Never getting "out of memory" on Win7 x64 / Geforce 9800 GT CL_MEM_OBJECT_ALLOCATION_FAILURE

Accelerated Computing CUDA CUDA Programming and Performance
1

Hello,

I tried to allocate more memory than my device is capable of and my kernels stopped working (produced only zeroes), but I never get any errors from OpenCL. I expected clCreateBuffer zu return a CL_MEM_OBJECT_ALLOCATION_FAILURE error code but this doesn’t happen.

What is the correct way to do error checking if I try to allocate a new block that doesn’t fit onto the device?

Is this a driver bug? Driver is 266.58

Thanks,

Nils

2

Just I wild guess, but does the notify function return any useful messages?

David

3

That was an awesome tip you gave me there, somehow I forgot to implement my empty notify function with anything useful, the messages reported there are of good help External Image