This is the mail archive of the java@gcc.gnu.org mailing list for the Java project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: java to RTL



Mohammad O'mara writes:

> all i need to know know if there is a possiblity to reach an RTL
> format of my JVM bytecodes ,

Gcc had an option to dump the RTL instead of the assembly code
(-dr). But you get RTL output for everything the compiler needs to
generate (helper functions, class initialization, etc.) If you run gcj
on a class file input along with the option -dr, I guess you would
achieve what you're looking for, unless I grossly misunderstood you.

I never really looked into the issue of whether this RTL is a suitable
input (after transformations) to a silicon compiler, but since it's
been defined for the purpose of implementing a compiler collection, I
guess it's not really related to, for example, the IEEE VHDL RTL
synthesis subset. You'd also had to experiment and find out what part
of the Java language can be synthesized. Java lacks certain things,
like the notion of what a signal or a port are, which are built into
hardware description languages used for synthesis -- I wonder if this
would be a problem as far as what you're trying to do.

> and back again from the optimized RTL to JVM instructions or simply
> a bytecode for that RTL ?

I don't know how hard this would be. I know it's said somewhere that
it's not doable to force gcc to re-ingest emitted RTL in order for it
do something with this input.

./A


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]