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 struct inline_functions_state {
  29    struct set *inlined;
  30    nir_builder builder;
  31    bool progress;
  32 };
  33
  34 static bool inline_function_impl(nir_function_impl *impl, struct set *inlined);
  35
  36 static bool
  37 rewrite_param_derefs_block(nir_block *block, void *void_state)
  38 {
  39    nir_call_instr *call = void_state;
  40
  41    nir_foreach_instr_safe(block, instr) {
  42       if (instr->type != nir_instr_type_intrinsic)
  43          continue;
  44
  45       nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
  46
  47       for (unsigned i = 0;
  48            i < nir_intrinsic_infos[intrin->intrinsic].num_variables; i++) {
  49          if (intrin->variables[i]->var->data.mode != nir_var_param)
  50             continue;
  51
  52          int param_idx = intrin->variables[i]->var->data.location;
  53
  54          nir_deref_var *call_deref;
  55          if (param_idx >= 0) {
  56             assert(param_idx < call->callee->num_params);
  57             call_deref = call->params[param_idx];
  58          } else {
  59             call_deref = call->return_deref;
  60          }
  61          assert(call_deref);
  62
  63          nir_deref_var *new_deref = nir_deref_as_var(nir_copy_deref(intrin, &call_deref->deref));
  64          nir_deref *new_tail = nir_deref_tail(&new_deref->deref);
  65          new_tail->child = intrin->variables[i]->deref.child;
  66          ralloc_steal(new_tail, new_tail->child);
  67          intrin->variables[i] = new_deref;
  68       }
  69    }
  70
  71    return true;
  72 }
  73
  74 static void
  75 lower_param_to_local(nir_variable *param, nir_function_impl *impl, bool write)
  76 {
  77    if (param->data.mode != nir_var_param)
  78       return;
  79
  80    nir_parameter_type param_type;
  81    if (param->data.location >= 0) {
  82       assert(param->data.location < impl->num_params);
  83       param_type = impl->function->params[param->data.location].param_type;
  84    } else {
  85       /* Return variable */
  86       param_type = nir_parameter_out;
  87    }
  88
  89    if ((write && param_type == nir_parameter_in) ||
  90        (!write && param_type == nir_parameter_out)) {
  91       /* In this case, we need a shadow copy.  Turn it into a local */
  92       param->data.mode = nir_var_local;
  93       exec_list_push_tail(&impl->locals, &param->node);
  94    }
  95 }
  96
  97 static bool
  98 lower_params_to_locals_block(nir_block *block, void *void_state)
  99 {
 100    nir_function_impl *impl = void_state;
 101
 102    nir_foreach_instr_safe(block, instr) {
 103       if (instr->type != nir_instr_type_intrinsic)
 104          continue;
 105
 106       nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
 107
 108       switch (intrin->intrinsic) {
 109       case nir_intrinsic_store_var:
 110          lower_param_to_local(intrin->variables[0]->var, impl, true);
 111          break;
 112
 113       case nir_intrinsic_copy_var:
 114          lower_param_to_local(intrin->variables[0]->var, impl, true);
 115          lower_param_to_local(intrin->variables[1]->var, impl, false);
 116          break;
 117
 118       case nir_intrinsic_load_var:
 119          /* All other intrinsics which access variables (image_load_store)
 120           * do so in a read-only fasion.
 121           */
 122          for (unsigned i = 0;
 123               i < nir_intrinsic_infos[intrin->intrinsic].num_variables; i++) {
 124             lower_param_to_local(intrin->variables[i]->var, impl, false);
 125          }
 126          break;
 127
 128       default:
 129          continue;
 130       }
 131    }
 132
 133    return true;
 134 }
 135
 136 static bool
 137 inline_functions_block(nir_block *block, void *void_state)
 138 {
 139    struct inline_functions_state *state = void_state;
 140
 141    nir_builder *b = &state->builder;
 142
 143    /* This is tricky.  We're iterating over instructions in a block but, as
 144     * we go, the block and its instruction list are being split into
 145     * pieces.  However, this *should* be safe since foreach_safe always
 146     * stashes the next thing in the iteration.  That next thing will
 147     * properly get moved to the next block when it gets split, and we
 148     * continue iterating there.
