[antlr-interest] Got GCJ to compile antlr and java parser/tree parser :)

[Terence Parr]

Ok, it was a pain in the ass, but with the knowledge that Ric Klaren 
had gotten antlr to compile, I spent a few hours today trying things 
out myself.

First, get ANTLR to compile.  Got it to compile out of the box:

for each package dir in antlr:

$ gcj --CLASSPATH=/home/parrt/org.antlr -c -g -O *.java

then link:

gcj -o /usr/local/bin/antlr --main=antlr.Tool -g *.o */*.o */*/*.o

After running strip on executable, it's like 1.3M. :)  Nice.

Getting antlr to run over a .g file is now easy:

$ antlr file.g

Hooray!  Now, to compile the resulting parser took me a while to get 
working.

I tried to make a library and all that, but couldn't make it work so I 
resorted to telling it to compile antlr each time along with the parser 
from file.g.  No biggie.

GCJ does not have swing so I had to yank out anything related to that 
like debug/misc/ASTFrame.* etc...  Then I jar'd it up (put in 
/tmp/antlr.jar) and then did this:

$ antlr java.g
$ antlr java.tree.g
$ gcj -c -g -O --classpath=/tmp/antlr.jar:. *.java
$ gcj -o jparse --main=Main -g *.o /tmp/antlr.jar

Lots of things like this come out:

warning: exception handler inside code that is being protected

but I ignored it and gave it the old smoke test.

ENVIRONMENT

My test box is a dual 1Ghz linux 9.x (can't remember which) with 1G RAM.

$ cat /proc/version
Linux version 2.4.20-6smp (bhcompile at porky.devel.redhat.com) (gcc 
version 3.2.2 20030222 (Red Hat Linux 3.2.2-5)) #1 SMP Thu Feb 27 
09:59:40 EST 2003

$ gcj --version
gcj (GCC) 3.2.2 20030222 (Red Hat Linux 3.2.2-5)

TIMING

Java VM parsing jdk 1.4 source using time program:

$ java Main /usr/local/java/java
real    0m28.868s
user    0m24.990s
sys     0m0.420s

GCJ compiled parser doing same thing:

$ ./jparse /usr/local/java/java
real    0m40.004s
user    0m38.870s
sys     0m0.470s

[Note there are some problems parsing the source, but I haven't checked 
what the error in my grammar is]

Yes, the GCJ version is SLOWER!  Could this be the garbage collector or 
slow IO library or something?

Ok, let's try on something smaller.  How about the antlr source.

$ time java Main /home/parrt/org.antlr/antlr
real    0m7.226s
user    0m4.750s
sys     0m0.090s

$ time ./jparse /home/parrt/org.antlr/antlr
real    0m5.794s
user    0m5.700s
sys     0m0.080s

In this case, GCJ wins.

When parsing 1.4 source, GCJ version needs 10M resident and java VM 
version needs 25M.

Well, that's all I have to report at the moment.  Can anybody comment, 
speculate?

Terence
--
Professor Comp. Sci., University of San Francisco
Creator, ANTLR Parser Generator, http://www.antlr.org
Co-founder, http://www.jguru.com
Co-founder, http://www.knowspam.net enjoy email again!
Co-founder, http://www.peerscope.com pure link sharing




 

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/