On Thu, 2019-01-24 at 10:22 -0800, Matt Turner wrote:
> On Wed, Jan 23, 2019 at 4:18 AM Iago Toral Quiroga <ito...@igalia.com
> > wrote:
> >
> > Commit c84ec70b3a72 implemented execution type promotion to 32-bit
> > for
> > conversions involving half-float registers, which empirical testing
> > suggested
> > was required, but it did not incorporate this change into the
> > assembly validator
> > logic. This commits adds that, preventing validation errors like
> > this:
> >
> > mov(16) g9<4>B g3<16,8,2>HF { align1 1H };
> > ERROR: Destination stride must be equal to the ratio of the sizes
> > of the
> > execution data type to the destination type
> >
> > Fixes: c84ec70b3a72 "intel/fs: Promote execution type to 32-bit
> > when any half-float conversion is needed."
> > ---
> > src/intel/compiler/brw_eu_validate.c | 27 ++++++++++++++--------
> > -----
>
> New rule: New restrictions (or relaxations) may not be added to
> brw_eu_validate.c without accompanying unit tests. I'll send a patch
> to add a comment to brw_eu_validate.c saying as much.
>
> Rationale: the reason I wrote brw_eu_validate.c was because I wasted
> a
> week debugging an issue where fulsim not only failed to inform me
> that
> one instruction was invalid but also incorrectly told me that one
> correct instruction *was* invalid. I would have been better off
> without such a tool.
>
> If the EU validator loses people's trust, then it's useless, but if
> it
> is incorrect it's worse than useless.
Makes sense to me.
Iago
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
