glsl: move some uniform linking code to new link_assign_uniform_storage()
[mesa.git] / src / compiler / glsl / linker.cpp
index 89cd9f795cc3d0d6aa44e7fdef8d5409248762fb..290811fd2b97ab9e0c53d2157847541f87a132fc 100644 (file)
@@ -71,7 +71,9 @@
 #include "glsl_parser_extras.h"
 #include "ir.h"
 #include "program.h"
-#include "program/hash_table.h"
+#include "program/prog_instruction.h"
+#include "util/set.h"
+#include "util/string_to_uint_map.h"
 #include "linker.h"
 #include "link_varyings.h"
 #include "ir_optimization.h"
@@ -179,48 +181,57 @@ private:
 };
 
 
-class geom_array_resize_visitor : public ir_hierarchical_visitor {
+class array_resize_visitor : public ir_hierarchical_visitor {
 public:
    unsigned num_vertices;
    gl_shader_program *prog;
+   gl_shader_stage stage;
 
-   geom_array_resize_visitor(unsigned num_vertices, gl_shader_program *prog)
+   array_resize_visitor(unsigned num_vertices,
+                        gl_shader_program *prog,
+                        gl_shader_stage stage)
    {
       this->num_vertices = num_vertices;
       this->prog = prog;
+      this->stage = stage;
    }
 
-   virtual ~geom_array_resize_visitor()
+   virtual ~array_resize_visitor()
    {
       /* empty */
    }
 
    virtual ir_visitor_status visit(ir_variable *var)
    {
-      if (!var->type->is_array() || var->data.mode != ir_var_shader_in)
+      if (!var->type->is_array() || var->data.mode != ir_var_shader_in ||
+          var->data.patch)
          return visit_continue;
 
       unsigned size = var->type->length;
 
-      /* Generate a link error if the shader has declared this array with an
-       * incorrect size.
-       */
-      if (size && size != this->num_vertices) {
-         linker_error(this->prog, "size of array %s declared as %u, "
-                      "but number of input vertices is %u\n",
-                      var->name, size, this->num_vertices);
-         return visit_continue;
-      }
+      if (stage == MESA_SHADER_GEOMETRY) {
+         /* Generate a link error if the shader has declared this array with
+          * an incorrect size.
+          */
+         if (!var->data.implicit_sized_array &&
+             size && size != this->num_vertices) {
+            linker_error(this->prog, "size of array %s declared as %u, "
+                         "but number of input vertices is %u\n",
+                         var->name, size, this->num_vertices);
+            return visit_continue;
+         }
 
-      /* Generate a link error if the shader attempts to access an input
-       * array using an index too large for its actual size assigned at link
-       * time.
-       */
-      if (var->data.max_array_access >= (int)this->num_vertices) {
-         linker_error(this->prog, "geometry shader accesses element %i of "
-                      "%s, but only %i input vertices\n",
-                      var->data.max_array_access, var->name, this->num_vertices);
-         return visit_continue;
+         /* Generate a link error if the shader attempts to access an input
+          * array using an index too large for its actual size assigned at
+          * link time.
+          */
+         if (var->data.max_array_access >= (int)this->num_vertices) {
+            linker_error(this->prog, "%s shader accesses element %i of "
+                         "%s, but only %i input vertices\n",
+                         _mesa_shader_stage_to_string(this->stage),
+                         var->data.max_array_access, var->name, this->num_vertices);
+            return visit_continue;
+         }
       }
 
       var->type = glsl_type::get_array_instance(var->type->fields.array,
@@ -249,144 +260,6 @@ public:
    }
 };
 
-class tess_eval_array_resize_visitor : public ir_hierarchical_visitor {
-public:
-   unsigned num_vertices;
-   gl_shader_program *prog;
-
-   tess_eval_array_resize_visitor(unsigned num_vertices, gl_shader_program *prog)
-   {
-      this->num_vertices = num_vertices;
-      this->prog = prog;
-   }
-
-   virtual ~tess_eval_array_resize_visitor()
-   {
-      /* empty */
-   }
-
-   virtual ir_visitor_status visit(ir_variable *var)
-   {
-      if (!var->type->is_array() || var->data.mode != ir_var_shader_in || var->data.patch)
-         return visit_continue;
-
-      var->type = glsl_type::get_array_instance(var->type->fields.array,
-                                                this->num_vertices);
-      var->data.max_array_access = this->num_vertices - 1;
-
-      return visit_continue;
-   }
-
-   /* Dereferences of input variables need to be updated so that their type
-    * matches the newly assigned type of the variable they are accessing. */
-   virtual ir_visitor_status visit(ir_dereference_variable *ir)
-   {
-      ir->type = ir->var->type;
-      return visit_continue;
-   }
-
-   /* Dereferences of 2D input arrays need to be updated so that their type
-    * matches the newly assigned type of the array they are accessing. */
-   virtual ir_visitor_status visit_leave(ir_dereference_array *ir)
-   {
-      const glsl_type *const vt = ir->array->type;
-      if (vt->is_array())
-         ir->type = vt->fields.array;
-      return visit_continue;
-   }
-};
-
-class barrier_use_visitor : public ir_hierarchical_visitor {
-public:
-   barrier_use_visitor(gl_shader_program *prog)
-      : prog(prog), in_main(false), after_return(false), control_flow(0)
-   {
-   }
-
-   virtual ~barrier_use_visitor()
-   {
-      /* empty */
-   }
-
-   virtual ir_visitor_status visit_enter(ir_function *ir)
-   {
-      if (strcmp(ir->name, "main") == 0)
-         in_main = true;
-
-      return visit_continue;
-   }
-
-   virtual ir_visitor_status visit_leave(ir_function *)
-   {
-      in_main = false;
-      after_return = false;
-      return visit_continue;
-   }
-
-   virtual ir_visitor_status visit_leave(ir_return *)
-   {
-      after_return = true;
-      return visit_continue;
-   }
-
-   virtual ir_visitor_status visit_enter(ir_if *)
-   {
-      ++control_flow;
-      return visit_continue;
-   }
-
-   virtual ir_visitor_status visit_leave(ir_if *)
-   {
-      --control_flow;
-      return visit_continue;
-   }
-
-   virtual ir_visitor_status visit_enter(ir_loop *)
-   {
-      ++control_flow;
-      return visit_continue;
-   }
-
-   virtual ir_visitor_status visit_leave(ir_loop *)
-   {
-      --control_flow;
-      return visit_continue;
-   }
-
-   /* FINISHME: `switch` is not expressed at the IR level -- it's already
-    * been lowered to a mess of `if`s. We'll correctly disallow any use of
-    * barrier() in a conditional path within the switch, but not in a path
-    * which is always hit.
-    */
-
-   virtual ir_visitor_status visit_enter(ir_call *ir)
-   {
-      if (ir->use_builtin && strcmp(ir->callee_name(), "barrier") == 0) {
-         /* Use of barrier(); determine if it is legal: */
-         if (!in_main) {
-            linker_error(prog, "Builtin barrier() may only be used in main");
-            return visit_stop;
-         }
-
-         if (after_return) {
-            linker_error(prog, "Builtin barrier() may not be used after return");
-            return visit_stop;
-         }
-
-         if (control_flow != 0) {
-            linker_error(prog, "Builtin barrier() may not be used inside control flow");
-            return visit_stop;
-         }
-      }
-      return visit_continue;
-   }
-
-private:
-   gl_shader_program *prog;
-   bool in_main, after_return;
-   int control_flow;
-};
-
 /**
  * Visitor that determines the highest stream id to which a (geometry) shader
  * emits vertices. It also checks whether End{Stream}Primitive is ever called.
@@ -653,7 +526,7 @@ link_invalidate_variable_locations(exec_list *ir)
  */
 static void
 analyze_clip_cull_usage(struct gl_shader_program *prog,
-                        struct gl_shader *shader,
+                        struct gl_linked_shader *shader,
                         struct gl_context *ctx,
                         GLuint *clip_distance_array_size,
                         GLuint *cull_distance_array_size)
@@ -661,7 +534,7 @@ analyze_clip_cull_usage(struct gl_shader_program *prog,
    *clip_distance_array_size = 0;
    *cull_distance_array_size = 0;
 
-   if (!prog->IsES && prog->Version >= 130) {
+   if (prog->Version >= (prog->IsES ? 300 : 130)) {
       /* From section 7.1 (Vertex Shader Special Variables) of the
        * GLSL 1.30 spec:
        *
@@ -669,13 +542,12 @@ analyze_clip_cull_usage(struct gl_shader_program *prog,
        *   gl_ClipVertex and gl_ClipDistance."
        *
        * This does not apply to GLSL ES shaders, since GLSL ES defines neither
-       * gl_ClipVertex nor gl_ClipDistance.
+       * gl_ClipVertex nor gl_ClipDistance. However with
+       * GL_EXT_clip_cull_distance, this functionality is exposed in ES 3.0.
        */
-      find_assignment_visitor clip_vertex("gl_ClipVertex");
       find_assignment_visitor clip_distance("gl_ClipDistance");
       find_assignment_visitor cull_distance("gl_CullDistance");
 
-      clip_vertex.run(shader->ir);
       clip_distance.run(shader->ir);
       cull_distance.run(shader->ir);
 
@@ -685,20 +557,26 @@ analyze_clip_cull_usage(struct gl_shader_program *prog,
        * a program to statically read or write both gl_ClipVertex and either
        * gl_ClipDistance or gl_CullDistance.
        *
-       * This does not apply to GLSL ES shaders, since GLSL ES defines neither
-       * gl_ClipVertex, gl_ClipDistance or gl_CullDistance.
+       * This does not apply to GLSL ES shaders, since GLSL ES doesn't define
+       * gl_ClipVertex.
        */
-      if (clip_vertex.variable_found() && clip_distance.variable_found()) {
-         linker_error(prog, "%s shader writes to both `gl_ClipVertex' "
-                      "and `gl_ClipDistance'\n",
-                      _mesa_shader_stage_to_string(shader->Stage));
-         return;
-      }
-      if (clip_vertex.variable_found() && cull_distance.variable_found()) {
-         linker_error(prog, "%s shader writes to both `gl_ClipVertex' "
-                      "and `gl_CullDistance'\n",
-                      _mesa_shader_stage_to_string(shader->Stage));
-         return;
+      if (!prog->IsES) {
+         find_assignment_visitor clip_vertex("gl_ClipVertex");
+
+         clip_vertex.run(shader->ir);
+
+         if (clip_vertex.variable_found() && clip_distance.variable_found()) {
+            linker_error(prog, "%s shader writes to both `gl_ClipVertex' "
+                         "and `gl_ClipDistance'\n",
+                         _mesa_shader_stage_to_string(shader->Stage));
+            return;
+         }
+         if (clip_vertex.variable_found() && cull_distance.variable_found()) {
+            linker_error(prog, "%s shader writes to both `gl_ClipVertex' "
+                         "and `gl_CullDistance'\n",
+                         _mesa_shader_stage_to_string(shader->Stage));
+            return;
+         }
       }
 
       if (clip_distance.variable_found()) {
@@ -743,7 +621,7 @@ analyze_clip_cull_usage(struct gl_shader_program *prog,
  */
 void
 validate_vertex_shader_executable(struct gl_shader_program *prog,
-                                  struct gl_shader *shader,
+                                  struct gl_linked_shader *shader,
                                   struct gl_context *ctx)
 {
    if (shader == NULL)
@@ -781,8 +659,8 @@ validate_vertex_shader_executable(struct gl_shader_program *prog,
       if (!find.variable_found()) {
         if (prog->IsES) {
           linker_warning(prog,
-                         "vertex shader does not write to `gl_Position'."
-                         "It's value is undefined. \n");
+                         "vertex shader does not write to `gl_Position'. "
+                         "Its value is undefined. \n");
         } else {
           linker_error(prog,
                        "vertex shader does not write to `gl_Position'. \n");
@@ -798,7 +676,7 @@ validate_vertex_shader_executable(struct gl_shader_program *prog,
 
 void
 validate_tess_eval_shader_executable(struct gl_shader_program *prog,
-                                     struct gl_shader *shader,
+                                     struct gl_linked_shader *shader,
                                      struct gl_context *ctx)
 {
    if (shader == NULL)
@@ -817,7 +695,7 @@ validate_tess_eval_shader_executable(struct gl_shader_program *prog,
  */
 void
 validate_fragment_shader_executable(struct gl_shader_program *prog,
-                                    struct gl_shader *shader)
+                                    struct gl_linked_shader *shader)
 {
    if (shader == NULL)
       return;
@@ -844,13 +722,13 @@ validate_fragment_shader_executable(struct gl_shader_program *prog,
  */
 void
 validate_geometry_shader_executable(struct gl_shader_program *prog,
-                                    struct gl_shader *shader,
+                                    struct gl_linked_shader *shader,
                                     struct gl_context *ctx)
 {
    if (shader == NULL)
       return;
 
-   unsigned num_vertices = vertices_per_prim(prog->Geom.InputType);
+   unsigned num_vertices = vertices_per_prim(shader->info.Geom.InputType);
    prog->Geom.VerticesIn = num_vertices;
 
    analyze_clip_cull_usage(prog, shader, ctx,
@@ -866,9 +744,11 @@ static void
 validate_geometry_shader_emissions(struct gl_context *ctx,
                                    struct gl_shader_program *prog)
 {
-   if (prog->_LinkedShaders[MESA_SHADER_GEOMETRY] != NULL) {
+   struct gl_linked_shader *sh = prog->_LinkedShaders[MESA_SHADER_GEOMETRY];
+
+   if (sh != NULL) {
       find_emit_vertex_visitor emit_vertex(ctx->Const.MaxVertexStreams - 1);
-      emit_vertex.run(prog->_LinkedShaders[MESA_SHADER_GEOMETRY]->ir);
+      emit_vertex.run(sh->ir);
       if (emit_vertex.error()) {
          linker_error(prog, "Invalid call %s(%d). Accepted values for the "
                       "stream parameter are in the range [0, %d].\n",
@@ -903,7 +783,7 @@ validate_geometry_shader_emissions(struct gl_context *ctx,
        * EmitStreamVertex() or EmitEndPrimitive() are called with a non-zero
        * stream.
        */
-      if (prog->Geom.UsesStreams && prog->Geom.OutputType != GL_POINTS) {
+      if (prog->Geom.UsesStreams && sh->info.Geom.OutputType != GL_POINTS) {
          linker_error(prog, "EmitStreamVertex(n) and EndStreamPrimitive(n) "
                       "with n>0 requires point output\n");
       }
@@ -965,236 +845,232 @@ validate_intrastage_arrays(struct gl_shader_program *prog,
  */
 void
 cross_validate_globals(struct gl_shader_program *prog,
-                      struct gl_shader **shader_list,
-                      unsigned num_shaders,
-                      bool uniforms_only)
+                       struct exec_list *ir, glsl_symbol_table *variables,
+                       bool uniforms_only)
 {
-   /* Examine all of the uniforms in all of the shaders and cross validate
-    * them.
-    */
-   glsl_symbol_table variables;
-   for (unsigned i = 0; i < num_shaders; i++) {
-      if (shader_list[i] == NULL)
-        continue;
-
-      foreach_in_list(ir_instruction, node, shader_list[i]->ir) {
-        ir_variable *const var = node->as_variable();
+   foreach_in_list(ir_instruction, node, ir) {
+      ir_variable *const var = node->as_variable();
 
