[llvm-commits] [Patch] Make sure isIdenticalTo takes into account the machine mem operands

Villmow, Micah Micah.Villmow at amd.com
Thu Sep 20 15:05:07 PDT 2012


Yeah, I wish I still had the test case laying around I could see if this is still a problem.

Micah

> -----Original Message-----
> From: Hal Finkel [mailto:hfinkel at anl.gov]
> Sent: Thursday, September 20, 2012 3:04 PM
> To: Villmow, Micah
> Cc: llvm-commits at cs.uiuc.edu
> Subject: Re: [llvm-commits] [Patch] Make sure isIdenticalTo takes into
> account the machine mem operands
> 
> On Thu, 20 Sep 2012 21:36:25 +0000
> "Villmow, Micah" <Micah.Villmow at amd.com> wrote:
> 
> > Some locations where this matters, IFConversions.cpp, it skips the
> > same instructions at the beginning of a true and false basic block,
> > but if the same machine instr with the same registers but point to a
> > different memory location, LLVM treats them as pointing to the same
> > memory location.
> 
> But if the preceding instructions are all the same, then how could the
> registers have different values to be loading or storing to different
> locations?
> 
>  -Hal
> 
> >
> > Sadly this change is from a year and a half ago, so I don't remember
> > the exact change that cause the bug to be triggered.
> >
> > > -----Original Message-----
> > > From: Hal Finkel [mailto:hfinkel at anl.gov]
> > > Sent: Thursday, September 20, 2012 2:18 PM
> > > To: Villmow, Micah
> > > Cc: llvm-commits at cs.uiuc.edu
> > > Subject: Re: [llvm-commits] [Patch] Make sure isIdenticalTo takes
> > > into account the machine mem operands
> > >
> > > Micah,
> > >
> > > Can you please explain the motivation for this? The memory operands
> > > are only informative (not actually directly encoded), right? Is only
> > > checking the sizes sufficient?
> > >
> > >  -Hal
> > >
> > > On Thu, 20 Sep 2012 15:19:53 +0000
> > > "Villmow, Micah" <Micah.Villmow at amd.com> wrote:
> > >
> > > > Ping!
> > > >
> > > > From: Villmow, Micah
> > > > Sent: Tuesday, September 18, 2012 10:35 AM
> > > > To: Villmow, Micah; llvm-commits at cs.uiuc.edu
> > > > Subject: RE: [Patch] Make sure isIdenticalTo takes into account
> > > > the machine mem operands
> > > >
> > > > Ping!
> > > >
> > > > From:
> > > > llvm-commits-bounces at cs.uiuc.edu<mailto:llvm-commits-bounces at cs.ui
> > > > uc.e
> > > > du> [mailto:llvm-commits-bounces at cs.uiuc.edu] On Behalf Of
> > > > du> Villmow,
> > > > Micah
> > > > Sent: Thursday, September 06, 2012 10:03 AM To:
> > > > llvm-commits at cs.uiuc.edu<mailto:llvm-commits at cs.uiuc.edu> Subject:
> > > > [llvm-commits] [Patch] Make sure isIdenticalTo takes into account
> > > > the machine mem operands
> > > >
> > > > In some cases, MI->isIdenticalTo can return true when everything
> > > > is equal but the mem operands. This makes sure that the machine
> > > > mem operands are taken into account when checking if a
> > > > MachineInstruction is equal to another.
> > > >
> > > > Thanks,
> > > > Micah
> > >
> > >
> > >
> > > --
> > > Hal Finkel
> > > Postdoctoral Appointee
> > > Leadership Computing Facility
> > > Argonne National Laboratory
> >
> >
> 
> 
> 
> --
> Hal Finkel
> Postdoctoral Appointee
> Leadership Computing Facility
> Argonne National Laboratory






More information about the llvm-commits mailing list