Enabling double precision

Hi there,

I’m trying to enable double precision through my kernels however I get an error thrown when I include the extension enabling pragma:
#pragma OPENCL EXTENSION cl_khr_fp64 : enable

I’m also using the C++ bindings, the error thrown is -45 during the kernel jit compilation but I have no idea what that error number actually refers to.

Without this pragma the kernel code compiles/runs without any issues,

What needs to be done to enable double precision?

Thanks in advance.

Ah apologies, just realised my gfx card does not support cuda model 1.3 for double precision (using a 8800gts).

On a related to note, is there an easy way to get meaning ful error strings from the error numbers thrown by the c++ bindings?