[Mesa-dev] [PATCH 2/2] anv/pipeline: don't pass constant view index in multiview

Caio Marcelo de Oliveira Filho caio.oliveira at intel.com
Wed Mar 14 23:16:56 UTC 2018


On Tue, Feb 27, 2018 at 12:13:52PM -0800, Jason Ekstrand wrote:
> > diff --git a/src/intel/vulkan/anv_nir_lower_multiview.c
> > b/src/intel/vulkan/anv_nir_lower_multiview.c
> > index d2aefdee62..365a70d757 100644
> > --- a/src/intel/vulkan/anv_nir_lower_multiview.c
> > +++ b/src/intel/vulkan/anv_nir_lower_multiview.c
> > @@ -72,7 +72,8 @@ build_view_index(struct lower_multiview_state *state)
> >        b->cursor = nir_before_block(nir_start_block(b->impl));
> >
> >        assert(state->view_mask != 0);
> > -      if (0 && _mesa_bitcount(state->view_mask) == 1) {
> > +      if (_mesa_bitcount(state->view_mask) == 1) {
> >
> 
> Yes, I think it's probably safe to turn this on now.  Originally, I had it
> commented out because I was afraid of not getting enough test coverage.
> For all I know, we still aren't getting enough test coverage but I think
> we've proven by now that the calculations below work.

There are now CTS tests covering multiview, including cases with a
single view set in the view_mask.


> > +      /* Unless there is only one possible view index (that would be set
> > +       * directly), pass it to the next stage. */
> >
> 
> With multi-line comments, we usually put the "*/" on its own line.

Just sent v2 fixing this.


Thanks,
Caio


More information about the mesa-dev mailing list