ac/nir/ngg: add an option to skip viewport-based culling

We can do W and face culling when we have multiple viewports, but not
frustum and small prim culling because those are dependent on the viewport.
When a shader writes the viewport index, the new option allows skipping
viewport-based culling while keeping W and face culling.

Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33482>

src/amd/common/nir/ac_nir.h

@@ -160,6 +160,10 @@ typedef struct {
    bool kill_layer;
    bool force_vrs;
    bool compact_primitives;
+   /* Skip culling dependent on the viewport state, which is frustum culling and small prim
+    * culling. Set this when the shader writes the viewport index.
+    */
+   bool skip_viewport_culling;
 
    /* VS */
    unsigned num_vertices_per_primitive;

src/amd/common/nir/ac_nir_cull.c

@@ -134,6 +134,7 @@ call_accept_func(nir_builder *b, nir_def *accepted, ac_nir_cull_accepted accept_
 
 static nir_def *
 ac_nir_cull_triangle(nir_builder *b,
+                     bool skip_viewport_culling,
                      nir_def *initially_accepted,
                      nir_def *pos[3][4],
                      position_w_info *w_info,
@@ -144,6 +145,11 @@ ac_nir_cull_triangle(nir_builder *b,
    accepted = nir_iand(b, accepted, nir_inot(b, w_info->all_w_negative_or_zero_or_nan));
    accepted = nir_iand(b, accepted, nir_inot(b, cull_face_triangle(b, pos, w_info)));
 
+   if (skip_viewport_culling) {
+      call_accept_func(b, accepted, accept_func, state);
+      return accepted;
+   }
+
    nir_def *bbox_accepted = NULL;
 
    nir_if *if_accepted = nir_push_if(b, accepted);
@@ -305,6 +311,7 @@ cull_small_primitive_line(nir_builder *b, nir_def *pos[3][4],
 
 static nir_def *
 ac_nir_cull_line(nir_builder *b,
+                 bool skip_viewport_culling,
                  nir_def *initially_accepted,
                  nir_def *pos[3][4],
                  position_w_info *w_info,
@@ -314,6 +321,11 @@ ac_nir_cull_line(nir_builder *b,
    nir_def *accepted = initially_accepted;
    accepted = nir_iand(b, accepted, nir_inot(b, w_info->all_w_negative_or_zero_or_nan));
 
+   if (skip_viewport_culling) {
+      call_accept_func(b, accepted, accept_func, state);
+      return accepted;
+   }
+
    nir_def *bbox_accepted = NULL;
 
    nir_if *if_accepted = nir_push_if(b, accepted);
@@ -336,6 +348,7 @@ ac_nir_cull_line(nir_builder *b,
 
 nir_def *
 ac_nir_cull_primitive(nir_builder *b,
+                      bool skip_viewport_culling,
                       nir_def *initially_accepted,
                       nir_def *pos[3][4],
                       unsigned num_vertices,
@@ -345,12 +358,15 @@ ac_nir_cull_primitive(nir_builder *b,
    position_w_info w_info = {0};
    analyze_position_w(b, pos, num_vertices, &w_info);
 
-   if (num_vertices == 3)
-      return ac_nir_cull_triangle(b, initially_accepted, pos, &w_info, accept_func, state);
-   else if (num_vertices == 2)
-      return ac_nir_cull_line(b, initially_accepted, pos, &w_info, accept_func, state);
-   else
+   if (num_vertices == 3) {
+      return ac_nir_cull_triangle(b, skip_viewport_culling, initially_accepted, pos, &w_info,
+                                  accept_func, state);
+   } else if (num_vertices == 2) {
+      return ac_nir_cull_line(b, skip_viewport_culling, initially_accepted, pos, &w_info,
+                              accept_func, state);
+   } else {
       unreachable("point culling not implemented");
+   }
 
    return NULL;
 }

src/amd/common/nir/ac_nir_helpers.h

@@ -146,6 +146,7 @@ ac_nir_map_io_location(unsigned location,
 
 nir_def *
 ac_nir_cull_primitive(nir_builder *b,
+                      bool skip_viewport_culling,
                       nir_def *initially_accepted,
                       nir_def *pos[3][4],
                       unsigned num_vertices,

src/amd/common/nir/ac_nir_lower_ngg.c

@@ -1288,7 +1288,7 @@ add_deferred_attribute_culling(nir_builder *b, nir_cf_list *original_extracted_c
          }
 
          /* See if the current primitive is accepted */
-         ac_nir_cull_primitive(b, accepted_by_clipdist, pos,
+         ac_nir_cull_primitive(b, s->options->skip_viewport_culling, accepted_by_clipdist, pos,
                                s->options->num_vertices_per_primitive,
                                cull_primitive_accepted, s);
       }

src/amd/common/nir/ac_nir_lower_ngg_gs.c

@@ -631,8 +631,9 @@ ngg_gs_cull_primitive(nir_builder *b, nir_def *tid_in_tg, nir_def *max_vtxcnt,
       /* TODO: support clipdist culling in GS */
       nir_def *accepted_by_clipdist = nir_imm_true(b);
 
-      nir_def *accepted = ac_nir_cull_primitive(
-         b, accepted_by_clipdist, pos, s->num_vertices_per_primitive, NULL, NULL);
+      nir_def *accepted = ac_nir_cull_primitive(b, s->options->skip_viewport_culling,
+                                                accepted_by_clipdist, pos,
+                                                s->num_vertices_per_primitive, NULL, NULL);
 
       nir_if *if_rejected = nir_push_if(b, nir_inot(b, accepted));
       {