glsl: add AoA support for atomic counters

[mesa.git] / src / glsl / builtin_variables.cpp

diff --git a/src/glsl/builtin_variables.cpp b/src/glsl/builtin_variables.cpp
index f185d939444c899ea0d3f1539b1d0242235ebda9..a6ad10505525e7c7563df2681296f6f520c45344 100644 (file)
--- a/src/glsl/builtin_variables.cpp
+++ b/src/glsl/builtin_variables.cpp
@@ -22,6 +22,8 @@
  */
 
 #include "ir.h"
+#include "ir_builder.h"
+#include "linker.h"
 #include "glsl_parser_extras.h"
 #include "glsl_symbol_table.h"
 #include "main/core.h"
@@ -29,6 +31,8 @@
 #include "program/prog_statevars.h"
 #include "program/prog_instruction.h"
 
+using namespace ir_builder;
+
 static const struct gl_builtin_uniform_element gl_NumSamples_elements[] = {
    {NULL, {STATE_NUM_SAMPLES, 0, 0}, SWIZZLE_XXXX}
 };
@@ -398,6 +402,7 @@ private:
 
    const glsl_type * const bool_t;
    const glsl_type * const int_t;
+   const glsl_type * const uint_t;
    const glsl_type * const float_t;
    const glsl_type * const vec2_t;
    const glsl_type * const vec3_t;
@@ -416,6 +421,7 @@ builtin_variable_generator::builtin_variable_generator(
    : instructions(instructions), state(state), symtab(state->symbols),
      compatibility(!state->is_version(140, 100)),
      bool_t(glsl_type::bool_type), int_t(glsl_type::int_type),
+     uint_t(glsl_type::uint_type),
      float_t(glsl_type::float_type), vec2_t(glsl_type::vec2_type),
      vec3_t(glsl_type::vec3_type), vec4_t(glsl_type::vec4_type),
      uvec3_t(glsl_type::uvec3_type),
@@ -674,14 +680,10 @@ builtin_variable_generator::generate_constants()
       if (!state->es_shader) {
          add_const("gl_MaxGeometryAtomicCounters",
                    state->Const.MaxGeometryAtomicCounters);
-
-        if (state->is_version(400, 0) ||
-             state->ARB_tessellation_shader_enable) {
-                add_const("gl_MaxTessControlAtomicCounters",
-                           state->Const.MaxTessControlAtomicCounters);
-                add_const("gl_MaxTessEvaluationAtomicCounters",
-                           state->Const.MaxTessEvaluationAtomicCounters);
-        }
+         add_const("gl_MaxTessControlAtomicCounters",
+                   state->Const.MaxTessControlAtomicCounters);
+         add_const("gl_MaxTessEvaluationAtomicCounters",
+                   state->Const.MaxTessEvaluationAtomicCounters);
       }
    }
 
@@ -1055,7 +1057,10 @@ builtin_variable_generator::generate_cs_special_vars()
 {
    add_system_value(SYSTEM_VALUE_LOCAL_INVOCATION_ID, uvec3_t,
                     "gl_LocalInvocationID");
-   /* TODO: finish this. */
+   add_system_value(SYSTEM_VALUE_WORK_GROUP_ID, uvec3_t, "gl_WorkGroupID");
+   add_system_value(SYSTEM_VALUE_NUM_WORK_GROUPS, uvec3_t, "gl_NumWorkGroups");
+   add_variable("gl_GlobalInvocationID", uvec3_t, ir_var_auto, 0);
+   add_variable("gl_LocalInvocationIndex", uint_t, ir_var_auto, 0);
 }
 
 
@@ -1206,3 +1211,84 @@ _mesa_glsl_initialize_variables(exec_list *instructions,
       break;
    }
 }
+
+
+/**
+ * Initialize compute shader variables with values that are derived from other
+ * compute shader variable.
+ */
+static void
+initialize_cs_derived_variables(gl_shader *shader,
+                                ir_function_signature *const main_sig)
+{
+   assert(shader->Stage == MESA_SHADER_COMPUTE);
+
+   ir_variable *gl_GlobalInvocationID =
+      shader->symbols->get_variable("gl_GlobalInvocationID");
+   assert(gl_GlobalInvocationID);
+   ir_variable *gl_WorkGroupID =
+      shader->symbols->get_variable("gl_WorkGroupID");
+   assert(gl_WorkGroupID);
+   ir_variable *gl_WorkGroupSize =
+      shader->symbols->get_variable("gl_WorkGroupSize");
+   if (gl_WorkGroupSize == NULL) {
+      void *const mem_ctx = ralloc_parent(shader->ir);
+      gl_WorkGroupSize = new(mem_ctx) ir_variable(glsl_type::uvec3_type,
+                                                  "gl_WorkGroupSize",
+                                                  ir_var_auto);
+      gl_WorkGroupSize->data.how_declared = ir_var_declared_implicitly;
+      gl_WorkGroupSize->data.read_only = true;
+      shader->ir->push_head(gl_WorkGroupSize);
+   }
+   ir_variable *gl_LocalInvocationID =
+      shader->symbols->get_variable("gl_LocalInvocationID");
+   assert(gl_LocalInvocationID);
+
+   /* gl_GlobalInvocationID =
+    *    gl_WorkGroupID * gl_WorkGroupSize + gl_LocalInvocationID
+    */
+   ir_instruction *inst =
+      assign(gl_GlobalInvocationID,
+             add(mul(gl_WorkGroupID, gl_WorkGroupSize),
+                 gl_LocalInvocationID));
+   main_sig->body.push_head(inst);
+
+   /* gl_LocalInvocationIndex =
+    *    gl_LocalInvocationID.z * gl_WorkGroupSize.x * gl_WorkGroupSize.y +
+    *    gl_LocalInvocationID.y * gl_WorkGroupSize.x +
+    *    gl_LocalInvocationID.x;
+    */
+   ir_expression *index_z =
+      mul(mul(swizzle_z(gl_LocalInvocationID), swizzle_x(gl_WorkGroupSize)),
+          swizzle_y(gl_WorkGroupSize));
+   ir_expression *index_y =
+      mul(swizzle_y(gl_LocalInvocationID), swizzle_x(gl_WorkGroupSize));
+   ir_expression *index_y_plus_z = add(index_y, index_z);
+   operand index_x(swizzle_x(gl_LocalInvocationID));
+   ir_expression *index_x_plus_y_plus_z = add(index_y_plus_z, index_x);
+   ir_variable *gl_LocalInvocationIndex =
+      shader->symbols->get_variable("gl_LocalInvocationIndex");
+   assert(gl_LocalInvocationIndex);
+   inst = assign(gl_LocalInvocationIndex, index_x_plus_y_plus_z);
+   main_sig->body.push_head(inst);
+}
+
+
+/**
+ * Initialize builtin variables with values based on other builtin variables.
+ * These are initialized in the main function.
+ */
+void
+_mesa_glsl_initialize_derived_variables(gl_shader *shader)
+{
+   /* We only need to set CS variables currently. */
+   if (shader->Stage != MESA_SHADER_COMPUTE)
+      return;
+
+   ir_function_signature *const main_sig =
+      _mesa_get_main_function_signature(shader);
+   if (main_sig == NULL)
+      return;
+
+   initialize_cs_derived_variables(shader, main_sig);
+}