3 ms·
Even if it's precompiled and not at runtime, we don't really know what the performance looks like, especially in hand rolled assembly where something that isn't
by auto 6y ago
Even if it's precompiled and not at runtime, we don't really know what the performance looks like, especially in hand rolled assembly where something that isn't a one-to-one cycle match could have obvious effects.
- diroussel 6y agoAlso, Rosetta 2 doesn't support AVX. So it would rely on there being an alternate code path which would be AOT transpiled.