[Mesa-dev] [RFC] ARB_gl_spirv and NIR backend for radeonsi

[Timothy Arceri]

On 01/06/17 02:45, Marek Olšák wrote:
> On Tue, May 23, 2017 at 12:07 AM, Timothy Arceri <tarceri at itsqueeze.com> wrote:
>> 3. The GLSL IR opts are both slow and not always optimal (possibly
>> transforming the code to something that's harder to opt later), but due to
>> uniform/varying optimisation requirements some optimisations are required
>> *before* we can do validation. With NIR we have an opportunity to do these
>> optimisations in NIR either by building a nir based linker for the final
>> linking stage (uniform/varying validation/location assignment) or by a
>> little bit of back and forth of information between NIR and GLSL IR. This is
>> something that can't really be done with LLVM/Gallium. I was working towards
>> this while at Collabora.
> 
> If you really want to have a NIR-based linker for all drivers, the
> easiest way to get there is to add NIR-to-TGSI (for Gallium) and
> NIR-to-MESA-IR (for old classic drivers).
> 
> Marek
> 

Well I'm not really working on this, I gave Elie a list of what was 
remaining to do, but not sure if he plans to work on it.

There idea of replacing MESA-IR with NIR (rather than a conversion pass) 
has come up a few times but nothing has come of it. I think Matt was 
looking at it at one point. IMO it would be safer/a better use of time 
to just branch off the old classic drivers ... but seems like that's off 
the tables for now.