We implemented a new method for reducing the execution time of Real 1024p radix-2 DIT FFT algorithm write in assembler on VisualDSP++ 3.5 (ADSP2181), and we compared the results of this method with the execution time of the Real 1024p radix-2 DIT FFT ( without BFP) that we found in VisualDSP++ 3.5 Help . But we can't compare these results with the Benchmark because we don’t know the kernel of this algorithm. Then we need the kernel assembly code source of Real 1024p radix-2 DIT FFT without block-floating point or some information about the kernel algorithm (Benchmarks).
↧