src/glsl/lower_output_reads.cpp

   1 /*
   2  * Copyright © 2012 Vincent Lejeune
   3  * Copyright © 2012 Intel Corporation
   4  *
   5  * Permission is hereby granted, free of charge, to any person obtaining a
   6  * copy of this software and associated documentation files (the "Software"),
   7  * to deal in the Software without restriction, including without limitation
   8  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
   9  * and/or sell copies of the Software, and to permit persons to whom the
  10  * Software is furnished to do so, subject to the following conditions:
  11  *
  12  * The above copyright notice and this permission notice (including the next
  13  * paragraph) shall be included in all copies or substantial portions of the
  14  * Software.
  15  *
  16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  21  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
  22  * DEALINGS IN THE SOFTWARE.
  23  */
  24
  25 #include "ir.h"
  26 #include "program/hash_table.h"
  27
  28 /**
  29  * \file lower_output_reads.cpp
  30  *
  31  * In GLSL, shader output variables (such as varyings) can be both read and
  32  * written.  However, on some hardware, reading an output register causes
  33  * trouble.
  34  *
  35  * This pass creates temporary shadow copies of every (used) shader output,
  36  * and replaces all accesses to use those instead.  It also adds code to the
  37  * main() function to copy the final values to the actual shader outputs.
  38  */
  39
  40 namespace {
  41
  42 class output_read_remover : public ir_hierarchical_visitor {
  43 protected:
  44    /**
  45     * A hash table mapping from the original ir_variable shader outputs
  46     * (ir_var_shader_out mode) to the new temporaries to be used instead.
  47     */
  48    hash_table *replacements;
  49
  50    void *mem_ctx;
  51 public:
  52    output_read_remover();
  53    ~output_read_remover();
  54    virtual ir_visitor_status visit(class ir_dereference_variable *);
  55    virtual ir_visitor_status visit_leave(class ir_emit_vertex *);
  56    virtual ir_visitor_status visit_leave(class ir_return *);
  57    virtual ir_visitor_status visit_leave(class ir_function_signature *);
  58 };
  59
  60 } /* anonymous namespace */
  61
  62 /**
  63  * Hash function for the output variables - computes the hash of the name.
  64  * NOTE: We're using the name string to ensure that the hash doesn't depend
  65  * on any random factors, otherwise the output_read_remover could produce
  66  * the random order of the assignments.
  67  *
  68  * NOTE: If you want to reuse this function please take into account that
  69  * generally the names of the variables are non-unique.
  70  */
  71 static unsigned
  72 hash_table_var_hash(const void *key)
  73 {
  74    const ir_variable * var = static_cast<const ir_variable *>(key);
  75    return hash_table_string_hash(var->name);
  76 }
  77
  78 output_read_remover::output_read_remover()
  79 {
  80    mem_ctx = ralloc_context(NULL);
  81    replacements =
  82       hash_table_ctor(0, hash_table_var_hash, hash_table_pointer_compare);
  83 }
  84
  85 output_read_remover::~output_read_remover()
  86 {
  87    hash_table_dtor(replacements);
  88    ralloc_free(mem_ctx);
  89 }
  90
  91 ir_visitor_status
  92 output_read_remover::visit(ir_dereference_variable *ir)
  93 {
  94    if (ir->var->data.mode != ir_var_shader_out)
  95       return visit_continue;
  96
  97    ir_variable *temp = (ir_variable *) hash_table_find(replacements, ir->var);
  98
  99    /* If we don't have an existing temporary, create one. */
 100    if (temp == NULL) {
 101       void *var_ctx = ralloc_parent(ir->var);
 102       temp = new(var_ctx) ir_variable(ir->var->type, ir->var->name,
 103                                       ir_var_temporary);
 104       hash_table_insert(replacements, temp, ir->var);
 105       ir->var->insert_after(temp);
 106    }
 107
 108    /* Update the dereference to use the temporary */
 109    ir->var = temp;
 110
 111    return visit_continue;
 112 }
 113
 114 /**
 115  * Create an assignment to copy a temporary value back to the actual output.
 116  */
 117 static ir_assignment *
 118 copy(void *ctx, ir_variable *output, ir_variable *temp)
 119 {
 120    ir_dereference_variable *lhs = new(ctx) ir_dereference_variable(output);
 121    ir_dereference_variable *rhs = new(ctx) ir_dereference_variable(temp);
 122    return new(ctx) ir_assignment(lhs, rhs);
 123 }
 124
 125 /** Insert a copy-back assignment before a "return" statement or a call to
 126  * EmitVertex().
 127  */
 128 static void
 129 emit_return_copy(const void *key, void *data, void *closure)
 130 {
 131    ir_return *ir = (ir_return *) closure;
 132    ir->insert_before(copy(ir, (ir_variable *) key, (ir_variable *) data));
 133 }
 134
 135 /** Insert a copy-back assignment at the end of the main() function */
 136 static void
 137 emit_main_copy(const void *key, void *data, void *closure)
 138 {
 139    ir_function_signature *sig = (ir_function_signature *) closure;
 140    sig->body.push_tail(copy(sig, (ir_variable *) key, (ir_variable *) data));
 141 }
 142
 143 ir_visitor_status
 144 output_read_remover::visit_leave(ir_return *ir)
 145 {
 146    hash_table_call_foreach(replacements, emit_return_copy, ir);
 147    return visit_continue;
 148 }
 149
 150 ir_visitor_status
 151 output_read_remover::visit_leave(ir_emit_vertex *ir)
 152 {
 153    hash_table_call_foreach(replacements, emit_return_copy, ir);
 154    hash_table_clear(replacements);
 155    return visit_continue;
 156 }
 157
 158 ir_visitor_status
 159 output_read_remover::visit_leave(ir_function_signature *sig)
 160 {
 161    if (strcmp(sig->function_name(), "main") != 0)
 162       return visit_continue;
 163
 164    hash_table_call_foreach(replacements, emit_main_copy, sig);
 165    return visit_continue;
 166 }
 167
 168 void
 169 lower_output_reads(exec_list *instructions)
 170 {
 171    output_read_remover v;
 172    visit_list_elements(&v, instructions);
 173 }