Re: [PATCH] ppc: Fix 'calli' when floating-point arguments are passed

[Paul Cercueil]

Hi Paulo,

Le mer., sept. 7 2022 at 10:38:14 -0300, Paulo César Pereira de 
Andrade <paulo.cesar.pereira.de.andrade@gmail.com> a écrit :
> Em qua., 7 de set. de 2022 às 10:30, Paul Cercueil
> <paul@crapouillou.net> escreveu:
> >  Le mer., sept. 7 2022 at 10:14:15 -0300, Paulo César Pereira de
> >  Andrade <paulo.cesar.pereira.de.andrade@gmail.com> a écrit :
> >  > Em qua., 7 de set. de 2022 às 09:46, Paul Cercueil
> >  > <paul@crapouillou.net> escreveu:
> >  >
> >  > [snip]
> >  >
> >  >>  Unfortunately it now starts to use my _R10 register as a
> >  >> temporary...
> >  >>  see attachment.
> >  >>
> >  >>  It seems to be detected properly as live at label L3, but it is 
> > not
> >  >>  detected as live at label L6, and it should be.
> >  >
> >  >   Sorry. In the previous patch I had just broken jit_live() in the
> >  > context
> >  > you use it.
> >  >   Should work now.
> >
> >  Looks like everything is working perfectly now :)
> >
> >  Thank you for the help and the quick responses.
>
>   No problems :) It is a holiday here, but it is in the middle of the
> week, and it is raining :)
>   I am doing some work to support simple atomic operations in
> lightning.
>   A simple jit_casr and jit_casi should be working for x86 and arm
> now.

I spoke too soon, I'm still getting problems :(

See the attachment.

The problem now is that r26 should be live at L2 but is not detected as 
such.
This causes Lightning to use r26 as a temporary for the andi line 34 
(lines 153-155 in the generated code).

Cheers,
-Paul

lightning_ppc3.txt
Description: Text document