Line data    Source code

       1              : // Dependency fusion reordering pass.
       2              : // Copyright (C) 2025-2026 Free Software Foundation, Inc.
       3              : //
       4              : // This file is part of GCC.
       5              : //
       6              : // GCC is free software; you can redistribute it and/or modify it under
       7              : // the terms of the GNU General Public License as published by the Free
       8              : // Software Foundation; either version 3, or (at your option) any later
       9              : // version.
      10              : //
      11              : // GCC is distributed in the hope that it will be useful, but WITHOUT ANY
      12              : // WARRANTY; without even the implied warranty of MERCHANTABILITY or
      13              : // FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
      14              : // for more details.
      15              : //
      16              : // You should have received a copy of the GNU General Public License
      17              : // along with GCC; see the file COPYING3.  If not see
      18              : // <http://www.gnu.org/licenses/>.
      19              : //
      20              : // This pass uses the RTL-SSA representation to detect def-use pairs that are
      21              : // macro-op-fusible in the current microarchitecture (using the
      22              : // macro_fusion_pair_p () target hook) and place them next to one another, if
      23              : // possible.
      24              : 
      25              : #define INCLUDE_ALGORITHM
      26              : #define INCLUDE_FUNCTIONAL
      27              : #define INCLUDE_MEMORY
      28              : #define INCLUDE_ARRAY
      29              : #include "config.h"
      30              : #include "system.h"
      31              : #include "coretypes.h"
      32              : #include "backend.h"
      33              : #include "rtl.h"
      34              : #include "df.h"
      35              : #include "rtl-ssa.h"
      36              : #include "print-rtl.h"
      37              : #include "tree-pass.h"
      38              : #include "cfgcleanup.h"
      39              : #include "target.h"
      40              : #include "dbgcnt.h"
      41              : 
      42              : namespace {
      43              : const pass_data pass_data_dep_fusion =
      44              : {
      45              :   RTL_PASS, // type
      46              :   "dep_fusion", // name
      47              :   OPTGROUP_NONE, // optinfo_flags
      48              :   TV_NONE, // tv_id
      49              :   0, // properties_required
      50              :   0, // properties_provided
      51              :   0, // properties_destroyed
      52              :   0, // todo_flags_start
      53              :   TODO_df_finish, // todo_flags_finish
      54              : };
      55              : 
      56              : class pass_dep_fusion : public rtl_opt_pass
      57              : {
      58              : public:
      59       571444 :   pass_dep_fusion (gcc::context *ctxt)
      60      1142888 :     : rtl_opt_pass (pass_data_dep_fusion, ctxt)
      61              :   {}
      62              : 
      63              :   // opt_pass methods:
      64       285722 :   opt_pass *clone () final override { return new pass_dep_fusion (m_ctxt); }
      65              :   bool gate (function *) final override;
      66              :   unsigned int execute (function *) final override;
      67              : };
      68              : 
      69              : bool
      70      2942740 : pass_dep_fusion::gate (function *)
      71              : {
      72      2942740 :   return optimize > 0 && flag_dep_fusion;
      73              : }
      74              : 
      75              : unsigned int
      76      1927962 : pass_dep_fusion::execute (function *fn)
      77              : {
      78              :   // If the target has no macro fusion, there is nothing to be done.
      79      1927962 :   if (!targetm.sched.macro_fusion_pair_p)
      80              :     return 0;
      81              : 
      82              :   // Initialization.
      83      1927962 :   calculate_dominance_info (CDI_DOMINATORS);
      84      1927962 :   df_analyze ();
      85      1927962 :   crtl->ssa = new rtl_ssa::function_info (fn);
      86              : 
      87      1927962 :   init_recog_no_volatile ();
      88              : 
      89      1927962 :   for (rtl_ssa::insn_info *insn = *crtl->ssa->nondebug_insns ().begin ();
      90    182035910 :        insn;
      91    180107948 :        insn = insn->next_nondebug_insn ())
      92              :     {
      93    180107948 :       if (!insn->can_be_optimized () || insn->num_defs () != 1)
      94    172240432 :        continue;
      95              : 
      96     74807784 :       rtl_ssa::set_info *def = single_set_info (insn);
      97     74807784 :       if (!def)
      98        21189 :        continue;
      99              : 
     100     74786595 :       rtl_ssa::use_info *use_insn = def->single_nondebug_insn_use ();
     101    141608105 :       if (!use_insn
     102     52306399 :           || !use_insn->insn ()->can_be_optimized ()
     103    120120932 :           || !targetm.sched.macro_fusion_pair_p (insn->rtl (),
     104              :                                                  use_insn->insn ()->rtl ()))
     105     66821510 :        continue;
     106              : 
     107      7965085 :       auto attempt = crtl->ssa->new_change_attempt ();
     108      7965085 :       rtl_ssa::insn_change change (use_insn->insn ());
     109              : 
     110      7966294 :       if (use_insn->insn () != insn->next_any_insn ())
     111              :         {
     112        97569 :           if (!can_move_insn_p (use_insn->insn ()))
     113        97569 :             continue;
     114              : 
     115            0 :           change.move_range = insn;
     116            0 :           if (!rtl_ssa::restrict_movement (change))
     117            0 :             continue;
     118              : 
     119            0 :           if (dump_file && (dump_flags & TDF_DETAILS))
     120              :             {
     121            0 :               fprintf (dump_file, "Moved a single-use instruction:\n");
     122            0 :               dump_insn_slim (dump_file, use_insn->insn ()->rtl ());
     123            0 :               fprintf (dump_file, "right after its definition:\n");
     124            0 :               dump_insn_slim (dump_file, insn->rtl ());
     125              :             }
     126              :         }
     127              : 
     128      7867516 :       SCHED_GROUP_P (use_insn->insn ()->rtl ()) = 1;
     129      7867516 :       confirm_change_group ();
     130      7867516 :       crtl->ssa->change_insn (change);
     131      7965085 :     }
     132              : 
     133              :   // Finalization.
     134      1927962 :   if (crtl->ssa->perform_pending_updates ())
     135            0 :     cleanup_cfg (0);
     136              : 
     137      1927962 :   delete crtl->ssa;
     138              : 
     139      1927962 :   init_recog ();
     140      1927962 :   free_dominance_info (CDI_DOMINATORS);
     141      1927962 :   return 0;
     142              : }
     143              : 
     144              : } // end namespace
     145              : 
     146              : // Create a new dep fusion pass instance.
     147              : 
     148              : rtl_opt_pass *
     149       285722 : make_pass_dep_fusion (gcc::context *ctxt)
     150              : {
     151       285722 :   return new pass_dep_fusion (ctxt);
     152              : }