src/compiler/nir/nir_inline_functions.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 #include "nir.h"
  25 #include "nir_builder.h"
  26 #include "nir_control_flow.h"
  27
  28 static bool inline_function_impl(nir_function_impl *impl, struct set *inlined);
  29
  30 static bool
  31 rewrite_param_derefs_block(nir_block *block, nir_call_instr *call)
  32 {
  33    nir_foreach_instr_safe(block, instr) {
  34       if (instr->type != nir_instr_type_intrinsic)
  35          continue;
  36
  37       nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
  38
  39       for (unsigned i = 0;
  40            i < nir_intrinsic_infos[intrin->intrinsic].num_variables; i++) {
  41          if (intrin->variables[i]->var->data.mode != nir_var_param)
  42             continue;
  43
  44          int param_idx = intrin->variables[i]->var->data.location;
  45
  46          nir_deref_var *call_deref;
  47          if (param_idx >= 0) {
  48             assert(param_idx < call->callee->num_params);
  49             call_deref = call->params[param_idx];
  50          } else {
  51             call_deref = call->return_deref;
  52          }
  53          assert(call_deref);
  54
  55          nir_deref_var *new_deref = nir_deref_as_var(nir_copy_deref(intrin, &call_deref->deref));
  56          nir_deref *new_tail = nir_deref_tail(&new_deref->deref);
  57          new_tail->child = intrin->variables[i]->deref.child;
  58          ralloc_steal(new_tail, new_tail->child);
  59          intrin->variables[i] = new_deref;
  60       }
  61    }
  62
  63    return true;
  64 }
  65
  66 static void
  67 lower_param_to_local(nir_variable *param, nir_function_impl *impl, bool write)
  68 {
  69    if (param->data.mode != nir_var_param)
  70       return;
  71
  72    nir_parameter_type param_type;
  73    if (param->data.location >= 0) {
  74       assert(param->data.location < impl->num_params);
  75       param_type = impl->function->params[param->data.location].param_type;
  76    } else {
  77       /* Return variable */
  78       param_type = nir_parameter_out;
  79    }
  80
  81    if ((write && param_type == nir_parameter_in) ||
  82        (!write && param_type == nir_parameter_out)) {
  83       /* In this case, we need a shadow copy.  Turn it into a local */
  84       param->data.mode = nir_var_local;
  85       exec_list_push_tail(&impl->locals, &param->node);
  86    }
  87 }
  88
  89 static bool
  90 lower_params_to_locals_block(nir_block *block, nir_function_impl *impl)
  91 {
  92    nir_foreach_instr_safe(block, instr) {
  93       if (instr->type != nir_instr_type_intrinsic)
  94          continue;
  95
  96       nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
  97
  98       switch (intrin->intrinsic) {
  99       case nir_intrinsic_store_var:
 100          lower_param_to_local(intrin->variables[0]->var, impl, true);
 101          break;
 102
 103       case nir_intrinsic_copy_var:
 104          lower_param_to_local(intrin->variables[0]->var, impl, true);
 105          lower_param_to_local(intrin->variables[1]->var, impl, false);
 106          break;
 107
 108       case nir_intrinsic_load_var:
 109          /* All other intrinsics which access variables (image_load_store)
 110           * do so in a read-only fasion.
 111           */
 112          for (unsigned i = 0;
 113               i < nir_intrinsic_infos[intrin->intrinsic].num_variables; i++) {
 114             lower_param_to_local(intrin->variables[i]->var, impl, false);
 115          }
 116          break;
 117
 118       default:
 119          continue;
 120       }
 121    }
 122
 123    return true;
 124 }
 125
 126 static bool
 127 inline_functions_block(nir_block *block, nir_builder *b,
 128                        struct set *inlined)
 129 {
 130    bool progress = false;
 131    /* This is tricky.  We're iterating over instructions in a block but, as
 132     * we go, the block and its instruction list are being split into
 133     * pieces.  However, this *should* be safe since foreach_safe always
 134     * stashes the next thing in the iteration.  That next thing will
 135     * properly get moved to the next block when it gets split, and we
 136     * continue iterating there.
