[parsec-users] output validation

Christian Bienia cbienia at CS.Princeton.EDU
Mon Jun 1 21:16:46 EDT 2009


You basically have two ways to verify that the program has been parallelized

1.) Use a thread checking tool. If any race conditions or other issues show
up then you know there's an issue in the program. But if the program passes
it does not mean it's correct.

2.) Check the output of the program. You can compare the output of the
serial version to that of the parallelized version of the program.
Unfortunately we don't have a tool that does that automatically, you need to
check manually. One issue is that floating point results vary across
machines, but you shouldn't run into that problem if you work on only one
machine. The workloads that produce a visual output can also be verified by
"looking at the output", but that will not reveal minor issues.


-----Original Message-----
From: parsec-users-bounces at lists.cs.princeton.edu
[mailto:parsec-users-bounces at lists.cs.princeton.edu] On Behalf Of Alin Jula
Sent: Monday, June 01, 2009 6:41 PM
To: parsec-users at lists.cs.princeton.edu
Subject: [parsec-users] output validation

Hi Chris,

How does one validate the output of a PARSEC benchmark parallelized with
a tool other than the ones provided (pThreads, OpenMP, TBB) ? For
example, supposed Black-Scholes was parallelized using a tool X. How
does one verify that the parallelization with X was correct and produced
the correct output?

Thanks a lot,
Alin Jula, Ph.D.
SAP Labs
T: 1-(650)-391-4552
F: 1-(650)-847-2646
3412 Hillview, 
Palo Alto, Ca, 94304

parsec-users mailing list
parsec-users at lists.cs.princeton.edu

More information about the parsec-users mailing list