> movq (JB_RBP*8)(%rdi),%r9 > movq (JB_PC*8)(%rdi),%rdx It saves a byte to make this: mov (JB_PC*8)(%rdi),%RDX_LP It makes no difference but is consistent to do it for the JB_RSP and JB_RBP loads too, so might as well. Thanks, Roland