[Mesa-dev] [PATCH 3/4] i965/fs: Use NIR for handling forced per-sample interpolation

[Jason Ekstrand]

---
 src/mesa/drivers/dri/i965/brw_fs.cpp | 40 +++---------------------------------
 src/mesa/drivers/dri/i965/brw_nir.c  |  9 ++++++--
 src/mesa/drivers/dri/i965/brw_nir.h  |  3 ++-
 3 files changed, 12 insertions(+), 40 deletions(-)

diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp b/src/mesa/drivers/dri/i965/brw_fs.cpp
index 9dbb699..a3fc839 100644
--- a/src/mesa/drivers/dri/i965/brw_fs.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs.cpp
@@ -6461,8 +6461,7 @@ move_interpolation_to_top(nir_shader *nir)
 static void
 brw_nir_set_default_interpolation(const struct gen_device_info *devinfo,
                                   struct nir_shader *nir,
-                                  bool api_flat_shade,
-                                  bool per_sample_interpolation)
+                                  bool api_flat_shade)
 {
    assert(nir->stage == MESA_SHADER_FRAGMENT);
 
@@ -6481,13 +6480,6 @@ brw_nir_set_default_interpolation(const struct gen_device_info *devinfo,
                                         : INTERP_MODE_SMOOTH;
       }
 
-      /* Apply 'sample' if necessary for API state. */
-      if (per_sample_interpolation &&
-          var->data.interpolation != INTERP_MODE_FLAT) {
-         var->data.centroid = false;
-         var->data.sample = true;
-      }
-
       /* On Ironlake and below, there is only one interpolation mode.
        * Centroid interpolation doesn't mean anything on this hardware --
        * there is no multisampling.
@@ -6497,32 +6489,6 @@ brw_nir_set_default_interpolation(const struct gen_device_info *devinfo,
          var->data.sample = false;
       }
    }
-
-   if (per_sample_interpolation) {
-      nir_foreach_block(block, nir_shader_get_entrypoint(nir)) {
-         nir_foreach_instr(instr, block) {
-            if (instr->type != nir_instr_type_intrinsic)
-               continue;
-
-            nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
-            if (intrin->intrinsic != nir_intrinsic_interp_var_at_centroid)
-               continue;
-
-            nir_variable *var = intrin->variables[0]->var;
-            if (var->data.interpolation == INTERP_MODE_FLAT)
-               continue;
-
-            /* The description of the interpolateAtCentroid intrinsic is that
-             * it interpolates the variable as if it had the "centroid"
-             * qualifier.  When executing with per_sample_interpolation, this
-             * is equivalent to having the "sample" qualifier.  Just convert
-             * it to a load_var instead.
-             */
-            assert(var->data.sample);
-            intrin->intrinsic = nir_intrinsic_load_var;
-         }
-      }
-   }
 }
 
 /**
@@ -6583,8 +6549,8 @@ brw_compile_fs(const struct brw_compiler *compiler, void *log_data,
    shader = brw_nir_apply_sampler_key(shader, compiler->devinfo, &key->tex,
                                       true);
    brw_nir_set_default_interpolation(compiler->devinfo, shader,
-                                     key->flat_shade, key->persample_interp);
-   brw_nir_lower_fs_inputs(shader);
+                                     key->flat_shade);
+   brw_nir_lower_fs_inputs(shader, key);
    brw_nir_lower_fs_outputs(shader);
    if (!key->multisample_fbo)
       NIR_PASS_V(shader, demote_sample_qualifiers);
diff --git a/src/mesa/drivers/dri/i965/brw_nir.c b/src/mesa/drivers/dri/i965/brw_nir.c
index 2273299..1a9ca8a 100644
--- a/src/mesa/drivers/dri/i965/brw_nir.c
+++ b/src/mesa/drivers/dri/i965/brw_nir.c
@@ -281,13 +281,18 @@ brw_nir_lower_tes_inputs(nir_shader *nir, const struct brw_vue_map *vue_map)
 }
 
 void
-brw_nir_lower_fs_inputs(nir_shader *nir)
+brw_nir_lower_fs_inputs(nir_shader *nir,
+                        const struct brw_wm_prog_key *key)
 {
    foreach_list_typed(nir_variable, var, node, &nir->inputs) {
       var->data.driver_location = var->data.location;
    }
 
-   nir_lower_io(nir, nir_var_shader_in, type_size_vec4, 0);
+   nir_lower_io_options lower_io_options = 0;
+   if (key->persample_interp)
+      lower_io_options |= nir_lower_io_force_sample_interpolation;
+
+   nir_lower_io(nir, nir_var_shader_in, type_size_vec4, lower_io_options);
 
    /* This pass needs actual constants */
    nir_opt_constant_folding(nir);
diff --git a/src/mesa/drivers/dri/i965/brw_nir.h b/src/mesa/drivers/dri/i965/brw_nir.h
index b025d55..51d2f63 100644
--- a/src/mesa/drivers/dri/i965/brw_nir.h
+++ b/src/mesa/drivers/dri/i965/brw_nir.h
@@ -106,7 +106,8 @@ void brw_nir_lower_vs_inputs(nir_shader *nir,
 void brw_nir_lower_vue_inputs(nir_shader *nir, bool is_scalar,
                               const struct brw_vue_map *vue_map);
 void brw_nir_lower_tes_inputs(nir_shader *nir, const struct brw_vue_map *vue);
-void brw_nir_lower_fs_inputs(nir_shader *nir);
+void brw_nir_lower_fs_inputs(nir_shader *nir,
+                             const struct brw_wm_prog_key *key);
 void brw_nir_lower_vue_outputs(nir_shader *nir, bool is_scalar);
 void brw_nir_lower_tcs_outputs(nir_shader *nir, const struct brw_vue_map *vue);
 void brw_nir_lower_fs_outputs(nir_shader *nir);
-- 
2.5.0.400.gff86faf