Message boards : Number crunching : why my credit is lower than everyone else?
Message board moderation

To post messages, you must log in.

AuthorMessage
wantamap

Send message
Joined: 5 Sep 05
Posts: 2
Credit: 10,383
RAC: 0
Message 19171 - Posted: 6 Mar 2008, 0:04:29 UTC
Last modified: 6 Mar 2008, 0:04:57 UTC

this has been constant for my computer, I am wondering why... Anyone has any ideas? thanks!

Here is one workunit:

http://lhcathome.cern.ch/lhcathome/workunit.php?wuid=2305189
ID: 19171 · Report as offensive     Reply Quote
Profile Alex

Send message
Joined: 2 Sep 04
Posts: 378
Credit: 10,765
RAC: 0
Message 19172 - Posted: 6 Mar 2008, 6:21:12 UTC

When this project started, they were getting different results from AMD versus intel pc's, and they would have three intel machines with one result, and one or two other machines with a similar result, but outside of the matching spec of the first group of three.
There's a feature where the second group of results gets half credit.

Your result has the 'invalid' tag, because it didn't completely validate.

Most likely it's because of errors in the simulation propagating over 1,000,000 turns instead of 100,000 turns.

Possible causes of floating point errors could be:
Your floating point processor is different.. you can test it with Testfloat to see if it passes the Ieee744 spec
http://www.math.utah.edu/~beebe/software/ieee/

Your PC could be heating up. I once had and IBM which didn't have the heat sink properly seated from the factory.. so it would act up after a couple of hours.

Your PC may be overclocked. Most software is integer based and uses boolean logic, but floating point errors rarely cause a program to exit with an error. In fact you could have a wrong result in the last digit when you calculate the square root of two, but nobody would notice unless they're a super math geek.
I'm not the LHC Alex. Just a number cruncher like everyone else here.
ID: 19172 · Report as offensive     Reply Quote
Profile The Gas Giant

Send message
Joined: 2 Sep 04
Posts: 309
Credit: 715,258
RAC: 0
Message 19173 - Posted: 6 Mar 2008, 10:04:14 UTC

Try running benchmarks when nothing else is happening on the computer.....
ID: 19173 · Report as offensive     Reply Quote
wantamap

Send message
Joined: 5 Sep 05
Posts: 2
Credit: 10,383
RAC: 0
Message 19177 - Posted: 6 Mar 2008, 23:53:58 UTC - in response to Message 19172.  
Last modified: 6 Mar 2008, 23:54:17 UTC

My results are valid results, not those invalid ones.

Take a look at this WU, everyone else is getting 69.35, but I only get 34.67! why?


http://lhcathome.cern.ch/lhcathome/workunit.php?wuid=2311068





When this project started, they were getting different results from AMD versus intel pc's, and they would have three intel machines with one result, and one or two other machines with a similar result, but outside of the matching spec of the first group of three.
There's a feature where the second group of results gets half credit.

Your result has the 'invalid' tag, because it didn't completely validate.

Most likely it's because of errors in the simulation propagating over 1,000,000 turns instead of 100,000 turns.

Possible causes of floating point errors could be:
Your floating point processor is different.. you can test it with Testfloat to see if it passes the Ieee744 spec
http://www.math.utah.edu/~beebe/software/ieee/

Your PC could be heating up. I once had and IBM which didn't have the heat sink properly seated from the factory.. so it would act up after a couple of hours.

Your PC may be overclocked. Most software is integer based and uses boolean logic, but floating point errors rarely cause a program to exit with an error. In fact you could have a wrong result in the last digit when you calculate the square root of two, but nobody would notice unless they're a super math geek.


ID: 19177 · Report as offensive     Reply Quote
Profile Alex

Send message
Joined: 2 Sep 04
Posts: 378
Credit: 10,765
RAC: 0
Message 19178 - Posted: 7 Mar 2008, 3:55:02 UTC - in response to Message 19177.  
Last modified: 7 Mar 2008, 4:18:03 UTC

The WU with the 34.67 credit granted has [url]http://lhcathome.cern.ch/lhcathome/result.php?resultid=11865322
[/url] as a result.

The validate state is listed as invalid.
Validate state Invalid
Claimed credit 67.1613128942793
Granted credit 34.6737190454881

You may want to give Testfloat a try.
http://www.jhauser.us/arithmetic/TestFloat.html



C:\\testfloat\\Test>testfloat -all -errorstop
...
Testing floatx80_eq_signaling.
46464 tests total.
46464 tests performed.
No errors found in floatx80_eq_signaling.
Testing floatx80_le_quiet.
46464 tests total.
46464 tests performed.
No errors found in floatx80_le_quiet.
Testing floatx80_lt_quiet.
46464 tests total.
46464 tests performed.
No errors found in floatx80_lt_quiet.



My results are valid results, not those invalid ones.

Take a look at this WU, everyone else is getting 69.35, but I only get 34.67! why?


http://lhcathome.cern.ch/lhcathome/workunit.php?wuid=2311068



I'm not the LHC Alex. Just a number cruncher like everyone else here.
ID: 19178 · Report as offensive     Reply Quote

Message boards : Number crunching : why my credit is lower than everyone else?


©2024 CERN