I peformed a simple test on the Jar file code in classpath. The test was mainly to
check that the Classpath and jazzlib code worked correctly.
I created a jar file with 162 files, consisting of .java source and .class files, and
also some html files and tar.gz archives.
I then wrote a java test program that extracted the files from this .jar file.
I ran this program with my JVM (kissme), using the jazzlib library. kissme does not
have a JIT.
I also ran it with IBM's JDK 1.3 for linux, with the JIT enabled and disabled.
I used md5sum and find to verify that the extracted files were correct.
I timed the process, you can see kissme is 20 to 40 times slower (user time) or 17 to
21 times slower (total time), I guess this is mostly because the zip handling code
wasn't compiled natively by a JIT or a C compiler.
KISSME
kissme -classpath kissme_test.JarTest 46.51s user 0.85s system 87% cpu 53.934 total
kissme -classpath kissme_test.JarTest 46.15s user 0.73s system 90% cpu 52.070 total
IBM JDK
/usr/local/jdk1.3/bin/java -cp .. kissme_test.JarTest 1.71s user 1.02s system 47% cpu
5.695 total
/usr/local/jdk1.3/bin/java -cp .. kissme_test.JarTest 1.56s user 0.89s system 66% cpu
3.694 total
IBM JDK - no JIT compiler
/usr/local/jdk1.3/bin/java -cp .. kissme_test.JarTest 0.93s user 0.67s system 66% cpu
2.412 total
/usr/local/jdk1.3/bin/java -cp .. kissme_test.JarTest 0.95s user 0.40s system 56% cpu
2.373 total
John Leuner
--
_______________________________________________
Classpath mailing list
[EMAIL PROTECTED]
http://mail.gnu.org/mailman/listinfo/classpath
