Le 27/02/12 20:27, stewart mackenzie a écrit :
Hello,
Yves
- may we do a non-backwards compatible release with Mozart2? ie fix
all the 'broken' bits in this fresh start.
At the VM level, that's what we are doing. At the language level I'd
like to do the same but I fear we would have a lot more heated debates,
we would also have to hack a lot more on the compiler.
So, I think we'll first have the new VM and a mostly compatible language
running on it and only after a revision of the language.
- how does one get oz to run on ARM etc? would we not use gcc just
after scala bootstrap stage, instead use llvm and target ARM, then
repeat compilation cycle till we get *.ozf compiler to run on ARM?
There are many ways to port the new VM to ARM. It mostly depends on how
fast your ARM is. If it's very fast, you could simply fo the whole build
on it. But building LLVM on an ARM processor is probably going to take
'some' time. On the other end, you could do the whole build on your host
machine. Just start with a normal build until you get the final oz
compiler. As a .ozf it is portable and could be copied to the ARM. Now,
rebuild the vm library and the emulator proper using a cross-compiler
(gcc or llvm based doesn't really matter) to obtain a VM running on ARM.
You could also avoid recompilation by going down to the .cc code
generation (i.e. running the generator and bootstrap-compiler) on the
host, cross-compiling the vm library, emulator and ozc1, and finally
finishing the compiler bootstrap on the ARM itself.
CMake is supposezd to have cross-compilation support but we haven't done
anything for that yet. So, for now, only the first way (everything on
the ARM) should work out-of-the-box.
Sebastien and Yves
- Thanks for the bytecode insight I really appreciate it.
You are welcome.
Kind regards
Stewart
Yves
_________________________________________________________________________________
mozart-hackers mailing list [email protected]
http://www.mozart-oz.org/mailman/listinfo/mozart-hackers
