What I can do is have standard forms for instructions, then just decode for those instead of handling each instruction. Basically the minimizer will have instruction formats to be used and such. This is only needed where the minimizer is, not the register code.


Stack operations like DUP will just call the stack handling stuff as needed although they will make a copy unless it is a cached local.


I can do stack caching on the stack as long as entries at higher positions of the stack only point to entries which are below. This will remove many pointless DUP operations and make the code more efficient. However the swap operations might end up being a bit more complex to handle.


I wonder if putting exception handlers last would work, but not really no.


By chance found a bug in the snapshot code where only cached objects on the stack were to be cleared, that would have been a really fun bug to fix in the future. Luckily it was caught.


So the next thing are branches which should not be difficult at all. Every point has a Java label so translation is simple.


Okay, I think there are too many graphics things. All of those graphics are so duplicated and pretty messy. Also it is a bit complicated as well... I think what I want to do, although it will add some slight overhead, is to make a graphics state and then have an enum of all the graphics functions. Whenever something is called for the enum it is literally just going to run whatever things. This means so that when dotted lines and such happen it can call that stuff directly rather than doing a branch and complicating things. I guess what I want is something that is much simpler to isolate graphics wise.


Okay so I moved over all the stuff, so can remove the old stuff and make it the default now.