, I was doing strictly single threaded tests (not even HT, as I said). (I think Brian was doing the same, one worker per HT.)

[Prime95]

Originally Posted by Dubslow

FWIW, I was doing strictly single threaded tests (not even HT, as I said). (I think Brian was doing the same, one worker per HT.)

Yes, the torture test is only single-threaded, but both multi-threaded and single-threaded use the new initialization code.

Build 3 is now uploaded. Let me know if I goofed on the fix.

[Dubslow]

Originally Posted by Prime95

Build 3 is now uploaded. Let me know if I goofed on the fix.

Well... at least it got to 4.3/4.6 MB before it hanged. Trying dabaichi's pause trick now.

[Jwb52z]

Why am I getting a Primenet Error 9 and Untrusted software versions currently excluded by Primenet when trying to use 27.6?

[Redarm]

Originally Posted by Prime95

Note: I'm somewhat surprised the problem didn't occur more often.

Had the same problem here, now fixed

[Dubslow]

Originally Posted by Jwb52z

Why am I getting a Primenet Error 9 and Untrusted software versions currently excluded by Primenet when trying to use 27.6?

Hmm... I reported a result last night just fine with 27.6. Are you using the new build (already)? If so, give it an hour for PrimeNet's database to catch up.

Edit: After a while the DL said "interrupted". I'll keep trying...
Edit2: It worked Testing now.
3: 5 minutes into the small FFT test, no errors... I'll let it run a half hour. Thanks to Brian for posting about it ('cause sometimes, it isn't hardware errors )
4: 30 minutes of Small FFT, no errors. Reverting to production

[Jwb52z]

I went back to 27.4 for now until it gets fixed better.

[BrianOC]

Originally Posted by Dubslow

Hmm... I reported a result last night just fine with 27.6. Are you using the new build (already)? If so, give it an hour for PrimeNet's database to catch up.

Edit: After a while the DL said "interrupted". I'll keep trying...
Edit2: It worked Testing now.
3: 5 minutes into the small FFT test, no errors... I'll let it run a half hour. Thanks to Brian for posting about it ('cause sometimes, it isn't hardware errors )

Yees not hardware failure huuh New bulid work i now tested no problems. Nice! Thanks for help all!

[Jwb52z]

I am currently on my new laptop and I think that the reason I may have gotten the error with 27.6, I have since switched back to 27.4 for now, is that 27.4 reports my i7 2670QM as having AVX processor cores. I remember vaguely reading something about that causing problems. I will try 27.6 again in a few days to see if something else I might have done wrong caused it to react with the errors I mentioned and then I'll report back.