Does using async/await add noticeable overhead compared to synchronous methods? How does the compiler generate the state machine behind the scenes?

When a method is marked async, the Roslyn compiler rewrites it into an IAsyncStateMachine struct. If the awaited task completes synchronously (like ValueTask.CompletedTask), it executes with zero allocation and near-zero overhead! If it yields, a heap frame is allocated to hold local variables until completion.

Using ValueTask<T> instead of Task<T> for methods that frequently complete synchronously (e.g. cache hits) reduces heap allocations to zero.