Currently we use a mix of load ImmPtr(0) and xorPtr(r,r) to zero the contents of a register. Given that these perform differently on different architectures it probably makes sense to simply have a MacroAssembler API to zero a register that does the most efficient thing for the target platform.