Re: [Qemu-devel] Crash due to invalid env->current_tb

[Paul Brook <paul@codesourcery.com>]

On Thursday 01 May 2008, Blue Swirl wrote:
> On 5/1/08, Paul Brook <paul@codesourcery.com> wrote:
> > > >                  T0 = gen_func();
> > > >
> >  > >  For 64bit target T0 is 64bits so "=a" does not work and "=A" is
> >  > > needed. The strange thing is that I need to throw away the upper
> >  > > 32bits because otherwise it won't work. gen_func is defined to
> >  > > return just long but T0 is unsigned long long, this seems
> >  > > inconsistent. The 'and' does not appear in 32bit targets so it does
> >  > > not harm there.
> >  >
> >  > This is because in this special case, T0 is not used as target CPU
> >  > temporary, but instead to return next TB address. On i386 this is 32
> >  > bits, so only EAX is needed. TCG does not touch EDX, so it contains
> >  > garbage. This also means that moving EDX to high word of T0 and then
> >  > throwing the high word away may be slightly wasteful.
> >
> > Do we need to use T0 at all here? Can't we just use a normal local
> > variable?
>
> I suspect T0 was used to gain extra performance, 

Really? I doubt it. Especially on x86 reserving a register for a fixed purpose 
is almost always a bad idea.

It was used with dyngen because there was no way of directly passing 
information between cpu_loop and generated code. However now we have a proper 
code generator there is no need for this.

> but in case of 64-bit  target on 32-bit host there is this unnecessary work.
>
> But does cpu-exec.c need to know about T0/T1/T2 at all?

I don't think so.

> Can we replace exec.h include with cpu.h one?

Currently we still need it to setup "env". However once we have proper TCG 
prologue/epilogue we should be able to pass env as an argument to gen_func, 
and have that do the setup for us.

Paul