[parsec-users] A problem about Parsec Benchmark

Christian Bienia cbienia at CS.Princeton.EDU
Mon Sep 15 12:12:07 EDT 2008


Hi Shinra,

 

The workload uses the pipelining model for two major reasons: (1) That's
what commercial workloads typically use and (2) the program first has to
fragment the input stream to create independent, parallel work units. Dedup
was modeled after a real-world software product which uses pipelining.

 

What type of problem do you have with dedup?

 

- Chris

 

 

From: parsec-users-bounces at lists.cs.princeton.edu
[mailto:parsec-users-bounces at lists.cs.princeton.edu] On Behalf Of Dennis
Yang
Sent: Monday, September 15, 2008 3:43 AM
To: parsec-users at lists.cs.princeton.edu
Subject: [parsec-users] A problem about Parsec Benchmark

 

Hello~ 

I have read the technical report and traced the code of "dedup" recently.

In the report, it mentioned that the input stream would be break up into
work units which can be processed "independently".

Does this mean that this application could be parallelized with data-level
parallelism? 

If so, may I ask the reason why it was been parallelized with pipelined
model?

Any help would be grateful!

Thanks a lot~

 

Shinra

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cs.princeton.edu/pipermail/parsec-users/attachments/20080915/a15236a9/attachment.html>


More information about the parsec-users mailing list