src/compiler/nir/nir_lower_io_to_temporaries.c

   1 /*
   2  * Copyright © 2015 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 DEALINGS
  21  * IN THE SOFTWARE.
  22  */
  23
  24 /*
  25  * Implements a pass that lowers output and/or input variables to a
  26  * temporary plus an output variable with a single copy at each exit
  27  * point of the shader and/or an input variable with a single copy
  28  * at the entrance point of the shader.  This way the output variable
  29  * is only ever written once and/or input is only read once, and there
  30  * are no indirect outut/input accesses.
  31  */
  32
  33 #include "nir.h"
  34 #include "nir_builder.h"
  35
  36 struct lower_io_state {
  37    nir_shader *shader;
  38    nir_function_impl *entrypoint;
  39    struct exec_list old_outputs;
  40    struct exec_list old_inputs;
  41 };
  42
  43 static void
  44 emit_copies(nir_builder *b, struct exec_list *dest_vars,
  45             struct exec_list *src_vars)
  46 {
  47    assert(exec_list_length(dest_vars) == exec_list_length(src_vars));
  48
  49    foreach_two_lists(dest_node, dest_vars, src_node, src_vars) {
  50       nir_variable *dest = exec_node_data(nir_variable, dest_node, node);
  51       nir_variable *src = exec_node_data(nir_variable, src_node, node);
  52
  53       /* No need to copy the contents of a non-fb_fetch_output output variable
  54        * to the temporary allocated for it, since its initial value is
  55        * undefined.
  56        */
  57       if (src->data.mode == nir_var_shader_out &&
  58           !src->data.fb_fetch_output)
  59          continue;
  60
  61       /* Can't copy the contents of the temporary back to a read-only
  62        * interface variable.  The value of the temporary won't have been
  63        * modified by the shader anyway.
  64        */
  65       if (dest->data.read_only)
  66          continue;
  67
  68       nir_copy_var(b, dest, src);
  69    }
  70 }
  71
  72 static void
  73 emit_output_copies_impl(struct lower_io_state *state, nir_function_impl *impl)
  74 {
  75    nir_builder b;
  76    nir_builder_init(&b, impl);
  77
  78    if (state->shader->info.stage == MESA_SHADER_GEOMETRY) {
  79       /* For geometry shaders, we have to emit the output copies right
  80        * before each EmitVertex call.
  81        */
  82       nir_foreach_block(block, impl) {
  83          nir_foreach_instr(instr, block) {
  84             if (instr->type != nir_instr_type_intrinsic)
  85                continue;
  86
  87             nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
  88             if (intrin->intrinsic == nir_intrinsic_emit_vertex) {
  89                b.cursor = nir_before_instr(&intrin->instr);
  90                emit_copies(&b, &state->shader->outputs, &state->old_outputs);
  91             }
  92          }
  93       }
  94    } else if (impl == state->entrypoint) {
  95       b.cursor = nir_before_block(nir_start_block(impl));
  96       emit_copies(&b, &state->old_outputs, &state->shader->outputs);
  97
  98       /* For all other shader types, we need to do the copies right before
  99        * the jumps to the end block.
 100        */
 101       set_foreach(impl->end_block->predecessors, block_entry) {
 102          struct nir_block *block = (void *)block_entry->key;
 103          b.cursor = nir_after_block_before_jump(block);
 104          emit_copies(&b, &state->shader->outputs, &state->old_outputs);
 105       }
 106    }
 107 }
 108
 109 static void
 110 emit_input_copies_impl(struct lower_io_state *state, nir_function_impl *impl)
 111 {
 112    if (impl == state->entrypoint) {
 113       nir_builder b;
 114       nir_builder_init(&b, impl);
 115       b.cursor = nir_before_block(nir_start_block(impl));
 116       emit_copies(&b, &state->old_inputs, &state->shader->inputs);
 117    }
 118 }
 119
 120 static nir_variable *
 121 create_shadow_temp(struct lower_io_state *state, nir_variable *var)
 122 {
 123    nir_variable *nvar = ralloc(state->shader, nir_variable);
 124    memcpy(nvar, var, sizeof *nvar);
 125
 126    /* The original is now the temporary */
 127    nir_variable *temp = var;
 128
 129    /* Reparent the name to the new variable */
 130    ralloc_steal(nvar, nvar->name);
 131
 132    assert(nvar->constant_initializer == NULL);
 133
 134    /* Give the original a new name with @<mode>-temp appended */
 135    const char *mode = (temp->data.mode == nir_var_shader_in) ? "in" : "out";
 136    temp->name = ralloc_asprintf(var, "%s@%s-temp", mode, nvar->name);
 137    temp->data.mode = nir_var_global;
 138    temp->data.read_only = false;
 139    temp->data.fb_fetch_output = false;
 140    temp->data.compact = false;
 141
 142    return nvar;
 143 }
 144
 145 void
 146 nir_lower_io_to_temporaries(nir_shader *shader, nir_function_impl *entrypoint,
 147                             bool outputs, bool inputs)
 148 {
 149    struct lower_io_state state;
 150
 151    if (shader->info.stage == MESA_SHADER_TESS_CTRL)
 152       return;
 153
 154    state.shader = shader;
 155    state.entrypoint = entrypoint;
 156
 157    if (inputs)
 158       exec_list_move_nodes_to(&shader->inputs, &state.old_inputs);
 159    else
 160       exec_list_make_empty(&state.old_inputs);
 161
 162    if (outputs)
 163       exec_list_move_nodes_to(&shader->outputs, &state.old_outputs);
 164    else
 165       exec_list_make_empty(&state.old_outputs);
 166
 167    /* Walk over all of the outputs turn each output into a temporary and
 168     * make a new variable for the actual output.
 169     */
 170    nir_foreach_variable(var, &state.old_outputs) {
 171       nir_variable *output = create_shadow_temp(&state, var);
 172       exec_list_push_tail(&shader->outputs, &output->node);
 173    }
 174
 175    /* and same for inputs: */
 176    nir_foreach_variable(var, &state.old_inputs) {
 177       nir_variable *input = create_shadow_temp(&state, var);
 178       exec_list_push_tail(&shader->inputs, &input->node);
 179    }
 180
 181    nir_foreach_function(function, shader) {
 182       if (function->impl == NULL)
 183          continue;
 184
 185       if (inputs)
 186          emit_input_copies_impl(&state, function->impl);
 187
 188       if (outputs)
 189          emit_output_copies_impl(&state, function->impl);
 190
 191       nir_metadata_preserve(function->impl, nir_metadata_block_index |
 192                                             nir_metadata_dominance);
 193    }
 194
 195    exec_list_append(&shader->globals, &state.old_inputs);
 196    exec_list_append(&shader->globals, &state.old_outputs);
 197
 198    nir_fixup_deref_modes(shader);
 199 }