Inlinekall

Inlinekall is a proposed compiler optimization technique that guides inlining decisions to expand callee bodies into their call sites. The aim is to reduce function call overhead and improve data locality without causing unbounded code growth. Inlinekall can operate in both ahead-of-time and just-in-time compilation contexts and is typically implemented as part of the inlining subsystem of a compiler or virtual machine.

How it works: Inlinekall uses a decision engine that evaluates runtime and static information such as hot paths from profiling data, code size budgets, callee side effects, and deoptimization costs. When the engine determines that inlining a callee will improve performance within a given constraint, the compiler expands the callee’s body into the caller. It supports guarded inlining, where guards are inserted to recover if assumptions fail, and deoptimization to revert to a slower path. Inlinekall often integrates with inline hints or attributes (for example, an inlinekall annotation) to bias decisions, and works with interprocedural analysis across modules when permitted by the language and toolchain.

Benefits and limitations: The technique can reduce function call overhead, improve instruction-cache locality, and enable further optimizations. However, it can increase code size and compile time, and mispredicted inlining can degrade performance. Practical use relies on careful tuning of thresholds and guard strategies.

Adoption: Inlinekall-type strategies appear in many modern runtimes as part of broader inlining and optimization frameworks, often alongside other techniques such as devirtualization and speculative optimization.