08:06
Okay so I am not going to dump my old code, I am just going to have a fork with an option that allows me to choose between different registerizers.
08:29
For the quick translation it is just going to be assumed that all registers are unique and there are no aliases.
08:33
I think what I can do is have a sane format for the register instructions so
I do not have duplicated code all over the place. For example instead of
handling ALOAD_2 it will just be ALOAD. I think this will greatly
simplify things. The same can be said for comparison types and all of the
various math operations as well. Just something where I can throw all the
stuff in and get less duplicated code in.
13:03
So it seems LDC got turned into FLOAT_NEG:[2147483647]/@1#LDC~:[2147483647]
and I wonder how that happened.
13:04
Was just using the wrong mnemonic get.
15:42
Okay so for the made exceptions, I could make the exceptions then just jump according to the cleanup states. Like I should have a variant but for the register enqueuing since with the super quick method that is all that matters.
16:30
So later processing of returns is never needed because all of them should be generated at the time they are called.