In order to solve the problems of huge volume backscattering signal data and time-consuming calculations when retrieving wind velocity in high repetition frequency coherent lidar, a parallel retrieval algorithm based on multicore Digital Signal Processor(DSP) was proposed. In this algorithm, a high performance DSPC-8681 card integrating four Texas Instruments(TI) TMS320C6678 processor with eight cores in each processor was employed. The digital data from Analog-to-Digital Convertor(ADC) was allocated equally to each core in each processor and processed in Single Instruction Multiply Data(SIMD) way. The wind velocity was retrieved by maximum likelihood discrete spectrum peak estimation. On the condition of 10 000 pulses accumulation, the time resolution of 1 s was realized. The performances of the developed coherent wind lidar are demonstrated by experiments, which are listed as follow: detection range in line of sight is 3 600 m; range resolution is 60 m; velocity range is ±30 m/s; time resolution is 1 s; velocity accuracy by rotating hard target is better than 0.48 m/s.