PDA

View Full Version : GeForce3 Paging Problem?



Husted
01-23-2002, 04:37 AM
Hi,

I'm currently having a strange paging problem which I hope you cold help me solve. The setup I use looks as follows: Intel P4 processor running Linux (kernel 2.4.7-10), GeForce3 64 MB board using the 1.0-2313 driver.

The application is a volume rendering program which uses a high amount of texture memory (each stack of slices takes approx. 64 MB of video memory). The application leaves all paging issues to the NVIDIA driver (i.e. no texture priorities are set).

Normally the application does not have a problem to render huge volumes. But if I start the application in a window and render a volume that requires paging, a switch to full-screen mode makes the application "hang".

A small investigation showed that the application didn't actually freeze: it just suddenly took a couple of minutes to render a single frame. This behavior is very strange as it only takes approx. 150 msec to render the same volume when then application is started in full-screen mode.

What seem more strange is that I can "wake up" the application by resizing the window to 95% of full-screen size. After this, it is possible to switch to full screen with no problems.

Does anyone have an idea on what this might be? I suspect that it is a problem with paging in the NVIDIA driver.

-- Niels

Husted
01-23-2002, 06:12 AM
Hi,

If I switching back to the 1.0-1541 drivers the problem disappears... Although this solves this problem - it is not a solution in general http://www.opengl.org/discussion_boards/ubb/smile.gif

-- Niels

Gorg
01-23-2002, 09:04 AM
Just write a small test app, that demonstrate the problem and send it to nvidia. They will be glad to fix it.

Husted
01-23-2002, 09:55 AM
Originally posted by Gorg:
Just write a small test app, that demonstrate the problem and send it to nvidia. They will be glad to fix it.

I was hoping that someone else had reported that bug so I wouldn't have to write a small test application http://www.opengl.org/discussion_boards/ubb/smile.gif.

Could someone from NVIDIA please let me know if such a bug has been reported previously?

-- Niels