Why is a 24-bit check used here instead of a 32-bit check?

[fumeboy]

goloader/mmap/mmap_darwin_arm64.go

Line 50 in 9fbc78a

if uintptr(unsafe.Pointer(&data[0]))-addr > 1<<24 {

[eh-steve]

Arm64v8 instructions don't use immediate 4 byte values next to the instructions like x86, instead the addresses/offsets are encoded as part of the 32-bit instruction so they generally have a maximum of 24 bits available for Branches (and 33-bits for ADRP+ADD relocs across 2 instructions)