Bluescreen long after closing the application

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi, I’m truly lost on this one,

I’m executing CUDA code. Everything seems to be working fine, no cuda errors or exceptions are thrown, and there is no problems on my device, but one of the co-developers reports that long after running and closing the application (successfully, with no errors), he gets a bluescreen with a code system_thread_exception. (We’re using CUDA toolkit 12.0, and he’s using MX 250 GPU (6.1 cc), and he might be using CUDA toolkit 12.1, if that makes a difference)

At first, we thought that it could be due to illegal memory access, but according to the forum “There is memory protection. All of the addresses are virtual; there’s no mapping from random addresses to other places in memory with sensitive data.”, and running the code with nsight debugger with all memory checks on returns no errors.

Secondly, google says that this is a common bluescreen for outdated drivers, but I don’t think that applies to developers, and he says that when he installed the CUDA toolkit, it updated his drivers.

The only other possibility that I can think of is that unfreed memory is somehow corrupting things? But that seems very far fetched, and I couldn’t find anything. Shouldn’t it be freed anyway after closing the app?

I would really appreciate a tip, am I missing anything?

2

Why do you think this error is associated with your program, or with CUDA in general?

1 Like
3

CUDA is only a part of the solution, and without executing the CUDA part, even if it’s present, everything works fine. We also had different times of bluescreen. Sometimes it occurs during the execution of the CUDA code, sometimes long after closing the application, but never in any other conditions.