[ros-users] Is the Gazebo simulation GPU bound?

[Steven Martin]

I haven't worked with Gazebo for a while now I have been using Openrave. Gazebo should run in approximately real time although this greatly depends on the complexity of the world. 

As far as I know ODE does not use any GPU acceleration. (It has been a while since I looked at the source code) So the only load on the GPU is from visualisation. 

I would suggest running Gazebo headless and using rviz for visualisation if this is suitable.

Steven Martin
steven.martin at qut.edu.au

PhD Student
Queensland University of Technology
Gardens Point, S1107

On 12/10/2010, at 5:59 PM, Ibrahim Awwal wrote:

>  Oh, specs might be useful. I've got a Core i5 laptop with the 
> integrated GPU that's on the Core i5, so probably Intel's best IGP. And 
> a Radeon 5650, but that's not working right now and hence the problem. I 
> can get access to a machine that works better pretty easily but it would 
> be nice to have it working well on my laptop. Is there any way I can 
> turn down the quality of the simulation or anything like that? 
> Wireframes doesn't seem to make a difference in performance. Thanks,
> 
> -Ibrahim Awwal
> 
> On 10/11/2010 10:54 PM, Ibrahim Awwal wrote:
>>   Hi guys,
>> I'm just wondering, how much does the Gazebo simulation stuff,
>> specifically for the PR2, depend on the GPU? I've got a laptop with
>> hybrid/switchable graphics, i.e. Intel integrated + ATI discrete, and
>> right now ATI's Linux drivers don't support switchable graphics
>> properly, so I'm stuck with the Intel integrated, and I get about 5fps
>> in the gazebo simulation. Also, I frequently get crashes where it says
>> something about OGRE running out of memory. Is it supposed to be much
>> faster, or even basically real time? Thanks,
>> 
>> -Ibrahim Awwal
>> _______________________________________________
>> ros-users mailing list
>> ros-users at code.ros.org
>> https://code.ros.org/mailman/listinfo/ros-users
> 
> _______________________________________________
> ros-users mailing list
> ros-users at code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users