https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122414
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Resolution|--- |INVALID
Status|UNCONFIRMED |RESOLVED
--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Martin Husemann from comment #3)
> Thanks for the documentation pointer and explanation.
>
> AFAICT full 64bit registers are saved on task switch always on a 64bit
> kernel (but I have to admit I have never run one on powerpc so far), so
> OS_MISSING_POWERPC64 should not be defined.
If that is the case the bug is/was in the use of _ARCH_PPC64 in sljit .
It looks like sljit was fixed already upstream with
https://github.com/zherczeg/sljit/commit/a3caf4956a81046e0b717c764f34a0cf34ead44d
.
So what GCC is doing is correct based on the definition of _ARCH_PPC64 and the
use of -mcpu=970 here.
