Home | History | Annotate | Download | only in glsl
      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 class output_read_remover : public ir_hierarchical_visitor {
     41 protected:
     42    /**
     43     * A hash table mapping from the original ir_variable shader outputs
     44     * (ir_var_out mode) to the new temporaries to be used instead.
     45     */
     46    hash_table *replacements;
     47 
     48    void *mem_ctx;
     49 public:
     50    output_read_remover();
     51    ~output_read_remover();
     52    virtual ir_visitor_status visit(class ir_dereference_variable *);
     53    virtual ir_visitor_status visit_leave(class ir_return *);
     54    virtual ir_visitor_status visit_leave(class ir_function_signature *);
     55 };
     56 
     57 /**
     58  * Hash function for the output variables - computes the hash of the name.
     59  * NOTE: We're using the name string to ensure that the hash doesn't depend
     60  * on any random factors, otherwise the output_read_remover could produce
     61  * the random order of the assignments.
     62  *
     63  * NOTE: If you want to reuse this function please take into account that
     64  * generally the names of the variables are non-unique.
     65  */
     66 static unsigned
     67 hash_table_var_hash(const void *key)
     68 {
     69    const ir_variable * var = static_cast<const ir_variable *>(key);
     70    return hash_table_string_hash(var->name);
     71 }
     72 
     73 output_read_remover::output_read_remover()
     74 {
     75    mem_ctx = ralloc_context(NULL);
     76    replacements =
     77       hash_table_ctor(0, hash_table_var_hash, hash_table_pointer_compare);
     78 }
     79 
     80 output_read_remover::~output_read_remover()
     81 {
     82    hash_table_dtor(replacements);
     83    ralloc_free(mem_ctx);
     84 }
     85 
     86 ir_visitor_status
     87 output_read_remover::visit(ir_dereference_variable *ir)
     88 {
     89    if (ir->var->mode != ir_var_out)
     90       return visit_continue;
     91 
     92    ir_variable *temp = (ir_variable *) hash_table_find(replacements, ir->var);
     93 
     94    /* If we don't have an existing temporary, create one. */
     95    if (temp == NULL) {
     96       void *var_ctx = ralloc_parent(ir->var);
     97       temp = new(var_ctx) ir_variable(ir->var->type, ir->var->name,
     98                                       ir_var_temporary);
     99       hash_table_insert(replacements, temp, ir->var);
    100    }
    101 
    102    /* Update the dereference to use the temporary */
    103    ir->var = temp;
    104 
    105    return visit_continue;
    106 }
    107 
    108 /**
    109  * Create an assignment to copy a temporary value back to the actual output.
    110  */
    111 static ir_assignment *
    112 copy(void *ctx, ir_variable *output, ir_variable *temp)
    113 {
    114    ir_dereference_variable *lhs = new(ctx) ir_dereference_variable(output);
    115    ir_dereference_variable *rhs = new(ctx) ir_dereference_variable(temp);
    116    return new(ctx) ir_assignment(lhs, rhs);
    117 }
    118 
    119 /** Insert a copy-back assignment before a "return" statement */
    120 static void
    121 emit_return_copy(const void *key, void *data, void *closure)
    122 {
    123    ir_return *ir = (ir_return *) closure;
    124    ir->insert_before(copy(ir, (ir_variable *) key, (ir_variable *) data));
    125 }
    126 
    127 /** Insert a copy-back assignment at the end of the main() function */
    128 static void
    129 emit_main_copy(const void *key, void *data, void *closure)
    130 {
    131    ir_function_signature *sig = (ir_function_signature *) closure;
    132    sig->body.push_tail(copy(sig, (ir_variable *) key, (ir_variable *) data));
    133 }
    134 
    135 ir_visitor_status
    136 output_read_remover::visit_leave(ir_return *ir)
    137 {
    138    hash_table_call_foreach(replacements, emit_return_copy, ir);
    139    return visit_continue;
    140 }
    141 
    142 ir_visitor_status
    143 output_read_remover::visit_leave(ir_function_signature *sig)
    144 {
    145    if (strcmp(sig->function_name(), "main") != 0)
    146       return visit_continue;
    147 
    148    hash_table_call_foreach(replacements, emit_main_copy, sig);
    149    return visit_continue;
    150 }
    151 
    152 void
    153 lower_output_reads(exec_list *instructions)
    154 {
    155    output_read_remover v;
    156    visit_list_elements(&v, instructions);
    157 }
    158