-        if (var == NULL)
-           continue;
+      if (var == NULL)
+         continue;
 
-        if (uniforms_only && (var->data.mode != ir_var_uniform && var->data.mode != ir_var_shader_storage))
-           continue;
+      if (uniforms_only && (var->data.mode != ir_var_uniform && var->data.mode != ir_var_shader_storage))
+         continue;
 
-         /* don't cross validate subroutine uniforms */
-         if (var->type->contains_subroutine())
-            continue;
+      /* don't cross validate subroutine uniforms */
+      if (var->type->contains_subroutine())
+         continue;
 
-        /* Don't cross validate temporaries that are at global scope.  These
-         * will eventually get pulled into the shaders 'main'.
-         */
-        if (var->data.mode == ir_var_temporary)
-           continue;
+      /* Don't cross validate temporaries that are at global scope.  These
+       * will eventually get pulled into the shaders 'main'.
+       */
+      if (var->data.mode == ir_var_temporary)
+         continue;
 
-        /* If a global with this name has already been seen, verify that the
-         * new instance has the same type.  In addition, if the globals have
-         * initializers, the values of the initializers must be the same.
-         */
-        ir_variable *const existing = variables.get_variable(var->name);
-        if (existing != NULL) {
-            /* Check if types match. Interface blocks have some special
-             * rules so we handle those elsewhere.
-             */
-           if (var->type != existing->type &&
-                !var->is_interface_instance()) {
-              if (!validate_intrastage_arrays(prog, var, existing)) {
-                  if (var->type->is_record() && existing->type->is_record()
-                      && existing->type->record_compare(var->type)) {
-                     existing->type = var->type;
-                  } else {
-                     /* If it is an unsized array in a Shader Storage Block,
-                      * two different shaders can access to different elements.
-                      * Because of that, they might be converted to different
-                      * sized arrays, then check that they are compatible but
-                      * ignore the array size.
-                      */
-                     if (!(var->data.mode == ir_var_shader_storage &&
-                           var->data.from_ssbo_unsized_array &&
-                           existing->data.mode == ir_var_shader_storage &&
-                           existing->data.from_ssbo_unsized_array &&
-                           var->type->gl_type == existing->type->gl_type)) {
-                        linker_error(prog, "%s `%s' declared as type "
-                                    "`%s' and type `%s'\n",
-                                    mode_string(var),
-                                    var->name, var->type->name,
-                                    existing->type->name);
-                        return;
-                     }
+      /* If a global with this name has already been seen, verify that the
+       * new instance has the same type.  In addition, if the globals have
+       * initializers, the values of the initializers must be the same.
+       */
+      ir_variable *const existing = variables->get_variable(var->name);
+      if (existing != NULL) {
+         /* Check if types match. Interface blocks have some special
+          * rules so we handle those elsewhere.
+          */
+         if (var->type != existing->type &&
+             !var->is_interface_instance()) {
+            if (!validate_intrastage_arrays(prog, var, existing)) {
+               if (var->type->is_record() && existing->type->is_record()
+                   && existing->type->record_compare(var->type)) {
+                   existing->type = var->type;
+               } else {
+                  /* If it is an unsized array in a Shader Storage Block,
+                   * two different shaders can access to different elements.
+                   * Because of that, they might be converted to different
+                   * sized arrays, then check that they are compatible but
+                   * ignore the array size.
+                   */
+                  if (!(var->data.mode == ir_var_shader_storage &&
+                        var->data.from_ssbo_unsized_array &&
+                        existing->data.mode == ir_var_shader_storage &&
+                        existing->data.from_ssbo_unsized_array &&
+                        var->type->gl_type == existing->type->gl_type)) {
+                     linker_error(prog, "%s `%s' declared as type "
+                                  "`%s' and type `%s'\n",
+                                  mode_string(var),
+                                  var->name, var->type->name,
+                                  existing->type->name);
+                     return;
                   }
-              }
-           }
-
-           if (var->data.explicit_location) {
-              if (existing->data.explicit_location
-                  && (var->data.location != existing->data.location)) {
-                    linker_error(prog, "explicit locations for %s "
-                                 "`%s' have differing values\n",
-                                 mode_string(var), var->name);
-                    return;
-              }
+               }
+            }
+         }
 
-              if (var->data.location_frac != existing->data.location_frac) {
-                    linker_error(prog, "explicit components for %s "
-                                 "`%s' have differing values\n",
-                                 mode_string(var), var->name);
-                    return;
-              }
+         if (var->data.explicit_location) {
+            if (existing->data.explicit_location
+                && (var->data.location != existing->data.location)) {
+               linker_error(prog, "explicit locations for %s "
+                            "`%s' have differing values\n",
+                            mode_string(var), var->name);
+               return;
+            }
 
-              existing->data.location = var->data.location;
-              existing->data.explicit_location = true;
-           } else {
-               /* Check if uniform with implicit location was marked explicit
-                * by earlier shader stage. If so, mark it explicit in this stage
-                * too to make sure later processing does not treat it as
-                * implicit one.
-                */
-               if (existing->data.explicit_location) {
-                 var->data.location = existing->data.location;
-                 var->data.explicit_location = true;
-               }
+            if (var->data.location_frac != existing->data.location_frac) {
+               linker_error(prog, "explicit components for %s `%s' have "
+                            "differing values\n", mode_string(var), var->name);
+               return;
             }
 
-            /* From the GLSL 4.20 specification:
-             * "A link error will result if two compilation units in a program
-             *  specify different integer-constant bindings for the same
-             *  opaque-uniform name.  However, it is not an error to specify a
-             *  binding on some but not all declarations for the same name"
+            existing->data.location = var->data.location;
+            existing->data.explicit_location = true;
+         } else {
+            /* Check if uniform with implicit location was marked explicit
+             * by earlier shader stage. If so, mark it explicit in this stage
+             * too to make sure later processing does not treat it as
+             * implicit one.
              */
-            if (var->data.explicit_binding) {
-               if (existing->data.explicit_binding &&
-                   var->data.binding != existing->data.binding) {
-                  linker_error(prog, "explicit bindings for %s "
-                               "`%s' have differing values\n",
-                               mode_string(var), var->name);
-                  return;
-               }
-
-               existing->data.binding = var->data.binding;
-               existing->data.explicit_binding = true;
+            if (existing->data.explicit_location) {
+               var->data.location = existing->data.location;
+               var->data.explicit_location = true;
             }
+         }
 
-            if (var->type->contains_atomic() &&
-                var->data.offset != existing->data.offset) {
-               linker_error(prog, "offset specifications for %s "
+         /* From the GLSL 4.20 specification:
+          * "A link error will result if two compilation units in a program
+          *  specify different integer-constant bindings for the same
+          *  opaque-uniform name.  However, it is not an error to specify a
+          *  binding on some but not all declarations for the same name"
+          */
+         if (var->data.explicit_binding) {
+            if (existing->data.explicit_binding &&
+                var->data.binding != existing->data.binding) {
+               linker_error(prog, "explicit bindings for %s "
                             "`%s' have differing values\n",
                             mode_string(var), var->name);
                return;
             }
 
-           /* Validate layout qualifiers for gl_FragDepth.
-            *
-            * From the AMD/ARB_conservative_depth specs:
-            *
-            *    "If gl_FragDepth is redeclared in any fragment shader in a
-            *    program, it must be redeclared in all fragment shaders in
-            *    that program that have static assignments to
-            *    gl_FragDepth. All redeclarations of gl_FragDepth in all
-            *    fragment shaders in a single program must have the same set
-            *    of qualifiers."
-            */
-           if (strcmp(var->name, "gl_FragDepth") == 0) {
-              bool layout_declared = var->data.depth_layout != ir_depth_layout_none;
-              bool layout_differs =
-                 var->data.depth_layout != existing->data.depth_layout;
-
-              if (layout_declared && layout_differs) {
-                 linker_error(prog,
-                              "All redeclarations of gl_FragDepth in all "
-                              "fragment shaders in a single program must have "
-                              "the same set of qualifiers.\n");
-              }
+            existing->data.binding = var->data.binding;
+            existing->data.explicit_binding = true;
+         }
 
-              if (var->data.used && layout_differs) {
-                 linker_error(prog,
-                              "If gl_FragDepth is redeclared with a layout "
-                              "qualifier in any fragment shader, it must be "
-                              "redeclared with the same layout qualifier in "
-                              "all fragment shaders that have assignments to "
-                              "gl_FragDepth\n");
-              }
-           }
+         if (var->type->contains_atomic() &&
+             var->data.offset != existing->data.offset) {
+            linker_error(prog, "offset specifications for %s "
+                         "`%s' have differing values\n",
+                         mode_string(var), var->name);
+            return;
+         }
 
-           /* Page 35 (page 41 of the PDF) of the GLSL 4.20 spec says:
-            *
-            *     "If a shared global has multiple initializers, the
-            *     initializers must all be constant expressions, and they
-            *     must all have the same value. Otherwise, a link error will
-            *     result. (A shared global having only one initializer does
-            *     not require that initializer to be a constant expression.)"
-            *
-            * Previous to 4.20 the GLSL spec simply said that initializers
-            * must have the same value.  In this case of non-constant
-            * initializers, this was impossible to determine.  As a result,
-            * no vendor actually implemented that behavior.  The 4.20
-            * behavior matches the implemented behavior of at least one other
-            * vendor, so we'll implement that for all GLSL versions.
-            */
-           if (var->constant_initializer != NULL) {
-              if (existing->constant_initializer != NULL) {
-                 if (!var->constant_initializer->has_value(existing->constant_initializer)) {
-                    linker_error(prog, "initializers for %s "
-                                 "`%s' have differing values\n",
-                                 mode_string(var), var->name);
-                    return;
-                 }
-              } else {
-                  /* If the first-seen instance of a particular uniform did
-                   * not have an initializer but a later instance does,
-                   * replace the former with the later.
-                   */
-                  variables.replace_variable(existing->name, var);
-              }
-           }
+         /* Validate layout qualifiers for gl_FragDepth.
+          *
+          * From the AMD/ARB_conservative_depth specs:
+          *
+          *    "If gl_FragDepth is redeclared in any fragment shader in a
+          *    program, it must be redeclared in all fragment shaders in
+          *    that program that have static assignments to
+          *    gl_FragDepth. All redeclarations of gl_FragDepth in all
+          *    fragment shaders in a single program must have the same set
+          *    of qualifiers."
+          */
+         if (strcmp(var->name, "gl_FragDepth") == 0) {
+            bool layout_declared = var->data.depth_layout != ir_depth_layout_none;
+            bool layout_differs =
+               var->data.depth_layout != existing->data.depth_layout;
 
-           if (var->data.has_initializer) {
-              if (existing->data.has_initializer
-                  && (var->constant_initializer == NULL
-                      || existing->constant_initializer == NULL)) {
-                 linker_error(prog,
-                              "shared global variable `%s' has multiple "
-                              "non-constant initializers.\n",
-                              var->name);
-                 return;
-              }
-           }
+            if (layout_declared && layout_differs) {
+               linker_error(prog,
+                            "All redeclarations of gl_FragDepth in all "
+                            "fragment shaders in a single program must have "
+                            "the same set of qualifiers.\n");
+            }
 
-           if (existing->data.invariant != var->data.invariant) {
-              linker_error(prog, "declarations for %s `%s' have "
-                           "mismatching invariant qualifiers\n",
-                           mode_string(var), var->name);
-              return;
-           }
-            if (existing->data.centroid != var->data.centroid) {
-               linker_error(prog, "declarations for %s `%s' have "
-                           "mismatching centroid qualifiers\n",
-                           mode_string(var), var->name);
-               return;
+            if (var->data.used && layout_differs) {
+               linker_error(prog,
+                            "If gl_FragDepth is redeclared with a layout "
+                            "qualifier in any fragment shader, it must be "
+                            "redeclared with the same layout qualifier in "
+                            "all fragment shaders that have assignments to "
+                            "gl_FragDepth\n");
             }
-            if (existing->data.sample != var->data.sample) {
-               linker_error(prog, "declarations for %s `%s` have "
-                            "mismatching sample qualifiers\n",
-                            mode_string(var), var->name);
-               return;
+         }
+
+         /* Page 35 (page 41 of the PDF) of the GLSL 4.20 spec says:
+          *
+          *     "If a shared global has multiple initializers, the
+          *     initializers must all be constant expressions, and they
+          *     must all have the same value. Otherwise, a link error will
+          *     result. (A shared global having only one initializer does
+          *     not require that initializer to be a constant expression.)"
+          *
+          * Previous to 4.20 the GLSL spec simply said that initializers
+          * must have the same value.  In this case of non-constant
+          * initializers, this was impossible to determine.  As a result,
+          * no vendor actually implemented that behavior.  The 4.20
+          * behavior matches the implemented behavior of at least one other
+          * vendor, so we'll implement that for all GLSL versions.
+          */
+         if (var->constant_initializer != NULL) {
+            if (existing->constant_initializer != NULL) {
+               if (!var->constant_initializer->has_value(existing->constant_initializer)) {
+                  linker_error(prog, "initializers for %s "
+                               "`%s' have differing values\n",
+                               mode_string(var), var->name);
+                  return;
+               }
+            } else {
+               /* If the first-seen instance of a particular uniform did
+                * not have an initializer but a later instance does,
+                * replace the former with the later.
+                */
+               variables->replace_variable(existing->name, var);
             }
-            if (existing->data.image_format != var->data.image_format) {
-               linker_error(prog, "declarations for %s `%s` have "
-                            "mismatching image format qualifiers\n",
-                            mode_string(var), var->name);
+         }
+
+         if (var->data.has_initializer) {
+            if (existing->data.has_initializer
+                && (var->constant_initializer == NULL
+                    || existing->constant_initializer == NULL)) {
+               linker_error(prog,
+                            "shared global variable `%s' has multiple "
+                            "non-constant initializers.\n",
+                            var->name);
                return;
             }
-        } else
-           variables.add_variable(var);
-      }
+         }
+
+         if (existing->data.invariant != var->data.invariant) {
+            linker_error(prog, "declarations for %s `%s' have "
+                         "mismatching invariant qualifiers\n",
+                         mode_string(var), var->name);
+            return;
+         }
+         if (existing->data.centroid != var->data.centroid) {
+            linker_error(prog, "declarations for %s `%s' have "
+                         "mismatching centroid qualifiers\n",
+                         mode_string(var), var->name);
+            return;
+         }
+         if (existing->data.sample != var->data.sample) {
+            linker_error(prog, "declarations for %s `%s` have "
+                         "mismatching sample qualifiers\n",
+                         mode_string(var), var->name);
+            return;
+         }
+         if (existing->data.image_format != var->data.image_format) {
+            linker_error(prog, "declarations for %s `%s` have "
+                         "mismatching image format qualifiers\n",
+                         mode_string(var), var->name);
+            return;
+         }
+
+         if (prog->IsES && existing->data.precision != var->data.precision) {
+            linker_error(prog, "declarations for %s `%s` have "
+                         "mismatching precision qualifiers\n",
+                         mode_string(var), var->name);
+            return;
+         }
+      } else
+         variables->add_variable(var);
    }
 }
 
