[PATCH, i386]: Fix secondary_reload_needed (was: Re: [PATCH, i386]: Improve STV conversion of shifts)

[Uros Bizjak <ubizjak@gmail.com>]

As usual with costing changes, the patch exposes latent problem. The
patched compiler tries to generate non-existing DImode move from mask
register to XMM register, and ICEs during reload [1]. Attached patch
tightens secondary_reload_needed condition and fixes the issue.

I'm bootstrapping and regression testing patch, and will submit a
formal submission later today.

[1] https://gcc.gnu.org/ml/gcc-regression/2019-08/msg00537.html

Uros.

On Thu, Aug 29, 2019 at 9:53 AM Uros Bizjak <ubizjak@gmail.com> wrote:
>
> On Wed, Aug 28, 2019 at 5:12 PM Uros Bizjak <ubizjak@gmail.com> wrote:
> >
> > Attached patch improves costing for STV shifts and corrects reject
> > condition for out of range shift count operands.
> >
> > 2019-08-28  Uroš Bizjak  <ubizjak@gmail.com>
> >
> >     * config/i386/i386-features.c
> >     (general_scalar_chain::compute_convert_gain):
> >     Correct cost for double-word shifts.
> >     (general_scalar_to_vector_candidate_p): Reject count operands
> >     greater or equal to mode bitsize.
> >
> > Bootstrapped and regression tested on x86_64-linux-gnu {,-m32}.
> >
> > Committed to mainline SVN.
>
> Ouch... I mixed up patches and actually committed the patch that
> removes maximum from cost of sse<->int moves.
>
> I can leave the patch for a day, so we can see the effects of the cost
> change, and if the patch creates problems, I'll revert it.
>
> Sorry for the mixup,
> Uros.