Stefan Eilemann wrote > Where does it say that? I couldn't find it on quick scan and don't > remember writing it. The PG is definitely wrong here, the default port and > device are undefined, which means 'use the default from the user > environment'.
So in the 1.4 programming guide in the pipe section it used to state that: *For the WGL window system, the device selects the nth GPU in the system. * and then under that (in the syntax description): *pipe # 1-n times* which is probably a little bit confusing. Maybe there should be a clarification about wgl (that it defaults to device 0 and that the device number actually defines the display rather than the GPU). > In "modern" (:P) Equalizer versions auto-configuration using hwsd will > enumerate the devices properly. It's been on my TODO list to port our code over to a more recent EQ but I've always had some trouble compiling more recent versions. I'll give it another shot after some deadlines, want some HWSD goodness :). > Depends. We really only support multi-GPU properly for Quadro cards, where > the device index is used to enumerate the GPUs using the > WGL_NV_gpu_affinity API. > > Yes, this is what I expect from the current implementation. A proper > implementation should auto-detect the GPU manufacturer and use a different > code path for AMD. Since we don't have the cards, nor a paying customer to > sponsor this effort, I can't do this myself. I would be happy to help you > implement it on your end, and merge the code into master. That's completely understandable, AMD's hardware isn't as prevalent as NVIDIAs. Additionally AMD does automatic GPU affinity for windowed opengl contexts (it will associate the context with the GPU on which the window that contains the context is first created according to our conversations with their hardware team and our benchmarks). They do have an affinity API for off-screen contexts but we haven't had a need to use it yet. I don't think this would be a problem is there was just a tad more clarity in the Programming guide (or maybe a special note about AMD users in windows?) > Thanks, merged. No problem. This + proper device selection should generally ensure correct affinity on AMD. > OS X should work, as I build regularly on it. Windows might be more > spotty, but I would happily merge any build fixes. > > Hopefully this year. I already upgraded to XCode 5, so I don't see what > else is in the way. I'm not sure why I'm having so many issues. I think there's some clang-5.0 specific stuff that needs to be added to the make files (for example, when compiling lunchbox, I get header problems since it's trying to include unordered_set from tr1/). I hacked those a bit and moved forward with the build but ended up having other issues and gave up on it after a couple of days. Thanks for the reply :) -Harris -- http://www.eyescale.ch https://github.com/Eyescale/ http://www.linkedin.com/in/eilemann _______________________________________________ eq-dev mailing list eq-dev@ http://www.equalizergraphics.com/cgi-bin/mailman/listinfo/eq-dev http://www.equalizergraphics.com signature.asc (210 bytes) <http://software.1713.n2.nabble.com/attachment/7584398/0/signature.asc> -- View this message in context: http://software.1713.n2.nabble.com/Bug-in-wgl-Pipe-pixel-viewport-detection-based-on-device-tp7584388p7584405.html Sent from the Equalizer - Parallel Rendering mailing list archive at Nabble.com. _______________________________________________ eq-dev mailing list [email protected] http://www.equalizergraphics.com/cgi-bin/mailman/listinfo/eq-dev http://www.equalizergraphics.com
