摘要：【中文】本发明一种大矩阵快速转置多核并行处理方法，包括以下步骤：步骤一：DSP每个内核利用EDMA将外部大存储器中需处理子矩阵A_i(N,M),i∈[0,x‑1]搬移至SRAM缓存；步骤二、x个内核并行处理，CPU利用优化的内联函数，对缓存数据进行转置，得到A^Τ_i(N,M),i∈[0,x‑1]，再通过EDMA将结果数据搬移至外部大存储器。本发明提高了数据处理速度。 【EN】The invention discloses a large matrix fast transposition multi-core parallel processing method, which comprises the following steps: the method comprises the following steps: each kernel of the DSP utilizes EDMA to process the sub-matrix A in the external large memory_i(N,M),i∈[0,x‑1]Moving to an SRAM cache; step two, x kernels are processed in parallel, and the CPU transposes the cache data by utilizing the optimized inline function to obtain A^Τ_i(N,M),i∈[0,x‑1]And then the result data is moved to an external large memory through the EDMA. The invention improves the data processing speed.