Your message dated Thu, 11 Feb 2010 16:02:09 +0100
with message-id <1265900529.18337.14.ca...@no>
and subject line not a bug
has caused the Debian Bug report #568093,
regarding java from openjdk very slow?
to be marked as done.
This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
Bug report if necessary, and/or fix the problem forthwith.
(NB: If you are a system administrator and have no idea what this
message is talking about, this may indicate a serious mail system
misconfiguration somewhere. Please contact [email protected]
immediately.)
--
568093: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=568093
Debian Bug Tracking System
Contact [email protected] with problems
--- Begin Message ---
Package: openjdk-6-jre-headless
Version: 6b17-1.7-1
Severity: normal
I am not sure if this is a limitation in openjdk, but while trying to
package jblas 1.0 (http://mikiobraun.github.com/jblas/ )
I recognized that it runs a lot slower under openjdk than under sun java.
If this is a known problem then sorry for my ignorance.
/usr/lib/jvm/java-1.6.0-openjdk/jre/bin/java -jar jblas-1.0.1.jar
Simple benchmark for jblas
Running sanity benchmarks.
checking vector addition... ok
checking matrix multiplication... ok
checking existence of dsyev...... ok
checking XERBLA... ok
Sanity checks passed.
Each benchmark will take about 5 seconds...
Running benchmark "Java matrix multiplication, double precision".
n = 10 : 574.8 MFLOPS (1437049 iterations in 5.0 seconds)
n = 100 : 1839.3 MFLOPS (4599 iterations in 5.0 seconds)
n = 1000 : 1176.4 MFLOPS (3 iterations in 5.1 seconds)
Running benchmark "Java matrix multiplication, single precision".
n = 10 : 605.4 MFLOPS (1513554 iterations in 5.0 seconds)
n = 100 : 1794.8 MFLOPS (4487 iterations in 5.0 seconds)
n = 1000 : 1960.5 MFLOPS (5 iterations in 5.1 seconds)
Running benchmark "ATLAS matrix multiplication, double precision".
n = 10 : 773.6 MFLOPS (1933990 iterations in 5.0 seconds)
n = 100 : 3597.3 MFLOPS (8994 iterations in 5.0 seconds)
n = 1000 : 4938.1 MFLOPS (13 iterations in 5.3 seconds)
Running benchmark "ATLAS matrix multiplication, single precision".
n = 10 : 785.8 MFLOPS (1964485 iterations in 5.0 seconds)
n = 100 : 6257.1 MFLOPS (15643 iterations in 5.0 seconds)
n = 1000 : 8217.2 MFLOPS (21 iterations in 5.1 seconds)
/usr/lib/jvm/java-6-sun-1.6.0.17/jre/bin/java -jar jblas-1.0.1.jar
Simple benchmark for jblas
Running sanity benchmarks.
checking vector addition... ok
-- org.jblas CONFIG BLAS native library not found in path. Copying native
library from the archive. Consider installing the library somewhere in the path
(for Windows: PATH, for Linux: LD_LIBRARY_PATH).
-- org.jblas CONFIG ArchFlavor native library not found in path. Copying native
library libjblas_arch_flavor from the archive. Consider installing the library
somewhere in the path (for Windows: PATH, for Linux: LD_LIBRARY_PATH).
-- org.jblas CONFIG Loading libjblas_arch_flavor.so from
/lib/static/Linux/amd64/.
-- org.jblas CONFIG Loading libjblas.so from /lib/static/Linux/amd64/sse3/.
checking matrix multiplication... ok
checking existence of dsyev...... ok
checking XERBLA... ok
Sanity checks passed.
Each benchmark will take about 5 seconds...
