Hot-code swapping would guarantee a really fast compile times and maybe even "hot reload" of program while it's running. Do not underestimate how much this helps with developer productivity. I'm so tired of the compile times with large C++ codebases...
Exactly! The idea is during development you use the super fast backend that supports hot code swapping. Then when you want to make a release build you use llvm and get all the optimization benefits of a longer compilation.
Comments
RE 3:
Hot-code swapping seems to require a lot of tricky code. Are there compelling use cases that justify the implementation effort?
The examples seem to be mostly focused on swapping out functions but I dot see how it could handle structural changes to data structures.
Hot-code swapping would guarantee a really fast compile times and maybe even "hot reload" of program while it's running. Do not underestimate how much this helps with developer productivity. I'm so tired of the compile times with large C++ codebases...
Exactly! The idea is during development you use the super fast backend that supports hot code swapping. Then when you want to make a release build you use llvm and get all the optimization benefits of a longer compilation.