Depth-First: Compiling Open Babel to Pure Java Bytecode with NestedVM: An Unsuccessful First Attempt

Compiling Open Babel to Pure Java Bytecode with NestedVM: An Unsuccessful First Attempt

Rich Apodaca — Mon, 19 Nov 2007 10:42:00 +0000

Wouldn't it be great to be able to compile code written in languages like FORTRAN, C, and C++ to Java bytecode? NestedVM - almost magically - can do just that. This article documents a failed first attempt to compile the popular cheminformatics toolkit Open Babel, which is written in C and C++, to pure Java bytecode with NestedVM.

A previous article described the successful compilation of the InChI toolkit, a C library, to a platform-independent executable jarfile.

The Problem

Open Babel is one of cheminformatics' most widely-used open source packages. It interconverts dozens of molecular languages, performs a host of cheminformatics analyses, and serves as a platform for many programs and Web services.

As useful as Open Babel is, it doesn't run directly on a Java Virtual Machine (JVM). Although an Open Babel JNI interface does exist, using it introduces a platform dependency, which in many cases is not acceptable. JNI is a great solution in some cases, but when maintaining a single version of a program is important, or when applets need to be used, or when code needs to work with unusual system configurations, it's a poor choice.

Our goal is to compile Open Babel's "babel" command-line utility into pure Java bytecode that can be run on any recent JVM without using JNI.

Overview of NestedVM

In a nutshell, NestedVM converts MIPS binaries to Java class files. In theory, this allows software written in any language that can be compiled to a MIPS binary to be run on a JVM.

To do this, NestedVM distributes two categories of tools: (1) a complete MIPS cross-compiler toolchain; and (2) a MIPS binary to Java bytecode compiler and accessories.

Building NestedVM

The preferred method to install NestedVM is to compile it from source found in the project repository. There are a number of prerequisites your system must meet in order to be able to do so. For now, this article assumes your system has all of them. Some of the following steps can be found in these instructions as well.

To obtain the source code from the NestedVM darcs repository:

$ darcs get --repo-name=nestedvm http://nestedvm.ibex.org

Then change into the nestedvm directory and build the main code:

$ cd nestedvm
$ make

On my machine, this step takes 10-15 minutes.

To make sure your build works, run the tests:

$ make test
...
1.574000e+00
-4.315000e+01l
-43
-4.315000e+01
4.315000e+01
Hello, World
7F
fabs(-2.24) = 2.34
Destructor!

NestedVM doesn't build the g++ compiler by default - it's something that needs to be done manually. Fortunately, it's not difficult to do:

$ make cxxtest
...
java -cp build tests.CXXTest
Test's constructor
Name: 0x50b40
Name: PKc
Is pointer: 1
Name: 0x50b3c
Name: i
Is pointer: 0
Hello, World from Test
Now throwing an exception
sayhi threw: const char *:Hello, Exception Handling!
Test's destructor

Finally, with all tools built, we need to set up our environment:

$ make env.sh
$ source env.sh

We're now ready to cross-compile Open Babel.

Cross-Compiling Open Babel

For this tutorial, we'll use the Open Babel 2.1.1 source distribution. Unpack the tarball and change into the directory.

Next, we'll need to set up our cross-compiler environment. Fortunately, NestedVM has made this easy. If you check your environment variables, you'll find that CXX and CC have both been set. All that remains is to notify the configure script that we'll be cross-compiling:

$ ./configure --host=mips-unknown-elf

Then we build the MIPS binaries:

$ make

Peeking into the tools directory, we can see all of the Open Babel command line tools have been built, including babel.

Unless you're running a MIPS machine, though, this binary won't be executable.

So far, it looks like everything worked. Although it didn't work the first time I tried it, the NestedVM team were most helpful.

Building the Java Class File

We're now ready for the final stage in the process, converting the MIPS binary to a Java class file. Again, NestedVM makes this simple:

$ cd tools
$ java org.ibex.nestedvm.Compiler -outfile Babel.class Babel babel
Exception in thread "main" java.lang.IllegalStateException: unresolved phantom target
        at org.ibex.classgen.MethodGen.resolveTarget(MethodGen.java:555)
        at org.ibex.classgen.MethodGen._generateCode(MethodGen.java:664)
        at org.ibex.classgen.MethodGen.generateCode(MethodGen.java:618)
        at org.ibex.classgen.MethodGen.dump(MethodGen.java:888)
        at org.ibex.classgen.ClassFile._dump(ClassFile.java:193)
        at org.ibex.classgen.ClassFile.dump(ClassFile.java:160)
        at org.ibex.nestedvm.ClassFileCompiler.__go(ClassFileCompiler.java:380)
        at org.ibex.nestedvm.ClassFileCompiler._go(ClassFileCompiler.java:72)
        at org.ibex.nestedvm.Compiler.go(Compiler.java:259)

Unfortunately, NestedVM has blown up with an exception. Although our target class file, Babel.class is now in our working directory, it is not complete and won't run.

What Went Wrong?

After bringing this problem to the NestedVM mailing list, it appears that this is a NestedVM bug.

