public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: Ayal Zaks <ZAKS@il.ibm.com>
To: Revital1 Eres <ERES@il.ibm.com>
Cc: abel@ispras.ru, gcc-patches@gcc.gnu.org,
	Kenneth.Zadeck@NaturalBridge.com,         volodyan@gmail.com
Subject: Re: [PING][PATCH] [modulo-sched] Change the ddg's construction
Date: Sun, 29 Jul 2007 22:21:00 -0000	[thread overview]
Message-ID: <OF14CA9093.91B5A0AE-ONC2257327.0070A1CF-C2257327.0073FAF9@il.ibm.com> (raw)
In-Reply-To: <OFE62D2EF3.8267B02E-ONC2257323.00727FE2-C2257327.0060B593@LocalDomain>

Revital1 Eres/Haifa/IBM wrote on 29/07/2007 20:36:19:

> Hello,
>
> This patch changes the ddg's construction as described in:
> http://gcc.gnu.org/ml/gcc-patches/2007-03/msg00711.html (attached is
> the version related to current trunk).
>
> This patch was bootstrapped and tested on ppc64 and x86_64.
>
> OK for mainline?
>

Kenny, thanks for the quick approval. Any follow-up changes needed for df,
now that df_ru_add_problem() is left without a caller?

Revital, this is ok with me too, with some minor comments below.

Ayal.

> Thanks,
> Revital
>
> 2007-07-29  Revital Eres  <eres@il.ibm.com>
>
>         * ddg.c (add_deps_for_def): Rename to...
>         (add deps): This.  Change implementation to use only reaching
>         def and def-use chains to construct the inter loop dependencies.
>         (add_deps_for_use): Remove function.
>         (build_inter_loop_deps): Call add_deps function to build the
>         inter loop dependencies.
>         * modulo-sched.c (sms_schedule): Build only
>         reaching def and def-use chains for the propose of the ddg
>         construction.
>
> [attachment "patch_ddg.txt" deleted by Ayal Zaks/Haifa/IBM]

> Index: ddg.c
> ===================================================================
> --- ddg.c (revision 126552)
> +++ ddg.c (working copy)
> @@ -224,131 +224,103 @@
>      add_edge_to_ddg (g, e);
>  }
>
> -
> -/* Given a downwards exposed register def RD, add inter-loop true
dependences
> -   for all its uses in the next iteration, and an output dependence to
the
> -   first def of the next iteration.  */
> +
> +/* Given a downwards exposed register def LAST_DEF (which is the last
> +   definition of that register in the bb), add inter-loop true
dependences
> +   for all its uses in the next iteration, an output dependence to the
      ^^^
      to

> +   first def of

the same register (possibly itself) in

> the next iteration and anti-dependences from its uses in
> +   the current iteration to the first definition in the next iteration.
> +   */
>  static void
> -add_deps_for_def (ddg_ptr g, struct df_ref *rd)
> +add_deps (ddg_ptr g, struct df_ref *last_def)

Better rename to something like add_cross_iteration_register_deps

>  {
> -  int regno = DF_REF_REGNO (rd);
> -  struct df_ru_bb_info *bb_info = DF_RU_BB_INFO (g->bb);
> +  int regno = DF_REF_REGNO (last_def);
>    struct df_link *r_use;
> -  int use_before_def = false;
> -  rtx def_insn = DF_REF_INSN (rd);
> -  ddg_node_ptr src_node = get_node_of_insn (g, def_insn);
> +  int has_use_in_bb_p = false;
> +  rtx def_insn = DF_REF_INSN (last_def);
> +  ddg_node_ptr last_def_node = get_node_of_insn (g, def_insn);
> +  ddg_node_ptr use_node;
> +  struct df_rd_bb_info *bb_info = DF_RD_BB_INFO (g->bb);
> +  struct df_ref *first_def = df_bb_regno_first_def_find (g->bb, regno);
>
> -  /* Create and inter-loop true dependence between RD and each of its
uses
> -     that is upwards exposed in RD's block.  */
> -  for (r_use = DF_REF_CHAIN (rd); r_use != NULL; r_use = r_use->next)
> +  gcc_assert (last_def_node && first_def);

Better split into two asserts.

