Re: [PATCH][IA64] Fix control speculation bug

[Vladimir Makarov <vmakarov at redhat dot com>]

Maxim Kuvyrkov wrote:
> Vladimir Makarov wrote:
> > Maxim Kuvyrkov wrote:
> > > Hello,
> > >
> > > This patch fixes ICE on virtually any case of using ia64 control 
> > > speculation.  The bug doesn't manifest itself because control 
> > > speculation by default is off.
> > >
> > > I don't have access to ia64 box at the moment, so I only checked it 
> > > on a cross-compiler.  Andrey, can you, possibly, test the patch?
> > >
> > > The patch is fairly simple and only constraints the scheduler to use 
> > > control speculation in lesser number of cases.
> > >
> > > OK for mainline and 4.3 branch?
> > Yes, the patch itself is OK.  But you should add also a test to check 
> > the control speculation for ia64.  I think it is easy to do because 
> > control speculation is completely broken now.
>
> There is a compilation test for control speculation in the second 
> patch I submitted (Subject: "Fix speculation support").  If you mean 
> an scan-assembler test, then that test can be fixed to check for ld.s 
> instruction in assembly; I doubt, though, it'll be stable.  Still, 
> fwiw, that test triggers control speculation on both trunk and 4.3 
> branch.
>
> Do you want me to add scan-assembler test?
No, thanks.  If the test in the mentioned email catches the bug, then 
you can ignore my asking for a test.   So you can  commit the patch as 
it was submitted.