However, the way babel works is to load its various language modules dynamically. It may be possible to fix the problem by producing a version of babel containing all of its modules in a single binary.

Although there is a major issue to be resolved, this tutorial illustrates the full process of compiling C++ code to Java bytecode using NestedVM.

"Compiling Open Babel to Pure Java Bytecode with NestedVM: An Unsuccessful First Attempt" by ¿¿¿Fortunately, it's not difficult to do:????

Thu, 18 Sep 2008 16:02:23 +0000

I found this error: alberto@alberto:~/nestedvm-2007-01-12$ make cxxtest make -C upstream tasks/build_gcc_step2 usr="/home/alberto/nestedvm-2007-01-12/upstream/install" \ MIPS_CFLAGS="-O3 -mmemcpy -ffunction-sections -fdata-sections -falign-functions=512 -fno-rename-registers -fno-schedule-insns -fno-delayed-branch -freduce-all-givs -march=mips1 -I. -Wall -Wno-unused" \ MIPS_PCFLAGS="-O3 -mmemcpy -ffunction-sections -fdata-sections -falign-functions=512 -fno-rename-registers -fno-schedule-insns -fno-delayed-branch -freduce-all-givs -march=mips1 -I. -Wall -Wno-unused --big-endian" \ MIPS_LDFLAGS="-march=mips1 --static -Wl,--gc-sections" make[1]: se ingresa al directorio `/home/alberto/nestedvm-2007-01-12/upstream' mkdir -p /home/alberto/nestedvm-2007-01-12/upstream/install mkdir -p build/gcc-obj && cd build/gcc-obj && \ echo | ../gcc-3.3.6/configure --prefix=/home/alberto/nestedvm-2007-01-12/upstream/install --target=mips-unknown-elf --disable-threads --with-gnu-ld --with-gnu-as --with-newlib=yes --enable-sjlj-exceptions --enable-languages="c" --enable-languages="c,c++,f77" && \ make TARGET_CFLAGS="-O3 -mmemcpy -ffunction-sections -fdata-sections -falign-functions=512 -fno-rename-registers -fno-schedule-insns -fno-delayed-branch -freduce-all-givs -march=mips1 -I. -Wall -Wno-unused" && \ make install Configuring for a i686-pc-linux-gnu host. Created "Makefile" in /home/alberto/nestedvm-2007-01-12/upstream/build/gcc-obj cc1: error: opción `memcpy' inválida *** The command 'gcc -o conftest -O3 -mmemcpy -ffunction-sections -fdata-sections -falign-functions=512 -fno-rename-registers -fno-schedule-insns -fno-delayed-branch -freduce-all-givs conftest.c' failed. *** You must set the environment variable CC to a working compiler. make[1]: *** [tasks/build_gcc_step2] Error 1 make[1]: se sale del directorio `/home/alberto/nestedvm-2007-01-12/upstream' make: *** [upstream/tasks/build_gcc_step2] Error 2 Any help???

"Compiling Open Babel to Pure Java Bytecode with NestedVM: An Unsuccessful First Attempt" by yash

Fri, 22 Feb 2008 12:42:33 +0000

hiiiiiiiii i want software for converting smi file to mol file

"Compiling Open Babel to Pure Java Bytecode with NestedVM: An Unsuccessful First Attempt" by Rich Apodaca

Mon, 26 Nov 2007 10:57:25 +0000

Zoli, I haven't tried converting 64-bit MIPS binaries or performing the conversion on a 64-bit OS, if that's what you mean.

I'd be interested in knowing if it works for you, though.

"Compiling Open Babel to Pure Java Bytecode with NestedVM: An Unsuccessful First Attempt" by Zoli

Mon, 26 Nov 2007 07:37:20 +0000

Can you make it in 64-bit?

"Compiling Open Babel to Pure Java Bytecode with NestedVM: An Unsuccessful First Attempt" by Brian Alliet

Tue, 20 Nov 2007 14:17:37 +0000

For the record, this bug has been fix. See the bug report above.

-Brian (brian@brianweb.net)

"Compiling Open Babel to Pure Java Bytecode with NestedVM: An Unsuccessful First Attempt" by Oleg

Tue, 20 Nov 2007 11:35:49 +0000

On my machine (SuSE linux x64) to statically compile I use the following config command ./configure --disable-dynamic-modules --enable-static=yes --enable-shared=no

it should work also for mips target.

Oleg. (OUrsu@salud.unm.edu)

"Compiling Open Babel to Pure Java Bytecode with NestedVM: An Unsuccessful First Attempt" by Rich Apodaca

Mon, 19 Nov 2007 17:26:35 +0000

Oleg, good question. It sure seems feasible, but I'm still trying to learn how to do it.

Do you know how to configure Open Babel to statically compile one large binary for the babel program?

"Compiling Open Babel to Pure Java Bytecode with NestedVM: An Unsuccessful First Attempt" by Oleg (OUrsu@salud.unm.edu)

Mon, 19 Nov 2007 14:45:32 +0000

what if you compile statically everything in one large MIPS binary file?

Oleg.