Ian Reddy

Progress Report

Last week I tried to get to the JAI source code, just to learn more about the native method invocations, I could not get hold of the source code except for the decompiled classes from the JAI package. I noticed that the JAI uses parameter blocks o pass to native methods. Then I came across some tools like toba, sumatra and harrisa. These programs promise to convert bytecodes to C. From their home page I knew that they were no longer being updated, basically they were abandoned projects. I tried to get them working, even downloaded the necessary JDK's for them, but the toba and sumatra won't build, I also tried to engineer some of the source code, but still no success. The Harissa would build, but the executable files would break.

I also looked at a compiler JC that was completely written in Java, and converts the bytecodes to C first an then executes them.(It is pretty old too).

The reason I wanted to do this, was If I could recognize a pattern that is capable of being accelerated, may be at the bytecode level or source code and if I could convert this pattern to C, and in line instruction sets.

As you had mentioned in class last week, as to automate the native method development, I just thought this would be of help, If I not only generated native wrappers, but also native code for existing java user defined methods.