Validation problems (update) - phase 1
log in

Advanced search

Message boards : News : Validation problems (update) - phase 1

Author Message
Profile valterc
Project administrator
Project tester
Send message
Joined: 30 Oct 13
Posts: 320
Credit: 16,279,659
RAC: 4,018
Italy
Message 281 - Posted: 9 Jan 2014, 17:58:21 UTC
Last modified: 13 Jan 2014, 15:14:13 UTC

The validation problem has been (theoretically) solved.

The problem: We noticed that results coming from Windows computers were slightly different than results coming from other machines, thus not validating, when compared with each other, even if the results are BOTH correct.

The solution: We enabled homogeneous redundancy for the above (two) classes of computers. This way we may just use the standard bitwise validator for doing the job.

Timing: Done. Let's check if it works.

All received workunits were checked for errors and eventually validated (with a bonus)

Profile valterc
Project administrator
Project tester
Send message
Joined: 30 Oct 13
Posts: 320
Credit: 16,279,659
RAC: 4,018
Italy
Message 282 - Posted: 9 Jan 2014, 18:04:04 UTC - in response to Message 281.

After a meeting I had with members of the application group we decided to modify the application code
- forget about 'scores' and md5 checksums
- limit the output to just one file.
(remember to modify the I/O profiles when doing the version upgrade)
- fix some annoying bugs ("writed" :)

they will also try, once again, to compile the Windows application in a gcc/mingw environment. Hopefully this will allow use to also forget about homogeneous redundancy. We marked this issue as 'low priority'.


Post to thread

Message boards : News : Validation problems (update) - phase 1


Main page · Your account · Message boards


Copyright © 2017 CNR-TN & UniTN