11:03
So I have an idea on how to do the JITConfig
, but not sure of how it would
be implemented and such.
11:04
Maybe a set of available options. I want something simple, but not something super complex such as the Linux configuration. Really the JIT is only concerned with options the JIT wants.
11:59
Ok, so like I thought of the other day, JITConfig
should not care about
anything. There would be things which use the config to do things.
18:12
Ok so with this design, the class decompiler and the HIL do not need to know anything about the JIT configuration.
18:17
Actually this is most effective. With my current design the considerations which are major at link time will happen at link time, this way at this early point in the JIT I do not really have to care about much at all.
20:32
I am going to make it so the JIT can be ran in parallel, I have a way where it can be done. Basically the JIT that turns the high level binary into machine code can be done in a way where multiple processes can be ran at the same time regardless of how fast input class processing is performed. Handling of JARs is a strictly linear process which should be rather fast, the compilation and optimization phase may take longer to perform. After initial work, I will want to make it where the later parts can run on as many CPUs as are available. This would definitely increase compilation speed.