[Mesa-dev] [PATCH] nir: extract out helper macros for running passes

Jason Ekstrand jason at jlekstrand.net
Fri Jan 1 12:40:08 PST 2016


On Wed, Dec 30, 2015 at 9:04 AM, Rob Clark <robdclark at gmail.com> wrote:

> On Wed, Dec 30, 2015 at 10:32 AM, Rob Clark <robdclark at gmail.com> wrote:
> > diff --git a/src/mesa/drivers/dri/i965/brw_nir.c
> b/src/mesa/drivers/dri/i965/brw_nir.c
> > index e031173..a8bed70 100644
> > --- a/src/mesa/drivers/dri/i965/brw_nir.c
> > +++ b/src/mesa/drivers/dri/i965/brw_nir.c
> > @@ -405,42 +405,8 @@ brw_nir_lower_uniforms(nir_shader *nir, bool
> is_scalar)
> >     }
> >  }
> >
> > -#include "util/debug.h"
> > -
> > -static bool
> > -should_clone_nir()
> > -{
> > -   static int should_clone = -1;
> > -   if (should_clone < 1)
>
> btw, EdB pointed out on irc..  shouldn't this be if (should_clone < 0) ?
>

Yes, it should.

With that,

Reviewed-by: Jason Ekstrand <jason.ekstraand at intel.com>

As I mentioned on IRC, I recently realized that Ken's mechanism for
enforcing metadata kind-of falls over in the face of multiple functions.
In particular, you may not modify one of the functions and therefore don't
need to invalidate metadata but if you invalidate so much as one function,
you return progress and it then asserts that all functions have invalidated
metadata.  Not sure what the best solution there is.  Maybe just have a
nir_metadata_preserve(impl, nir_metadata_all) and require you to call
that?  I don't really like that either because someone will use
nir_metadata_all when they make a change that doesn't invalidate anything
and then someone will add metadata that does get invalidated and we have a
problem.  Thoughts?


>
> BR,
> -R
>
> > -      should_clone = env_var_as_boolean("NIR_TEST_CLONE", false);
> > -
> > -   return should_clone;
> > -}
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/mesa-dev/attachments/20160101/de0b15d2/attachment.html>


More information about the mesa-dev mailing list