I was reading the cuda reference manual, and saw something very interesting in the description for cudaGetDeviceCount().
It seems CUDA will eventually have an automatic emulation mode whenever there is no CUDA hardware available (a most sensible feature). Any word on when will that happen?
P.S. I love how nvidia documents features that don’t exist yet. No, really, it’s very helpful since you won’t have to re-learn things later.
Of course, the feature exists. I use CUDA 2.0. When I run my application on development machine, it lists Device emulator as Device 0.
If I run the same app on a machine with a GPU installed, it again lists 1 device that stands for the GPU and NO emulator is listed.
Does it really? Of course I only try running cuda on machines that have cards, but I’ve never heard of something like this on the forums.