src/glsl/lower_vec_index_to_cond_assign.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 /**
  25  * \file lower_vec_index_to_cond_assign.cpp
  26  *
  27  * Turns indexing into vector types to a series of conditional moves
  28  * of each channel's swizzle into a temporary.
  29  *
  30  * Most GPUs don't have a native way to do this operation, and this
  31  * works around that.  For drivers using both this pass and
  32  * ir_vec_index_to_swizzle, there's a risk that this pass will happen
  33  * before sufficient constant folding to find that the array index is
  34  * constant.  However, we hope that other optimization passes,
  35  * particularly constant folding of assignment conditions and copy
  36  * propagation, will result in the same code in the end.
  37  */
  38
  39 #include "ir.h"
  40 #include "ir_visitor.h"
  41 #include "ir_optimization.h"
  42 #include "glsl_types.h"
  43
  44 /**
  45  * Visitor class for replacing expressions with ir_constant values.
  46  */
  47
  48 class ir_vec_index_to_cond_assign_visitor : public ir_hierarchical_visitor {
  49 public:
  50    ir_vec_index_to_cond_assign_visitor()
  51    {
  52       progress = false;
  53    }
  54
  55    ir_rvalue *convert_vec_index_to_cond_assign(void *mem_ctx,
  56                                                ir_rvalue *orig_vector,
  57                                                ir_rvalue *orig_index,
  58                                                const glsl_type *type);
  59
  60    ir_rvalue *convert_vector_extract_to_cond_assign(ir_rvalue *ir);
  61
  62    virtual ir_visitor_status visit_enter(ir_expression *);
  63    virtual ir_visitor_status visit_enter(ir_swizzle *);
  64    virtual ir_visitor_status visit_leave(ir_assignment *);
  65    virtual ir_visitor_status visit_enter(ir_return *);
  66    virtual ir_visitor_status visit_enter(ir_call *);
  67    virtual ir_visitor_status visit_enter(ir_if *);
  68
  69    bool progress;
  70 };
  71
  72 ir_rvalue *
  73 ir_vec_index_to_cond_assign_visitor::convert_vec_index_to_cond_assign(void *mem_ctx,
  74                                                                       ir_rvalue *orig_vector,
  75                                                                       ir_rvalue *orig_index,
  76                                                                       const glsl_type *type)
  77 {
  78    ir_assignment *assign, *value_assign;
  79    ir_variable *index, *var, *value;
  80    ir_dereference *deref, *deref_value;
  81    unsigned i;
  82
  83
  84    exec_list list;
  85
  86    /* Store the index to a temporary to avoid reusing its tree. */
  87    index = new(base_ir) ir_variable(glsl_type::int_type,
  88                                     "vec_index_tmp_i",
  89                                     ir_var_temporary);
  90    list.push_tail(index);
  91    deref = new(base_ir) ir_dereference_variable(index);
  92    assign = new(base_ir) ir_assignment(deref, orig_index, NULL);
  93    list.push_tail(assign);
  94
  95    /* Store the value inside a temp, thus avoiding matrixes duplication */
  96    value = new(base_ir) ir_variable(orig_vector->type, "vec_value_tmp",
  97                                     ir_var_temporary);
  98    list.push_tail(value);
  99    deref_value = new(base_ir) ir_dereference_variable(value);
 100    value_assign = new(base_ir) ir_assignment(deref_value, orig_vector);
 101    list.push_tail(value_assign);
 102
 103    /* Temporary where we store whichever value we swizzle out. */
 104    var = new(base_ir) ir_variable(type, "vec_index_tmp_v",
 105                                   ir_var_temporary);
 106    list.push_tail(var);
 107
 108    /* Generate a single comparison condition "mask" for all of the components
 109     * in the vector.
 110     */
 111    ir_rvalue *const cond_deref =
 112       compare_index_block(&list, index, 0,
 113                           orig_vector->type->vector_elements,
 114                           mem_ctx);
 115
 116    /* Generate a conditional move of each vector element to the temp. */
 117    for (i = 0; i < orig_vector->type->vector_elements; i++) {
 118       ir_rvalue *condition_swizzle =
 119          new(base_ir) ir_swizzle(cond_deref->clone(mem_ctx, NULL),
 120                                  i, 0, 0, 0, 1);
 121
 122       /* Just clone the rest of the deref chain when trying to get at the
 123        * underlying variable.