@@ -1205,8 +1081,14 @@ cross_validate_globals(struct gl_shader_program *prog,
 void
 cross_validate_uniforms(struct gl_shader_program *prog)
 {
-   cross_validate_globals(prog, prog->_LinkedShaders,
-                          MESA_SHADER_STAGES, true);
+   glsl_symbol_table variables;
+   for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
+      if (prog->_LinkedShaders[i] == NULL)
+        continue;
+
+      cross_validate_globals(prog, prog->_LinkedShaders[i]->ir, &variables,
+                             true);
+   }
 }
 
 /**
@@ -1236,7 +1118,7 @@ interstage_cross_validate_uniform_blocks(struct gl_shader_program *prog,
    }
 
    for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
-      struct gl_shader *sh = prog->_LinkedShaders[i];
+      struct gl_linked_shader *sh = prog->_LinkedShaders[i];
 
       InterfaceBlockStageIndex[i] = new int[max_num_buffer_blocks];
       for (unsigned int j = 0; j < max_num_buffer_blocks; j++)
@@ -1281,7 +1163,7 @@ interstage_cross_validate_uniform_blocks(struct gl_shader_program *prog,
          int stage_index = InterfaceBlockStageIndex[i][j];
 
         if (stage_index != -1) {
-           struct gl_shader *sh = prog->_LinkedShaders[i];
+           struct gl_linked_shader *sh = prog->_LinkedShaders[i];
 
             blks[j].stageref |= (1 << i);
 
@@ -1310,7 +1192,7 @@ interstage_cross_validate_uniform_blocks(struct gl_shader_program *prog,
  * Populates a shaders symbol table with all global declarations
  */
 static void
-populate_symbol_table(gl_shader *sh)
+populate_symbol_table(gl_linked_shader *sh)
 {
    sh->symbols = new(sh) glsl_symbol_table;
 
@@ -1348,12 +1230,12 @@ populate_symbol_table(gl_shader *sh)
  *                     should be added.
  */
 void
-remap_variables(ir_instruction *inst, struct gl_shader *target,
+remap_variables(ir_instruction *inst, struct gl_linked_shader *target,
                hash_table *temps)
 {
    class remap_visitor : public ir_hierarchical_visitor {
    public:
-        remap_visitor(struct gl_shader *target,
+        remap_visitor(struct gl_linked_shader *target,
                    hash_table *temps)
       {
         this->target = target;
@@ -1365,7 +1247,8 @@ remap_variables(ir_instruction *inst, struct gl_shader *target,
       virtual ir_visitor_status visit(ir_dereference_variable *ir)
       {
         if (ir->var->data.mode == ir_var_temporary) {
-           ir_variable *var = (ir_variable *) hash_table_find(temps, ir->var);
+           hash_entry *entry = _mesa_hash_table_search(temps, ir->var);
+           ir_variable *var = entry ? (ir_variable *) entry->data : NULL;
 
            assert(var != NULL);
            ir->var = var;
@@ -1388,7 +1271,7 @@ remap_variables(ir_instruction *inst, struct gl_shader *target,
       }
 
    private:
-      struct gl_shader *target;
+      struct gl_linked_shader *target;
       glsl_symbol_table *symbols;
       exec_list *instructions;
       hash_table *temps;
@@ -1423,13 +1306,13 @@ remap_variables(ir_instruction *inst, struct gl_shader *target,
  */
 exec_node *
 move_non_declarations(exec_list *instructions, exec_node *last,
-                     bool make_copies, gl_shader *target)
+                     bool make_copies, gl_linked_shader *target)
 {
    hash_table *temps = NULL;
 
    if (make_copies)
-      temps = hash_table_ctor(0, hash_table_pointer_hash,
-                             hash_table_pointer_compare);
+      temps = _mesa_hash_table_create(NULL, _mesa_hash_pointer,
+                                      _mesa_key_pointer_equal);
 
    foreach_in_list_safe(ir_instruction, inst, instructions) {
       if (inst->as_function())
@@ -1448,7 +1331,7 @@ move_non_declarations(exec_list *instructions, exec_node *last,
         inst = inst->clone(target, NULL);
 
         if (var != NULL)
-           hash_table_insert(temps, inst, var);
+           _mesa_hash_table_insert(temps, var, inst);
         else
            remap_variables(inst, target, temps);
       } else {
@@ -1460,7 +1343,7 @@ move_non_declarations(exec_list *instructions, exec_node *last,
    }
 
    if (make_copies)
-      hash_table_dtor(temps);
+      _mesa_hash_table_destroy(temps, NULL);
 
    return last;
 }
@@ -1475,22 +1358,25 @@ class array_sizing_visitor : public ir_hierarchical_visitor {
 public:
    array_sizing_visitor()
       : mem_ctx(ralloc_context(NULL)),
-        unnamed_interfaces(hash_table_ctor(0, hash_table_pointer_hash,
-                                           hash_table_pointer_compare))
+        unnamed_interfaces(_mesa_hash_table_create(NULL, _mesa_hash_pointer,
+                                                   _mesa_key_pointer_equal))
    {
    }
 
    ~array_sizing_visitor()
    {
-      hash_table_dtor(this->unnamed_interfaces);
+      _mesa_hash_table_destroy(this->unnamed_interfaces, NULL);
       ralloc_free(this->mem_ctx);
    }
 
    virtual ir_visitor_status visit(ir_variable *var)
    {
       const glsl_type *type_without_array;
+      bool implicit_sized_array = var->data.implicit_sized_array;
       fixup_type(&var->type, var->data.max_array_access,
-                 var->data.from_ssbo_unsized_array);
+                 var->data.from_ssbo_unsized_array,
+                 &implicit_sized_array);
+      var->data.implicit_sized_array = implicit_sized_array;
       type_without_array = var->type->without_array();
       if (var->type->is_interface()) {
          if (interface_contains_unsized_arrays(var->type)) {
@@ -1514,13 +1400,17 @@ public:
          /* Store a pointer to the variable in the unnamed_interfaces
           * hashtable.
           */
-         ir_variable **interface_vars = (ir_variable **)
-            hash_table_find(this->unnamed_interfaces, ifc_type);
+         hash_entry *entry =
+               _mesa_hash_table_search(this->unnamed_interfaces,
+                                       ifc_type);
+
+         ir_variable **interface_vars = entry ? (ir_variable **) entry->data : NULL;
+
          if (interface_vars == NULL) {
             interface_vars = rzalloc_array(mem_ctx, ir_variable *,
                                            ifc_type->length);
-            hash_table_insert(this->unnamed_interfaces, interface_vars,
-                              ifc_type);
+            _mesa_hash_table_insert(this->unnamed_interfaces, ifc_type,
+                                    interface_vars);
          }
          unsigned index = ifc_type->field_index(var->name);
          assert(index < ifc_type->length);
@@ -1548,11 +1438,12 @@ private:
     * it with a sized array whose size is determined by max_array_access.
     */
    static void fixup_type(const glsl_type **type, unsigned max_array_access,
-                          bool from_ssbo_unsized_array)
+                          bool from_ssbo_unsized_array, bool *implicit_sized)
    {
       if (!from_ssbo_unsized_array && (*type)->is_unsized_array()) {
          *type = glsl_type::get_array_instance((*type)->fields.array,
                                                max_array_access + 1);
+         *implicit_sized = true;
          assert(*type != NULL);
       }
    }
@@ -1601,15 +1492,17 @@ private:
       memcpy(fields, type->fields.structure,
              num_fields * sizeof(*fields));
       for (unsigned i = 0; i < num_fields; i++) {
+         bool implicit_sized_array = fields[i].implicit_sized_array;
          /* If SSBO last member is unsized array, we don't replace it by a sized
           * array.
           */
          if (is_ssbo && i == (num_fields - 1))
             fixup_type(&fields[i].type, max_ifc_array_access[i],
-                       true);
+                       true, &implicit_sized_array);
          else
             fixup_type(&fields[i].type, max_ifc_array_access[i],
-                       false);
+                       false, &implicit_sized_array);
+         fields[i].implicit_sized_array = implicit_sized_array;
       }
       glsl_interface_packing packing =
          (glsl_interface_packing) type->interface_packing;
@@ -1672,42 +1565,43 @@ private:
 static void
 link_xfb_stride_layout_qualifiers(struct gl_context *ctx,
                                   struct gl_shader_program *prog,
-                                 struct gl_shader *linked_shader,
+                                 struct gl_linked_shader *linked_shader,
                                  struct gl_shader **shader_list,
                                  unsigned num_shaders)
 {
    for (unsigned i = 0; i < MAX_FEEDBACK_BUFFERS; i++) {
-      linked_shader->TransformFeedback.BufferStride[i] = 0;
+      linked_shader->info.TransformFeedback.BufferStride[i] = 0;
    }
 
    for (unsigned i = 0; i < num_shaders; i++) {
       struct gl_shader *shader = shader_list[i];
 
       for (unsigned j = 0; j < MAX_FEEDBACK_BUFFERS; j++) {
-         if (shader->TransformFeedback.BufferStride[j]) {
-           if (linked_shader->TransformFeedback.BufferStride[j] != 0 &&
-                shader->TransformFeedback.BufferStride[j] != 0 &&
-               linked_shader->TransformFeedback.BufferStride[j] !=
-                   shader->TransformFeedback.BufferStride[j]) {
+         if (shader->info.TransformFeedback.BufferStride[j]) {
+           if (linked_shader->info.TransformFeedback.BufferStride[j] != 0 &&
+                shader->info.TransformFeedback.BufferStride[j] != 0 &&
+               linked_shader->info.TransformFeedback.BufferStride[j] !=
+                   shader->info.TransformFeedback.BufferStride[j]) {
               linker_error(prog,
                             "intrastage shaders defined with conflicting "
                             "xfb_stride for buffer %d (%d and %d)\n", j,
-                            linked_shader->TransformFeedback.BufferStride[j],
-                           shader->TransformFeedback.BufferStride[j]);
+                            linked_shader->
+                               info.TransformFeedback.BufferStride[j],
+                           shader->info.TransformFeedback.BufferStride[j]);
               return;
            }
 
-            if (shader->TransformFeedback.BufferStride[j])
-              linked_shader->TransformFeedback.BufferStride[j] =
-                  shader->TransformFeedback.BufferStride[j];
+            if (shader->info.TransformFeedback.BufferStride[j])
+              linked_shader->info.TransformFeedback.BufferStride[j] =
+                  shader->info.TransformFeedback.BufferStride[j];
          }
       }
    }
 
    for (unsigned j = 0; j < MAX_FEEDBACK_BUFFERS; j++) {
-      if (linked_shader->TransformFeedback.BufferStride[j]) {
+      if (linked_shader->info.TransformFeedback.BufferStride[j]) {
          prog->TransformFeedback.BufferStride[j] =
-            linked_shader->TransformFeedback.BufferStride[j];
+            linked_shader->info.TransformFeedback.BufferStride[j];
 
          /* We will validate doubles at a later stage */
          if (prog->TransformFeedback.BufferStride[j] % 4) {
@@ -1736,11 +1630,11 @@ link_xfb_stride_layout_qualifiers(struct gl_context *ctx,
  */
 static void
 link_tcs_out_layout_qualifiers(struct gl_shader_program *prog,
-                             struct gl_shader *linked_shader,
+                             struct gl_linked_shader *linked_shader,
                              struct gl_shader **shader_list,
                              unsigned num_shaders)
 {
-   linked_shader->TessCtrl.VerticesOut = 0;
+   linked_shader->info.TessCtrl.VerticesOut = 0;
 
    if (linked_shader->Stage != MESA_SHADER_TESS_CTRL)
       return;
@@ -1758,16 +1652,18 @@ link_tcs_out_layout_qualifiers(struct gl_shader_program *prog,
    for (unsigned i = 0; i < num_shaders; i++) {
       struct gl_shader *shader = shader_list[i];
 
-      if (shader->TessCtrl.VerticesOut != 0) {
-        if (linked_shader->TessCtrl.VerticesOut != 0 &&
-            linked_shader->TessCtrl.VerticesOut != shader->TessCtrl.VerticesOut) {
+      if (shader->info.TessCtrl.VerticesOut != 0) {
+        if (linked_shader->info.TessCtrl.VerticesOut != 0 &&
+            linked_shader->info.TessCtrl.VerticesOut !=
+             shader->info.TessCtrl.VerticesOut) {
            linker_error(prog, "tessellation control shader defined with "
                         "conflicting output vertex count (%d and %d)\n",
-                        linked_shader->TessCtrl.VerticesOut,
-                        shader->TessCtrl.VerticesOut);
+                        linked_shader->info.TessCtrl.VerticesOut,
+                        shader->info.TessCtrl.VerticesOut);
            return;
         }
-        linked_shader->TessCtrl.VerticesOut = shader->TessCtrl.VerticesOut;
+        linked_shader->info.TessCtrl.VerticesOut =
+            shader->info.TessCtrl.VerticesOut;
       }
    }
 
@@ -1775,12 +1671,11 @@ link_tcs_out_layout_qualifiers(struct gl_shader_program *prog,
     * since we already know we're in the right type of shader program
     * for doing it.
     */
-   if (linked_shader->TessCtrl.VerticesOut == 0) {
+   if (linked_shader->info.TessCtrl.VerticesOut == 0) {
       linker_error(prog, "tessellation control shader didn't declare "
                   "vertices out layout qualifier\n");
       return;
    }
-   prog->TessCtrl.VerticesOut = linked_shader->TessCtrl.VerticesOut;
 }
 
 
@@ -1792,14 +1687,14 @@ link_tcs_out_layout_qualifiers(struct gl_shader_program *prog,
  */
 static void
 link_tes_in_layout_qualifiers(struct gl_shader_program *prog,
-                               struct gl_shader *linked_shader,
+                               struct gl_linked_shader *linked_shader,
                                struct gl_shader **shader_list,
                                unsigned num_shaders)
 {
-   linked_shader->TessEval.PrimitiveMode = PRIM_UNKNOWN;
-   linked_shader->TessEval.Spacing = 0;
-   linked_shader->TessEval.VertexOrder = 0;
-   linked_shader->TessEval.PointMode = -1;
+   linked_shader->info.TessEval.PrimitiveMode = PRIM_UNKNOWN;
+   linked_shader->info.TessEval.Spacing = 0;
+   linked_shader->info.TessEval.VertexOrder = 0;
+   linked_shader->info.TessEval.PointMode = -1;
 
    if (linked_shader->Stage != MESA_SHADER_TESS_EVAL)
       return;
@@ -1821,44 +1716,50 @@ link_tes_in_layout_qualifiers(struct gl_shader_program *prog,
    for (unsigned i = 0; i < num_shaders; i++) {
       struct gl_shader *shader = shader_list[i];
 
-      if (shader->TessEval.PrimitiveMode != PRIM_UNKNOWN) {
-        if (linked_shader->TessEval.PrimitiveMode != PRIM_UNKNOWN &&
-            linked_shader->TessEval.PrimitiveMode != shader->TessEval.PrimitiveMode) {
+      if (shader->info.TessEval.PrimitiveMode != PRIM_UNKNOWN) {
+        if (linked_shader->info.TessEval.PrimitiveMode != PRIM_UNKNOWN &&
+            linked_shader->info.TessEval.PrimitiveMode !=
+             shader->info.TessEval.PrimitiveMode) {
            linker_error(prog, "tessellation evaluation shader defined with "
                         "conflicting input primitive modes.\n");
            return;
         }
-        linked_shader->TessEval.PrimitiveMode = shader->TessEval.PrimitiveMode;
+        linked_shader->info.TessEval.PrimitiveMode = shader->info.TessEval.PrimitiveMode;
       }
 
