Open roman955b opened 3 years ago
If you only want to generate "valid code" that can run correctly on the target HW, you don't need the target HW.
But to generate "fastest code" for that target HW, Antares need to access that HW since it will evaluate a number of different implementations and finally pick up one of the most efficient codes.
To optimize code for the target HW, you need to install the SDK for that HW. For example, CUDA SDK is needed to make nvcc
command work.
In order to use Antares, does one need to have access to the target HW or is it enough to fill in one of those cfg files? What is required by the targeted HW if one wants to optimize for it?