From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jeffrey A Law To: Bernd Schmidt Cc: meissner@cygnus.com, toon@moene.indiv.nluug.nl, egcs@cygnus.com Subject: Re: Reload patch to improve 386 code Date: Mon, 07 Sep 1998 19:09:00 -0000 Message-id: <29448.905198888@hurl.cygnus.com> References: X-SW-Source: 1998-09/msg00401.html In message < Pine.GSO.4.02A.9809071547250.23612-100000@matlock.informatik.rwth-aachen.de >you write: > It really shouldn't. Inheritance only looks at the insns that reload is > already done with, and the one currently being processed. For those > which reload has completed, the register life information calculated by > my patch is never referenced again, so the inheritance code can do > whatever it wants to them. Ok. We're probably OK then. > > We actually have most of the stuff to implement this now. We just need > > to set the alias set for such MEMs so that they have a different set > > than all the MEMs created before reload. > > We'd need to be a bit careful about this, since some of these MEMs might be > replaced (for example) in instruction splitting later on, and the special > "this MEM made by reload" alias set must be preserved during such operation s. Actually if it is not preserved, then we're still OK since the alias set would be set to zero, indicating that we don't know what alias set the split mems belong to. While nonoptimal, it will work until we find a way to preserve the alias set in the split MEMs (which may be trivial in this case, I don't know). One of the things we need to do throughout the compiler is propagate the alias set information through to more places. I suspect that during the various optimization phases we're losing some alias set info provided by the front end. jeff