-      if (shader->TessEval.Spacing != 0) {
-        if (linked_shader->TessEval.Spacing != 0 &&
-            linked_shader->TessEval.Spacing != shader->TessEval.Spacing) {
+      if (shader->info.TessEval.Spacing != 0) {
+        if (linked_shader->info.TessEval.Spacing != 0 &&
+            linked_shader->info.TessEval.Spacing !=
+             shader->info.TessEval.Spacing) {
            linker_error(prog, "tessellation evaluation shader defined with "
                         "conflicting vertex spacing.\n");
            return;
         }
-        linked_shader->TessEval.Spacing = shader->TessEval.Spacing;
+        linked_shader->info.TessEval.Spacing = shader->info.TessEval.Spacing;
       }
 
-      if (shader->TessEval.VertexOrder != 0) {
-        if (linked_shader->TessEval.VertexOrder != 0 &&
-            linked_shader->TessEval.VertexOrder != shader->TessEval.VertexOrder) {
+      if (shader->info.TessEval.VertexOrder != 0) {
+        if (linked_shader->info.TessEval.VertexOrder != 0 &&
+            linked_shader->info.TessEval.VertexOrder !=
+             shader->info.TessEval.VertexOrder) {
            linker_error(prog, "tessellation evaluation shader defined with "
                         "conflicting ordering.\n");
            return;
         }
-        linked_shader->TessEval.VertexOrder = shader->TessEval.VertexOrder;
+        linked_shader->info.TessEval.VertexOrder =
+            shader->info.TessEval.VertexOrder;
       }
 
-      if (shader->TessEval.PointMode != -1) {
-        if (linked_shader->TessEval.PointMode != -1 &&
-            linked_shader->TessEval.PointMode != shader->TessEval.PointMode) {
+      if (shader->info.TessEval.PointMode != -1) {
+        if (linked_shader->info.TessEval.PointMode != -1 &&
+            linked_shader->info.TessEval.PointMode !=
+             shader->info.TessEval.PointMode) {
            linker_error(prog, "tessellation evaluation shader defined with "
                         "conflicting point modes.\n");
            return;
         }
-        linked_shader->TessEval.PointMode = shader->TessEval.PointMode;
+        linked_shader->info.TessEval.PointMode =
+            shader->info.TessEval.PointMode;
       }
 
    }
@@ -1867,25 +1768,21 @@ link_tes_in_layout_qualifiers(struct gl_shader_program *prog,
     * since we already know we're in the right type of shader program
     * for doing it.
     */
-   if (linked_shader->TessEval.PrimitiveMode == PRIM_UNKNOWN) {
+   if (linked_shader->info.TessEval.PrimitiveMode == PRIM_UNKNOWN) {
       linker_error(prog,
                   "tessellation evaluation shader didn't declare input "
                   "primitive modes.\n");
       return;
    }
-   prog->TessEval.PrimitiveMode = linked_shader->TessEval.PrimitiveMode;
 
-   if (linked_shader->TessEval.Spacing == 0)
-      linked_shader->TessEval.Spacing = GL_EQUAL;
-   prog->TessEval.Spacing = linked_shader->TessEval.Spacing;
+   if (linked_shader->info.TessEval.Spacing == 0)
+      linked_shader->info.TessEval.Spacing = GL_EQUAL;
 
-   if (linked_shader->TessEval.VertexOrder == 0)
-      linked_shader->TessEval.VertexOrder = GL_CCW;
-   prog->TessEval.VertexOrder = linked_shader->TessEval.VertexOrder;
+   if (linked_shader->info.TessEval.VertexOrder == 0)
+      linked_shader->info.TessEval.VertexOrder = GL_CCW;
 
-   if (linked_shader->TessEval.PointMode == -1)
-      linked_shader->TessEval.PointMode = GL_FALSE;
-   prog->TessEval.PointMode = linked_shader->TessEval.PointMode;
+   if (linked_shader->info.TessEval.PointMode == -1)
+      linked_shader->info.TessEval.PointMode = GL_FALSE;
 }
 
 
@@ -1895,15 +1792,16 @@ link_tes_in_layout_qualifiers(struct gl_shader_program *prog,
  * and propagates them to the linked FS and linked shader program.
  */
 static void
-link_fs_input_layout_qualifiers(struct gl_shader_program *prog,
-                               struct gl_shader *linked_shader,
+link_fs_inout_layout_qualifiers(struct gl_shader_program *prog,
+                               struct gl_linked_shader *linked_shader,
                                struct gl_shader **shader_list,
                                unsigned num_shaders)
 {
-   linked_shader->redeclares_gl_fragcoord = false;
-   linked_shader->uses_gl_fragcoord = false;
-   linked_shader->origin_upper_left = false;
-   linked_shader->pixel_center_integer = false;
+   linked_shader->info.redeclares_gl_fragcoord = false;
+   linked_shader->info.uses_gl_fragcoord = false;
+   linked_shader->info.origin_upper_left = false;
+   linked_shader->info.pixel_center_integer = false;
+   linked_shader->info.BlendSupport = 0;
 
    if (linked_shader->Stage != MESA_SHADER_FRAGMENT ||
        (prog->Version < 150 && !prog->ARB_fragment_coord_conventions_enable))
@@ -1917,12 +1815,12 @@ link_fs_input_layout_qualifiers(struct gl_shader_program *prog,
        *    it must be redeclared in all the fragment shaders in that program
        *    that have a static use gl_FragCoord."
        */
-      if ((linked_shader->redeclares_gl_fragcoord
-           && !shader->redeclares_gl_fragcoord
-           && shader->uses_gl_fragcoord)
-          || (shader->redeclares_gl_fragcoord
-              && !linked_shader->redeclares_gl_fragcoord
-              && linked_shader->uses_gl_fragcoord)) {
+      if ((linked_shader->info.redeclares_gl_fragcoord
+           && !shader->info.redeclares_gl_fragcoord
+           && shader->info.uses_gl_fragcoord)
+          || (shader->info.redeclares_gl_fragcoord
+              && !linked_shader->info.redeclares_gl_fragcoord
+              && linked_shader->info.uses_gl_fragcoord)) {
              linker_error(prog, "fragment shader defined with conflicting "
                          "layout qualifiers for gl_FragCoord\n");
       }
@@ -1932,9 +1830,12 @@ link_fs_input_layout_qualifiers(struct gl_shader_program *prog,
        *   "All redeclarations of gl_FragCoord in all fragment shaders in a
        *    single program must have the same set of qualifiers."
        */
-      if (linked_shader->redeclares_gl_fragcoord && shader->redeclares_gl_fragcoord
-          && (shader->origin_upper_left != linked_shader->origin_upper_left
-          || shader->pixel_center_integer != linked_shader->pixel_center_integer)) {
+      if (linked_shader->info.redeclares_gl_fragcoord &&
+          shader->info.redeclares_gl_fragcoord &&
+          (shader->info.origin_upper_left !=
+           linked_shader->info.origin_upper_left ||
+           shader->info.pixel_center_integer !=
+           linked_shader->info.pixel_center_integer)) {
          linker_error(prog, "fragment shader defined with conflicting "
                       "layout qualifiers for gl_FragCoord\n");
       }
@@ -1944,16 +1845,22 @@ link_fs_input_layout_qualifiers(struct gl_shader_program *prog,
        * are multiple redeclarations, all the fields except uses_gl_fragcoord
        * are already known to be the same.
        */
-      if (shader->redeclares_gl_fragcoord || shader->uses_gl_fragcoord) {
-         linked_shader->redeclares_gl_fragcoord =
-            shader->redeclares_gl_fragcoord;
-         linked_shader->uses_gl_fragcoord = linked_shader->uses_gl_fragcoord
-            || shader->uses_gl_fragcoord;
-         linked_shader->origin_upper_left = shader->origin_upper_left;
-         linked_shader->pixel_center_integer = shader->pixel_center_integer;
+      if (shader->info.redeclares_gl_fragcoord ||
+          shader->info.uses_gl_fragcoord) {
+         linked_shader->info.redeclares_gl_fragcoord =
+            shader->info.redeclares_gl_fragcoord;
+         linked_shader->info.uses_gl_fragcoord =
+            linked_shader->info.uses_gl_fragcoord ||
+            shader->info.uses_gl_fragcoord;
+         linked_shader->info.origin_upper_left =
+            shader->info.origin_upper_left;
+         linked_shader->info.pixel_center_integer =
+            shader->info.pixel_center_integer;
       }
 
-      linked_shader->EarlyFragmentTests |= shader->EarlyFragmentTests;
+      linked_shader->info.EarlyFragmentTests |=
+         shader->info.EarlyFragmentTests;
+      linked_shader->info.BlendSupport |= shader->info.BlendSupport;
    }
 }
 
@@ -1964,14 +1871,14 @@ link_fs_input_layout_qualifiers(struct gl_shader_program *prog,
  */
 static void
 link_gs_inout_layout_qualifiers(struct gl_shader_program *prog,
-                               struct gl_shader *linked_shader,
+                               struct gl_linked_shader *linked_shader,
                                struct gl_shader **shader_list,
                                unsigned num_shaders)
 {
-   linked_shader->Geom.VerticesOut = 0;
-   linked_shader->Geom.Invocations = 0;
-   linked_shader->Geom.InputType = PRIM_UNKNOWN;
-   linked_shader->Geom.OutputType = PRIM_UNKNOWN;
+   linked_shader->info.Geom.VerticesOut = -1;
+   linked_shader->info.Geom.Invocations = 0;
+   linked_shader->info.Geom.InputType = PRIM_UNKNOWN;
+   linked_shader->info.Geom.OutputType = PRIM_UNKNOWN;
 
    /* No in/out qualifiers defined for anything but GLSL 1.50+
     * geometry shaders so far.
@@ -1992,48 +1899,52 @@ link_gs_inout_layout_qualifiers(struct gl_shader_program *prog,
    for (unsigned i = 0; i < num_shaders; i++) {
       struct gl_shader *shader = shader_list[i];
 
-      if (shader->Geom.InputType != PRIM_UNKNOWN) {
-        if (linked_shader->Geom.InputType != PRIM_UNKNOWN &&
-            linked_shader->Geom.InputType != shader->Geom.InputType) {
+      if (shader->info.Geom.InputType != PRIM_UNKNOWN) {
+        if (linked_shader->info.Geom.InputType != PRIM_UNKNOWN &&
+            linked_shader->info.Geom.InputType !=
+             shader->info.Geom.InputType) {
            linker_error(prog, "geometry shader defined with conflicting "
                         "input types\n");
            return;
         }
-        linked_shader->Geom.InputType = shader->Geom.InputType;
+        linked_shader->info.Geom.InputType = shader->info.Geom.InputType;
       }
 
-      if (shader->Geom.OutputType != PRIM_UNKNOWN) {
-        if (linked_shader->Geom.OutputType != PRIM_UNKNOWN &&
-            linked_shader->Geom.OutputType != shader->Geom.OutputType) {
+      if (shader->info.Geom.OutputType != PRIM_UNKNOWN) {
+        if (linked_shader->info.Geom.OutputType != PRIM_UNKNOWN &&
+            linked_shader->info.Geom.OutputType !=
+             shader->info.Geom.OutputType) {
            linker_error(prog, "geometry shader defined with conflicting "
                         "output types\n");
            return;
         }
-        linked_shader->Geom.OutputType = shader->Geom.OutputType;
+        linked_shader->info.Geom.OutputType = shader->info.Geom.OutputType;
       }
 
-      if (shader->Geom.VerticesOut != 0) {
-        if (linked_shader->Geom.VerticesOut != 0 &&
-            linked_shader->Geom.VerticesOut != shader->Geom.VerticesOut) {
+      if (shader->info.Geom.VerticesOut != -1) {
+        if (linked_shader->info.Geom.VerticesOut != -1 &&
+            linked_shader->info.Geom.VerticesOut !=
+             shader->info.Geom.VerticesOut) {
            linker_error(prog, "geometry shader defined with conflicting "
                         "output vertex count (%d and %d)\n",
-                        linked_shader->Geom.VerticesOut,
-                        shader->Geom.VerticesOut);
+                        linked_shader->info.Geom.VerticesOut,
+                        shader->info.Geom.VerticesOut);
            return;
         }
-        linked_shader->Geom.VerticesOut = shader->Geom.VerticesOut;
+        linked_shader->info.Geom.VerticesOut = shader->info.Geom.VerticesOut;
       }
 
-      if (shader->Geom.Invocations != 0) {
-        if (linked_shader->Geom.Invocations != 0 &&
-            linked_shader->Geom.Invocations != shader->Geom.Invocations) {
+      if (shader->info.Geom.Invocations != 0) {
+        if (linked_shader->info.Geom.Invocations != 0 &&
+            linked_shader->info.Geom.Invocations !=
+             shader->info.Geom.Invocations) {
            linker_error(prog, "geometry shader defined with conflicting "
                         "invocation count (%d and %d)\n",
-                        linked_shader->Geom.Invocations,
-                        shader->Geom.Invocations);
+                        linked_shader->info.Geom.Invocations,
+                        shader->info.Geom.Invocations);
            return;
         }
-        linked_shader->Geom.Invocations = shader->Geom.Invocations;
+        linked_shader->info.Geom.Invocations = shader->info.Geom.Invocations;
       }
    }
 
@@ -2041,31 +1952,26 @@ link_gs_inout_layout_qualifiers(struct gl_shader_program *prog,
     * since we already know we're in the right type of shader program
     * for doing it.
     */
-   if (linked_shader->Geom.InputType == PRIM_UNKNOWN) {
+   if (linked_shader->info.Geom.InputType == PRIM_UNKNOWN) {
       linker_error(prog,
                   "geometry shader didn't declare primitive input type\n");
       return;
    }
-   prog->Geom.InputType = linked_shader->Geom.InputType;
 
-   if (linked_shader->Geom.OutputType == PRIM_UNKNOWN) {
+   if (linked_shader->info.Geom.OutputType == PRIM_UNKNOWN) {
       linker_error(prog,
                   "geometry shader didn't declare primitive output type\n");
       return;
    }
-   prog->Geom.OutputType = linked_shader->Geom.OutputType;
 
-   if (linked_shader->Geom.VerticesOut == 0) {
+   if (linked_shader->info.Geom.VerticesOut == -1) {
       linker_error(prog,
                   "geometry shader didn't declare max_vertices\n");
       return;
    }
-   prog->Geom.VerticesOut = linked_shader->Geom.VerticesOut;
-
-   if (linked_shader->Geom.Invocations == 0)
-      linked_shader->Geom.Invocations = 1;
 
-   prog->Geom.Invocations = linked_shader->Geom.Invocations;
+   if (linked_shader->info.Geom.Invocations == 0)
+      linked_shader->info.Geom.Invocations = 1;
 }
 
 
@@ -2076,12 +1982,12 @@ link_gs_inout_layout_qualifiers(struct gl_shader_program *prog,
  */
 static void
 link_cs_input_layout_qualifiers(struct gl_shader_program *prog,
-                                struct gl_shader *linked_shader,
+                                struct gl_linked_shader *linked_shader,
                                 struct gl_shader **shader_list,
                                 unsigned num_shaders)
 {
    for (int i = 0; i < 3; i++)
-      linked_shader->Comp.LocalSize[i] = 0;
+      linked_shader->info.Comp.LocalSize[i] = 0;
 
