Multiple Fourier transform method on an SIMD computer

We implement a block tridiagonal Toeplitz linear system solver called MFT on an SIMD massively parallel computer, the MasPar MP-2. MP-2 is suitable to the linear solver which has rich data-parallelism. We examine five implementation methods to identify actual problems that hamper high-performance computing. We observe that good data allocation method is important to achieve maximum throughput.