Instruction Combining
While we can't fully implement this without gas profiling and knowledge of instruction costs, we can do some basic combinations like folding successive adds to the same register into one add, etc.
relevant to optimize things within what are currently asm blocks: #855
We can't optimize inside asm blocks, so we would need to introduce intrinsics for a lot of basic opcodes.
folding successive adds to the same register into one add,
I'm not sure I understand this. Can you post a small example? Also any other patterns that need to be optimized, which you have in mind. That'll help me plan this better.
Closing this due to lack of information.
Related #5270.