    /* This function is called for all shader stages, but it only has an effect
     * for compute shaders.
@@ -2102,19 +2008,21 @@ link_cs_input_layout_qualifiers(struct gl_shader_program *prog,
    for (unsigned sh = 0; sh < num_shaders; sh++) {
       struct gl_shader *shader = shader_list[sh];
 
-      if (shader->Comp.LocalSize[0] != 0) {
-         if (linked_shader->Comp.LocalSize[0] != 0) {
+      if (shader->info.Comp.LocalSize[0] != 0) {
+         if (linked_shader->info.Comp.LocalSize[0] != 0) {
             for (int i = 0; i < 3; i++) {
-               if (linked_shader->Comp.LocalSize[i] !=
-                   shader->Comp.LocalSize[i]) {
+               if (linked_shader->info.Comp.LocalSize[i] !=
+                   shader->info.Comp.LocalSize[i]) {
                   linker_error(prog, "compute shader defined with conflicting "
                                "local sizes\n");
                   return;
                }
             }
          }
-         for (int i = 0; i < 3; i++)
-            linked_shader->Comp.LocalSize[i] = shader->Comp.LocalSize[i];
+         for (int i = 0; i < 3; i++) {
+            linked_shader->info.Comp.LocalSize[i] =
+               shader->info.Comp.LocalSize[i];
+         }
       }
    }
 
@@ -2122,12 +2030,12 @@ link_cs_input_layout_qualifiers(struct gl_shader_program *prog,
     * since we already know we're in the right type of shader program
     * for doing it.
     */
-   if (linked_shader->Comp.LocalSize[0] == 0) {
+   if (linked_shader->info.Comp.LocalSize[0] == 0) {
       linker_error(prog, "compute shader didn't declare local size\n");
       return;
    }
    for (int i = 0; i < 3; i++)
-      prog->Comp.LocalSize[i] = linked_shader->Comp.LocalSize[i];
+      prog->Comp.LocalSize[i] = linked_shader->info.Comp.LocalSize[i];
 }
 
 
@@ -2138,7 +2046,7 @@ link_cs_input_layout_qualifiers(struct gl_shader_program *prog,
  * If this function is supplied a single shader, it is cloned, and the new
  * shader is returned.
  */
-static struct gl_shader *
+static struct gl_linked_shader *
 link_intrastage_shaders(void *mem_ctx,
                        struct gl_context *ctx,
                        struct gl_shader_program *prog,
@@ -2152,7 +2060,13 @@ link_intrastage_shaders(void *mem_ctx,
 
    /* Check that global variables defined in multiple shaders are consistent.
     */
-   cross_validate_globals(prog, shader_list, num_shaders, false);
+   glsl_symbol_table variables;
+   for (unsigned i = 0; i < num_shaders; i++) {
+      if (shader_list[i] == NULL)
+        continue;
+      cross_validate_globals(prog, shader_list[i]->ir, &variables, false);
+   }
+
    if (!prog->LinkStatus)
       return NULL;
 
@@ -2163,14 +2077,6 @@ link_intrastage_shaders(void *mem_ctx,
    if (!prog->LinkStatus)
       return NULL;
 
-   /* Link up uniform blocks defined within this stage. */
-   link_uniform_blocks(mem_ctx, ctx, prog, shader_list, num_shaders,
-                       &ubo_blocks, &num_ubo_blocks, &ssbo_blocks,
-                       &num_ssbo_blocks);
-
-   if (!prog->LinkStatus)
-      return NULL;
-
    /* Check that there is only a single definition of each function signature
     * across all shaders.
     */
@@ -2192,14 +2098,13 @@ link_intrastage_shaders(void *mem_ctx,
               continue;
 
            foreach_in_list(ir_function_signature, sig, &f->signatures) {
-              if (!sig->is_defined || sig->is_builtin())
+              if (!sig->is_defined)
                  continue;
 
               ir_function_signature *other_sig =
                  other->exact_matching_signature(NULL, &sig->parameters);
 
-              if ((other_sig != NULL) && other_sig->is_defined
-                  && !other_sig->is_builtin()) {
+              if (other_sig != NULL && other_sig->is_defined) {
                  linker_error(prog, "function `%s' is multiply defined\n",
                               f->name);
                  return NULL;
@@ -2218,7 +2123,7 @@ link_intrastage_shaders(void *mem_ctx,
     */
    gl_shader *main = NULL;
    for (unsigned i = 0; i < num_shaders; i++) {
-      if (_mesa_get_main_function_signature(shader_list[i]) != NULL) {
+      if (_mesa_get_main_function_signature(shader_list[i]->symbols)) {
         main = shader_list[i];
         break;
       }
@@ -2230,29 +2135,11 @@ link_intrastage_shaders(void *mem_ctx,
       return NULL;
    }
 
-   gl_shader *linked = ctx->Driver.NewShader(NULL, 0, main->Type);
+   gl_linked_shader *linked = ctx->Driver.NewShader(shader_list[0]->Stage);
    linked->ir = new(linked) exec_list;
    clone_ir_list(mem_ctx, linked->ir, main->ir);
 
-   /* Copy ubo blocks to linked shader list */
-   linked->UniformBlocks =
-      ralloc_array(linked, gl_uniform_block *, num_ubo_blocks);
-   ralloc_steal(linked, ubo_blocks);
-   for (unsigned i = 0; i < num_ubo_blocks; i++) {
-      linked->UniformBlocks[i] = &ubo_blocks[i];
-   }
-   linked->NumUniformBlocks = num_ubo_blocks;
-
-   /* Copy ssbo blocks to linked shader list */
-   linked->ShaderStorageBlocks =
-      ralloc_array(linked, gl_uniform_block *, num_ssbo_blocks);
-   ralloc_steal(linked, ssbo_blocks);
-   for (unsigned i = 0; i < num_ssbo_blocks; i++) {
-      linked->ShaderStorageBlocks[i] = &ssbo_blocks[i];
-   }
-   linked->NumShaderStorageBlocks = num_ssbo_blocks;
-
-   link_fs_input_layout_qualifiers(prog, linked, shader_list, num_shaders);
+   link_fs_inout_layout_qualifiers(prog, linked, shader_list, num_shaders);
    link_tcs_out_layout_qualifiers(prog, linked, shader_list, num_shaders);
    link_tes_in_layout_qualifiers(prog, linked, shader_list, num_shaders);
    link_gs_inout_layout_qualifiers(prog, linked, shader_list, num_shaders);
@@ -2266,7 +2153,7 @@ link_intrastage_shaders(void *mem_ctx,
     * copy of the original shader that contained the main function).
     */
    ir_function_signature *const main_sig =
-      _mesa_get_main_function_signature(linked);
+      _mesa_get_main_function_signature(linked->symbols);
 
    /* Move any instructions other than variable declarations or function
     * declarations into main.
@@ -2283,45 +2170,45 @@ link_intrastage_shaders(void *mem_ctx,
                                              insertion_point, true, linked);
    }
 
-   /* Check if any shader needs built-in functions. */
-   bool need_builtins = false;
-   for (unsigned i = 0; i < num_shaders; i++) {
-      if (shader_list[i]->uses_builtin_functions) {
-         need_builtins = true;
-         break;
-      }
+   if (!link_function_calls(prog, linked, shader_list, num_shaders)) {
+      _mesa_delete_linked_shader(ctx, linked);
+      return NULL;
    }
 
-   bool ok;
-   if (need_builtins) {
-      /* Make a temporary array one larger than shader_list, which will hold
-       * the built-in function shader as well.
-       */
-      gl_shader **linking_shaders = (gl_shader **)
-         calloc(num_shaders + 1, sizeof(gl_shader *));
-
-      ok = linking_shaders != NULL;
-
-      if (ok) {
-         memcpy(linking_shaders, shader_list, num_shaders * sizeof(gl_shader *));
-         _mesa_glsl_initialize_builtin_functions();
-         linking_shaders[num_shaders] = _mesa_glsl_get_builtin_function_shader();
+   /* Make a pass over all variable declarations to ensure that arrays with
+    * unspecified sizes have a size specified.  The size is inferred from the
+    * max_array_access field.
+    */
+   array_sizing_visitor v;
+   v.run(linked->ir);
+   v.fixup_unnamed_interface_types();
 
-         ok = link_function_calls(prog, linked, linking_shaders, num_shaders + 1);
+   /* Link up uniform blocks defined within this stage. */
+   link_uniform_blocks(mem_ctx, ctx, prog, linked, &ubo_blocks,
+                       &num_ubo_blocks, &ssbo_blocks, &num_ssbo_blocks);
 
-         free(linking_shaders);
-      } else {
-         _mesa_error_no_memory(__func__);
-      }
-   } else {
-      ok = link_function_calls(prog, linked, shader_list, num_shaders);
+   if (!prog->LinkStatus) {
+      _mesa_delete_linked_shader(ctx, linked);
+      return NULL;
    }
 
+   /* Copy ubo blocks to linked shader list */
+   linked->UniformBlocks =
+      ralloc_array(linked, gl_uniform_block *, num_ubo_blocks);
+   ralloc_steal(linked, ubo_blocks);
+   for (unsigned i = 0; i < num_ubo_blocks; i++) {
+      linked->UniformBlocks[i] = &ubo_blocks[i];
+   }
+   linked->NumUniformBlocks = num_ubo_blocks;
 
-   if (!ok) {
-      _mesa_delete_shader(ctx, linked);
-      return NULL;
+   /* Copy ssbo blocks to linked shader list */
+   linked->ShaderStorageBlocks =
+      ralloc_array(linked, gl_uniform_block *, num_ssbo_blocks);
+   ralloc_steal(linked, ssbo_blocks);
+   for (unsigned i = 0; i < num_ssbo_blocks; i++) {
+      linked->ShaderStorageBlocks[i] = &ssbo_blocks[i];
    }
+   linked->NumShaderStorageBlocks = num_ssbo_blocks;
 
    /* At this point linked should contain all of the linked IR, so
     * validate it to make sure nothing went wrong.
@@ -2330,8 +2217,9 @@ link_intrastage_shaders(void *mem_ctx,
 
    /* Set the size of geometry shader input arrays */
    if (linked->Stage == MESA_SHADER_GEOMETRY) {
-      unsigned num_vertices = vertices_per_prim(prog->Geom.InputType);
-      geom_array_resize_visitor input_resize_visitor(num_vertices, prog);
+      unsigned num_vertices = vertices_per_prim(linked->info.Geom.InputType);
+      array_resize_visitor input_resize_visitor(num_vertices, prog,
+                                                MESA_SHADER_GEOMETRY);
       foreach_in_list(ir_instruction, ir, linked->ir) {
          ir->accept(&input_resize_visitor);
       }
@@ -2340,22 +2228,6 @@ link_intrastage_shaders(void *mem_ctx,
    if (ctx->Const.VertexID_is_zero_based)
       lower_vertex_id(linked);
 
-   /* Validate correct usage of barrier() in the tess control shader */
-   if (linked->Stage == MESA_SHADER_TESS_CTRL) {
-      barrier_use_visitor visitor(prog);
-      foreach_in_list(ir_instruction, ir, linked->ir) {
-         ir->accept(&visitor);
-      }
-   }
-
-   /* Make a pass over all variable declarations to ensure that arrays with
-    * unspecified sizes have a size specified.  The size is inferred from the
-    * max_array_access field.
-    */
-   array_sizing_visitor v;
-   v.run(linked->ir);
-   v.fixup_unnamed_interface_types();
-
    return linked;
 }
 
@@ -2455,23 +2327,24 @@ resize_tes_inputs(struct gl_context *ctx,
    if (prog->_LinkedShaders[MESA_SHADER_TESS_EVAL] == NULL)
       return;
 
-   gl_shader *const tcs = prog->_LinkedShaders[MESA_SHADER_TESS_CTRL];
-   gl_shader *const tes = prog->_LinkedShaders[MESA_SHADER_TESS_EVAL];
+   gl_linked_shader *const tcs = prog->_LinkedShaders[MESA_SHADER_TESS_CTRL];
+   gl_linked_shader *const tes = prog->_LinkedShaders[MESA_SHADER_TESS_EVAL];
 
    /* If no control shader is present, then the TES inputs are statically
     * sized to MaxPatchVertices; the actual size of the arrays won't be
     * known until draw time.
     */
    const int num_vertices = tcs
-      ? tcs->TessCtrl.VerticesOut
+      ? tcs->info.TessCtrl.VerticesOut
       : ctx->Const.MaxPatchVertices;
 
-   tess_eval_array_resize_visitor input_resize_visitor(num_vertices, prog);
+   array_resize_visitor input_resize_visitor(num_vertices, prog,
+                                             MESA_SHADER_TESS_EVAL);
    foreach_in_list(ir_instruction, ir, tes->ir) {
       ir->accept(&input_resize_visitor);
    }
 
-   if (tcs) {
+   if (tcs || ctx->Const.LowerTESPatchVerticesIn) {
       /* Convert the gl_PatchVerticesIn system value into a constant, since
        * the value is known at this point.
        */
@@ -2480,9 +2353,22 @@ resize_tes_inputs(struct gl_context *ctx,
          if (var && var->data.mode == ir_var_system_value &&
              var->data.location == SYSTEM_VALUE_VERTICES_IN) {
             void *mem_ctx = ralloc_parent(var);
-            var->data.mode = ir_var_auto;
             var->data.location = 0;
-            var->constant_value = new(mem_ctx) ir_constant(num_vertices);
+            var->data.explicit_location = false;
+            if (tcs) {
+               var->data.mode = ir_var_auto;
+               var->constant_value = new(mem_ctx) ir_constant(num_vertices);
+            } else {
+               var->data.mode = ir_var_uniform;
+               var->data.how_declared = ir_var_hidden;
+               var->allocate_state_slots(1);
+               ir_state_slot *slot0 = &var->get_state_slots()[0];
+               slot0->swizzle = SWIZZLE_XXXX;
+               slot0->tokens[0] = STATE_INTERNAL;
+               slot0->tokens[1] = STATE_TES_PATCH_VERTICES_IN;
+               for (int i = 2; i < STATE_LENGTH; i++)
+                  slot0->tokens[i] = 0;
+            }
          }
       }
    }
@@ -2523,6 +2409,7 @@ find_available_slots(unsigned used_mask, unsigned needed_count)
 /**
  * Assign locations for either VS inputs or FS outputs
  *
+ * \param mem_ctx       Temporary ralloc context used for linking
  * \param prog          Shader program whose variables need locations assigned
  * \param constants     Driver specific constant values for the program.
  * \param target_index  Selector for the program target to receive location
@@ -2534,7 +2421,8 @@ find_available_slots(unsigned used_mask, unsigned needed_count)
  * error is emitted to the shader link log and false is returned.
  */
 bool
-assign_attribute_or_color_locations(gl_shader_program *prog,
+assign_attribute_or_color_locations(void *mem_ctx,
+                                    gl_shader_program *prog,
                                     struct gl_constants *constants,
                                     unsigned target_index)
 {
@@ -2555,7 +2443,7 @@ assign_attribute_or_color_locations(gl_shader_program *prog,
    assert((target_index == MESA_SHADER_VERTEX)
          || (target_index == MESA_SHADER_FRAGMENT));
 
