[Mesa-dev] [PATCH v3 30/44] i965/fs: Handle nir shared variable store intrinsic

[Iago Toral]

On Tue, 2015-12-01 at 13:00 -0800, Jordan Justen wrote:
> On 2015-12-01 04:45:05, Iago Toral wrote:
> > On Tue, 2015-12-01 at 00:19 -0800, Jordan Justen wrote:
> > > Signed-off-by: Jordan Justen <jordan.l.justen at intel.com>
> > > ---
> > >  src/mesa/drivers/dri/i965/brw_fs_nir.cpp | 52 ++++++++++++++++++++++++++++++++
> > >  1 file changed, 52 insertions(+)
> > > 
> > > diff --git a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
> > > index 12a8b59..6cbb0e2 100644
> > > --- a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
> > > +++ b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
> > > @@ -2472,6 +2472,58 @@ fs_visitor::nir_emit_intrinsic(const fs_builder &bld, nir_intrinsic_instr *instr
> > >        break;
> > >     }
> > >  
> > > +   case nir_intrinsic_store_shared_indirect:
> > > +      has_indirect = true;
> > > +      /* fallthrough */
> > > +   case nir_intrinsic_store_shared: {
> > > +      assert(devinfo->gen >= 7);
> > > +
> > > +      /* Block index */
> > > +      fs_reg surf_index = brw_imm_ud(GEN7_BTI_SLM);
> > > +
> > > +      /* Offset */
> > > +      fs_reg offset_reg = vgrf(glsl_type::uint_type);
> > > +      unsigned const_offset_bytes = 0;
> > > +      if (has_indirect) {
> > > +         bld.MOV(offset_reg, get_nir_src(instr->src[1]));
> > > +      } else {
> > > +         const_offset_bytes = instr->const_index[0];
> > > +         bld.MOV(offset_reg, brw_imm_ud(const_offset_bytes));
> > > +      }
> > > +
> > > +      /* Value */
> > > +      fs_reg val_reg = get_nir_src(instr->src[0]);
> > > +
> > > +      /* Writemask */
> > > +      unsigned writemask = instr->const_index[1];
> > > +
> > > +      /* Write each component present in the writemask */
> > 
> > I made a comment in v2 that this loop is based on early ssbo code that
> > was not optimized (it always emits a write for each component). The
> > current implementation for ssbo store is better and I think it is only a
> > matter of copying the same loop here, since the implementation is the
> > same as in the case of ssbos.
> 
> Noted (in v2 :)
> 
> http://lists.freedesktop.org/archives/mesa-dev/2015-November/101866.html
> 
> -Jordan

Ah, sorry I had missed your reply. I have just reviewed the v4 version
with the optimized version.

Iago

> > 
> > > +      unsigned skipped_channels = 0;
> > > +      for (int i = 0; i < instr->num_components; i++) {
> > > +         int component_mask = 1 << i;
> > > +         if (writemask & component_mask) {
> > > +            if (skipped_channels) {
> > > +               if (!has_indirect) {
> > > +                  const_offset_bytes += 4 * skipped_channels;
> > > +                  bld.MOV(offset_reg, brw_imm_ud(const_offset_bytes));
> > > +               } else {
> > > +                  bld.ADD(offset_reg, offset_reg,
> > > +                           brw_imm_ud(4 * skipped_channels));
> > > +               }
> > > +               skipped_channels = 0;
> > > +            }
> > > +
> > > +            emit_untyped_write(bld, surf_index, offset_reg,
> > > +                               offset(val_reg, bld, i),
> > > +                               1 /* dims */, 1 /* size */,
> > > +                               BRW_PREDICATE_NONE);
> > > +         }
> > > +
> > > +         skipped_channels++;
> > > +      }
> > > +      break;
> > > +   }
> > > +
> > >     case nir_intrinsic_load_input_indirect:
> > >        has_indirect = true;
> > >        /* fallthrough */
> > 
> > 
>