From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-patches-return-338002-listarch-gcc-patches=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 12842 invoked by alias); 11 Mar 2013 09:10:27 -0000
Received: (qmail 12830 invoked by uid 22791); 11 Mar 2013 09:10:25 -0000
X-SWARE-Spam-Status: No, hits=-5.0 required=5.0	tests=AWL,BAYES_00,DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,FREEMAIL_FROM,KHOP_RCVD_TRUST,KHOP_THREADED,RCVD_IN_DNSWL_LOW,RCVD_IN_HOSTKARMA_YE
X-Spam-Check-By: sourceware.org
Received: from mail-we0-f177.google.com (HELO mail-we0-f177.google.com) (74.125.82.177)    by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Mon, 11 Mar 2013 09:10:09 +0000
Received: by mail-we0-f177.google.com with SMTP id d7so3245798wer.22        for <gcc-patches@gcc.gnu.org>; Mon, 11 Mar 2013 02:10:08 -0700 (PDT)
MIME-Version: 1.0
X-Received: by 10.194.77.129 with SMTP id s1mr17488461wjw.17.1362993007384; Mon, 11 Mar 2013 02:10:07 -0700 (PDT)
Received: by 10.194.56.100 with HTTP; Mon, 11 Mar 2013 02:10:07 -0700 (PDT)
In-Reply-To: <CABu31nPKUkhm3BprYx=V_TQCjLwRg9RwaT1O89e6FJM3GBNu8A@mail.gmail.com>
References: <CABu31nPKUkhm3BprYx=V_TQCjLwRg9RwaT1O89e6FJM3GBNu8A@mail.gmail.com>
Date: Mon, 11 Mar 2013 09:10:00 -0000
Message-ID: <CAFiYyc1wJhthngtr1Pa5gm4u8+Hphs2Z+48_ob-QQ3VEgghZSg@mail.gmail.com>
Subject: Re: [patch] make gcse.c respect -fno-gcse-lm
From: Richard Biener <richard.guenther@gmail.com>
To: Steven Bosscher <stevenb.gcc@gmail.com>
Cc: GCC Patches <gcc-patches@gcc.gnu.org>
Content-Type: text/plain; charset=ISO-8859-1
X-IsSubscribed: yes
Mailing-List: contact gcc-patches-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-patches.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-patches/>
List-Post: <mailto:gcc-patches@gcc.gnu.org>
List-Help: <mailto:gcc-patches-help@gcc.gnu.org>
Sender: gcc-patches-owner@gcc.gnu.org
X-SW-Source: 2013-03/txt/msg00394.txt.bz2

On Mon, Mar 11, 2013 at 12:17 AM, Steven Bosscher <stevenb.gcc@gmail.com> wrote:
> Hello,
>
> RTL PRE has an option to disable load motion. This option works fine,
> except that all analysis for load motion is still performed.
>
> This patch stops gcse.c from recording memory sets for -fno-gcse-lm,
> and conservatively returns true in oprs_unchanged_p for any MEM.
>
> Bootstrapped&tested on {x86_64,powerpc64}-unknown-linux-gnu. OK?

Ok.

Thanks,
Richard.

> Ciao!
> Steven
>
>
>         * gcse.c (oprs_unchanged_p): Respect flag_gcse_lm.
>         (record_last_mem_set_info): Likewise.
>
> Index: gcse.c
> ===================================================================
> --- gcse.c      (revision 196576)
> +++ gcse.c      (working copy)
> @@ -890,8 +890,9 @@ oprs_unchanged_p (const_rtx x, const_rtx
>        }
>
>      case MEM:
> -      if (load_killed_in_block_p (current_bb, DF_INSN_LUID (insn),
> -                                 x, avail_p))
> +      if (! flag_gcse_lm
> +         || load_killed_in_block_p (current_bb, DF_INSN_LUID (insn),
> +                                    x, avail_p))
>         return 0;
>        else
>         return oprs_unchanged_p (XEXP (x, 0), insn, avail_p);
> @@ -1471,10 +1472,14 @@ canon_list_insert (rtx dest ATTRIBUTE_UN
>  static void
>  record_last_mem_set_info (rtx insn)
>  {
> -  int bb = BLOCK_FOR_INSN (insn)->index;
> +  int bb;
> +
> +  if (! flag_gcse_lm)
> +    return;
>
>    /* load_killed_in_block_p will handle the case of calls clobbering
>       everything.  */
> +  bb = BLOCK_FOR_INSN (insn)->index;
>    modify_mem_list[bb].safe_push (insn);
>    bitmap_set_bit (modify_mem_list_set, bb);