prefix_xor64
def prefix_xor64(x: UInt64) -> UInt64 Per-bit prefix XOR over 64 bits.
Equivalent to clmul(x, ~UInt64(0)).low_64() (one PCLMULQDQ on x86
or one PMULL on ARM64 with CLMUL). The shift-based form below
compiles to six xor+shl pairs, ~6 cycles on a modern OoO core
-- well within the per-chunk budget for 64-byte stage 1 work
units, and free of any ISA feature gating.