Running benchmark "Java matrix multiplication, double precision".
n = 10 : 699.1 MFLOPS (1747749 iterations in 5.0 seconds)
n = 100 : 1805.5 MFLOPS (4514 iterations in 5.0 seconds)
n = 1000 : 1282.9 MFLOPS (4 iterations in 6.2 seconds)
Running benchmark "Java matrix multiplication, single precision".
n = 10 : 700.3 MFLOPS (1750641 iterations in 5.0 seconds)
n = 100 : 1842.9 MFLOPS (4608 iterations in 5.0 seconds)
n = 1000 : 1940.2 MFLOPS (5 iterations in 5.2 seconds)
Running benchmark "ATLAS matrix multiplication, double precision".
n = 10 : 692.2 MFLOPS (1730600 iterations in 5.0 seconds)
n = 100 : 4753.7 MFLOPS (11885 iterations in 5.0 seconds)
n = 1000 : 9129.3 MFLOPS (23 iterations in 5.0 seconds)
Running benchmark "ATLAS matrix multiplication, single precision".
n = 10 : 678.4 MFLOPS (1696019 iterations in 5.0 seconds)
n = 100 : 9011.1 MFLOPS (22528 iterations in 5.0 seconds)
n = 1000 : 14637.5 MFLOPS (37 iterations in 5.1 seconds)
-- System Information:
Debian Release: squeeze/sid
APT prefers stable
APT policy: (700, 'stable'), (650, 'testing'), (600, 'unstable'), (500,
'oldstable'), (1, 'experimental')
Architecture: amd64 (x86_64)
Kernel: Linux 2.6.32.3-sonne (SMP w/2 CPU cores; PREEMPT)
Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash
Versions of packages openjdk-6-jre-headless depends on:
ii ca-certificates-java 20091021 Common CA certificates (JKS keysto
ii dpkg 1.15.5.6 Debian package management system
ii java-common 0.34 Base of all Java packages
ii libc6 2.10.2-5 Embedded GNU C Library: Shared lib
ii libcups2 1.4.2-7 Common UNIX Printing System(tm) -
ii libfreetype6 2.3.11-1 FreeType 2 font engine, shared lib
ii libgcc1 1:4.4.3-1 GCC support library
ii libjpeg62 6b-15 The Independent JPEG Group's JPEG
ii liblcms1 1.18.dfsg-1.2+b1 Color management library
ii libnss3-1d 3.12.5-2 Network Security Service libraries
ii openjdk-6-jre-lib 6b17-1.7-1 OpenJDK Java runtime (architecture
ii rhino 1.7R2-3 JavaScript engine written in Java
ii tzdata-java 2010b-1 time zone and daylight-saving time
ii zlib1g 1:1.2.3.4.dfsg-3 compression library - runtime
Versions of packages openjdk-6-jre-headless recommends:
ii icedtea-6-jre-cacao 6b17-1.7-1 Alternative JVM for OpenJDK, using
Versions of packages openjdk-6-jre-headless suggests:
ii libnss-mdns 0.10-3.1 NSS module for Multicast DNS name
pn sun-java6-fonts <none> (no description available)
ii ttf-baekmuk 2.2-2 Baekmuk series TrueType fonts
ii ttf-dejavu-core 2.30-2 Vera font family derivate with add
ii ttf-indic-fonts 1:0.5.6 Metapackage for free Indian langua
ii ttf-kochi-gothic 20030809-6 Kochi Subst Gothic Japanese TrueTy
ii ttf-sazanami-gothic 20040629-8 Sazanami Gothic Japanese TrueType
ii ttf-sazanami-mincho 20040629-8 Sazanami Mincho Japanese TrueType
ii ttf-wqy-zenhei 0.8.38-1 "WenQuanYi Zen Hei" A Hei-Ti Style
-- no debconf information
--- End Message ---
--- Begin Message ---
the reason is simple: jblas (from the web) used atlas + sse3 + core2 duo
optimized code. I here had only plain atlas installed. Installing the
sse3 core2 stuff from experimental speeds up things drastically and
performance between openjdk and sunjava is up to par.
Soeren
--
For the one fact about the future of which we can be certain is that it
will be utterly fantastic. -- Arthur C. Clarke, 1962
signature.asc
Description: This is a digitally signed message part
--- End Message ---
