[parsec-users] data footprint of facesim, raytrace, vips

Christian Bienia cbienia at CS.Princeton.EDU
Thu Jul 8 15:18:31 EDT 2010


Hi Mu-Tien,

 

Here are the answers to your questions:

 

1.)    The three simulation inputs for facesim are the same because it is
very challenging to scale the face mesh in a reasonable way.

2.)    The main way to scale the runtime of raytrace is by changing the
resolution of the output screen. That does not require different input data.

3.)    Vips has the ability to only load the parts of an image which are
necessary to do the computations. That way the program can work on images
that are larger than main memory.

 

Best,

Chris

 

 

From: parsec-users-bounces at lists.cs.princeton.edu
[mailto:parsec-users-bounces at lists.cs.princeton.edu] On Behalf Of Mu-Tien
Chang
Sent: Wednesday, July 07, 2010 2:26 PM
To: parsec-users at lists.cs.princeton.edu
Subject: [parsec-users] data footprint of facesim, raytrace, vips

 

Hi All,

 

I just did some simulation using pin tool to get data footprint of parsec
workloads. Generally with larger input, I get larger data footprint.
Exceptions are facesim, raytrace, and vips. For facesim, I wonder what is
the difference between running simsmall, simmedium, and simlarge? I diff
these three inputs but turns out they are identical. For raytrace and vips,
does simlarge contains lots of replicated data blocks so that data footprint
does not appear larger?

 

Thanks a lot,

 

Mu-Tien

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cs.princeton.edu/pipermail/parsec-users/attachments/20100708/b0797560/attachment.htm>


More information about the parsec-users mailing list