How does the compiler transform an async method into a state machine, and why does this matter for understanding performance?

The compiler rewrites an async method into a compiler-generated struct or class implementing IAsyncStateMachine, with a numeric state field tracking progress and each 'await' becoming a suspend/resume point — this transformation has real allocation costs (especially for a class-based state machine when the method captures local variables or is a class-based async method), which is why hot-path async methods sometimes prefer ValueTask over Task to reduce allocations.

// Conceptual illustration of what the compiler generates behind the scenes:
// class MethodNameStateMachine : IAsyncStateMachine {
//   int state;
//   TaskAwaiter awaiter;
//   void MoveNext() { /* resumes execution from the last await point */ }
// }