[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [RFC] 1 pass architecture and sorted writing.
From: |
Hideki IWAMOTO |
Subject: |
Re: [RFC] 1 pass architecture and sorted writing. |
Date: |
Mon, 17 May 2010 21:20:25 +0900 |
Hi.
On Thu, 13 May 2010 18:44:12 +0900, Shigio YAMAGUCHI wrote...
> (2) Sorted writing
> The (1) brings big GRTAGS and decrease at the writing speed.
> To avoid it, the records written to the tag files are sorted beforehand.
> This function becomes effective if DBOP_SORTED_WRITE flag is specified
> for function dbop_open(). It uses external POSIX sort command.
When C lolale is not used, the performance might be worse than CVS version.
You should restore
set_env("LC_ALL", "C");
deleted in gtags.c revision 1.183.
http://cvs.savannah.gnu.org/viewvc/global/gtags/gtags.c?root=global&r1=1.182&r2=1.183
The elapsed time measured on Fedora core 4 is as follows.
CPU: Pentium D (dual core)
Input: linux-2.6.33 source
-------------- -------------- --------------------------------
patched version
B-tree cache CVS version ---------------- ---------------
LANG=ja_JP.UTF-8 LANG=C
-------------- -------------- ---------------- ---------------
32MiB 107.09 sec 76.25 sec 45.27 sec
64MiB 70.25 sec 78.61 sec 44.65 sec
128MiB 68.02 sec 79.64 sec 45.61 sec
-------------- -------------- ---------------- ---------------
==== CVS version ====
$ foreach cachesize ( 33554432 67108864 134217728 )
foreach? rm -fr linux-2.6.33; tar xfj ~/download/linux/linux-2.6.33.tar.bz2;
sync
foreach? (cd linux-2.6.33; time env GTAGSCACHE=$cachesize gtags)
foreach? end
47.422u 6.380s 1:47.09 50.2% 0+0k 0+0io 2pf+0w
47.858u 4.424s 1:10.25 74.4% 0+0k 0+0io 2pf+0w
48.131u 3.660s 1:08.02 76.1% 0+0k 0+0io 2pf+0w
==== patched version ====
$ foreach locale ( ja_JP.UTF-8 C )
foreach? foreach cachesize ( 33554432 67108864 134217728 )
foreach? rm -fr linux-2.6.33; tar xfj ~/download/linux/linux-2.6.33.tar.bz2;
sync
foreach? (cd linux-2.6.33; time env LANG=$locale GTAGSCACHE=$cachesize gtags)
foreach? end
foreach? end
76.860u 4.736s 1:16.25 107.0% 0+0k 0+0io 5pf+0w
79.456u 4.524s 1:18.61 106.8% 0+0k 0+0io 0pf+0w
79.912u 4.360s 1:19.64 105.8% 0+0k 0+0io 2pf+0w
39.106u 4.232s 0:45.27 95.7% 0+0k 0+0io 2pf+0w
38.946u 4.620s 0:44.65 97.5% 0+0k 0+0io 2pf+0w
39.074u 4.592s 0:45.61 95.7% 0+0k 0+0io 3pf+0w
----
Hideki IWAMOTO address@hidden