Might be a long shot, but I figure I'll ask here anyway...

I'm building a GL display list from a file (openGL). It works great, displays very fast, all is good. However, when I want to load a new file, it takes a good 2 seconds to parse the file & compile the display list.

What I'd like to do is cache the data so I don't have to parse the files after they are all loaded. I could do this in main memory, but figure that if it stays in the graphics card's memory, it will be even faster. What I don't know is how to see how much memory I'm using in a display list or on the graphics card.

Is there a way to determine this mem usage of an openGL display list? Is there a way to see how much memory is in use on the graphics card? Preferably via code rather than a system call.

Maybe the card driver has an entry in /proc somewhere? Either that or look for an ioctl for the device (if there is one.)
ta0kira

I do not know of any methods to achieve the details you want in a program but external programs will let you view this sort of info. Such as gDEBugger, nvidia tools(PerfKit), ati tools(PerfStudio).
Why are you using display lists and not VBO's etc?

http://www.gremedy.com/
http://developer.nvidia.com/object/nvperfkit_home.html
http://ati.amd.com/developer/gpups/index.html

Thanks for the responses...

- while there are a few entries in /proc for the nvidia card, none seem to reflect memory usage.

- gDEBugger - this seems promising, but doesn't run on my machine - a lib issue. I'll ask my admin about it today (::grumble:: about not having admin privileges)

- nvPerfkit requires you to run its own driver- that driver is outdated and I couldn't install it anyway (see grumble above )

- the ati perf studio requires windows and an ati card.

- display lists vs vbo? I could get long winded, but the short answer is "because I chose to". This is not trying to sound nasty, just saying that both methods do similar things in very different ways, and both have advantages over the other in certain cases. Several of the advatages of VBOs don't apply in my case (shared vertexes, for example), so I went the display list route. I haven't done an A vs B, but the display lists are running quite fast, so much so that I didn't see a need to optimize getting data to the screen any further... just need to get it off of the disk faster and less frequently.