 137     */
 138    nir_foreach_instr_safe(block, instr) {
 139       if (instr->type != nir_instr_type_call)
 140          continue;
 141
 142       progress = true;
 143
 144       nir_call_instr *call = nir_instr_as_call(instr);
 145       assert(call->callee->impl);
 146
 147       inline_function_impl(call->callee->impl, inlined);
 148
 149       nir_function_impl *callee_copy =
 150          nir_function_impl_clone(call->callee->impl);
 151       callee_copy->function = call->callee;
 152
 153       /* Add copies of all in parameters */
 154       assert(call->num_params == callee_copy->num_params);
 155
 156       exec_list_append(&b->impl->locals, &callee_copy->locals);
 157       exec_list_append(&b->impl->registers, &callee_copy->registers);
 158
 159       b->cursor = nir_before_instr(&call->instr);
 160
 161       /* We now need to tie the two functions together using the
 162        * parameters.  There are two ways we do this: One is to turn the
 163        * parameter into a local variable and do a shadow-copy.  The other
 164        * is to treat the parameter as a "proxy" and rewrite derefs to use
 165        * the actual variable that comes from the call instruction.  We
 166        * implement both schemes.  The first is needed in the case where we
 167        * have an in parameter that we write or similar.  The second case is
 168        * needed for handling things such as images and uniforms properly.
 169        */
 170
 171       /* Figure out when we need to lower to a shadow local */
 172       nir_foreach_block(block, callee_copy) {
 173          lower_params_to_locals_block(block, callee_copy);
 174       }
 175
 176       for (unsigned i = 0; i < callee_copy->num_params; i++) {
 177          nir_variable *param = callee_copy->params[i];
 178
 179          if (param->data.mode == nir_var_local &&
 180              call->callee->params[i].param_type != nir_parameter_out) {
 181             nir_copy_deref_var(b, nir_deref_var_create(b->shader, param),
 182                                   call->params[i]);
 183          }
 184       }
 185
 186       nir_foreach_block(block, callee_copy) {
 187          rewrite_param_derefs_block(block, call);
 188       }
 189
 190       /* Pluck the body out of the function and place it here */
 191       nir_cf_list body;
 192       nir_cf_list_extract(&body, &callee_copy->body);
 193       nir_cf_reinsert(&body, b->cursor);
 194
 195       b->cursor = nir_before_instr(&call->instr);
 196
 197       /* Add copies of all out parameters and the return */
 198       assert(call->num_params == callee_copy->num_params);
 199       for (unsigned i = 0; i < callee_copy->num_params; i++) {
 200          nir_variable *param = callee_copy->params[i];
 201
 202          if (param->data.mode == nir_var_local &&
 203              call->callee->params[i].param_type != nir_parameter_in) {
 204             nir_copy_deref_var(b, call->params[i],
 205                                   nir_deref_var_create(b->shader, param));
 206          }
 207       }
 208       if (!glsl_type_is_void(call->callee->return_type) &&
 209           callee_copy->return_var->data.mode == nir_var_local) {
 210          nir_copy_deref_var(b, call->return_deref,
 211                                nir_deref_var_create(b->shader,
 212                                                     callee_copy->return_var));
 213       }
 214
 215       nir_instr_remove(&call->instr);
 216    }
 217
 218    return progress;
 219 }
 220
 221 static bool
 222 inline_function_impl(nir_function_impl *impl, struct set *inlined)
 223 {
 224    if (_mesa_set_search(inlined, impl))
 225       return false; /* Already inlined */
 226
 227    nir_builder b;
 228    nir_builder_init(&b, impl);
 229
 230    bool progress = false;
 231    nir_foreach_block_safe(block, impl) {
 232       progress |= inline_functions_block(block, &b, inlined);
 233    }
 234
 235    if (progress) {
 236       /* SSA and register indices are completely messed up now */
 237       nir_index_ssa_defs(impl);
 238       nir_index_local_regs(impl);
 239
 240       nir_metadata_preserve(impl, nir_metadata_none);
 241    }
 242
 243    _mesa_set_add(inlined, impl);
 244
 245    return progress;
 246 }
 247
 248 bool
 249 nir_inline_functions(nir_shader *shader)
 250 {
 251    struct set *inlined = _mesa_set_create(NULL, _mesa_hash_pointer,
 252                                           _mesa_key_pointer_equal);
 253    bool progress = false;
 254
 255    nir_foreach_function(shader, function) {
 256       if (function->impl)
 257          progress = inline_function_impl(function->impl, inlined) || progress;
 258    }
 259
 260    _mesa_set_destroy(inlined, NULL);
 261
 262    return progress;
 263 }