# XiaoVM Spec

## Allowed instructions

* LOADSTRING

* LOADNUMBER

* LOADSPECIAL

* CALL

It's easy and fun to add a new function to XiaoVM.

For the specific number these names represent, read processor.h.

## Builtins

* io::print

## Namespace

You may wonder what the "::" means. In fact it's nothing more than a string... There isn't built-in namespace support and thus this is a simulation.

## Lifespan

LOAD* will create new objects from static data on the heap, and give them the reference count 1.

After CALL is done, reference counts of all the involved objects will automatically be decreased.

## "Stack"?

If you read the source you may find I used the term, "stack", in "processor.[ch]". To clarify my point, I have to say it's NOT real stack. See the examples and you'll find why. I have no better names so I simply pick up "stack" and put it there.

## Limitations

* I didn't pay attention to portability at first, so the bytecode is not portable. This is easy-to-fix because the source code is very short.

* Valgrind reports some "definite" memory leaks.

* The size of any number or length of strings can't be larger than sizeof(char). This is a problem in the compiler but I'm too lazy to fix it. Maybe later I will try 'long', or even 'long long'.

## Advantages

* It's an extensible framework

* The "assemble language" consists of very simple concepts and functions

* The source code is short and easy-to-read

* The C API is clean and neat

* It's strongly typed!

* It's also very easy to extend with the C API

## Disadvantages

* The source code is short, thus the function is limited

* Who will expect such VM?! NO ONE WOULD LIKE TO USE THIS! THUS...

* THIS IS NOT STRONGLY MAINTAINED. IT'S JUST A TOY!