In software transactional memory, transactions are aborted forcefully because correctness, which must be retried as new transactions immediately. After aborted and retried, transactions may conflict again, leading to more aborts and degrading the overall throughput. An abort manage algorithm that control all aborted transactions was presented. The results showed that forcefully aborted transactions are not same as the new and normal aborted ones, which avoid conflict and give chance to the others. After a while, these transactions restart. So the forcefully aborted transactions avoid conflict and should be helped to access the share memory.