 149     */
 150    nir_foreach_instr_safe(block, instr) {
 151       if (instr->type != nir_instr_type_call)
 152          continue;
 153
 154       state->progress = true;
 155
 156       nir_call_instr *call = nir_instr_as_call(instr);
 157       assert(call->callee->impl);
 158
 159       inline_function_impl(call->callee->impl, state->inlined);
 160
 161       nir_function_impl *callee_copy =
 162          nir_function_impl_clone(call->callee->impl);
 163       callee_copy->function = call->callee;
 164
 165       /* Add copies of all in parameters */
 166       assert(call->num_params == callee_copy->num_params);
 167
 168       exec_list_append(&b->impl->locals, &callee_copy->locals);
 169       exec_list_append(&b->impl->registers, &callee_copy->registers);
 170
 171       b->cursor = nir_before_instr(&call->instr);
 172
 173       /* We now need to tie the two functions together using the
 174        * parameters.  There are two ways we do this: One is to turn the
 175        * parameter into a local variable and do a shadow-copy.  The other
 176        * is to treat the parameter as a "proxy" and rewrite derefs to use
 177        * the actual variable that comes from the call instruction.  We
 178        * implement both schemes.  The first is needed in the case where we
 179        * have an in parameter that we write or similar.  The second case is
 180        * needed for handling things such as images and uniforms properly.
 181        */
 182
 183       /* Figure out when we need to lower to a shadow local */
 184       nir_foreach_block(callee_copy, lower_params_to_locals_block, callee_copy);
 185       for (unsigned i = 0; i < callee_copy->num_params; i++) {
 186          nir_variable *param = callee_copy->params[i];
 187
 188          if (param->data.mode == nir_var_local &&
 189              call->callee->params[i].param_type != nir_parameter_out) {
 190             nir_copy_deref_var(b, nir_deref_var_create(b->shader, param),
 191                                   call->params[i]);
 192          }
 193       }
 194
 195       nir_foreach_block(callee_copy, rewrite_param_derefs_block, call);
 196
 197       /* Pluck the body out of the function and place it here */
 198       nir_cf_list body;
 199       nir_cf_list_extract(&body, &callee_copy->body);
 200       nir_cf_reinsert(&body, b->cursor);
 201
 202       b->cursor = nir_before_instr(&call->instr);
 203
 204       /* Add copies of all out parameters and the return */
 205       assert(call->num_params == callee_copy->num_params);
 206       for (unsigned i = 0; i < callee_copy->num_params; i++) {
 207          nir_variable *param = callee_copy->params[i];
 208
 209          if (param->data.mode == nir_var_local &&
 210              call->callee->params[i].param_type != nir_parameter_in) {
 211             nir_copy_deref_var(b, call->params[i],
 212                                   nir_deref_var_create(b->shader, param));
 213          }
 214       }
 215       if (!glsl_type_is_void(call->callee->return_type) &&
 216           callee_copy->return_var->data.mode == nir_var_local) {
 217          nir_copy_deref_var(b, call->return_deref,
 218                                nir_deref_var_create(b->shader,
 219                                                     callee_copy->return_var));
 220       }
 221
 222       nir_instr_remove(&call->instr);
 223    }
 224
 225    return true;
 226 }
 227
 228 static bool
 229 inline_function_impl(nir_function_impl *impl, struct set *inlined)
 230 {
 231    if (_mesa_set_search(inlined, impl))
 232       return false; /* Already inlined */
 233
 234    struct inline_functions_state state;
 235
 236    state.inlined = inlined;
 237    state.progress = false;
 238    nir_builder_init(&state.builder, impl);
 239
 240    nir_foreach_block(impl, inline_functions_block, &state);
 241
 242    if (state.progress) {
 243       /* SSA and register indices are completely messed up now */
 244       nir_index_ssa_defs(impl);
 245       nir_index_local_regs(impl);
 246
 247       nir_metadata_preserve(impl, nir_metadata_none);
 248    }
 249
 250    _mesa_set_add(inlined, impl);
 251
 252    return state.progress;
 253 }
 254
 255 bool
 256 nir_inline_functions(nir_shader *shader)
 257 {
 258    struct set *inlined = _mesa_set_create(NULL, _mesa_hash_pointer,
 259                                           _mesa_key_pointer_equal);
 260    bool progress = false;
 261
 262    nir_foreach_function(shader, function) {
 263       if (function->impl)
 264          progress = inline_function_impl(function->impl, inlined) || progress;
 265    }
 266
 267    _mesa_set_destroy(inlined, NULL);
 268
 269    return progress;
 270 }