[visit-users] Local rendering, and how do I know it's using the graphics card? [...]

Paul Melis paul.melis at surfsara.nl
Fri Jul 11 04:24:35 EDT 2014


It's a shame that VTK isn't geared towards using modern OpenGL 
techniques, like buffer objects. The speedup in rendering of a model 
that doesn't change geometry every frame, even compared to display 
lists, is enormous. I have a simple benchmark tool that just renders a 
bunch of triangles, one light source (fixed-function, not even using 
shaders). For a model of 5M triangles rendered at 1920x1080 on an NVidia 
GTX780 I get:

immediate mode
13.4 Mtris/sec

display list
41.6 Mtris/sec

vertex + normal arrays
34.1 Mtris/sec

vertex buffer object
1260.8 Mtris/sec

llvmpipe 10.2.2 w/ display lists (using 12 threads)
0.7 Mtris/sec

Regards,
Paul



On 07/10/2014 05:02 PM, Brugger, Eric wrote:
> Fred,
>
> Thanks for the feedback. I wouldn't have thought that display lists would make that large a difference running completely locally. I will need to keep that in mind if others encounter similar situations.
>
> Eric
>
> -----Original Message-----
> From: Fred Morris [mailto:m3047 at m3047.net]
> Sent: Wednesday, July 09, 2014 9:47 PM
> To: VisIt software users community
> Subject: Re: [visit-users] Local rendering, and how do I know it's using the graphics card? [...]
>
> Thanks Eric! "Use display lists" -> "Always" made a big difference.
>
> On Wednesday 09 July 2014 07:18, Brugger, Eric wrote:
>> If you are running completely locally you should be using the graphics card.
> This appears to be true...
>
>> You can change the rendering options with the "Rendering" window,
>> which can be brought up by going to "Options->Rendering". You would
>> need to go to the "Advanced" tab and change "Use scalable rendering" to "Always"
>> to force it to always render in software. If you changed it to "Never"
>> then it would always use the graphics card.
> As anticipated, this was a huge change for the worse (it also didn't actually render until I stopped whatever motion I was making).
>
>> You could also change the "Use display lists" on the "Basic" tab and
>> see if that makes a difference.
> "Use display lists" -> "Always" made a HUGE difference. I'm getting 10-20 FPS now.
>
>> It might be useful to run VisIt with "-debug 5" and look at the
>> viewer.5.vlog file and see what it says about the graphics system.
> It didn't have much to say. About the only thing which seemed at all graphics-related was "Forcing GL context initialization...":
>
> Xfer::Update: Sending: opcode=2, name=SimulationCommand
> Xfer::Process: Opcode=48, len=19, type=SimulationCommandRPC Executing SimulationCommandRPC: INTERNALSYNC
> Xfer::Update: Sending: opcode=49, name=NonBlockingRPC::CompletionData (from
> SimulationCommandRPC)
> Xfer::Process: Opcode=42, len=14288, type=SetWinAnnotAttsRPC Executing SetWinAnnotAttsRPC 1018x948 Creating new VisWindow for id=0 Forcing GL context initialization...
> Unknown if transparencies exist; recalculating (this requires global communication).
> Xfer::Update: Sending: opcode=43, name=VisItRPC::RPCReply (from
> SetWinAnnotAttsRPC)
> Xfer::Process: Opcode=64, len=3747, type=ExpressionList
> Xfer::Process: Opcode=4, len=211, type=ReadRPC Executing ReadRPC Skipping already loaded database plugin SimV2 version 1.0
> NetworkManager::AddDB: Adding velocity to the name stack
> Xfer::Update: Sending: opcode=5, name=VisItRPC::RPCReply (from ReadRPC)
> Xfer::Process: Opcode=11, len=28, type=PreparePlotRPC Executing PreparePlotRPC: Pseudocolor_1.0 Loaded full plot plugin Pseudocolor version 1.0
> Xfer::Update: Sending: opcode=12, name=VisItRPC::RPCReply (from
> PreparePlotRPC)
> Xfer::Process: Opcode=10, len=260, type=MakePlotRPC Executing MakePlotRPC: Pseudocolor_1.0
> NetworkManager::MakePlot(): Leaving NetworkManager::MakePlot().
> Xfer::Update: Sending: opcode=13, name=MakePlotRPC::NetworkID (from
> MakePlotRPC)
> Xfer::Process: Opcode=26, len=3, type=ExecuteRPC Executing ExecuteRPC with respondWithNullDataObject = 0
> PseudocolorPlot: Did not hit target topological dimension: 2 Guide function indicated that we should do the first Update on pipeline 1.
> Entered update for avtActualExtentsFilter
>
>> HTH,
>>
>> Eric
> Thanks again....
>
> --
>
> Fred
>
> --
> VisIt Users Wiki: http://visitusers.org/ Frequently Asked Questions for VisIt: http://visit.llnl.gov/FAQ.html To Unsubscribe: send a blank email to visit-users-unsubscribe at elist.ornl.gov
> More Options: https://elist.ornl.gov/mailman/listinfo/visit-users
> --
> VisIt Users Wiki: http://visitusers.org/
> Frequently Asked Questions for VisIt: http://visit.llnl.gov/FAQ.html
> To Unsubscribe: send a blank email to visit-users-unsubscribe at elist.ornl.gov
> More Options: https://elist.ornl.gov/mailman/listinfo/visit-users



More information about the visit-users mailing list