Using OpenSUSE 11.1 with the 2.3 Cuda tools and SDK on an i7/X58 system with a GTX 260 – When trying to profile a simple test application with “cudaprof”, I randomly but not infrequently (say 1/5 profile runs of the same code) will have X11 lockup to the point that I must physically reboot the machine. The previous X11 log (Xorg.0.log.old) ends with something like:
… lots of stuff …
(II) 3rd Button detected: disabling emulate3Button
(WW) Sep 29 17:36:25 NVIDIA(0): WAIT (2, 6, 0x8000, 0x000020a0, 0x000022e8)
(WW) Sep 29 17:36:32 NVIDIA(0): WAIT (1, 6, 0x8000, 0x000020a0, 0x000022e8)
[mi] EQ overflowing. The server is probably stuck in an infinite loop.
[mi] EQ overflowing. The server is probably stuck in an infinite loop.
[mi] EQ overflowing. The server is probably stuck in an infinite loop.
… repeats until rebooted …
which typically indicates that the GPU has crashed. Outside of cudaprof (which works fine until it crashes!), the application works correctly and never has a problem. Has anyone else seen this, or can confirm that cudaprof for Linux is in fact stable? On a side but maybe related note, my X11 display sometimes “locks up” for 5 seconds or so, but always comes back. I have seen plenty of comments regarding this phenomena on the net and they always blame the display driver (190.18), and hope the problem will be fixed in a future release. The X11 log will generally have warning similar to the above, but the infinite loop does not occur.
tnx … John G. Shaw