glsl2: Add flags to enable variable index lowering

5 files changed, 82 insertions, 8 deletions

diff --git a/src/glsl/ir_optimization.h b/src/glsl/ir_optimization.h
index 5c8dc17d33..6a37e167fe 100644
--- a/src/glsl/ir_optimization.h
+++ b/src/glsl/ir_optimization.h
@@ -56,5 +56,6 @@ bool do_tree_grafting(exec_list *instructions);
 bool do_vec_index_to_cond_assign(exec_list *instructions);
 bool do_vec_index_to_swizzle(exec_list *instructions);
 bool lower_noise(exec_list *instructions);
-bool lower_variable_index_to_cond_assign(exec_list *instructions);
+bool lower_variable_index_to_cond_assign(exec_list *instructions,
+    bool lower_input, bool lower_output, bool lower_temp, bool lower_uniform);
 bool optimize_redundant_jumps(exec_list *instructions);
diff --git a/src/glsl/lower_variable_index_to_cond_assign.cpp b/src/glsl/lower_variable_index_to_cond_assign.cpp
index 095592b9f7..2044199d27 100644
--- a/src/glsl/lower_variable_index_to_cond_assign.cpp
+++ b/src/glsl/lower_variable_index_to_cond_assign.cpp
@@ -210,18 +210,56 @@ struct switch_generator
 
 class variable_index_to_cond_assign_visitor : public ir_rvalue_visitor {
 public:
-   variable_index_to_cond_assign_visitor()
+   variable_index_to_cond_assign_visitor(bool lower_input,
+					 bool lower_output,
+					 bool lower_temp,
+					 bool lower_uniform)
    {
       this->progress = false;
+      this->lower_inputs = lower_input;
+      this->lower_outputs = lower_output;
+      this->lower_temps = lower_temp;
+      this->lower_uniforms = lower_uniform;
    }
 
    bool progress;
+   bool lower_inputs;
+   bool lower_outputs;
+   bool lower_temps;
+   bool lower_uniforms;
 
    bool is_array_or_matrix(const ir_instruction *ir) const
    {
       return (ir->type->is_array() || ir->type->is_matrix());
    }
 
+   bool needs_lowering(ir_dereference_array *deref) const
+   {
+      if (deref == NULL || deref->array_index->as_constant()
+	  || !is_array_or_matrix(deref->array))
+	 return false;
+
+      if (deref->array->ir_type == ir_type_constant)
+	 return this->lower_temps;
+
+      const ir_variable *const var = deref->array->variable_referenced();
+      switch (var->mode) {
+      case ir_var_auto:
+      case ir_var_temporary:
+	 return this->lower_temps;
+      case ir_var_uniform:
+	 return this->lower_uniforms;
+      case ir_var_in:
+	 return (var->location == -1) ? this->lower_temps : this->lower_inputs;
+      case ir_var_out:
+	 return (var->location == -1) ? this->lower_temps : this->lower_outputs;
+      case ir_var_inout:
+	 return this->lower_temps;
+      }
+
+      assert(!"Should not get here.");
+   }
+
    ir_variable *convert_dereference_array(ir_dereference_array *orig_deref,
 					  ir_rvalue* value)
    {
@@ -276,8 +314,7 @@ public:
          return;
 
       ir_dereference_array* orig_deref = (*pir)->as_dereference_array();
-      if (orig_deref && !orig_deref->array_index->as_constant()
-	  && is_array_or_matrix(orig_deref->array)) {
+      if (needs_lowering(orig_deref)) {
          ir_variable* var = convert_dereference_array(orig_deref, 0);
          assert(var);
          *pir = new(talloc_parent(base_ir)) ir_dereference_variable(var);
@@ -292,8 +329,7 @@ public:
 
       ir_dereference_array *orig_deref = ir->lhs->as_dereference_array();
 
