What we do: 1. Inline functions (not really an "op...

# language-proposals

[dmitry.petrov]

What we do:
1. Inline functions (not really an "optimization", but) - avoid unnecessary heap allocation and class/method generation for lambdas. Inlining is performed on bytecode-level and allows more extensive bytecode-level transformations.
2. Some expressions (for, when, string templates / string concatenation) have specific code generation schemes (e.g., for over a range, when by constant selectors, etc) that produce "optimized" bytecode.
3. A few functions and properties are implemented as compiler intrinsics (yet again, a particular bytecode generation scheme provided by compiler).
Sometimes it can be treated as an "optimization". E.g., A::class.java doesn't create an intermediate KClass object.
4. Bytecode postprocessing steps eliminate redundant boxing/unboxing, redundant null checks, and dead code. Such transformations are intraprocedural (that is, within a context of a single method), so, as mentioned above, inline functions provide additional benefits here.