[parsec-users] Cannot achieve declared speedup for Dedup

Christian Bienia cbienia at CS.Princeton.EDU
Mon Oct 20 15:09:49 EDT 2008


Hi Nikolay,

 

The numbers in the report are for an ideal machine. Many PARSEC applications
will quickly become memory bound in practice, but a speedup of 1.3x for
dedup does seem very low. I get a speedup of ~3.5x on our 8-way machine. It
might be that your test machine simply has very small caches or highly
constrained memory bandwidth. Have you also tried it on another machine?

 

- Chris

 

 

From: parsec-users-bounces at lists.cs.princeton.edu
[mailto:parsec-users-bounces at lists.cs.princeton.edu] On Behalf Of Nikolay
Kurtov
Sent: Monday, October 20, 2008 9:25 AM
To: parsec-users at lists.cs.princeton.edu
Subject: [parsec-users] Cannot achieve declared speedup for Dedup

 

In parsec-report.pdf it is written that Dedup is extremely good scalable.
For instance, on 4 threads it runs ~3.5 faster then on 1 thread.
I tried hard but all I could achieve is 1.3x speedup on 4 threads.

Am I missing something???
What do you think about the situation?


-- 
Regards,
Nikolay

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cs.princeton.edu/pipermail/parsec-users/attachments/20081020/cc9f43ec/attachment.html>


More information about the parsec-users mailing list