[Mesa-dev] [PATCH 2/4] nir: Silence unused "options" warning in algebraic passes.

[Dylan Baker]

Quoting Dylan Baker (2016-04-08 14:54:38)
> Quoting Kenneth Graunke (2016-04-08 11:48:01)
> > On Friday, April 8, 2016 11:18:46 AM PDT Dylan Baker wrote: > 
> > > Quoting Eduardo Lima Mitev (2016-04-08 08:26:22)
> > > > On 04/08/2016 01:35 AM, Kenneth Graunke wrote:
> > > > > Some passes may not refer to options->..., at which point the compiler
> > > > > will warn about an unused variable. Just cast to void unconditionally
> > > > > to shut it up.
> > > > >
> > > > > Signed-off-by: Kenneth Graunke <kenneth at whitecape.org>
> > > > > ---
> > > > > src/compiler/nir/nir_algebraic.py | 1 +
> > > > > 1 file changed, 1 insertion(+)
> > > > >
> > > > > diff --git a/src/compiler/nir/nir_algebraic.py b/src/compiler/nir/
> > nir_algebraic.py
> > > > > index d05564f..53a7907 100644
> > > > > --- a/src/compiler/nir/nir_algebraic.py
> > > > > +++ b/src/compiler/nir/nir_algebraic.py
> > > > > @@ -291,6 +291,7 @@ ${pass_name}(nir_shader *shader)
> > > > > bool progress = false;
> > > > > bool condition_flags[${len(condition_list)}];
> > > > > const nir_shader_compiler_options *options = shader->options;
> > > > > + (void) options;
> > > > >
> > > >
> > > > Hmm, don't like this very much. I suppose since it is generated code, it
> > > > can not be done per block where options is unused.
> > > > For respect to other people's right to have clean build logs, patch is:
> > > >
> > > > Reviewed-by: Eduardo Lima Mitev <elima at igalia.com>
> > > >
> > > > > % for index, condition in enumerate(condition_list):
> > > > > condition_flags[${index}] = ${condition};
> > > > >
> > > >
> > >
> > > I'm not familiar with the code, but looking at it options must be used
> > > in the condition_list argument?
> >  
> > Right, we create the 'options' temporary so you can write rules like:
> >  
> > (('~fmin', ('fmax', a, 0.0), 1.0), ('fsat', a), '!options->lower_fsat'),
> >  
> > where !options->lower_fsat becomes part of the condition_list. However,
> > conditions are optional. If you have an algebraic rule list which doesn't
> > include any conditions referring to options->..., the variable will be
> > unused. (For example, the list in patch 4 meets this criteria.)
> >  
> 
> So what about wrapping the definitions of options in a conditional?
> % if 'options' in conditions[-1]:
> const nir_shader_compiler_options *options = shader->options;
> % endif
> 
> This only covers that options is the last argument, which seems to be
> what nir_opt_algebraic does, although I don't know that is actually a
> hard requirement, or just a coincidence.
> 
> Dylan

Never mind. Jason and I talked about this and this might possibly work,
by accident. Please ignore.

Dylan
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 473 bytes
Desc: signature
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20160408/252a2494/attachment.sig>