src/glsl/loop_controls.cpp

   1 /*
   2  * Copyright © 2010 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 #include <limits.h>
  25 #include "main/compiler.h"
  26 #include "glsl_types.h"
  27 #include "loop_analysis.h"
  28 #include "ir_hierarchical_visitor.h"
  29
  30 /**
  31  * Find an initializer of a variable outside a loop
  32  *
  33  * Works backwards from the loop to find the pre-loop value of the variable.
  34  * This is used, for example, to find the initial value of loop induction
  35  * variables.
  36  *
  37  * \param loop  Loop where \c var is an induction variable
  38  * \param var   Variable whose initializer is to be found
  39  *
  40  * \return
  41  * The \c ir_rvalue assigned to the variable outside the loop.  May return
  42  * \c NULL if no initializer can be found.
  43  */
  44 ir_rvalue *
  45 find_initial_value(ir_loop *loop, ir_variable *var)
  46 {
  47    for (exec_node *node = loop->prev;
  48         !node->is_head_sentinel();
  49         node = node->prev) {
  50       ir_instruction *ir = (ir_instruction *) node;
  51
  52       switch (ir->ir_type) {
  53       case ir_type_call:
  54       case ir_type_loop:
  55       case ir_type_loop_jump:
  56       case ir_type_return:
  57       case ir_type_if:
  58          return NULL;
  59
  60       case ir_type_function:
  61       case ir_type_function_signature:
  62          assert(!"Should not get here.");
  63          return NULL;
  64
  65       case ir_type_assignment: {
  66          ir_assignment *assign = ir->as_assignment();
  67          ir_variable *assignee = assign->lhs->whole_variable_referenced();
  68
  69          if (assignee == var)
  70             return (assign->condition != NULL) ? NULL : assign->rhs;
  71
  72          break;
  73       }
  74
  75       default:
  76          break;
  77       }
  78    }
  79
  80    return NULL;
  81 }
  82
  83
  84 int
  85 calculate_iterations(ir_rvalue *from, ir_rvalue *to, ir_rvalue *increment,
  86                      enum ir_expression_operation op)
  87 {
  88    if (from == NULL || to == NULL || increment == NULL)
  89       return -1;
  90
  91    void *mem_ctx = ralloc_context(NULL);
  92
  93    ir_expression *const sub =
  94       new(mem_ctx) ir_expression(ir_binop_sub, from->type, to, from);
  95
  96    ir_expression *const div =
  97       new(mem_ctx) ir_expression(ir_binop_div, sub->type, sub, increment);
  98
  99    ir_constant *iter = div->constant_expression_value();
 100
 101    if (iter == NULL)
 102       return -1;
 103
 104    if (!iter->type->is_integer()) {
 105       ir_rvalue *cast =
 106          new(mem_ctx) ir_expression(ir_unop_f2i, glsl_type::int_type, iter,
 107                                     NULL);
 108
 109       iter = cast->constant_expression_value();
 110    }
 111
 112    int iter_value = iter->get_int_component(0);
 113
 114    /* Make sure that the calculated number of iterations satisfies the exit
 115     * condition.  This is needed to catch off-by-one errors and some types of
 116     * ill-formed loops.  For example, we need to detect that the following
 117     * loop does not have a maximum iteration count.
 118     *
 119     *    for (float x = 0.0; x != 0.9; x += 0.2)
 120     *        ;
 121     */
 122    const int bias[] = { -1, 0, 1 };
 123    bool valid_loop = false;
 124
 125    for (unsigned i = 0; i < Elements(bias); i++) {
 126       /* Increment may be of type int, uint or float. */
 127       switch (increment->type->base_type) {
 128       case GLSL_TYPE_INT:
 129          iter = new(mem_ctx) ir_constant(iter_value + bias[i]);
 130          break;
 131       case GLSL_TYPE_UINT:
 132          iter = new(mem_ctx) ir_constant(unsigned(iter_value + bias[i]));
 133          break;
 134       case GLSL_TYPE_FLOAT:
 135          iter = new(mem_ctx) ir_constant(float(iter_value + bias[i]));
 136          break;
 137       default:
 138           unreachable(!"Unsupported type for loop iterator.");
 139       }
 140
 141       ir_expression *const mul =
 142          new(mem_ctx) ir_expression(ir_binop_mul, increment->type, iter,
 143                                     increment);
 144
 145       ir_expression *const add =
 146          new(mem_ctx) ir_expression(ir_binop_add, mul->type, mul, from);
 147
 148       ir_expression *const cmp =
 149          new(mem_ctx) ir_expression(op, glsl_type::bool_type, add, to);
 150
 151       ir_constant *const cmp_result = cmp->constant_expression_value();
 152
 153       assert(cmp_result != NULL);
 154       if (cmp_result->get_bool_component(0)) {
 155          iter_value += bias[i];
 156          valid_loop = true;
 157          break;
 158       }
 159    }
 160
 161    ralloc_free(mem_ctx);
 162    return (valid_loop) ? iter_value : -1;
 163 }
 164
 165 namespace {
 166
 167 class loop_control_visitor : public ir_hierarchical_visitor {
 168 public:
 169    loop_control_visitor(loop_state *state)
 170    {
 171       this->state = state;
 172       this->progress = false;
 173    }
 174
 175    virtual ir_visitor_status visit_leave(ir_loop *ir);
 176
 177    loop_state *state;
 178
 179    bool progress;
 180 };
 181
 182 } /* anonymous namespace */
 183
 184 ir_visitor_status
 185 loop_control_visitor::visit_leave(ir_loop *ir)
 186 {
 187    loop_variable_state *const ls = this->state->get(ir);
 188
 189    /* If we've entered a loop that hasn't been analyzed, something really,
 190     * really bad has happened.
 191     */
 192    if (ls == NULL) {
 193       assert(ls != NULL);
 194       return visit_continue;
 195    }
 196
 197    if (ls->limiting_terminator != NULL) {
 198       /* If the limiting terminator has an iteration count of zero, then we've
 199        * proven that the loop cannot run, so delete it.
 200        */
 201       int iterations = ls->limiting_terminator->iterations;
 202       if (iterations == 0) {
 203          ir->remove();
 204          this->progress = true;
 205          return visit_continue;
 206       }
 207    }
 208
 209    /* Remove the conditional break statements associated with all terminators
 210     * that are associated with a fixed iteration count, except for the one
 211     * associated with the limiting terminator--that one needs to stay, since
 212     * it terminates the loop.  Exception: if the loop still has a normative
 213     * bound, then that terminates the loop, so we don't even need the limiting
 214     * terminator.
 215     */
 216    foreach_in_list(loop_terminator, t, &ls->terminators) {
 217       if (t->iterations < 0)
 218          continue;
 219
 220       if (t != ls->limiting_terminator) {
 221          t->ir->remove();
 222
 223          assert(ls->num_loop_jumps > 0);
 224          ls->num_loop_jumps--;
 225
 226          this->progress = true;
 227       }
 228    }
 229
 230    return visit_continue;
 231 }
 232
 233
 234 bool
 235 set_loop_controls(exec_list *instructions, loop_state *ls)
 236 {
 237    loop_control_visitor v(ls);
 238
 239    v.run(instructions);
 240
 241    return v.progress;
 242 }