A better question would be why OpenGL should be in any way involved with networking and transferring of pixel data?I want to stream pixels from my hefty uber-SLI multi core desktop to my iPad. Why should I need to haggle with the server GUI (if any) for that?
I wasn't suggesting OpenGL is concerning with networking or pixel data transfer.Originally Posted by Alfonse Reinheart
I'm pointing out that with CUDA (for example) I can allocate GPU resources, do the computation, put the result in a file or on a web server, without have GLX or WGL getting in my way.
So, again, I suggest is unfortunately tied to some obsolete design assumptions, unfortunately.
I would like to run OpenGL tests over ssh merely looking if the final pixels match some specific md5sum hash.
X11 or windows desktop is irrelevant for that, just adds various ways for tests to fail for no good reason.
- Nigel
---
Regal - as OpenGL ought to be
Obsolete? OpenGL is for rendering; that's its job. Drawing things.I'm pointing out that with CUDA (for example) I can allocate GPU resources, do the computation, put the result in a file or on a web server, without have GLX or WGL getting in my way.
So, again, I suggest is unfortunately tied to some obsolete design assumptions, unfortunately.
Then do that; nobody's stopping you. Just create a window and minimize it. In GLX land, you don't even need to create a window to create a pbuffer context.I would like to run OpenGL tests over ssh merely looking if the final pixels match some specific md5sum hash.
Indeed. Drawing pixels. Into a window. Or perhaps into a .png file. Or streaming h264 to an iPad. But I shouldn't need a GUI API to get a workable OpenGL context.
And it shouldn't matter who's logged into the machine or if there is any actual windowing system running.
Creating an extraneous window, then going to the trouble of minimizing it - obsolete.
- Nigel
---
Regal - as OpenGL ought to be
Posting Permissions
