[Bosinski]

I guess 128-bits is the max-width found on Arm/Neon. My question would be if wider registers > 138-bit on Intel/Amd will be used if avail ?

[kenarsa]

100% of AVX/AVX2 have 256 SIMD registers, and Arm also has non-NEON registers, which are 256/512. of course, this requires runtime detection, but that is ok because even for native code, people need to do that.