[Mesa-dev] [RFC 31/31] nir: Add a bool to float32 lowering pass

Jason Ekstrand jason at jlekstrand.net
Mon Oct 22 23:42:38 UTC 2018

On Mon, Oct 22, 2018 at 6:20 PM Alyssa Rosenzweig <alyssa at rosenzweig.io>

> For what it's worth, Midgard has real integers (including int32
> support), using hardware-level D3D10 boolean conventions. I'm trying to
> wrap my head around how this interacts with 5d85a0a.

Right.  Sorry.  I forgot who's working on what chip these days.  Maybe Ilia
can CC the right person.


> I'm tempted to think the standard lower_bool_to_int32 pass would work,
> with an emulated b2f instruction in the backend IR level, rather than
> burdening NIR with those details.
> I'll have to look at the patch set closer to understand the impact.
> ---
> > +nir_lower_bool_to_int32_impl(nir_function_impl *impl)
> > +nir_lower_bool_to_int32(nir_shader *shader)
> > +      if (function->impl &&
> nir_lower_bool_to_int32_impl(function->impl))
> I'm guessing these were intended to read float32?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20181022/b55594f9/attachment.html>

More information about the mesa-dev mailing list