Re: [PULL 3/5] tcg: add perfmap and jitdump

[Richard Henderson]

On 6/30/23 14:39, Ilya Leoshkevich wrote:
> On Thu, 2023-06-29 at 14:59 +0200, Richard Henderson wrote:
> > On 6/29/23 13:31, Philippe Mathieu-Daudé wrote:
> > > > diff --git a/tcg/tcg.c b/tcg/tcg.c
> > > > index da91779890..9b7df71e7a 100644
> > > > --- a/tcg/tcg.c
> > > > +++ b/tcg/tcg.c
> > > > @@ -61,6 +61,7 @@
> > > >    #include "exec/log.h"
> > > >    #include "tcg/tcg-ldst.h"
> > > >    #include "tcg-internal.h"
> > > > +#include "accel/tcg/perf.h"
> > >
> > > Is it OK to include an header from QEMU's accel/tcg/ here?
> > > I thought we wanted to keep tcg/ kinda independant (or maybe
> > > this is already too late and this isn't a concern anymore).
> >
> > It's not ideal, no.  Perf really should live in tcg/.
> >
> >
> > r~
>
> This would require to somehow get rid of this:
>
> #if defined(TARGET_I386)
>              q[insn].address -= tb->cs_base;
> #endif
>
> I'll try to come up with a patch.

Just drop it?  Did you really want EIP instead of the full virtual address?
It only makes a difference for 16-bit mode anyway.


r~