TCC driver and RDC: cannot restart after crash

Accelerated Computing CUDA CUDA Programming and Performance
1

Here is a problem I have just encountered.

I am running the latest TCC driver with a Tesla 2050, on Windows Server 2008 RC2

As I have no physical access to the machine i need to run everything through Remote Desktop. (aka i was not able to test the behavior if I am directly on the machine)

Whenever I crash a program because of an error, I then cannot relaunch anything CUDA without a reboot, may it be my own code or any of the CUDA examples from the SDK, it hangs forever. More precisely, the cuContextCreate function does not return.

As anyone seen this behavior? I am using CUDA 3.2 RC, and have tested with both drivers available for my platform… always the same behavior. I had also tested with 3.1 and the older driver, same issue.

2

Here is a problem I have just encountered.

I am running the latest TCC driver with a Tesla 2050, on Windows Server 2008 RC2

As I have no physical access to the machine i need to run everything through Remote Desktop. (aka i was not able to test the behavior if I am directly on the machine)

Whenever I crash a program because of an error, I then cannot relaunch anything CUDA without a reboot, may it be my own code or any of the CUDA examples from the SDK, it hangs forever. More precisely, the cuContextCreate function does not return.

As anyone seen this behavior? I am using CUDA 3.2 RC, and have tested with both drivers available for my platform… always the same behavior. I had also tested with 3.1 and the older driver, same issue.

3

Hmmmm, okay, that’s odd. I’ll look into it.

4

Hmmmm, okay, that’s odd. I’ll look into it.