Gcj binaries performance (arm-linux)

Mon Apr 28 12:11:00 GMT 2008

Hi all,
I'm doing some tests to compare gcj performance vs some java virtual machine (I'm using JamVM at the moment).
My target is a PXA270 with 128MB of Ram.
The program I use to test performance is a simple BubbleSort (order an array of 1000 elements for 100 times).
Here are the result I am getting:
Binary compiled with arm-linux-gcj: 45720 ms
Virtual machine (JamVM): 40500 ms
I'm quite surprised of this result, I thought that a binary compiled with gcj would be much faster than a java virtual machine.
I've also tried to pass some options (e.g. -march=, -mtune, -mcpu...) when compiling with gcj but results are not changing.
Any idea to improve performance?
Should I try to use some optimization options? (I see a long list when doing --help=opt)
Here is my configure:
Target: arm-linux-gnueabi
Configured with: ../configure --prefix=/root/tmp/f8/2install --with-sysroot=/root/tmp/f8/rootfs-f8 --disable-libssp --disable-libgomp --disable-libmudflap --enable-libgcj --disable-bootstrap --disable-multilib --disable-sjlj-exceptions --target=arm-linux-gnueabi --enable-languages=c,c++,java --with-build-time-tools=/root/tmp/f8/binutils/arm-linux-gnueabi/bin --with-ld=/root/tmp/f8/binutils/arm-linux-gnueabi/bin/ld --with-nm=/root/tmp/f8/binutils/arm-linux-gnueabi/bin/nm --with-strip=/root/tmp/f8/binutils/arm-linux-gnueabi/bin/strip --with-as=/root/tmp/f8/binutils/arm-linux-gnueabi/bin/as --with-float=soft
Thread model: posix
gcc version 4.3.0 (GCC)
Thank you,
Francesco