Do you know this project?

[egonzalez.ergio]

http://jwik.codeplex.com/

[genom2]

Yes, I knew it. But - as far as I could see - it has nothing which inspired me for HaikuVM. (For some unknown reasons even source code is not accessable for me.)

[egonzalez.ergio]

This is the lastest code:

https://mega.co.nz/#!YAEDGDTY!KWLDB1Gydhavcpp02TG6DBpaUAUdlbTc-Uh0mxDEItE

[genom2]

Thank you for showing the access path to jwik. As stated in the jwik C code it's "... work based on NanoVM ...". I like the NanoVM as a ground breaking work. But it's a bit clumbsy to use. (So might be jwik.)

In NanoVM you have to enable whole groups of bytecode by hand in C code if your JAVA program wants to use them. E.g. you have to #define:

NVM_USE_FLOAT (vs. VM_INCLUDE_FLOAT in jwik)
NVM_USE_EXTSTACKOPS (vs. VM_INCLUDE_EXTSTACKOPS in jwik)
NVM_USE_ARRAY (vs. VM_INCLUDE_ARRAY in jwik)

To mention only a few groups.
In contrast HaikuVM (automatic) includes each single bytecode (only) if your program needs it. As a result HaikuVM tends to have a much smaller code footprint.

NanoVM does only support about half of the official JAVA 6 bytecodes. (Missing real interesting ones like ATHROW, MONITORENTER, MONITOREXIT, MULTIANEWARRAY.)
In contrast HaikuVM supports all JAVA 6 bytecodes.

NanoVM does arithmetic e.g. 3+4*5 on the JAVA stack roughly like this:

push(3); push(4); push(5); push(pop()*pop()); push(pop()+pop());

In contrast HaikuVM uses a special C variable called top to mangage the top of stack. Arithmetic is done like this:

push(3); push(4); top=5; top*=pop(); top+=pop()

As a result HaikuVM needs less then half of stack operations. This and other optimizations lead to a speedup of factor 2.5.

For jwik I like the document "JwiK™ vs. Arduino". I wish I could write documents clear and as easy to understand like this one.