> should also be easier for CPU designers to handle efficiently I'm always dubious of claims like this. I think in general the architecture would need to turn

> should also be easier for CPU designers to handle efficiently I'm always dubious of claims like this. I think in general the architecture would need to turn it back into the highest SIMD code it supports at instruction decode.