CAJ | 학술논문

数字信号处理器的内存较小，而且数字信号处理领域的应用往往是数据密集型，这要求在设计数字信号处理应用算法时既要考虑时间复杂度又要兼顾算法的空间复杂度。为此提出了一种原位的逆序算法；针对数字信号处理器比较高的内存访问并行度，设计了部分逆序的原位高效 FFT算法；并在魂芯 DSP 平台上实现了该算法框架。实验表明，与非原位 FFT 算法相比，该原位算法的空间复杂度大幅降低而时间效率的损失在可接受范围之内。
수자신호처리기적내존교소，이차수자신호처리영역적응용왕왕시수거밀집형，저요구재설계수자신호처리응용산법시기요고필시간복잡도우요겸고산법적공간복잡도。위차제출료일충원위적역서산법；침대수자신호처리기비교고적내존방문병행도，설계료부분역서적원위고효 FFT산법；병재혼심 DSP 평태상실현료해산법광가。실험표명，여비원위 FFT 산법상비，해원위산법적공간복잡도대폭강저이시간효솔적손실재가접수범위지내。
The on‐chip memory in DSP is small ,and applications for DSP are often data‐intensive ,which requires that space complexity as well as time complexity must be considered when algorithms are designed .So a in‐place bit reverse algorithm was proposed .Then ,to take advantage of memory bandwidth offered by DSP ,an effective in‐place FFT algorithm with part bit reverse was designed and implemented on BWDSP . Experiment result shows that , compared with the out‐of‐place FFT algorithm , its space complexity is significantly reduced ,while the loss of time efficiency for the proposed in‐place FFT algorithm is acceptable .