> +
> +  /* Create inter-loop true dependences and anti dependences.  */
> +  for (r_use = DF_REF_CHAIN (last_def); r_use != NULL; r_use =
r_use->next)
>      {
> -      if (bitmap_bit_p (bb_info->gen, r_use->ref->id))
> - {
> -   rtx use_insn = DF_REF_INSN (r_use->ref);
> -   ddg_node_ptr dest_node = get_node_of_insn (g, use_insn);
> +      rtx use_insn = DF_REF_INSN (r_use->ref);
>
> -   gcc_assert (src_node && dest_node);
> +      if (BLOCK_FOR_INSN (use_insn) != g->bb)
> + continue;
>
> -   /* Any such upwards exposed use appears before the rd def.  */
> -   use_before_def = true;
> -   create_ddg_dep_no_link (g, src_node, dest_node, TRUE_DEP,
> +      use_node = get_node_of_insn (g, use_insn);
> +      gcc_assert (use_node);
> +      has_use_in_bb_p = true;
> +      if (use_node->cuid <= last_def_node->cuid)
> + {
> +   /* Add true deps from last_def to it's uses in the next
> +      iteration.  Any such upwards exposed use appears before
> +      the last_def def.  */
> +   create_ddg_dep_no_link (g, last_def_node, use_node, TRUE_DEP,
>        REG_DEP, 1);
>   }
> -    }
> +      else
> + {
> +   /* Add anti deps from last_def's uses in the current iteration
> +      to the first def in the next iteration.  We must not
                                                     ^^^^do

> +      add ANTI dep when there is an intra-loop TRUE dep in the
> +      opposite direction

, but use regmoves to fix such disregarded ANTI deps when broken

> + .  If the first_def reaches the USE then
> +      there is such a dep.  */
> +   ddg_node_ptr first_def_node = get_node_of_insn (g,
> +         first_def->insn);
>
> -  /* Create an inter-loop output dependence between RD (which is the
> -     last def in its block, being downwards exposed) and the first def
> -     in its block.  Avoid creating a self output dependence.  Avoid
creating
> -     an output dependence if there is a dependence path between the two
defs
> -     starting with a true dependence followed by an anti dependence
(i.e. if
> -     there is a use between the two defs.  */
> -  if (! use_before_def)
> +   gcc_assert (first_def_node);
> +
> +   if (!bitmap_bit_p (bb_info->gen, first_def->id))

Simply check instead if first_def is not last_def?

> +     create_ddg_dep_no_link (g, use_node, first_def_node, ANTI_DEP,
> +        REG_DEP, 1);
> + }
> +    }
> +  /* Create an inter-loop output dependence between LAST_DEF (which is
the
> +     last def in its block, being downwards exposed) and the first def
in
> +     its block.  Avoid creating a self output dependence.  Avoid
creating
> +     an output dependence if there is a dependence path between the two
> +     defs starting with a true dependence to a use which can be in the
> +     next iteration; followed by an anti dependence of that use to the
> +     first def (i.e. if there is a use between the two defs.)  */
> +  if (!has_use_in_bb_p)
>      {
> -      struct df_ref *def = df_bb_regno_first_def_find (g->bb, regno);
> -      int i;
>        ddg_node_ptr dest_node;
>
> -      if (!def || rd->id == def->id)
> +      if (last_def->id == first_def->id)
>   return;
>
> -      /* Check if there are uses after RD.  */
> -      for (i = src_node->cuid + 1; i < g->num_nodes; i++)
> -  if (df_find_use (g->nodes[i].insn, DF_REF_REG (rd)))
> -    return;
> -
> -      dest_node = get_node_of_insn (g, def->insn);
> -      create_ddg_dep_no_link (g, src_node, dest_node, OUTPUT_DEP,
REG_DEP, 1);
> +      dest_node = get_node_of_insn (g, first_def->insn);
> +      gcc_assert (dest_node);
> +      create_ddg_dep_no_link (g, last_def_node, dest_node,
> +         OUTPUT_DEP, REG_DEP, 1);
>      }
>  }
> -
> -/* Given a register USE, add an inter-loop anti dependence to the first
> -   (nearest BLOCK_BEGIN) def of the next iteration, unless USE is
followed
> -   by a def in the block.  */
> -static void
> -add_deps_for_use (ddg_ptr g, struct df_ref *use)
> -{
> -  int i;
> -  int regno = DF_REF_REGNO (use);
> -  struct df_ref *first_def = df_bb_regno_first_def_find (g->bb, regno);
> -  ddg_node_ptr use_node;
> -  ddg_node_ptr def_node;
> -  struct df_rd_bb_info *bb_info;
> -
> -  bb_info = DF_RD_BB_INFO (g->bb);
> -
> -  if (!first_def)
> -    return;
> -
> -  use_node = get_node_of_insn (g, use->insn);
> -  def_node = get_node_of_insn (g, first_def->insn);
> -
> -  gcc_assert (use_node && def_node);
> -
> -  /* Make sure there are no defs after USE.  */
> -  for (i = use_node->cuid + 1; i < g->num_nodes; i++)
> -     if (df_find_def (g->nodes[i].insn, DF_REF_REG (use)))
> -       return;
> -  /* We must not add ANTI dep when there is an intra-loop TRUE dep in
> -     the opposite direction. If the first_def reaches the USE then there
is
> -     such a dep.  */
> -  if (! bitmap_bit_p (bb_info->gen, first_def->id))
> -    create_ddg_dep_no_link (g, use_node, def_node, ANTI_DEP, REG_DEP,
1);
> -}
> -
>  /* Build inter-loop dependencies, by looking at DF analysis backwards.
*/
>  static void
>  build_inter_loop_deps (ddg_ptr g)
>  {
> -  unsigned rd_num, u_num;
> +  unsigned rd_num;
>    struct df_rd_bb_info *rd_bb_info;
> -  struct df_ru_bb_info *ru_bb_info;
>    bitmap_iterator bi;
>
>    rd_bb_info = DF_RD_BB_INFO (g->bb);
>
> -  /* Find inter-loop output and true deps by connecting downward exposed
defs
> -     to the first def of the BB and to upwards exposed uses.  */
> +  /* Find inter-loop