-   gl_shader *const sh = prog->_LinkedShaders[target_index];
+   gl_linked_shader *const sh = prog->_LinkedShaders[target_index];
    if (sh == NULL)
       return true;
 
@@ -2601,7 +2489,12 @@ assign_attribute_or_color_locations(gl_shader_program *prog,
    } to_assign[32];
    assert(max_index <= 32);
 
+   /* Temporary array for the set of attributes that have locations assigned.
+    */
+   ir_variable *assigned[16];
+
    unsigned num_attr = 0;
+   unsigned assigned_attr = 0;
 
    foreach_in_list(ir_instruction, node, sh->ir) {
       ir_variable *const var = node->as_variable();
@@ -2632,18 +2525,36 @@ assign_attribute_or_color_locations(gl_shader_program *prog,
       } else if (target_index == MESA_SHADER_FRAGMENT) {
         unsigned binding;
         unsigned index;
+         const char *name = var->name;
+         const glsl_type *type = var->type;
+
+         while (type) {
+            /* Check if there's a binding for the variable name */
+            if (prog->FragDataBindings->get(binding, name)) {
+               assert(binding >= FRAG_RESULT_DATA0);
+               var->data.location = binding;
+               var->data.is_unmatched_generic_inout = 0;
+
+               if (prog->FragDataIndexBindings->get(index, name)) {
+                  var->data.index = index;
+               }
+               break;
+            }
 
-        if (prog->FragDataBindings->get(binding, var->name)) {
-           assert(binding >= FRAG_RESULT_DATA0);
-           var->data.location = binding;
-            var->data.is_unmatched_generic_inout = 0;
+            /* If not, but it's an array type, look for name[0] */
+            if (type->is_array()) {
+               name = ralloc_asprintf(mem_ctx, "%s[0]", name);
+               type = type->fields.array;
+               continue;
+            }
 
-           if (prog->FragDataIndexBindings->get(index, var->name)) {
-              var->data.index = index;
-           }
-        }
+            break;
+         }
       }
 
+      if (strcmp(var->name, "gl_LastFragData") == 0)
+         continue;
+
       /* From GL4.5 core spec, section 15.2 (Shader Execution):
        *
        *     "Output binding assignments will cause LinkProgram to fail:
@@ -2763,18 +2674,62 @@ assign_attribute_or_color_locations(gl_shader_program *prog,
             * attribute overlaps any previously allocated bits.
             */
            if ((~(use_mask << attr) & used_locations) != used_locations) {
-               if (target_index == MESA_SHADER_FRAGMENT ||
-                   (prog->IsES && prog->Version >= 300)) {
-                  linker_error(prog,
-                               "overlapping location is assigned "
-                               "to %s `%s' %d %d %d\n", string,
-                               var->name, used_locations, use_mask, attr);
+               if (target_index == MESA_SHADER_FRAGMENT && !prog->IsES) {
+                  /* From section 4.4.2 (Output Layout Qualifiers) of the GLSL
+                   * 4.40 spec:
+                   *
+                   *    "Additionally, for fragment shader outputs, if two
+                   *    variables are placed within the same location, they
+                   *    must have the same underlying type (floating-point or
+                   *    integer). No component aliasing of output variables or
+                   *    members is allowed.
+                   */
+                  for (unsigned i = 0; i < assigned_attr; i++) {
+                     unsigned assigned_slots =
+                        assigned[i]->type->count_attribute_slots(false);
+                    unsigned assig_attr =
+                        assigned[i]->data.location - generic_base;
+                    unsigned assigned_use_mask = (1 << assigned_slots) - 1;
+
+                     if ((assigned_use_mask << assig_attr) &
+                         (use_mask << attr)) {
+
+                        const glsl_type *assigned_type =
+                           assigned[i]->type->without_array();
+                        const glsl_type *type = var->type->without_array();
+                        if (assigned_type->base_type != type->base_type) {
+                           linker_error(prog, "types do not match for aliased"
+                                        " %ss %s and %s\n", string,
+                                        assigned[i]->name, var->name);
+                           return false;
+                        }
+
+                        unsigned assigned_component_mask =
+                           ((1 << assigned_type->vector_elements) - 1) <<
+                           assigned[i]->data.location_frac;
+                        unsigned component_mask =
+                           ((1 << type->vector_elements) - 1) <<
+                           var->data.location_frac;
+                        if (assigned_component_mask & component_mask) {
+                           linker_error(prog, "overlapping component is "
+                                        "assigned to %ss %s and %s "
+                                        "(component=%d)\n",
+                                        string, assigned[i]->name, var->name,
+                                        var->data.location_frac);
+                           return false;
+                        }
+                     }
+                  }
+               } else if (target_index == MESA_SHADER_FRAGMENT ||
+                          (prog->IsES && prog->Version >= 300)) {
+                  linker_error(prog, "overlapping location is assigned "
+                               "to %s `%s' %d %d %d\n", string, var->name,
+                               used_locations, use_mask, attr);
                   return false;
                } else {
-                  linker_warning(prog,
-                                 "overlapping location is assigned "
-                                 "to %s `%s' %d %d %d\n", string,
-                                 var->name, used_locations, use_mask, attr);
+                  linker_warning(prog, "overlapping location is assigned "
+                                 "to %s `%s' %d %d %d\n", string, var->name,
+                                 used_locations, use_mask, attr);
                }
            }
 
@@ -2800,10 +2755,13 @@ assign_attribute_or_color_locations(gl_shader_program *prog,
              * issue (3) of the GL_ARB_vertex_attrib_64bit behavior, this
              * is optional behavior, but it seems preferable.
              */
-            if (var->type->without_array()->is_dual_slot_double())
+            if (var->type->without_array()->is_dual_slot())
                double_storage_locations |= (use_mask << attr);
         }
 
+         assigned[assigned_attr] = var;
+         assigned_attr++;
+
         continue;
       }
 
@@ -2874,7 +2832,7 @@ assign_attribute_or_color_locations(gl_shader_program *prog,
       to_assign[i].var->data.is_unmatched_generic_inout = 0;
       used_locations |= (use_mask << location);
 
-      if (to_assign[i].var->type->without_array()->is_dual_slot_double())
+      if (to_assign[i].var->type->without_array()->is_dual_slot())
          double_storage_locations |= (use_mask << location);
    }
 
@@ -2903,8 +2861,8 @@ assign_attribute_or_color_locations(gl_shader_program *prog,
  * unmatch flag if found so we don't optimise them away.
  */
 static void
-match_explicit_outputs_to_inputs(gl_shader *producer,
-                                 gl_shader *consumer)
+match_explicit_outputs_to_inputs(gl_linked_shader *producer,
+                                 gl_linked_shader *consumer)
 {
    glsl_symbol_table parameters;
    ir_variable *explicit_locations[MAX_VARYINGS_INCL_PATCH][4] =
@@ -3008,7 +2966,7 @@ check_resources(struct gl_context *ctx, struct gl_shader_program *prog)
    unsigned total_shader_storage_blocks = 0;
 
    for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
-      struct gl_shader *sh = prog->_LinkedShaders[i];
+      struct gl_linked_shader *sh = prog->_LinkedShaders[i];
 
       if (sh == NULL)
         continue;
@@ -3102,7 +3060,7 @@ static void
 link_calculate_subroutine_compat(struct gl_shader_program *prog)
 {
    for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
-      struct gl_shader *sh = prog->_LinkedShaders[i];
+      struct gl_linked_shader *sh = prog->_LinkedShaders[i];
       int count;
       if (!sh)
          continue;
@@ -3140,7 +3098,7 @@ static void
 check_subroutine_resources(struct gl_shader_program *prog)
 {
    for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
-      struct gl_shader *sh = prog->_LinkedShaders[i];
+      struct gl_linked_shader *sh = prog->_LinkedShaders[i];
 
       if (sh) {
          if (sh->NumSubroutineUniformRemapTable > MAX_SUBROUTINE_UNIFORM_LOCATIONS)
@@ -3163,7 +3121,7 @@ check_image_resources(struct gl_context *ctx, struct gl_shader_program *prog)
       return;
 
    for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
-      struct gl_shader *sh = prog->_LinkedShaders[i];
+      struct gl_linked_shader *sh = prog->_LinkedShaders[i];
 
       if (sh) {
          if (sh->NumImages > ctx->Const.Program[i].MaxImageUniforms)
@@ -3267,7 +3225,7 @@ reserve_explicit_locations(struct gl_shader_program *prog,
 
 static bool
 reserve_subroutine_explicit_locations(struct gl_shader_program *prog,
-                                      struct gl_shader *sh,
+                                      struct gl_linked_shader *sh,
                                       ir_variable *var)
 {
    unsigned slots = var->type->uniform_locations();
@@ -3342,7 +3300,7 @@ check_explicit_uniform_locations(struct gl_context *ctx,
 
    unsigned entries_total = 0;
    for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
-      struct gl_shader *sh = prog->_LinkedShaders[i];
+      struct gl_linked_shader *sh = prog->_LinkedShaders[i];
 
       if (!sh)
          continue;
@@ -3470,15 +3428,15 @@ should_add_buffer_variable(struct gl_shader_program *shProg,
 }
 
 static bool
-add_program_resource(struct gl_shader_program *prog, GLenum type,
-                     const void *data, uint8_t stages)
+add_program_resource(struct gl_shader_program *prog,
+                     struct set *resource_set,
+                     GLenum type, const void *data, uint8_t stages)
 {
    assert(data);
 
    /* If resource already exists, do not add it again. */
-   for (unsigned i = 0; i < prog->NumProgramResourceList; i++)
-      if (prog->ProgramResourceList[i].Data == data)
-         return true;
+   if (_mesa_set_search(resource_set, data))
+      return true;
 
    prog->ProgramResourceList =
       reralloc(prog,
@@ -3500,6 +3458,8 @@ add_program_resource(struct gl_shader_program *prog, GLenum type,
 
    prog->NumProgramResourceList++;
 
+   _mesa_set_add(resource_set, data);
+
    return true;
 }
 
@@ -3547,7 +3507,7 @@ build_stageref(struct gl_shader_program *shProg, const char *name,
    assert(MESA_SHADER_STAGES < 8);
 
    for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
-      struct gl_shader *sh = shProg->_LinkedShaders[i];
+      struct gl_linked_shader *sh = shProg->_LinkedShaders[i];
       if (!sh)
          continue;
 
@@ -3594,7 +3554,8 @@ static gl_shader_variable *
 create_shader_variable(struct gl_shader_program *shProg,
                        const ir_variable *in,
                        const char *name, const glsl_type *type,
-                       bool use_implicit_location, int location)
+                       bool use_implicit_location, int location,
+                       const glsl_type *outermost_struct_type)
 {
    gl_shader_variable *out = ralloc(shProg, struct gl_shader_variable);
    if (!out)
@@ -3606,6 +3567,18 @@ create_shader_variable(struct gl_shader_program *shProg,
    if (in->data.mode == ir_var_system_value &&
        in->data.location == SYSTEM_VALUE_VERTEX_ID_ZERO_BASE) {
       out->name = ralloc_strdup(shProg, "gl_VertexID");
+   } else if ((in->data.mode == ir_var_shader_out &&
+               in->data.location == VARYING_SLOT_TESS_LEVEL_OUTER) ||
+              (in->data.mode == ir_var_system_value &&
+               in->data.location == SYSTEM_VALUE_TESS_LEVEL_OUTER)) {
+      out->name = ralloc_strdup(shProg, "gl_TessLevelOuter");
+      type = glsl_type::get_array_instance(glsl_type::float_type, 4);
+   } else if ((in->data.mode == ir_var_shader_out &&
+               in->data.location == VARYING_SLOT_TESS_LEVEL_INNER) ||
+              (in->data.mode == ir_var_system_value &&
+               in->data.location == SYSTEM_VALUE_TESS_LEVEL_INNER)) {
+      out->name = ralloc_strdup(shProg, "gl_TessLevelInner");
+      type = glsl_type::get_array_instance(glsl_type::float_type, 2);
    } else {
       out->name = ralloc_strdup(shProg, name);
    }
@@ -3637,19 +3610,26 @@ create_shader_variable(struct gl_shader_program *shProg,
    }
 
    out->type = type;
+   out->outermost_struct_type = outermost_struct_type;
+   out->interface_type = in->get_interface_type();
    out->component = in->data.location_frac;
    out->index = in->data.index;
    out->patch = in->data.patch;
    out->mode = in->data.mode;
+   out->interpolation = in->data.interpolation;
+   out->explicit_location = in->data.explicit_location;
+   out->precision = in->data.precision;
 
    return out;
 }
 
 static bool
-add_shader_variable(struct gl_shader_program *shProg, unsigned stage_mask,
+add_shader_variable(struct gl_shader_program *shProg, struct set *resource_set,
+                    unsigned stage_mask,
                     GLenum programInterface, ir_variable *var,
                     const char *name, const glsl_type *type,
-                    bool use_implicit_location, int location)
+                    bool use_implicit_location, int location,
+                    const glsl_type *outermost_struct_type = NULL)
 {
    const bool is_vertex_input =
       programInterface == GL_PROGRAM_INPUT &&
@@ -3666,13 +3646,18 @@ add_shader_variable(struct gl_shader_program *shProg, unsigned stage_mask,
        *     structure member to enumerate is itself a structure or array,
        *     these enumeration rules are applied recursively."
        */
+      if (outermost_struct_type == NULL)
+         outermost_struct_type = type;
+
       unsigned field_location = location;
       for (unsigned i = 0; i < type->length; i++) {
          const struct glsl_struct_field *field = &type->fields.structure[i];
          char *field_name = ralloc_asprintf(shProg, "%s.%s", name, field->name);
-         if (!add_shader_variable(shProg, stage_mask, programInterface,
+         if (!add_shader_variable(shProg, resource_set,
+                                  stage_mask, programInterface,
                                   var, field_name, field->type,
-                                  use_implicit_location, field_location))
+                                  use_implicit_location, field_location,
+                                  outermost_struct_type))
             return false;
 
          field_location +=
@@ -3706,17 +3691,20 @@ add_shader_variable(struct gl_shader_program *shProg, unsigned stage_mask,
        */
       gl_shader_variable *sha_v =
          create_shader_variable(shProg, var, prefixed_name, type,
-                                use_implicit_location, location);
+                                use_implicit_location, location,
+                                outermost_struct_type);
       if (!sha_v)
          return false;
 
-      return add_program_resource(shProg, programInterface, sha_v, stage_mask);
+      return add_program_resource(shProg, resource_set,
+                                  programInterface, sha_v, stage_mask);
    }
    }
 }
 
 static bool
 add_interface_variables(struct gl_shader_program *shProg,
+                        struct set *resource_set,
                         unsigned stage, GLenum programInterface)
 {
    exec_list *ir = shProg->_LinkedShaders[stage]->ir;
@@ -3747,6 +3735,9 @@ add_interface_variables(struct gl_shader_program *shProg,
          continue;
       };
 
+      if (var->data.patch)
+         loc_bias = int(VARYING_SLOT_PATCH0);
+
       /* Skip packed varyings, packed varyings are handled separately
        * by add_packed_varyings.
        */
@@ -3763,7 +3754,8 @@ add_interface_variables(struct gl_shader_program *shProg,
          (stage == MESA_SHADER_VERTEX && var->data.mode == ir_var_shader_in) ||
          (stage == MESA_SHADER_FRAGMENT && var->data.mode == ir_var_shader_out);
 
