public inbox for gcc-prs@sourceware.org
help / color / mirror / Atom feed
From: Jan Hubicka <jh@suse.cz>
To: nobody@gcc.gnu.org
Cc: gcc-prs@gcc.gnu.org,
Subject: Re: c/7344: performance regression on huge case statements
Date: Fri, 11 Oct 2002 08:06:00 -0000	[thread overview]
Message-ID: <20021011150601.11025.qmail@sources.redhat.com> (raw)

The following reply was made to PR c/7344; it has been noted by GNATS.

From: Jan Hubicka <jh@suse.cz>
To: Jan Hubicka <jh@suse.cz>
Cc: Nathanael Nerode <neroden@twcny.rr.com>, gcc-gnats@gcc.gnu.org,
	rschiele@uni-mannheim.de, gcc-bugs@gcc.gnu.org, nobody@gcc.gnu.org,
	gcc-patches@gcc.gnu.org
Subject: Re: c/7344: performance regression on huge case statements
Date: Fri, 11 Oct 2002 17:01:48 +0200

 Hi,
 this second patch makes CFG construction to take 0.88 second instead of
 29 seconds it did originally.
 
 Honza
 
 Fri Oct 11 16:50:55 CEST 2002  Jan Hubicka  <jh@suse.cz>
 	* cfgbuild.c (make_edges): Create edge cache when we do have
 	large jumptable.
 	(do_tablejump): Note size of maximal jumptable.
 	* function.c (prepare_function_start): Zero out size.
 	* function.h (function): Add max_jumptable_ents.
 Index: cfgbuild.c
 ===================================================================
 RCS file: /cvs/gcc/egcs/gcc/cfgbuild.c,v
 retrieving revision 1.27
 diff -c -3 -p -r1.27 cfgbuild.c
 *** cfgbuild.c	5 Aug 2002 18:46:32 -0000	1.27
 --- cfgbuild.c	11 Oct 2002 14:49:54 -0000
 *************** make_edges (label_value_list, min, max, 
 *** 292,298 ****
     /* Heavy use of computed goto in machine-generated code can lead to
        nearly fully-connected CFGs.  In that case we spend a significant
        amount of time searching the edge lists for duplicates.  */
 !   if (forced_labels || label_value_list)
       {
         edge_cache = sbitmap_vector_alloc (last_basic_block, last_basic_block);
         sbitmap_vector_zero (edge_cache, last_basic_block);
 --- 292,298 ----
     /* Heavy use of computed goto in machine-generated code can lead to
        nearly fully-connected CFGs.  In that case we spend a significant
        amount of time searching the edge lists for duplicates.  */
 !   if (forced_labels || label_value_list || cfun->max_jumptable_ents > 100)
       {
         edge_cache = sbitmap_vector_alloc (last_basic_block, last_basic_block);
         sbitmap_vector_zero (edge_cache, last_basic_block);
 Index: expr.c
 ===================================================================
 RCS file: /cvs/gcc/egcs/gcc/expr.c,v
 retrieving revision 1.486
 diff -c -3 -p -r1.486 expr.c
 *** expr.c	7 Oct 2002 07:37:09 -0000	1.486
 --- expr.c	11 Oct 2002 14:49:57 -0000
 *************** do_tablejump (index, mode, range, table_
 *** 10941,10946 ****
 --- 10930,10938 ----
        enum machine_mode mode;
   {
     rtx temp, vector;
 + 
 +   if (range > cfun->max_jumptable_ents)
 +     cfun->max_jumptable_ents = range;
   
     /* Do an unsigned comparison (in the proper mode) between the index
        expression and the value which represents the length of the range.
 Index: function.c
 ===================================================================
 RCS file: /cvs/gcc/egcs/gcc/function.c,v
 retrieving revision 1.385
 diff -c -3 -p -r1.385 function.c
 *** function.c	11 Oct 2002 01:28:24 -0000	1.385
 --- function.c	11 Oct 2002 14:49:59 -0000
 *************** prepare_function_start ()
 *** 6295,6300 ****
 --- 6295,6302 ----
   
     cfun->function_frequency = FUNCTION_FREQUENCY_NORMAL;
   
 +   cfun->max_jumptable_ents = 0;
 + 
     (*lang_hooks.function.init) (cfun);
     if (init_machine_status)
       cfun->machine = (*init_machine_status) ();
 Index: function.h
 ===================================================================
 RCS file: /cvs/gcc/egcs/gcc/function.h,v
 retrieving revision 1.88
 diff -c -3 -p -r1.88 function.h
 *** function.h	22 Sep 2002 14:09:33 -0000	1.88
 --- function.h	11 Oct 2002 14:49:59 -0000
 *************** struct function GTY(())
 *** 500,505 ****
 --- 500,509 ----
          (set only when profile feedback is available).  */
       FUNCTION_FREQUENCY_HOT
     } function_frequency;
 + 
 +   /* Maximal number of entities in the single jumptable.  Used to estimate
 +      final flowgraph size.  */
 +   int max_jumptable_ents;
   };
   
   /* The function currently being compiled.  */


             reply	other threads:[~2002-10-11 15:06 UTC|newest]

Thread overview: 27+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2002-10-11  8:06 Jan Hubicka [this message]
  -- strict thread matches above, loose matches on Subject: below --
2003-04-08 13:15 hubicka
2003-01-08  2:00 bangerth
2002-10-18  6:56 Robert Schiele
2002-10-18  6:16 Nathanael Nerode
2002-10-17  6:46 Robert Schiele
2002-10-17  6:36 Jan Hubicka
2002-10-17  6:36 Nathanael Nerode
2002-10-17  6:26 Michael Matz
2002-10-17  6:06 Nathanael Nerode
2002-10-15 12:16 Jan Hubicka
2002-10-15 11:56 Kaveh R. Ghazi
2002-10-11 13:26 Richard Henderson
2002-10-11 12:56 Jan Hubicka
2002-10-11 10:26 Richard Henderson
2002-10-11  7:26 Jan Hubicka
2002-10-11  2:26 Jan Hubicka
2002-10-10 19:56 Nathanael Nerode
2002-10-10 19:26 Daniel Berlin
2002-10-10 19:16 Daniel Berlin
2002-10-10 19:16 Daniel Berlin
2002-10-10 19:16 Nathanael Nerode
2002-10-10 19:06 Nathanael Nerode
2002-10-10 19:06 Nathanael Nerode
2002-10-10 18:56 Nathanael Nerode
2002-07-17 13:16 Gerald Pfeifer
2002-07-17 10:46 rschiele

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=20021011150601.11025.qmail@sources.redhat.com \
    --to=jh@suse.cz \
    --cc=gcc-prs@gcc.gnu.org \
    --cc=nobody@gcc.gnu.org \
    /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).