Re: [Qemu-devel] Incorrect handling of PPC64 rldcl insn

[Aurelien Jarno <aurelien@aurel32.net>]

On Tue, May 07, 2013 at 12:14:47AM +0200, Alexander Graf wrote:
> 
> On 06.05.2013, at 20:13, Torbjorn Granlund wrote:
> 
> > Alexander Graf <agraf@suse.de> writes:
> > 
> >  Thanks a lot for the bug report and test case! Please CC qemu-ppc
> >  whenever you find issues or have patches for PPC. That makes filtering
> >  for important mails a lot easier.
> > 
> > Would that make my complaints be considered more or less important?  :-)
> > 
> >  Does the patch below fix the issue for you?
> > 
> > It indeed does.  (I actually tried that already, but I cannot follow the
> > data flow into these functions, so cannot tell if that patch is
> > sufficient.  
> 
> Yes, it is. It's a leftover bug from converting the code to TCG I assume.

Yes, looks like I am the culprit here.

> > This bug indicates complete non-testing status of these
> > insns, which are mainstream enough to be generated by gcc.  I suppose
> > there will likely be more such fundamental errors if more instructions
> > are also completely untested.)
> 
> There's a certain chance that happens, yes. We don't have instruction test suites for the PPC target.
> 

We have the Gwenole Beauschene testsuite for the PPC32 target, even if
it doesn't work when compiled on a recent distribution, one has to use
the old binary. It currently passes, so the PPC32 and Altivec
instructions should be fine.

On the contrary, the PPC64 instructions are untested, and there are
likely a few bugs like this one left, especially on complex
instructions.

-- 
Aurelien Jarno                          GPG: 1024D/F1BCDB73
aurelien@aurel32.net                 http://www.aurel32.net