 124        */
 125       ir_rvalue *swizzle =
 126          new(base_ir) ir_swizzle(deref_value->clone(mem_ctx, NULL),
 127                                  i, 0, 0, 0, 1);
 128
 129       deref = new(base_ir) ir_dereference_variable(var);
 130       assign = new(base_ir) ir_assignment(deref, swizzle, condition_swizzle);
 131       list.push_tail(assign);
 132    }
 133
 134    /* Put all of the new instructions in the IR stream before the old
 135     * instruction.
 136     */
 137    base_ir->insert_before(&list);
 138
 139    this->progress = true;
 140    return new(base_ir) ir_dereference_variable(var);
 141 }
 142
 143 ir_rvalue *
 144 ir_vec_index_to_cond_assign_visitor::convert_vector_extract_to_cond_assign(ir_rvalue *ir)
 145 {
 146    ir_expression *const expr = ir->as_expression();
 147
 148    if (expr == NULL || expr->operation != ir_binop_vector_extract)
 149       return ir;
 150
 151    return convert_vec_index_to_cond_assign(ralloc_parent(ir),
 152                                            expr->operands[0],
 153                                            expr->operands[1],
 154                                            ir->type);
 155 }
 156
 157 ir_visitor_status
 158 ir_vec_index_to_cond_assign_visitor::visit_enter(ir_expression *ir)
 159 {
 160    unsigned int i;
 161
 162    for (i = 0; i < ir->get_num_operands(); i++) {
 163       ir->operands[i] = convert_vector_extract_to_cond_assign(ir->operands[i]);
 164    }
 165
 166    return visit_continue;
 167 }
 168
 169 ir_visitor_status
 170 ir_vec_index_to_cond_assign_visitor::visit_enter(ir_swizzle *ir)
 171 {
 172    /* Can't be hit from normal GLSL, since you can't swizzle a scalar (which
 173     * the result of indexing a vector is.  But maybe at some point we'll end up
 174     * using swizzling of scalars for vector construction.
 175     */
 176    ir->val = convert_vector_extract_to_cond_assign(ir->val);
 177
 178    return visit_continue;
 179 }
 180
 181 ir_visitor_status
 182 ir_vec_index_to_cond_assign_visitor::visit_leave(ir_assignment *ir)
 183 {
 184    ir->rhs = convert_vector_extract_to_cond_assign(ir->rhs);
 185
 186    if (ir->condition) {
 187       ir->condition = convert_vector_extract_to_cond_assign(ir->condition);
 188    }
 189
 190    return visit_continue;
 191 }
 192
 193 ir_visitor_status
 194 ir_vec_index_to_cond_assign_visitor::visit_enter(ir_call *ir)
 195 {
 196    foreach_iter(exec_list_iterator, iter, *ir) {
 197       ir_rvalue *param = (ir_rvalue *)iter.get();
 198       ir_rvalue *new_param = convert_vector_extract_to_cond_assign(param);
 199
 200       if (new_param != param) {
 201          param->replace_with(new_param);
 202       }
 203    }
 204
 205    return visit_continue;
 206 }
 207
 208 ir_visitor_status
 209 ir_vec_index_to_cond_assign_visitor::visit_enter(ir_return *ir)
 210 {
 211    if (ir->value) {
 212       ir->value = convert_vector_extract_to_cond_assign(ir->value);
 213    }
 214
 215    return visit_continue;
 216 }
 217
 218 ir_visitor_status
 219 ir_vec_index_to_cond_assign_visitor::visit_enter(ir_if *ir)
 220 {
 221    ir->condition = convert_vector_extract_to_cond_assign(ir->condition);
 222
 223    return visit_continue;
 224 }
 225
 226 bool
 227 do_vec_index_to_cond_assign(exec_list *instructions)
 228 {
 229    ir_vec_index_to_cond_assign_visitor v;
 230
 231    visit_list_elements(&v, instructions);
 232
 233    return v.progress;
 234 }