pub fn device_compute_capability() -> Result<(u32, u32)>Expand description
(major, minor) compute capability of the calling context’s device.
Requires a current CUDA context, exactly like sm_count_cu next door, so
it is only safe to call from a thread that has one. --check-kernels runs
it after the backend is up, which is such a thread. The preflight does
not — see device_compute_capability_of.