On Thu, 18 Jan 2018, Jakub Jelinek wrote: > On Thu, Jan 18, 2018 at 09:18:21AM +0100, Richard Biener wrote: > > On Thu, 18 Jan 2018, Jakub Jelinek wrote: > > > > > On Thu, Jan 18, 2018 at 09:10:03AM +0100, Richard Biener wrote: > > > > On Wed, 17 Jan 2018, Jakub Jelinek wrote: > > > > > > > > > Hi! > > > > > > > > > > PR83882 complains that PR81715 testcase fails on callee copies > > > > > parameter > > > > > targets. The following patch ought to fix that, but I have only > > > > > bootstrapped/regtested it on x86_64-linux and i686-linux + on the > > > > > testcase > > > > > with hppa. > > > > > > > > Looks reasonable. > > > > > > > > > John, do you think you could test this on hppa without the callee > > > > > copies > > > > > default change? > > > > > > > > > > Or should we not care anymore if there aren't any similar targets > > > > > left? > > > > > > > > How's that communicated to the middle-end anyways? > > > > > > You mean the callee copies stuff? gimplify_parameters creates new > > > temporaries, assigns the parameter to them, set DECL_VALUE_EXPR on the > > > parameters to the temporary and at the end of function clears > > > DECL_VALUE_EXPR and with this patch adds a CLOBBER in a cleanup as well. > > > > No, I meant whether the target wants this or not. > > Ah, that. Through targetm.calls.callee_copies hook.
So there are quite a few more callee copies targets then, notably unconditionally epiphany, mmix, mn10300 and v850 and conditionally a few more, for example mips. Richard.
