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: Mon, 30 Jul 2007 21:10:00 -0000	[thread overview]
Message-ID: <OF17F488BC.1FC95BA6-ONC2257328.007046C4-C2257328.0070CC91@il.ibm.com> (raw)
In-Reply-To: <OF3CA27F68.A9D59391-ONC2257328.003BB6AF-C2257328.0069A46E@LocalDomain>

Revital1 Eres/Haifa/IBM wrote on 30/07/2007 22:13:53:

>
> > 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 for the comments.  Regarding the following -
>
> > > -  /* 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.

> I'll address this in a separate patch which means we currently do not
> exclude uses that appears inside a REG_EQUAL or REG_EQUIV note; I've
> added ??? mark to revisit this issue later on.
>
> Attached is the patch including the comments.
>
> OK to mainline?
>

OK.

Instead of checking
> +          if (INSN_UID (first_def->insn) != INSN_UID (last_def->insn))
for anti-deps you could simply check
> +      if (last_def->id == first_def->id)
as done for output-deps.

Ayal.


> Thanks,
> Revital

> 2007-07-30  Revital Eres  <eres@il.ibm.com>
>
>         * ddg.c (add_deps_for_def): Rename to...
>         (add_cross_iteration_register_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_cross_iteration_register_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_30.txt" deleted by Ayal Zaks/Haifa/IBM]

       reply	other threads:[~2007-07-30 20:59 UTC|newest]

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