@lisyarus > Instead, you have to either build the constant from 16-bit patches (there's an instruction that loads a 16-bit constant with an offset of 0, 16, 32, or 48 bits), or put the constant into constant memory and load it from there (that's what I do for floating-point constants, by the way).you can also one-shot runs of repeated bit patterns as a bit immediate, which is surprisingly effective https://kddnewton.com/2022/08/11/aarch64-bitmask-immediates.html
