.56ns per operation using "optimized" writer but without touching the ring buffer.
1 file changed