This paper examines a new measure of bidirectional sieve in hybrid parallelism based on SMP Cluster, which is the individual computational units joined together by the communication network, are usually shared-memory systems with one or more multicore processors. to high-efficiency optimization, we propose average allocate data into nodes and generate double-ended queues (deque) in order to Intra-Node simultaneously sifting out primes from deque's head and tail. Each node also create a FIFO queue as dynamic data buffer to ache temporary sieve from another nodes and Each node can exploit queue and deque to optimize sieve. Algorithms to combine these estimates and produce a huge speedup and efficiency are presented and justified. Our proposed model is verified through experimental study.