-      if (!add_shader_variable(shProg, 1 << stage, programInterface,
+      if (!add_shader_variable(shProg, resource_set,
+                               1 << stage, programInterface,
                                var, var->name, var->type, vs_input_or_fs_output,
                                var->data.location - loc_bias))
          return false;
@@ -3772,9 +3764,10 @@ add_interface_variables(struct gl_shader_program *shProg,
 }
 
 static bool
-add_packed_varyings(struct gl_shader_program *shProg, int stage, GLenum type)
+add_packed_varyings(struct gl_shader_program *shProg, struct set *resource_set,
+                    int stage, GLenum type)
 {
-   struct gl_shader *sh = shProg->_LinkedShaders[stage];
+   struct gl_linked_shader *sh = shProg->_LinkedShaders[stage];
    GLenum iface;
 
    if (!sh || !sh->packed_varyings)
@@ -3797,7 +3790,8 @@ add_packed_varyings(struct gl_shader_program *shProg, int stage, GLenum type)
          if (type == iface) {
             const int stage_mask =
                build_stageref(shProg, var->name, var->data.mode);
-            if (!add_shader_variable(shProg, stage_mask,
+            if (!add_shader_variable(shProg, resource_set,
+                                     stage_mask,
                                      iface, var, var->name, var->type, false,
                                      var->data.location - VARYING_SLOT_VAR0))
                return false;
@@ -3808,9 +3802,9 @@ add_packed_varyings(struct gl_shader_program *shProg, int stage, GLenum type)
 }
 
 static bool
-add_fragdata_arrays(struct gl_shader_program *shProg)
+add_fragdata_arrays(struct gl_shader_program *shProg, struct set *resource_set)
 {
-   struct gl_shader *sh = shProg->_LinkedShaders[MESA_SHADER_FRAGMENT];
+   struct gl_linked_shader *sh = shProg->_LinkedShaders[MESA_SHADER_FRAGMENT];
 
    if (!sh || !sh->fragdata_arrays)
       return true;
@@ -3820,7 +3814,7 @@ add_fragdata_arrays(struct gl_shader_program *shProg)
       if (var) {
          assert(var->data.mode == ir_var_shader_out);
 
-         if (!add_shader_variable(shProg,
+         if (!add_shader_variable(shProg, resource_set,
                                   1 << MESA_SHADER_FRAGMENT,
                                   GL_PROGRAM_OUTPUT, var, var->name, var->type,
                                   true, var->data.location - FRAG_RESULT_DATA0))
@@ -4005,12 +3999,12 @@ calculate_array_size_and_stride(struct gl_shader_program *shProg,
       }
    }
 
-   for (unsigned i = 0; i < shProg->NumShaders; i++) {
-      if (shProg->Shaders[i] == NULL)
+   for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
+      const gl_linked_shader *sh = shProg->_LinkedShaders[i];
+      if (sh == NULL)
          continue;
 
-      const gl_shader *stage = shProg->Shaders[i];
-      foreach_in_list(ir_instruction, node, stage->ir) {
+      foreach_in_list(ir_instruction, node, sh->ir) {
          ir_variable *var = node->as_variable();
          if (!var || !var->get_interface_type() ||
              var->data.mode != ir_var_shader_storage)
@@ -4073,29 +4067,38 @@ build_program_resource_list(struct gl_context *ctx,
    if (input_stage == MESA_SHADER_STAGES && output_stage == 0)
       return;
 
+   struct set *resource_set = _mesa_set_create(NULL,
+                                               _mesa_hash_pointer,
+                                               _mesa_key_pointer_equal);
+
    /* Program interface needs to expose varyings in case of SSO. */
    if (shProg->SeparateShader) {
-      if (!add_packed_varyings(shProg, input_stage, GL_PROGRAM_INPUT))
+      if (!add_packed_varyings(shProg, resource_set,
+                               input_stage, GL_PROGRAM_INPUT))
          return;
 
-      if (!add_packed_varyings(shProg, output_stage, GL_PROGRAM_OUTPUT))
+      if (!add_packed_varyings(shProg, resource_set,
+                               output_stage, GL_PROGRAM_OUTPUT))
          return;
    }
 
-   if (!add_fragdata_arrays(shProg))
+   if (!add_fragdata_arrays(shProg, resource_set))
       return;
 
    /* Add inputs and outputs to the resource list. */
-   if (!add_interface_variables(shProg, input_stage, GL_PROGRAM_INPUT))
+   if (!add_interface_variables(shProg, resource_set,
+                                input_stage, GL_PROGRAM_INPUT))
       return;
 
-   if (!add_interface_variables(shProg, output_stage, GL_PROGRAM_OUTPUT))
+   if (!add_interface_variables(shProg, resource_set,
+                                output_stage, GL_PROGRAM_OUTPUT))
       return;
 
    /* Add transform feedback varyings. */
    if (shProg->LinkedTransformFeedback.NumVarying > 0) {
       for (int i = 0; i < shProg->LinkedTransformFeedback.NumVarying; i++) {
-         if (!add_program_resource(shProg, GL_TRANSFORM_FEEDBACK_VARYING,
+         if (!add_program_resource(shProg, resource_set,
+                                   GL_TRANSFORM_FEEDBACK_VARYING,
                                    &shProg->LinkedTransformFeedback.Varyings[i],
                                    0))
          return;
@@ -4106,7 +4109,8 @@ build_program_resource_list(struct gl_context *ctx,
    for (unsigned i = 0; i < ctx->Const.MaxTransformFeedbackBuffers; i++) {
       if ((shProg->LinkedTransformFeedback.ActiveBuffers >> i) & 1) {
          shProg->LinkedTransformFeedback.Buffers[i].Binding = i;
-         if (!add_program_resource(shProg, GL_TRANSFORM_FEEDBACK_BUFFER,
+         if (!add_program_resource(shProg, resource_set,
+                                   GL_TRANSFORM_FEEDBACK_BUFFER,
                                    &shProg->LinkedTransformFeedback.Buffers[i],
                                    0))
          return;
@@ -4141,28 +4145,28 @@ build_program_resource_list(struct gl_context *ctx,
          calculate_array_size_and_stride(shProg, &shProg->UniformStorage[i]);
       }
 
-      if (!add_program_resource(shProg, type,
+      if (!add_program_resource(shProg, resource_set, type,
                                 &shProg->UniformStorage[i], stageref))
          return;
    }
 
    /* Add program uniform blocks. */
    for (unsigned i = 0; i < shProg->NumUniformBlocks; i++) {
-      if (!add_program_resource(shProg, GL_UNIFORM_BLOCK,
+      if (!add_program_resource(shProg, resource_set, GL_UNIFORM_BLOCK,
           &shProg->UniformBlocks[i], 0))
          return;
    }
 
    /* Add program shader storage blocks. */
    for (unsigned i = 0; i < shProg->NumShaderStorageBlocks; i++) {
-      if (!add_program_resource(shProg, GL_SHADER_STORAGE_BLOCK,
+      if (!add_program_resource(shProg, resource_set, GL_SHADER_STORAGE_BLOCK,
           &shProg->ShaderStorageBlocks[i], 0))
          return;
    }
 
    /* Add atomic counter buffers. */
    for (unsigned i = 0; i < shProg->NumAtomicBuffers; i++) {
-      if (!add_program_resource(shProg, GL_ATOMIC_COUNTER_BUFFER,
+      if (!add_program_resource(shProg, resource_set, GL_ATOMIC_COUNTER_BUFFER,
                                 &shProg->AtomicBuffers[i], 0))
          return;
    }
@@ -4179,13 +4183,14 @@ build_program_resource_list(struct gl_context *ctx,
 
          type = _mesa_shader_stage_to_subroutine_uniform((gl_shader_stage)j);
          /* add shader subroutines */
-         if (!add_program_resource(shProg, type, &shProg->UniformStorage[i], 0))
+         if (!add_program_resource(shProg, resource_set,
+                                   type, &shProg->UniformStorage[i], 0))
             return;
       }
    }
 
    for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
-      struct gl_shader *sh = shProg->_LinkedShaders[i];
+      struct gl_linked_shader *sh = shProg->_LinkedShaders[i];
       GLuint type;
 
       if (!sh)
@@ -4193,10 +4198,13 @@ build_program_resource_list(struct gl_context *ctx,
 
       type = _mesa_shader_stage_to_subroutine((gl_shader_stage)i);
       for (unsigned j = 0; j < sh->NumSubroutineFunctions; j++) {
-         if (!add_program_resource(shProg, type, &sh->SubroutineFunctions[j], 0))
+         if (!add_program_resource(shProg, resource_set,
+                                   type, &sh->SubroutineFunctions[j], 0))
             return;
       }
    }
+
+   _mesa_set_destroy(resource_set, NULL);
 }
 
 /**
@@ -4237,7 +4245,7 @@ static void
 link_assign_subroutine_types(struct gl_shader_program *prog)
 {
    for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
-      gl_shader *sh = prog->_LinkedShaders[i];
+      gl_linked_shader *sh = prog->_LinkedShaders[i];
 
       if (sh == NULL)
          continue;
@@ -4343,7 +4351,7 @@ disable_varying_optimizations_for_sso(struct gl_shader_program *prog)
       return;
 
    for (unsigned stage = 0; stage < MESA_SHADER_STAGES; stage++) {
-      gl_shader *sh = prog->_LinkedShaders[stage];
+      gl_linked_shader *sh = prog->_LinkedShaders[stage];
       if (!sh)
          continue;
 
@@ -4368,6 +4376,224 @@ disable_varying_optimizations_for_sso(struct gl_shader_program *prog)
    }
 }
 
+static bool
+link_varyings_and_uniforms(unsigned first, unsigned last,
+                           unsigned num_explicit_uniform_locs,
+                           struct gl_context *ctx,
+                           struct gl_shader_program *prog, void *mem_ctx)
+{
+   bool has_xfb_qualifiers = false;
+   unsigned num_tfeedback_decls = 0;
+   char **varying_names = NULL;
+   tfeedback_decl *tfeedback_decls = NULL;
+
+   /* Mark all generic shader inputs and outputs as unpaired. */
+   for (unsigned i = MESA_SHADER_VERTEX; i <= MESA_SHADER_FRAGMENT; i++) {
+      if (prog->_LinkedShaders[i] != NULL) {
+         link_invalidate_variable_locations(prog->_LinkedShaders[i]->ir);
+      }
+   }
+
+   unsigned prev = first;
+   for (unsigned i = prev + 1; i <= MESA_SHADER_FRAGMENT; i++) {
+      if (prog->_LinkedShaders[i] == NULL)
+         continue;
+
+      match_explicit_outputs_to_inputs(prog->_LinkedShaders[prev],
+                                       prog->_LinkedShaders[i]);
+      prev = i;
+   }
+
+   if (!assign_attribute_or_color_locations(mem_ctx, prog, &ctx->Const,
+                                            MESA_SHADER_VERTEX)) {
+      return false;
+   }
+
+   if (!assign_attribute_or_color_locations(mem_ctx, prog, &ctx->Const,
+                                            MESA_SHADER_FRAGMENT)) {
+      return false;
+   }
+
+   /* From the ARB_enhanced_layouts spec:
+    *
+    *    "If the shader used to record output variables for transform feedback
+    *    varyings uses the "xfb_buffer", "xfb_offset", or "xfb_stride" layout
+    *    qualifiers, the values specified by TransformFeedbackVaryings are
+    *    ignored, and the set of variables captured for transform feedback is
+    *    instead derived from the specified layout qualifiers."
+    */
+   for (int i = MESA_SHADER_FRAGMENT - 1; i >= 0; i--) {
+      /* Find last stage before fragment shader */
+      if (prog->_LinkedShaders[i]) {
+         has_xfb_qualifiers =
+            process_xfb_layout_qualifiers(mem_ctx, prog->_LinkedShaders[i],
+                                          &num_tfeedback_decls,
+                                          &varying_names);
+         break;
+      }
+   }
+
+   if (!has_xfb_qualifiers) {
+      num_tfeedback_decls = prog->TransformFeedback.NumVarying;
+      varying_names = prog->TransformFeedback.VaryingNames;
+   }
+
+   if (num_tfeedback_decls != 0) {
+      /* From GL_EXT_transform_feedback:
+       *   A program will fail to link if:
+       *
+       *   * the <count> specified by TransformFeedbackVaryingsEXT is
+       *     non-zero, but the program object has no vertex or geometry
+       *     shader;
+       */
+      if (first >= MESA_SHADER_FRAGMENT) {
+         linker_error(prog, "Transform feedback varyings specified, but "
+                      "no vertex, tessellation, or geometry shader is "
+                      "present.\n");
+         return false;
+      }
+
+      tfeedback_decls = ralloc_array(mem_ctx, tfeedback_decl,
+                                     num_tfeedback_decls);
+      if (!parse_tfeedback_decls(ctx, prog, mem_ctx, num_tfeedback_decls,
+                                 varying_names, tfeedback_decls))
+         return false;
+   }
+
+   /* If there is no fragment shader we need to set transform feedback.
+    *
+    * For SSO we also need to assign output locations.  We assign them here
+    * because we need to do it for both single stage programs and multi stage
+    * programs.
+    */
+   if (last < MESA_SHADER_FRAGMENT &&
+       (num_tfeedback_decls != 0 || prog->SeparateShader)) {
+      const uint64_t reserved_out_slots =
+         reserved_varying_slot(prog->_LinkedShaders[last], ir_var_shader_out);
+      if (!assign_varying_locations(ctx, mem_ctx, prog,
+                                    prog->_LinkedShaders[last], NULL,
+                                    num_tfeedback_decls, tfeedback_decls,
+                                    reserved_out_slots))
+         return false;
+   }
+
+   if (last <= MESA_SHADER_FRAGMENT) {
+      /* Remove unused varyings from the first/last stage unless SSO */
+      remove_unused_shader_inputs_and_outputs(prog->SeparateShader,
+                                              prog->_LinkedShaders[first],
+                                              ir_var_shader_in);
+      remove_unused_shader_inputs_and_outputs(prog->SeparateShader,
+                                              prog->_LinkedShaders[last],
+                                              ir_var_shader_out);
+
+      /* If the program is made up of only a single stage */
+      if (first == last) {
+         gl_linked_shader *const sh = prog->_LinkedShaders[last];
+
+         do_dead_builtin_varyings(ctx, NULL, sh, 0, NULL);
+         do_dead_builtin_varyings(ctx, sh, NULL, num_tfeedback_decls,
+                                  tfeedback_decls);
+
+         if (prog->SeparateShader) {
+            const uint64_t reserved_slots =
+               reserved_varying_slot(sh, ir_var_shader_in);
+
+            /* Assign input locations for SSO, output locations are already
+             * assigned.
+             */
+            if (!assign_varying_locations(ctx, mem_ctx, prog,
+                                          NULL /* producer */,
+                                          sh /* consumer */,
+                                          0 /* num_tfeedback_decls */,
+                                          NULL /* tfeedback_decls */,
+                                          reserved_slots))
+               return false;
+         }
+      } else {
+         /* Linking the stages in the opposite order (from fragment to vertex)
+          * ensures that inter-shader outputs written to in an earlier stage
+          * are eliminated if they are (transitively) not used in a later
+          * stage.
+          */
+         int next = last;
+         for (int i = next - 1; i >= 0; i--) {
+            if (prog->_LinkedShaders[i] == NULL && i != 0)
+               continue;
+
+            gl_linked_shader *const sh_i = prog->_LinkedShaders[i];
+            gl_linked_shader *const sh_next = prog->_LinkedShaders[next];
+
+            const uint64_t reserved_out_slots =
+               reserved_varying_slot(sh_i, ir_var_shader_out);
+            const uint64_t reserved_in_slots =
+               reserved_varying_slot(sh_next, ir_var_shader_in);
+
+            do_dead_builtin_varyings(ctx, sh_i, sh_next,
+                      next == MESA_SHADER_FRAGMENT ? num_tfeedback_decls : 0,
+                      tfeedback_decls);
+
+            if (!assign_varying_locations(ctx, mem_ctx, prog, sh_i, sh_next,
+                      next == MESA_SHADER_FRAGMENT ? num_tfeedback_decls : 0,
+                      tfeedback_decls,
+                      reserved_out_slots | reserved_in_slots))
+               return false;
+
+            /* This must be done after all dead varyings are eliminated. */
+            if (sh_i != NULL) {
+               unsigned slots_used = _mesa_bitcount_64(reserved_out_slots);
+               if (!check_against_output_limit(ctx, prog, sh_i, slots_used)) {
+                  return false;
+               }
+            }
+
+            unsigned slots_used = _mesa_bitcount_64(reserved_in_slots);
+            if (!check_against_input_limit(ctx, prog, sh_next, slots_used))
+               return false;
+
+            next = i;
+         }
+      }
+   }
+
+   if (!store_tfeedback_info(ctx, prog, num_tfeedback_decls, tfeedback_decls,
+                             has_xfb_qualifiers))
+      return false;
+
+   update_array_sizes(prog);
+   link_assign_uniform_locations(prog, ctx, num_explicit_uniform_locs);
+   link_assign_atomic_counter_resources(ctx, prog);
+
+   link_calculate_subroutine_compat(prog);
+   check_resources(ctx, prog);
+   check_subroutine_resources(prog);
+   check_image_resources(ctx, prog);
+   link_check_atomic_counter_resources(ctx, prog);
+
+   if (!prog->LinkStatus)
+      return false;
+
+   for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
+      if (prog->_LinkedShaders[i] == NULL)
+         continue;
+
+      const struct gl_shader_compiler_options *options =
+         &ctx->Const.ShaderCompilerOptions[i];
+
+      if (options->LowerBufferInterfaceBlocks)
+         lower_ubo_reference(prog->_LinkedShaders[i],
+                             options->ClampBlockIndicesToArrayBounds);
+
+      if (options->LowerShaderSharedVariables)
+         lower_shared_reference(prog->_LinkedShaders[i],
+                                &prog->Comp.SharedSize);
+
+      lower_vector_derefs(prog->_LinkedShaders[i]);
+      do_vec_index_to_swizzle(prog->_LinkedShaders[i]->ir);
+   }
+
+   return true;
+}
+
 void
 link_shaders(struct gl_context *ctx, struct gl_shader_program *prog)
 {
@@ -4394,11 +4620,7 @@ link_shaders(struct gl_context *ctx, struct gl_shader_program *prog)
       return;
    }
 
