This is the mail archive of the java@gcc.gnu.org mailing list for the Java project.

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]
Other format:	[Raw text]

Re: Slow recursive functions

From: Anthony Green <green at redhat dot com>
To: Andrew Haley <aph at redhat dot com>
Cc: Christian Mayrhuber <christian dot mayrhuber at gmx dot net>, java at gcc dot gnu dot org
Date: Fri, 15 Apr 2005 06:57:24 -0700
Subject: Re: Slow recursive functions
Organization: Red Hat, Inc.
References: <425ED948.2070108@gmx.net> <16991.39643.410627.812606@cuddles.cambridge.redhat.com>

On Fri, 2005-04-15 at 11:43 +0100, Andrew Haley wrote:
> That, and the calling convention.  We use the x86 system calling
> convention throughout gcj, and this is slower than passing args in
> registers.  There's also the possibility that some JITs might be
> optimized for this kind of benchmark.

Here are some x86 Linux results.  java is 1.5, gcj is 4.0ish.  Times are
best of three run real times.

Original Benchmark

0m10.825s java -server
0m36.277s gcj -O3 -march=i686 -mtune=pentium4
0m41.259s gcj -O3
0m44.348s java
0m50.497s gcj -O2 -march=i686 -mtune=pentium4
1m3.868s  gcj -O2

Modified Benchmark (Ack method is private)

0m9.288s  gcj -O3 -march=i686 -mtune=pentium4
0m9.431s  gcj -O3
0m11.409s java -server
0m40.931s java
0m43.026s gcj -O2 -march=i686 -mtune=pentium4
0m44.210s gcj -O2

I think there are a number of well known optimizations we could
implement for gcj to produce the best "modified" results from the
original sources.  But I'm really skeptical they would make any
difference on real word programs.

AG

References:
- Slow recursive functions
  - From: Christian Mayrhuber
- Re: Slow recursive functions
  - From: Andrew Haley

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]