On 5/6/24 09:31, Paolo Bonzini wrote:
The comment deals with the former, the removal with the latter.The idea of the comment is that after SCAS/CMPS you have CC_OP_SUB*, so in principle you may expect that you need to set CC_OP_DYNAMIC explicitly at the end of a REPZ/REPNZ, which is where the CX != 0 and CX == 0 paths join. But this is not necessary, because there is nothing after that instruction - the translation block ends. So I guess the comment could instead be placed at the end of the function? /* * Only one iteration is done at a time, so the translation * block has ended unconditionally at this point and there * is no control flow junction - no need to set CC_OP_DYNAMIC. */ What do you think?
Just before gen_jmp_rel_csize? Yes, that seems good. r~