-   unsigned num_tfeedback_decls = 0;
    unsigned int num_explicit_uniform_locs = 0;
-   bool has_xfb_qualifiers = false;
-   char **varying_names = NULL;
-   tfeedback_decl *tfeedback_decls = NULL;
 
    void *mem_ctx = ralloc_context(NULL); // temporary linker context
 
@@ -4427,7 +4649,7 @@ link_shaders(struct gl_context *ctx, struct gl_shader_program *prog)
         goto done;
       }
 
-      if (prog->Shaders[i]->ARB_fragment_coord_conventions_enable) {
+      if (prog->Shaders[i]->info.ARB_fragment_coord_conventions_enable) {
          prog->ARB_fragment_coord_conventions_enable = true;
       }
 
@@ -4502,8 +4724,9 @@ link_shaders(struct gl_context *ctx, struct gl_shader_program *prog)
    }
 
    for (unsigned int i = 0; i < MESA_SHADER_STAGES; i++) {
-      if (prog->_LinkedShaders[i] != NULL)
-        _mesa_delete_shader(ctx, prog->_LinkedShaders[i]);
+      if (prog->_LinkedShaders[i] != NULL) {
+        _mesa_delete_linked_shader(ctx, prog->_LinkedShaders[i]);
+      }
 
       prog->_LinkedShaders[i] = NULL;
    }
@@ -4512,13 +4735,13 @@ link_shaders(struct gl_context *ctx, struct gl_shader_program *prog)
     */
    for (int stage = 0; stage < MESA_SHADER_STAGES; stage++) {
       if (num_shaders[stage] > 0) {
-         gl_shader *const sh =
+         gl_linked_shader *const sh =
             link_intrastage_shaders(mem_ctx, ctx, prog, shader_list[stage],
                                     num_shaders[stage]);
 
          if (!prog->LinkStatus) {
             if (sh)
-               _mesa_delete_shader(ctx, sh);
+               _mesa_delete_linked_shader(ctx, sh);
             goto done;
          }
 
@@ -4541,11 +4764,11 @@ link_shaders(struct gl_context *ctx, struct gl_shader_program *prog)
          }
          if (!prog->LinkStatus) {
             if (sh)
-               _mesa_delete_shader(ctx, sh);
+               _mesa_delete_linked_shader(ctx, sh);
             goto done;
          }
 
-         _mesa_reference_shader(ctx, &prog->_LinkedShaders[stage], sh);
+         prog->_LinkedShaders[stage] = sh;
       }
    }
 
@@ -4616,8 +4839,7 @@ link_shaders(struct gl_context *ctx, struct gl_shader_program *prog)
    }
 
    /* Cross-validate uniform blocks between shader stages */
-   validate_interstage_uniform_blocks(prog, prog->_LinkedShaders,
-                                      MESA_SHADER_STAGES);
+   validate_interstage_uniform_blocks(prog, prog->_LinkedShaders);
    if (!prog->LinkStatus)
       goto done;
 
@@ -4633,7 +4855,7 @@ link_shaders(struct gl_context *ctx, struct gl_shader_program *prog)
     * This rule also applies to GLSL ES 3.00.
     */
    if (max_version >= (prog->IsES ? 300 : 130)) {
-      struct gl_shader *sh = prog->_LinkedShaders[MESA_SHADER_FRAGMENT];
+      struct gl_linked_shader *sh = prog->_LinkedShaders[MESA_SHADER_FRAGMENT];
       if (sh) {
         lower_discard_flow(sh->ir);
       }
@@ -4675,7 +4897,8 @@ link_shaders(struct gl_context *ctx, struct gl_shader_program *prog)
                                     ctx->Const.NativeIntegers))
         ;
 
-      lower_const_arrays_to_uniforms(prog->_LinkedShaders[i]->ir);
+      lower_const_arrays_to_uniforms(prog->_LinkedShaders[i]->ir, i);
+      propagate_invariance(prog->_LinkedShaders[i]->ir);
    }
 
    /* Validation for special cases where we allow sampler array indexing
@@ -4691,176 +4914,10 @@ link_shaders(struct gl_context *ctx, struct gl_shader_program *prog)
    /* Check and validate stream emissions in geometry shaders */
    validate_geometry_shader_emissions(ctx, prog);
 
-   /* Mark all generic shader inputs and outputs as unpaired. */
-   for (unsigned i = MESA_SHADER_VERTEX; i <= MESA_SHADER_FRAGMENT; i++) {
-      if (prog->_LinkedShaders[i] != NULL) {
-         link_invalidate_variable_locations(prog->_LinkedShaders[i]->ir);
-      }
-   }
-
-   prev = first;
-   for (unsigned i = prev + 1; i <= MESA_SHADER_FRAGMENT; i++) {
-      if (prog->_LinkedShaders[i] == NULL)
-         continue;
-
-      match_explicit_outputs_to_inputs(prog->_LinkedShaders[prev],
-                                       prog->_LinkedShaders[i]);
-      prev = i;
-   }
-
-   if (!assign_attribute_or_color_locations(prog, &ctx->Const,
-                                            MESA_SHADER_VERTEX)) {
-      goto done;
-   }
-
-   if (!assign_attribute_or_color_locations(prog, &ctx->Const,
-                                            MESA_SHADER_FRAGMENT)) {
-      goto done;
-   }
-
-   /* From the ARB_enhanced_layouts spec:
-    *
-    *    "If the shader used to record output variables for transform feedback
-    *    varyings uses the "xfb_buffer", "xfb_offset", or "xfb_stride" layout
-    *    qualifiers, the values specified by TransformFeedbackVaryings are
-    *    ignored, and the set of variables captured for transform feedback is
-    *    instead derived from the specified layout qualifiers."
-    */
-   for (int i = MESA_SHADER_FRAGMENT - 1; i >= 0; i--) {
-      /* Find last stage before fragment shader */
-      if (prog->_LinkedShaders[i]) {
-         has_xfb_qualifiers =
-            process_xfb_layout_qualifiers(mem_ctx, prog->_LinkedShaders[i],
-                                          &num_tfeedback_decls,
-                                          &varying_names);
-         break;
-      }
-   }
-
-   if (!has_xfb_qualifiers) {
-      num_tfeedback_decls = prog->TransformFeedback.NumVarying;
-      varying_names = prog->TransformFeedback.VaryingNames;
-   }
-
-   if (num_tfeedback_decls != 0) {
-      /* From GL_EXT_transform_feedback:
-       *   A program will fail to link if:
-       *
-       *   * the <count> specified by TransformFeedbackVaryingsEXT is
-       *     non-zero, but the program object has no vertex or geometry
-       *     shader;
-       */
-      if (first >= MESA_SHADER_FRAGMENT) {
-         linker_error(prog, "Transform feedback varyings specified, but "
-                      "no vertex, tessellation, or geometry shader is "
-                      "present.\n");
-         goto done;
-      }
-
-      tfeedback_decls = ralloc_array(mem_ctx, tfeedback_decl,
-                                     num_tfeedback_decls);
-      if (!parse_tfeedback_decls(ctx, prog, mem_ctx, num_tfeedback_decls,
-                                 varying_names, tfeedback_decls))
-         goto done;
-   }
-
-   /* If there is no fragment shader we need to set transform feedback.
-    *
-    * For SSO we also need to assign output locations.  We assign them here
-    * because we need to do it for both single stage programs and multi stage
-    * programs.
-    */
-   if (last < MESA_SHADER_FRAGMENT &&
-       (num_tfeedback_decls != 0 || prog->SeparateShader)) {
-      if (!assign_varying_locations(ctx, mem_ctx, prog,
-                                    prog->_LinkedShaders[last], NULL,
-                                    num_tfeedback_decls, tfeedback_decls))
-         goto done;
-   }
-
-   if (last <= MESA_SHADER_FRAGMENT) {
-      /* Remove unused varyings from the first/last stage unless SSO */
-      remove_unused_shader_inputs_and_outputs(prog->SeparateShader,
-                                              prog->_LinkedShaders[first],
-                                              ir_var_shader_in);
-      remove_unused_shader_inputs_and_outputs(prog->SeparateShader,
-                                              prog->_LinkedShaders[last],
-                                              ir_var_shader_out);
-
-      /* If the program is made up of only a single stage */
-      if (first == last) {
-
-         gl_shader *const sh = prog->_LinkedShaders[last];
-         if (prog->SeparateShader) {
-            /* Assign input locations for SSO, output locations are already
-             * assigned.
-             */
-            if (!assign_varying_locations(ctx, mem_ctx, prog,
-                                          NULL /* producer */,
-                                          sh /* consumer */,
-                                          0 /* num_tfeedback_decls */,
-                                          NULL /* tfeedback_decls */))
-               goto done;
-         }
-
-         do_dead_builtin_varyings(ctx, NULL, sh, 0, NULL);
-         do_dead_builtin_varyings(ctx, sh, NULL, num_tfeedback_decls,
-                                  tfeedback_decls);
-      } else {
-         /* Linking the stages in the opposite order (from fragment to vertex)
-          * ensures that inter-shader outputs written to in an earlier stage
-          * are eliminated if they are (transitively) not used in a later
-          * stage.
-          */
-         int next = last;
-         for (int i = next - 1; i >= 0; i--) {
-            if (prog->_LinkedShaders[i] == NULL && i != 0)
-               continue;
-
-            gl_shader *const sh_i = prog->_LinkedShaders[i];
-            gl_shader *const sh_next = prog->_LinkedShaders[next];
-
-            if (!assign_varying_locations(ctx, mem_ctx, prog, sh_i, sh_next,
-                      next == MESA_SHADER_FRAGMENT ? num_tfeedback_decls : 0,
-                      tfeedback_decls))
-               goto done;
-
-            do_dead_builtin_varyings(ctx, sh_i, sh_next,
-                      next == MESA_SHADER_FRAGMENT ? num_tfeedback_decls : 0,
-                      tfeedback_decls);
-
-            /* This must be done after all dead varyings are eliminated. */
-            if (sh_i != NULL) {
-               if (!check_against_output_limit(ctx, prog, sh_i)) {
-                  goto done;
-               }
-            }
-            if (!check_against_input_limit(ctx, prog, sh_next))
-               goto done;
-
-            next = i;
-         }
-      }
-   }
-
-   if (!store_tfeedback_info(ctx, prog, num_tfeedback_decls, tfeedback_decls,
-                             has_xfb_qualifiers))
-      goto done;
-
-   update_array_sizes(prog);
-   link_assign_uniform_locations(prog, ctx->Const.UniformBooleanTrue,
-                                 num_explicit_uniform_locs,
-                                 ctx->Const.MaxUserAssignableUniformLocations);
-   link_assign_atomic_counter_resources(ctx, prog);
    store_fragdepth_layout(prog);
 
-   link_calculate_subroutine_compat(prog);
-   check_resources(ctx, prog);
-   check_subroutine_resources(prog);
-   check_image_resources(ctx, prog);
-   link_check_atomic_counter_resources(ctx, prog);
-
-   if (!prog->LinkStatus)
+   if(!link_varyings_and_uniforms(first, last, num_explicit_uniform_locs, ctx,
+                                  prog, mem_ctx))
       goto done;
 
    /* OpenGL ES < 3.1 requires that a vertex shader and a fragment shader both
@@ -4898,24 +4955,6 @@ link_shaders(struct gl_context *ctx, struct gl_shader_program *prog)
       }
    }
 
-   for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
-      if (prog->_LinkedShaders[i] == NULL)
-        continue;
-
-      const struct gl_shader_compiler_options *options =
-         &ctx->Const.ShaderCompilerOptions[i];
-
-      if (options->LowerBufferInterfaceBlocks)
-         lower_ubo_reference(prog->_LinkedShaders[i]);
-
-      if (options->LowerShaderSharedVariables)
-         lower_shared_reference(prog->_LinkedShaders[i],
-                                &prog->Comp.SharedSize);
-
-      lower_vector_derefs(prog->_LinkedShaders[i]);
-      do_vec_index_to_swizzle(prog->_LinkedShaders[i]->ir);
-   }
-
 done:
    for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
       free(shader_list[i]);