Manta Interactive Ray Tracer Development Mailing List

Text archives Help


Re: [MANTA] strange threaded behavior after last commit


Chronological Thread 
  • From: Aaron Knoll <knolla@cs.utah.edu>
  • To: James Bigler <bigler@cs.utah.edu>
  • Cc: manta@sci.utah.edu
  • Subject: Re: [MANTA] strange threaded behavior after last commit
  • Date: Wed, 24 May 2006 10:20:02 -0600

So I'm at rev. 1083, and I get the following:

% ./bin/manta -np 1 -bench -imagedisplay null
Benchmark completed in 8.34647 seconds (100 frames, 11.9811 frames per second)
% ./bin/manta -np 2 -bench -imagedisplay null
Benchmark completed in 8.2654 seconds (100 frames, 12.0986 frames per second)
% ./bin/manta -np 3 -bench -imagedisplay null
Benchmark completed in 8.39504 seconds (100 frames, 11.9118 frames per second)

Very odd. What I notice most of all is that even with -np 1, both processors are running at full capacity.

-Aaron

On May 24, 2006, at 9:46 AM, James Bigler wrote:

I was on revision 1081 and things seemed fine.

I just updated to 1083 and I get the right scaling for arachne and muzzle for the default scene.

% bin/manta -np 1 -bench -imagedisplay null
Benchmark completed in 7.29303 seconds (100 frames, 13.7117 frames per second)
% bin/manta -np 2 -bench -imagedisplay null
Benchmark completed in 3.68707 seconds (100 frames, 27.1218 frames per second)
% bin/manta -np 3 -bench -imagedisplay null
Benchmark completed in 2.46603 seconds (100 frames, 40.551 frames per second)

James

James Bigler wrote:
What machine is this on and what is the scene you are rendering?
Do you know what version you were coming from?
Also keep in mind you may have pulled in changes from the SCIRun/ Core libraries.
James
Aaron Knoll wrote:
Before I updated with James's last commit, when I specified -np 2 I got roughly a 2x speed increase. Now, for some reason, I get a slight speed decrease. This seems to happen for all scenes -- the default manta scene as well as my isovolume work. Has anyone else noticed anything like this?

-Aaron






Archive powered by MHonArc 2.6.16.

Top of page