register

> output, true and anti deps.  */

>    EXECUTE_IF_SET_IN_BITMAP (rd_bb_info->gen, 0, rd_num, bi)
> -    {
> -      struct df_ref *rd = DF_DEFS_GET (rd_num);
> +  {
> +    struct df_ref *rd = DF_DEFS_GET (rd_num);
>
> -      add_deps_for_def (g, rd);
> -    }
> -
> -  ru_bb_info = DF_RU_BB_INFO (g->bb);
> -
> -  /* Find inter-loop anti deps.  We are interested in uses of the block
that
> -     appear below all defs; this implies that these uses are killed.  */
> -  EXECUTE_IF_SET_IN_BITMAP (ru_bb_info->kill, 0, u_num, bi)
> -    {
> -      struct df_ref *use = DF_USES_GET (u_num);
> -      if (!(DF_REF_FLAGS (use) & DF_REF_IN_NOTE))

Please comment why we no longer check and exclude such uses.

> - /* We are interested in uses of this BB.  */
> - if (BLOCK_FOR_INSN (use->insn) == g->bb)
> -   add_deps_for_use (g, use);
> -    }
> +    add_deps (g, rd);
> +  }
>  }
>
> +
>  /* Given two nodes, analyze their RTL insns and add inter-loop mem deps
>     to ddg G.  */
>  static void
> Index: modulo-sched.c
> ===================================================================
> --- modulo-sched.c (revision 126552)
> +++ modulo-sched.c (working copy)
> @@ -914,9 +914,8 @@
>    /* Init Data Flow analysis, to be used in interloop dep calculation.
*/
>    df_set_flags (DF_LR_RUN_DCE);
>    df_rd_add_problem ();
> -  df_ru_add_problem ();
>    df_note_add_problem ();
> -  df_chain_add_problem (DF_DU_CHAIN + DF_UD_CHAIN);
> +  df_chain_add_problem (DF_DU_CHAIN);
>    df_analyze ();
>    regstat_compute_calls_crossed ();
>    sched_init ();

       reply	other threads:[~2007-07-29 21:06 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <OFE62D2EF3.8267B02E-ONC2257323.00727FE2-C2257327.0060B593@LocalDomain>
2007-07-29 22:21 ` Ayal Zaks [this message]
2007-07-29 22:23   ` Kenneth Zadeck
2007-07-30 19:35   ` Revital1 Eres
2007-07-30 19:55     ` Kenneth Zadeck
     [not found] <OF3CA27F68.A9D59391-ONC2257328.003BB6AF-C2257328.0069A46E@LocalDomain>
2007-07-30 21:10 ` Ayal Zaks
     [not found] <OF17F488BC.1FC95BA6-ONC2257328.007046C4-C2257328.0070CC91@LocalDomain>
2007-07-30 20:59 ` Revital1 Eres
2007-07-29 18:55 Revital1 Eres
2007-07-29 19:55 ` Kenneth Zadeck

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=OF14CA9093.91B5A0AE-ONC2257327.0070A1CF-C2257327.0073FAF9@il.ibm.com \
    --to=zaks@il.ibm.com \
    --cc=ERES@il.ibm.com \
    --cc=Kenneth.Zadeck@NaturalBridge.com \
    --cc=abel@ispras.ru \
    --cc=gcc-patches@gcc.gnu.org \
    --cc=volodyan@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).