The invention belongs to the technical field of parallel optimization of computer multi-core structures, and particularly relates to a thread level parallel data optimization method and device for a shared memory multi-core structure, and the method comprises the steps: detecting and judging whether all cyclic carrying dependencies of to-be-optimized data have privatization advantages or not; Anddata privatization processing is carried out according to a detection judgment result. According to the method, for the circular parallel thread in the multi-core hardware structure, firstly, the advantages of privatization processing are detected, then the circular carrying dependence is eliminated through privatization of the data, and then multithread parallel of a program is achieved; The dataprivatization is further divided into scalar privatization and array privatization; The advantages of the multi-core hardware structure of shared storage are fully utilized, the thread execution efficiency is improved, the utilization rate of the performance of a processor and even a computer system is increased, the operation expenditure is reduced, and the method has important guiding significance for the thread-level parallel processing technology of the shared memory multi-core structure.