-      if (orig_deref && !orig_deref->array_index->as_constant()
-	  && is_array_or_matrix(orig_deref->array)) {
+      if (needs_lowering(orig_deref)) {
          convert_dereference_array(orig_deref, ir->rhs);
          ir->remove();
          this->progress = true;
@@ -304,9 +340,16 @@ public:
 };
 
 bool
-lower_variable_index_to_cond_assign(exec_list *instructions)
+lower_variable_index_to_cond_assign(exec_list *instructions,
+				    bool lower_input,
+				    bool lower_output,
+				    bool lower_temp,
+				    bool lower_uniform)
 {
-   variable_index_to_cond_assign_visitor v;
+   variable_index_to_cond_assign_visitor v(lower_input,
+					   lower_output,
+					   lower_temp,
+					   lower_uniform);
 
    visit_list_elements(&v, instructions);
 
diff --git a/src/mesa/drivers/dri/i965/brw_context.c b/src/mesa/drivers/dri/i965/brw_context.c
index de78400d42..d08538ec20 100644
--- a/src/mesa/drivers/dri/i965/brw_context.c
+++ b/src/mesa/drivers/dri/i965/brw_context.c
@@ -116,6 +116,13 @@ GLboolean brwCreateContext( int api,
       ctx->ShaderCompilerOptions[i].EmitNVTempInitialization = GL_TRUE;
       ctx->ShaderCompilerOptions[i].EmitNoNoise = GL_TRUE;
       ctx->ShaderCompilerOptions[i].EmitNoMainReturn = GL_TRUE;
+      ctx->ShaderCompilerOptions[i].EmitNoIndirectInput = GL_TRUE;
+      ctx->ShaderCompilerOptions[i].EmitNoIndirectOutput = GL_TRUE;
+
+      ctx->ShaderCompilerOptions[i].EmitNoIndirectUniform =
+	 (i == MESA_SHADER_FRAGMENT);
+      ctx->ShaderCompilerOptions[i].EmitNoIndirectTemp =
+	 (i == MESA_SHADER_FRAGMENT);
    }
 
    ctx->Const.VertexProgram.MaxNativeInstructions = (16 * 1024);
diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h
index 864805af0e..fdf8100c8c 100644
--- a/src/mesa/main/mtypes.h
+++ b/src/mesa/main/mtypes.h
@@ -2191,6 +2191,16 @@ struct gl_shader_compiler_options
    GLboolean EmitNoMainReturn;            /**< Emit CONT/RET opcodes? */
    GLboolean EmitNoNoise;                 /**< Emit NOISE opcodes? */
 
+   /**
+    * \name Forms of indirect addressing the driver cannot do.
+    */
+   /*@{*/
+   GLboolean EmitNoIndirectInput;   /**< No indirect addressing of inputs */
+   GLboolean EmitNoIndirectOutput;  /**< No indirect addressing of outputs */
+   GLboolean EmitNoIndirectTemp;    /**< No indirect addressing of temps */
+   GLboolean EmitNoIndirectUniform; /**< No indirect addressing of constants */
+   /*@}*/
+
    GLuint MaxUnrollIterations;
 
    struct gl_sl_pragmas DefaultPragmas; /**< Default #pragma settings */
diff --git a/src/mesa/program/ir_to_mesa.cpp b/src/mesa/program/ir_to_mesa.cpp
index 227dfdbd50..29b92821f6 100644
--- a/src/mesa/program/ir_to_mesa.cpp
+++ b/src/mesa/program/ir_to_mesa.cpp
@@ -2745,6 +2745,19 @@ _mesa_ir_link_shader(GLcontext *ctx, struct gl_shader_program *prog)
 	 if (options->EmitNoNoise)
 	    progress = lower_noise(ir) || progress;
 
+	 /* If there are forms of indirect addressing that the driver
+	  * cannot handle, perform the lowering pass.
+	  */
+	 if (options->EmitNoIndirectInput || options->EmitNoIndirectOutput
+	     || options->EmitNoIndirectTemp || options->EmitNoIndirectUniform)
+	   progress =
+	     lower_variable_index_to_cond_assign(ir,
+						 options->EmitNoIndirectInput,
+						 options->EmitNoIndirectOutput,
+						 options->EmitNoIndirectTemp,
+						 options->EmitNoIndirectUniform)
+	     || progress;
+
 	 progress = do_vec_index_to_cond_assign(ir) || progress;
       } while (progress);