Definitely true; there are some overheads that could be low-hanging fruit to optimize, but the real solution is most likely to use a tiered-compilation system. ie only invoke the full LLVM code generator once a function has been called 10,000 times. It's definitely an open question as to how to get faster compilation times out of it; we have a simple LLVM interpreter but since LLVM isn't designed for interpretation it's pretty slow.
I think you can look at the people who are interested in adding LLVM tiers to their JITs (ex Apple, Facebook) to see that we're not the only ones that think there can be a place for "good but expensive" code generation in a JIT.
Comments
Definitely true; there are some overheads that could be low-hanging fruit to optimize, but the real solution is most likely to use a tiered-compilation system. ie only invoke the full LLVM code generator once a function has been called 10,000 times. It's definitely an open question as to how to get faster compilation times out of it; we have a simple LLVM interpreter but since LLVM isn't designed for interpretation it's pretty slow.
I think you can look at the people who are interested in adding LLVM tiers to their JITs (ex Apple, Facebook) to see that we're not the only ones that think there can be a place for "good but expensive" code generation in a JIT.
Like this?
https://github.com/dropbox/pyston#compilation-tiers