- Handle the case where the CPU supports AVX, but we are running on an hypervisor with AVX disabled/not supported. - Faster (2x) scalarmult_base() when using the ref10 implementation