On 03/13/2012 12:41 AM, Andreas Schwab wrote: > Andreas Schwab <sch...@linux-m68k.org> writes: > >> Ian Lance Taylor <i...@google.com> writes: >> >>> Andreas Schwab <sch...@linux-m68k.org> writes: >>> >>>> Ian Lance Taylor <i...@google.com> writes: >>>> >>>>> But it also looks like the pattern should use a match_scratch. >>>> >>>> It is also used as input in operand 2. >>> >>> Sorry, I missed that. >> >> That appears not to be an issue actually, there is already one use of >> match_scratch together with a matching constraint in *cmpdi_internal. >> But then, using match_scratch instead of match_operand doesn't really >> fix the bug either (it only helps a simplified test case, but not the >> original one). > > It doesn't actually change anything (I was confused because 4.7/4.8 no > longer generates the overlapping output for the simplified testcase).
Does 4.7 still have the failure at all? I've checked with the 4.6 branch, and regrename gets confused because there's a REG_DEAD note for the register, and another REG_UNUSED for the same reg. As far as I remember, it used to be the case that there should not be a REG_DEAD note for a register that gets set in the insn, but maybe df changed the rules? Or maybe it was a df bug in 4.6? Bernd
