i965: Handle TCS outputs and TES inputs.

TCS outputs and TES inputs both refer to a common "patch URB entry"
shared across all invocations.  First, there are some number of
per-patch entries.  Then, there are per-vertex entries accessed via
an offset for the variable and a stride times the vertex index.

Because these calculations need to be done in both the vec4 and scalar
backends, it's simpler to just compute the offset calculations in NIR.
It doesn't necessarily make much sense to use per-vertex intrinsics
afterwards, but that at least means we don't lose the per-patch vs.
per-vertex information.

v2: Use is_input/is_output helpers (suggested by Jordan Justen).

Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Jordan Justen <[email protected]>

1 files changed, 112 insertions, 2 deletions

diff --git a/src/mesa/drivers/dri/i965/brw_nir.c b/src/mesa/drivers/dri/i965/brw_nir.c
index 2cce4aa946b..6fb1d62a5ca 100644
--- a/src/mesa/drivers/dri/i965/brw_nir.c
+++ b/src/mesa/drivers/dri/i965/brw_nir.c
@@ -135,6 +135,61 @@ remap_inputs_with_vue_map(nir_block *block, void *closure)
    return true;
 }
 
+struct remap_patch_urb_offsets_state {
+   nir_builder b;
+   struct brw_vue_map vue_map;
+};
+
+static bool
+remap_patch_urb_offsets(nir_block *block, void *closure)
+{
+   struct remap_patch_urb_offsets_state *state = closure;
+
+   nir_foreach_instr_safe(block, instr) {
+      if (instr->type != nir_instr_type_intrinsic)
+         continue;
+
+      nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
+
+      gl_shader_stage stage = state->b.shader->stage;
+
+      if ((stage == MESA_SHADER_TESS_CTRL && is_output(intrin)) ||
+          (stage == MESA_SHADER_TESS_EVAL && is_input(intrin))) {
+         int vue_slot = state->vue_map.varying_to_slot[intrin->const_index[0]];
+         assert(vue_slot != -1);
+         intrin->const_index[0] = vue_slot;
+
+         nir_src *vertex = nir_get_io_vertex_index_src(intrin);
+         if (vertex) {
+            nir_const_value *const_vertex = nir_src_as_const_value(*vertex);
+            if (const_vertex) {
+               intrin->const_index[0] += const_vertex->u[0] *
+                                         state->vue_map.num_per_vertex_slots;
+            } else {
+               state->b.cursor = nir_before_instr(&intrin->instr);
+
+               /* Multiply by the number of per-vertex slots. */
+               nir_ssa_def *vertex_offset =
+                  nir_imul(&state->b,
+                           nir_ssa_for_src(&state->b, *vertex, 1),
+                           nir_imm_int(&state->b,
+                                       state->vue_map.num_per_vertex_slots));
+
+               /* Add it to the existing offset */
+               nir_src *offset = nir_get_io_offset_src(intrin);
+               nir_ssa_def *total_offset =
+                  nir_iadd(&state->b, vertex_offset,
+                           nir_ssa_for_src(&state->b, *offset, 1));
+
+               nir_instr_rewrite_src(&intrin->instr, offset,
+                                     nir_src_for_ssa(total_offset));
+            }
+         }
+      }
+   }
+   return true;
+}
+
 static void
 brw_nir_lower_inputs(nir_shader *nir,
                      const struct brw_device_info *devinfo,
@@ -225,6 +280,31 @@ brw_nir_lower_inputs(nir_shader *nir,
       }
       break;
    }
+   case MESA_SHADER_TESS_EVAL: {
+      struct remap_patch_urb_offsets_state state;
+      brw_compute_tess_vue_map(&state.vue_map,
+                               nir->info.inputs_read & ~VARYING_BIT_PRIMITIVE_ID,
+                               nir->info.patch_inputs_read);
+
+      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);
+
+      /* This pass needs actual constants */
+      nir_opt_constant_folding(nir);
+
+      nir_foreach_overload(nir, overload) {
+         if (overload->impl) {
+            nir_builder_init(&params.b, overload->impl);
+            nir_foreach_block(overload->impl, add_const_offset_to_base, &params);
+            nir_builder_init(&state.b, overload->impl);
+            nir_foreach_block(overload->impl, remap_patch_urb_offsets, &state);
+         }
+      }
+      break;
+   }
    case MESA_SHADER_FRAGMENT:
       assert(is_scalar);
       nir_assign_var_locations(&nir->inputs, &nir->num_inputs,
@@ -240,7 +320,9 @@ brw_nir_lower_inputs(nir_shader *nir,
 }
 
 static void
-brw_nir_lower_outputs(nir_shader *nir, bool is_scalar)
+brw_nir_lower_outputs(nir_shader *nir,
+                      const struct brw_device_info *devinfo,
+                      bool is_scalar)
 {
    switch (nir->stage) {
    case MESA_SHADER_VERTEX:
@@ -255,6 +337,34 @@ brw_nir_lower_outputs(nir_shader *nir, bool is_scalar)
             var->data.driver_location = var->data.location;
       }
       break;
+   case MESA_SHADER_TESS_CTRL: {
+      struct add_const_offset_to_base_params params = {
+         .mode = nir_var_shader_out
+      };
+
+      struct remap_patch_urb_offsets_state state;
+      brw_compute_tess_vue_map(&state.vue_map, nir->info.outputs_written,
+                               nir->info.patch_outputs_written);
+
+      nir_foreach_variable(var, &nir->outputs) {
+         var->data.driver_location = var->data.location;
+      }
+
+      nir_lower_io(nir, nir_var_shader_out, type_size_vec4);
+
+      /* This pass needs actual constants */
+      nir_opt_constant_folding(nir);
+
+      nir_foreach_overload(nir, overload) {
+         if (overload->impl) {
+            nir_builder_init(&params.b, overload->impl);
+            nir_foreach_block(overload->impl, add_const_offset_to_base, &params);
+            nir_builder_init(&state.b, overload->impl);
+            nir_foreach_block(overload->impl, remap_patch_urb_offsets, &state);
+         }
+      }
+      break;
+   }
    case MESA_SHADER_FRAGMENT:
       nir_assign_var_locations(&nir->outputs, &nir->num_outputs,
                                type_size_scalar);
@@ -422,7 +532,7 @@ brw_lower_nir(nir_shader *nir,
    (void)progress;
 
    OPT_V(brw_nir_lower_inputs, devinfo, is_scalar);
-   OPT_V(brw_nir_lower_outputs, is_scalar);
+   OPT_V(brw_nir_lower_outputs, devinfo, is_scalar);
    OPT_V(brw_nir_lower_uniforms, is_scalar);
    OPT_V(nir_lower_io, nir_var_all, is_scalar ? type_size_scalar : type_size_vec4);