On Wed, Sep 25, 2019 at 4:18 PM Martin Jambor <mjam...@suse.cz> wrote:
>
> Hi,
>
> On Tue, Sep 24 2019, Martin Jambor wrote:
> >
> >
> > It is the correct thing to do, sorry for the breakage. I have to run
> > now but will prepare a patch tomorrow.
> >
>
> and here it is. The patch fixes the thinko explained in my email
> yesterday - basically the test for locally_unused was intended for
> unused aggregates which have however not been marked as such yet and
> going this way for unsplitable but unused register-type parameters may
> cause problems in some cases, if they are for example big SVE vectors.
>
> Passed bootstrap and testing on x86_64-linux. OK for trunk?
OK.
Richard.
> Thanks,
>
> Martin
>
>
> 2019-09-25 Martin Jambor <mjam...@suse.cz>
>
> * ipa-sra.c (process_scan_results): Fix continue condition.
> ---
> gcc/ipa-sra.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/gcc/ipa-sra.c b/gcc/ipa-sra.c
> index 0ccebbd4607..b35fff69472 100644
> --- a/gcc/ipa-sra.c
> +++ b/gcc/ipa-sra.c
> @@ -2239,7 +2239,7 @@ process_scan_results (cgraph_node *node, struct
> function *fun,
> desc_index++, parm = DECL_CHAIN (parm))
> {
> gensum_param_desc *desc = &(*param_descriptions)[desc_index];
> - if (!desc->locally_unused && !desc->split_candidate)
> + if (!desc->split_candidate)
> continue;
>
> if (flag_checking)
> --
> 2.23.0
>
