mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > PrimeNet

Reply
 
Thread Tools
Old 2018-11-22, 13:40   #1
Chuck
 
Chuck's Avatar
 
May 2011
Orange Park, FL

15438 Posts
Default Prime95 29.5 build 4 not working for me

This version looked like it was going to be OK, but when it got to the end of the PRP test it indicated an error and then started outputting a continuous stream of errors for the next queued test.

I'm going back to 29.4 again.
Attached Thumbnails
Click image for larger version

Name:	Clip0003.jpg
Views:	70
Size:	528.4 KB
ID:	19310  
Chuck is offline   Reply With Quote
Old 2018-11-22, 14:53   #2
GP2
 
GP2's Avatar
 
Sep 2003

29·89 Posts
Default

Quote:
Originally Posted by Chuck View Post
This version looked like it was going to be OK, but when it got to the end of the PRP test it indicated an error and then started outputting a continuous stream of errors for the next queued test.

I'm going back to 29.4 again.
Can you restart from the savefile? Maybe only part of the memory is corrupted, and the savefile is salvageable.

How many cores are you using? The problem might be worse if you have many cores. I am running it on a few dozen 2-core virtual machines (with hyperthreading enabled, because it benchmarks better that way for very small PRP exponents below 10M), and so far no problems.
GP2 is offline   Reply With Quote
Old 2018-11-22, 15:19   #3
Chuck
 
Chuck's Avatar
 
May 2011
Orange Park, FL

86710 Posts
Default

Quote:
Originally Posted by GP2 View Post
Can you restart from the savefile? Maybe only part of the memory is corrupted, and the savefile is salvageable.
No I erased everything and restarted.
Chuck is offline   Reply With Quote
Old 2018-11-22, 15:24   #4
GP2
 
GP2's Avatar
 
Sep 2003

A1516 Posts
Default

Quote:
Originally Posted by Chuck View Post
No I erased everything and restarted.
Can you recover it somehow, from your Recycle bin?

It's PRP, so Gerbicz error checking will make the result extremely reliable, if you can salvage it and complete it. I can run a double-check on it if you want.

And if it can't be restarted from the savefile, send the savefile to George and maybe it might at least be helpful for debugging purposes.


Edit: hmmm, from your attachment, the PRP test of M87883681 died at iteration 238824 [100%]. I saw the "100%" and figured the test was nearly complete, but that's probably more memory corruption, since the top part of the attachment clearly shows the test just started.

Last fiddled with by GP2 on 2018-11-22 at 15:28
GP2 is offline   Reply With Quote
Old 2018-11-22, 23:33   #5
Chuck
 
Chuck's Avatar
 
May 2011
Orange Park, FL

36316 Posts
Default

Quote:
Originally Posted by GP2 View Post
Edit: hmmm, from your attachment, the PRP test of M87883681 died at iteration 238824 [100%]. I saw the "100%" and figured the test was nearly complete, but that's probably more memory corruption, since the top part of the attachment clearly shows the test just started.
Yes it had only run a short time so nothing was lost.
Chuck is offline   Reply With Quote
Old 2018-11-22, 23:56   #6
GP2
 
GP2's Avatar
 
Sep 2003

29×89 Posts
Default

Quote:
Originally Posted by Chuck View Post
Yes it had only run a short time so nothing was lost.
Out of curiosity, what does your local.txt file looks like (how many cores and workers) and what CPU you are using?
GP2 is offline   Reply With Quote
Old 2018-11-23, 03:30   #7
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

2×43×83 Posts
Default

Quote:
Originally Posted by GP2 View Post
Out of curiosity, what does your local.txt file looks like (how many cores and workers) and what CPU you are using?
Yes, I think this might be important for trying to reproduce the problem. It looks like the exact same stack variable got roached. This increases the chance we can reproduce the issue.
Prime95 is online now   Reply With Quote
Old 2018-11-23, 04:20   #8
kladner
 
kladner's Avatar
 
"Kieren"
Jul 2011
In My Own Galaxy!

270416 Posts
Default I have sudden onset-problems with v295b4

I went to work this morning, with M48405113, a DC, near completion. When I got home in late afternoon, I found a 1st time LL running. I realized that when I changed to 295b4, I had not changed the default work type to DC. But worktodo.txt and Status showed a DC assignment ahead of the LL.

More investigation showed tightly packed errors almost instantly after the Jacobi check on the M50852xxx DC. I tried trashing the save files, and got the same error stream which was taking place in single digit iterations.

There are a string of P95 outputs in the ZIP file. I tried to section and label the experiments. Eventually, I commented the entire worktodo.txt, got new assignments, which came in the 48M range. Two of them started and proceeded without problems. The 87.9M LL runs fine.

I have tried with a new directory with a newly extracted P95. Much of it is outlined by the notes to the window copy string.

This is an i7-6700K, running at 4300mhz at first, later at 4200mhz, with no change on the problem Exponent. 16GB RAM, 3200Mhz, relaxed timings. I will provide other data on request.

The system has a spotty record, mostly because I mess with it too much.
See .txt attachment.
NOTE that I do not care about this assignment. I will give the exponent to anyone who wants it, especially if they can make it run.
Attached Files
File Type: zip P95 errors on startup.zip (1.3 KB, 27 views)
File Type: txt recent results.txt (6.3 KB, 30 views)

Last fiddled with by kladner on 2018-11-23 at 04:38 Reason: 'm'
kladner is offline   Reply With Quote
Old 2018-11-23, 05:55   #9
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

157428 Posts
Default

Quote:
Originally Posted by kladner View Post
This is an i7-6700K,
I may have accidentally broken AVX FFTs. I'll look at it.

Edit: Yes, the 2688K AVX FFT is broken.

Edit2: AVX FFT lengths that are divisible by 7 are broken. I have a fix and will gen 29.5b5 soon.

Last fiddled with by Prime95 on 2018-11-23 at 06:34
Prime95 is online now   Reply With Quote
Old 2018-11-23, 12:56   #10
kladner
 
kladner's Avatar
 
"Kieren"
Jul 2011
In My Own Galaxy!

270416 Posts
Default

Quote:
Originally Posted by Prime95 View Post
I may have accidentally broken AVX FFTs. I'll look at it.

Edit: Yes, the 2688K AVX FFT is broken.

Edit2: AVX FFT lengths that are divisible by 7 are broken. I have a fix and will gen 29.5b5 soon.
Thanks, George.
kladner is offline   Reply With Quote
Old 2018-11-23, 14:21   #11
Chuck
 
Chuck's Avatar
 
May 2011
Orange Park, FL

3·172 Posts
Default

Quote:
Originally Posted by GP2 View Post
Out of curiosity, what does your local.txt file looks like (how many cores and workers) and what CPU you are using?
Here they are.
Attached Thumbnails
Click image for larger version

Name:	cpu.jpg
Views:	57
Size:	165.3 KB
ID:	19317  
Attached Files
File Type: txt local.txt (632 Bytes, 34 views)
Chuck is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Hibernate cycle when working with Prime95? iamue Information & Answers 4 2017-08-09 05:15
Prime95 64-bit v27.9 Build 1 BSOD storm5510 Software 9 2013-12-19 04:21
Log Out not working? cheesehead Forum Feedback 1 2012-03-19 17:13
Prime95 has Stopped working SeeD419 Software 2 2011-07-27 19:05
Help restarting prime95 version 25.6 build 6 stars10250 Software 9 2008-08-25 20:26

All times are UTC. The time now is 18:20.

Sat Oct 24 18:20:55 UTC 2020 up 44 days, 15:31, 0 users, load averages: 1.93, 2.03, 1.95

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.