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 namespace {
  45
  46 /**
  47  * Visitor class for replacing expressions with ir_constant values.
  48  */
  49
  50 class ir_vec_index_to_cond_assign_visitor : public ir_hierarchical_visitor {
  51 public:
  52    ir_vec_index_to_cond_assign_visitor()
  53    {
  54       progress = false;
  55    }
  56
  57    ir_rvalue *convert_vec_index_to_cond_assign(void *mem_ctx,
  58                                                ir_rvalue *orig_vector,
  59                                                ir_rvalue *orig_index,
  60                                                const glsl_type *type);
  61
  62    ir_rvalue *convert_vector_extract_to_cond_assign(ir_rvalue *ir);
  63
  64    virtual ir_visitor_status visit_enter(ir_expression *);
  65    virtual ir_visitor_status visit_enter(ir_swizzle *);
  66    virtual ir_visitor_status visit_leave(ir_assignment *);
  67    virtual ir_visitor_status visit_enter(ir_return *);
  68    virtual ir_visitor_status visit_enter(ir_call *);
  69    virtual ir_visitor_status visit_enter(ir_if *);
  70
  71    bool progress;
  72 };
  73
  74 } /* anonymous namespace */
  75
  76 ir_rvalue *
  77 ir_vec_index_to_cond_assign_visitor::convert_vec_index_to_cond_assign(void *mem_ctx,
  78                                                                       ir_rvalue *orig_vector,
  79                                                                       ir_rvalue *orig_index,
  80                                                                       const glsl_type *type)
  81 {
  82    ir_assignment *assign, *value_assign;
  83    ir_variable *index, *var, *value;
  84    ir_dereference *deref, *deref_value;
  85    unsigned i;
  86
  87
  88    exec_list list;
  89
  90    /* Store the index to a temporary to avoid reusing its tree. */
  91    index = new(base_ir) ir_variable(glsl_type::int_type,
  92                                     "vec_index_tmp_i",
  93                                     ir_var_temporary);
  94    list.push_tail(index);
  95    deref = new(base_ir) ir_dereference_variable(index);
  96    assign = new(base_ir) ir_assignment(deref, orig_index, NULL);
  97    list.push_tail(assign);
  98
  99    /* Store the value inside a temp, thus avoiding matrixes duplication */
 100    value = new(base_ir) ir_variable(orig_vector->type, "vec_value_tmp",
 101                                     ir_var_temporary);
 102    list.push_tail(value);
 103    deref_value = new(base_ir) ir_dereference_variable(value);
 104    value_assign = new(base_ir) ir_assignment(deref_value, orig_vector);
 105    list.push_tail(value_assign);
 106
 107    /* Temporary where we store whichever value we swizzle out. */
 108    var = new(base_ir) ir_variable(type, "vec_index_tmp_v",
 109                                   ir_var_temporary);
 110    list.push_tail(var);
 111
 112    /* Generate a single comparison condition "mask" for all of the components
 113     * in the vector.
 114     */
 115    ir_rvalue *const cond_deref =
 116       compare_index_block(&list, index, 0,
 117                           orig_vector->type->vector_elements,
 118                           mem_ctx);
 119
 120    /* Generate a conditional move of each vector element to the temp. */
 121    for (i = 0; i < orig_vector->type->vector_elements; i++) {
 122       ir_rvalue *condition_swizzle =
 123          new(base_ir) ir_swizzle(cond_deref->clone(mem_ctx, NULL),
 124                                  i, 0, 0, 0, 1);
 125
 126       /* Just clone the rest of the deref chain when trying to get at the
 127        * underlying variable.
 128        */
 129       ir_rvalue *swizzle =
 130          new(base_ir) ir_swizzle(deref_value->clone(mem_ctx, NULL),
 131                                  i, 0, 0, 0, 1);
 132
 133       deref = new(base_ir) ir_dereference_variable(var);
 134       assign = new(base_ir) ir_assignment(deref, swizzle, condition_swizzle);
 135       list.push_tail(assign);
 136    }
 137
 138    /* Put all of the new instructions in the IR stream before the old
 139     * instruction.
 140     */
 141    base_ir->insert_before(&list);
 142
 143    this->progress = true;
 144    return new(base_ir) ir_dereference_variable(var);
 145 }
 146
 147 ir_rvalue *
 148 ir_vec_index_to_cond_assign_visitor::convert_vector_extract_to_cond_assign(ir_rvalue *ir)
 149 {
 150    ir_expression *const expr = ir->as_expression();
 151
 152    if (expr == NULL || expr->operation != ir_binop_vector_extract)
 153       return ir;
 154
 155    return convert_vec_index_to_cond_assign(ralloc_parent(ir),
 156                                            expr->operands[0],
 157                                            expr->operands[1],
 158                                            ir->type);
 159 }
 160
 161 ir_visitor_status
 162 ir_vec_index_to_cond_assign_visitor::visit_enter(ir_expression *ir)
 163 {
 164    unsigned int i;
 165
 166    for (i = 0; i < ir->get_num_operands(); i++) {
 167       ir->operands[i] = convert_vector_extract_to_cond_assign(ir->operands[i]);
 168    }
 169
 170    return visit_continue;
 171 }
 172
 173 ir_visitor_status
 174 ir_vec_index_to_cond_assign_visitor::visit_enter(ir_swizzle *ir)
 175 {
 176    /* Can't be hit from normal GLSL, since you can't swizzle a scalar (which
 177     * the result of indexing a vector is.  But maybe at some point we'll end up
 178     * using swizzling of scalars for vector construction.
 179     */
 180    ir->val = convert_vector_extract_to_cond_assign(ir->val);
 181
 182    return visit_continue;
 183 }
 184
 185 ir_visitor_status
 186 ir_vec_index_to_cond_assign_visitor::visit_leave(ir_assignment *ir)
 187 {
 188    ir->rhs = convert_vector_extract_to_cond_assign(ir->rhs);
 189
 190    if (ir->condition) {
 191       ir->condition = convert_vector_extract_to_cond_assign(ir->condition);
 192    }
 193
 194    return visit_continue;
 195 }
 196
 197 ir_visitor_status
 198 ir_vec_index_to_cond_assign_visitor::visit_enter(ir_call *ir)
 199 {
 200    foreach_iter(exec_list_iterator, iter, *ir) {
 201       ir_rvalue *param = (ir_rvalue *)iter.get();
 202       ir_rvalue *new_param = convert_vector_extract_to_cond_assign(param);
 203
 204       if (new_param != param) {
 205          param->replace_with(new_param);
 206       }
 207    }
 208
 209    return visit_continue;
 210 }
 211
 212 ir_visitor_status
 213 ir_vec_index_to_cond_assign_visitor::visit_enter(ir_return *ir)
 214 {
 215    if (ir->value) {
 216       ir->value = convert_vector_extract_to_cond_assign(ir->value);
 217    }
 218
 219    return visit_continue;
 220 }
 221
 222 ir_visitor_status
 223 ir_vec_index_to_cond_assign_visitor::visit_enter(ir_if *ir)
 224 {
 225    ir->condition = convert_vector_extract_to_cond_assign(ir->condition);
 226
 227    return visit_continue;
 228 }
 229
 230 bool
 231 do_vec_index_to_cond_assign(exec_list *instructions)
 232 {
 233    ir_vec_index_to_cond_assign_visitor v;
 234
 235    visit_list_elements(&v, instructions);
 236
 237    return v.progress;
 238 }