Double precision and GPULib

A few users have asked about double precision calculations using GPULib. I thought I'd try to clear up some common questions.

First, you may notice when running the GPULib unit tests in IDL (using 'make check') that the double precision tests fail. There are two reasons for this. First, not all CUDA-enabled hardware is capable of double precision calculations -- currently, the GTX 200 series are the only cards that can do this; check the NVIDIA Web site for details on your specific card.

Secondly, you may need to update your video card drivers. We've had some issues with the 177.x series of CUDA drivers, but the 180.x series is doing much better. I'd recommend upgrading to the most recent drivers available from NVIDIA in any case, as they are constantly improving both features and performance. The downloads are available at http://www.nvidia.com/object/cuda_get.html