DISCLAIMER: These notes are from the defunct k8 project which precedes SquirrelJME. The notes for SquirrelJME start on 2016/02/26! The k8 project was effectively a Java SE 8 operating system and as such all of the notes are in the context of that scope. That project is no longer my goal as SquirrelJME is the spiritual successor to it.
After splitting the parts of the code, recompilation does seem much faster as less dependencies and less code is compiled in two independent packages. I figure for the SSA register allocation that it is segment (basic block) local so it is easier to generate code for (live and not live, etc.).
In another note, LLVM's basic blocks have it so invoke of external functions is at the end of a basic block. Currently for me I have it so the invoke of methods is not on a terminating block. Based on reading GCC does what I do in that function calls do not mean the end of a block. Also Wikipedia states that any instruction that can throw an exception is the end of a block, but that would mean (in terms of Java) that anything can throw an exception for the most part, which would mean very small blocks.
I suppose register allocation is the hardest part of SSA, since it is NP- complete. Means it can take decades to find the optimal solution with current algorithms. I could go for sub-optimal and remove the NP-complete taking forever nature of it however.