nir: wire shading rate variables

v2: Fixup comment about bits in nir_intrinsics.py

v3: Use varying for primitive shading rate builtin (samuel)

v4: Reoder switch alphabetically
    Make divergence of frag_shading_rate an option

v5: Remove stage check for frag_shading_rate in divergence (Samuel)

v6: s/frag_shading_rate_per_subgroup/single_frag_shading_rate_per_subgroup/ (Jason)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7795>

src/compiler/nir/nir.c

@@ -2247,6 +2247,8 @@ nir_intrinsic_from_system_value(gl_system_value val)
       return nir_intrinsic_load_ray_geometry_index;
    case SYSTEM_VALUE_RAY_INSTANCE_CUSTOM_INDEX:
       return nir_intrinsic_load_ray_instance_custom_index;
+   case SYSTEM_VALUE_FRAG_SHADING_RATE:
+      return nir_intrinsic_load_frag_shading_rate;
    default:
       unreachable("system value does not directly correspond to intrinsic");
    }
@@ -2382,6 +2384,8 @@ nir_system_value_from_intrinsic(nir_intrinsic_op intrin)
       return SYSTEM_VALUE_RAY_GEOMETRY_INDEX;
    case nir_intrinsic_load_ray_instance_custom_index:
       return SYSTEM_VALUE_RAY_INSTANCE_CUSTOM_INDEX;
+   case nir_intrinsic_load_frag_shading_rate:
+      return SYSTEM_VALUE_FRAG_SHADING_RATE;
    default:
       unreachable("intrinsic doesn't produce a system value");
    }

src/compiler/nir/nir.h

@@ -3070,6 +3070,7 @@ typedef enum {
    nir_divergence_single_patch_per_tcs_subgroup = (1 << 1),
    nir_divergence_single_patch_per_tes_subgroup = (1 << 2),
    nir_divergence_view_index_uniform = (1 << 3),
+   nir_divergence_single_frag_shading_rate_per_subgroup = (1 << 4),
 } nir_divergence_options;
 
 typedef struct nir_shader_compiler_options {

src/compiler/nir/nir_divergence_analysis.c

@@ -136,6 +136,9 @@ visit_intrinsic(nir_shader *shader, nir_intrinsic_instr *instr)
       break;
 
    /* Intrinsics with divergence depending on shader stage and hardware */
+   case nir_intrinsic_load_frag_shading_rate:
+      is_divergent = !(options & nir_divergence_single_frag_shading_rate_per_subgroup);
+      break;
    case nir_intrinsic_load_input:
       is_divergent = instr->src[0].ssa->divergent;
       if (stage == MESA_SHADER_FRAGMENT)

src/compiler/nir/nir_gather_info.c

@@ -465,6 +465,7 @@ gather_intrinsic_info(nir_intrinsic_instr *instr, nir_shader *shader,
    case nir_intrinsic_load_draw_id:
    case nir_intrinsic_load_invocation_id:
    case nir_intrinsic_load_frag_coord:
+   case nir_intrinsic_load_frag_shading_rate:
    case nir_intrinsic_load_point_coord:
    case nir_intrinsic_load_line_coord:
    case nir_intrinsic_load_front_face:

src/compiler/nir/nir_intrinsics.py

@@ -943,6 +943,15 @@ store("global", [1], [WRITE_MASK, ACCESS, ALIGN_MUL, ALIGN_OFFSET])
 # src[] = { value, offset }.
 store("scratch", [1], [ALIGN_MUL, ALIGN_OFFSET, WRITE_MASK])
 
+# A bit field to implement SPIRV FragmentShadingRateKHR
+# bit | name              | description
+#   0 | Vertical2Pixels   | Fragment invocation covers 2 pixels vertically
+#   1 | Vertical4Pixels   | Fragment invocation covers 4 pixels vertically
+#   2 | Horizontal2Pixels | Fragment invocation covers 2 pixels horizontally
+#   3 | Horizontal4Pixels | Fragment invocation covers 4 pixels horizontally
+intrinsic("load_frag_shading_rate", dest_comp=1, bit_sizes=[32],
+          flags=[CAN_ELIMINATE, CAN_REORDER])
+
 # IR3-specific version of most SSBO intrinsics. The only different
 # compare to the originals is that they add an extra source to hold
 # the dword-offset, which is needed by the backend code apart from

src/compiler/nir/nir_opt_peephole_select.c

@@ -113,6 +113,7 @@ block_check_for_allowed_instrs(nir_block *block, unsigned *count,
          case nir_intrinsic_load_subgroup_id:
          case nir_intrinsic_load_subgroup_invocation:
          case nir_intrinsic_load_num_subgroups:
+         case nir_intrinsic_load_frag_shading_rate:
             if (!alu_ok)
                return false;
             break;

src/compiler/shader_enums.c

@@ -304,6 +304,7 @@ gl_system_value_name(gl_system_value sysval)
      ENUM(SYSTEM_VALUE_RAY_GEOMETRY_INDEX),
      ENUM(SYSTEM_VALUE_GS_HEADER_IR3),
      ENUM(SYSTEM_VALUE_TCS_HEADER_IR3),
+     ENUM(SYSTEM_VALUE_FRAG_SHADING_RATE),
    };
    STATIC_ASSERT(ARRAY_SIZE(names) == SYSTEM_VALUE_MAX);
    return NAME(sysval);

src/compiler/shader_enums.h

@@ -288,7 +288,8 @@ typedef enum
    VARYING_SLOT_BOUNDING_BOX1, /* Only appears as TCS output. */
    VARYING_SLOT_VIEW_INDEX,
    VARYING_SLOT_VIEWPORT_MASK, /* Does not appear in FS */
-   VARYING_SLOT_VAR0, /* First generic varying slot */
+   VARYING_SLOT_PRIMITIVE_SHADING_RATE = VARYING_SLOT_FACE, /* Does not appear in FS. */
+   VARYING_SLOT_VAR0 = 32, /* First generic varying slot */
    /* the remaining are simply for the benefit of gl_varying_slot_name()
     * and not to be construed as an upper bound:
     */
@@ -708,6 +709,11 @@ typedef enum
    SYSTEM_VALUE_GS_HEADER_IR3,
    SYSTEM_VALUE_TCS_HEADER_IR3,
 
+   /**
+    * Fragment shading rate used for KHR_fragment_shading_rate (Vulkan).
+    */
+   SYSTEM_VALUE_FRAG_SHADING_RATE,
+
    SYSTEM_VALUE_MAX             /**< Number of values */
 } gl_system_value;