Practice makes perfection

I find it easier than x86 SIMD[^1], but the lack of support for vector operations for 32 bits makes me very grumpy.

[^1] Though, to be honest, x86 has great automatic SIMD tools so I never had to try too hard.

Just LLM it