A parallel implementation for optimal lambda-calculus reduction