Re: Jit Implementation

Hey guys


I just want to say a massive thank you! :) I understand this now, i
really didn't know you could use memory like that literaly like a
function pointer but it makes sense, i see the mmap is probably a
safer approach to make sure the memory is executable etc. I've found
many of the projects relating to this un-readable and now i understand
why, since if your going to spend your time translating your
instructions for many architectures you might have a virtual
instruction set and then translate into memory.


Its got my thinking about my language project and my idea was to make
it all seem like an interpreter, but in the backend have a virtual
instruction set and output bytecode where i might be able to preform
optimizations and then load it in properly like a VM and jit away
easier. But as i write this i see this is how most big projects work
these days and it makes total sense.


I always had it in my head that jit's had to output formal object code
into files and execute them like normaly ahead of compiled programs,
since many developers don't take about this area too much i guess
because its trivial when you know how its just knowing in the first
place. I think i need more time to get more used to more instruction
sets to define a virtual instruction set properly, but i really like
the parrot project so i might just take inspiration from ( aka use )
PASM :).


--Phil

Return to the comp.compilers page.
Search the comp.compilers archives again.