Oleg Endo <oleg.e...@t-online.de> wrote:
> The attached patch is the same as the last proposed patch in the PR.
> Re-tested it on sh-sim against rev 187914 with -msoft-atomic and
> -mhard-atomic enabled, as well with 'make info dvi pdf'.
>
> I was hoping that after recent reload patches the R0 spill failure
> problem
> (see http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50751#c29 ) might go
> away, but it didn't. The gcc.c-torture/compile/sync-1.c test still
> fails when -mhard-atomic is enabled.
> Actually this patch does not introduce the spill failure problem, the
> sync-1 test case just makes it obvious. The actual cause comes from the
> patches of PR 50751 and (as far as I understand) some missing cases in
> reload, which should be addressed separately, I think.
>
> OK to apply?
OK.
Regards,
kaz
