src/glsl/opt_flip_matrices.cpp

   1 /*
   2  * Copyright © 2013 Intel Corporation
   3  *
   4  * Permission is hereby granted, free of charge, to any person obtaining a
   5  * copy of this software and associated documentation files (the "Software"),
   6  * to deal in the Software without restriction, including without limitation
   7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
   8  * and/or sell copies of the Software, and to permit persons to whom the
   9  * Software is furnished to do so, subject to the following conditions:
  10  *
  11  * The above copyright notice and this permission notice (including the next
  12  * paragraph) shall be included in all copies or substantial portions of the
  13  * Software.
  14  *
  15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  18  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  20  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
  21  * DEALINGS IN THE SOFTWARE.
  22  */
  23
  24 /**
  25  * \file opt_flip_matrices.cpp
  26  *
  27  * Convert (matrix * vector) operations to (vector * matrixTranspose),
  28  * which can be done using dot products rather than multiplies and adds.
  29  * On some hardware, this is more efficient.
  30  *
  31  * This currently only does the conversion for built-in matrices which
  32  * already have transposed equivalents.  Namely, gl_ModelViewProjectionMatrix
  33  * and gl_TextureMatrix.
  34  */
  35 #include "ir.h"
  36 #include "ir_optimization.h"
  37 #include "main/macros.h"
  38
  39 namespace {
  40 class matrix_flipper : public ir_hierarchical_visitor {
  41 public:
  42    matrix_flipper(exec_list *instructions)
  43    {
  44       progress = false;
  45       mvp_transpose = NULL;
  46       texmat_transpose = NULL;
  47
  48       foreach_list(n, instructions) {
  49          ir_instruction *ir = (ir_instruction *) n;
  50          ir_variable *var = ir->as_variable();
  51          if (!var)
  52             continue;
  53          if (strcmp(var->name, "gl_ModelViewProjectionMatrixTranspose") == 0)
  54             mvp_transpose = var;
  55          if (strcmp(var->name, "gl_TextureMatrixTranspose") == 0)
  56             texmat_transpose = var;
  57       }
  58    }
  59
  60    ir_visitor_status visit_enter(ir_expression *ir);
  61
  62    bool progress;
  63
  64 private:
  65    ir_variable *mvp_transpose;
  66    ir_variable *texmat_transpose;
  67 };
  68 }
  69
  70 ir_visitor_status
  71 matrix_flipper::visit_enter(ir_expression *ir)
  72 {
  73    if (ir->operation != ir_binop_mul ||
  74        !ir->operands[0]->type->is_matrix() ||
  75        !ir->operands[1]->type->is_vector())
  76       return visit_continue;
  77
  78    ir_variable *mat_var = ir->operands[0]->variable_referenced();
  79    if (!mat_var)
  80       return visit_continue;
  81
  82    if (mvp_transpose &&
  83        strcmp(mat_var->name, "gl_ModelViewProjectionMatrix") == 0) {
  84 #ifndef NDEBUG
  85       ir_dereference_variable *deref = ir->operands[0]->as_dereference_variable();
  86       assert(deref && deref->var == mat_var);
  87 #endif
  88
  89       void *mem_ctx = ralloc_parent(ir);
  90
  91       ir->operands[0] = ir->operands[1];
  92       ir->operands[1] = new(mem_ctx) ir_dereference_variable(mvp_transpose);
  93
  94       progress = true;
  95    } else if (texmat_transpose &&
  96               strcmp(mat_var->name, "gl_TextureMatrix") == 0) {
  97       ir_dereference_array *array_ref = ir->operands[0]->as_dereference_array();
  98       assert(array_ref != NULL);
  99       ir_dereference_variable *var_ref = array_ref->array->as_dereference_variable();
 100       assert(var_ref && var_ref->var == mat_var);
 101
 102       ir->operands[0] = ir->operands[1];
 103       ir->operands[1] = array_ref;
 104
 105       var_ref->var = texmat_transpose;
 106
 107       texmat_transpose->max_array_access =
 108          MAX2(texmat_transpose->max_array_access, mat_var->max_array_access);
 109
 110       progress = true;
 111    }
 112
 113    return visit_continue;
 114 }
 115
 116 bool
 117 opt_flip_matrices(struct exec_list *instructions)
 118 {
 119    matrix_flipper v(instructions);
 120
 121    visit_list_elements(&v, instructions);
 122
 123    return v.progress;
 124 }