Comment by bigyabai

Both incur a performance hit. ARM NEON isn't fully analogous to modern AVX or SSE, so even a 1:1 native port will compile down to more bytecode than x86. This issue is definitely exacerbated when translating, but inherent to any comparison of the two.