im running on a linux box (ubuntu 7.10) with GeForce 6600. when I run the example applications in the cuda sdk it says ‘Using device 0: Device Emulation (CPU)’ but the results are not as expected, is there a way to get the cpu emulation to work correctly and is it generally a reliable way for testing cuda code?
It probably is working correctly (it gives different error values than what the example apps running on the GPU expect), and no, it is not. --deviceemu serializes all operations, and as a result, you miss any race conditions you may have in your code. Considering the nature of parallel programming, that’s potentially a pretty big deal.