桑红石, 高伟. 高效可配置浮点FFT处理器设计[J]. 微电子学与计算机, 2012, 29(4): 36-40.
引用本文: 桑红石, 高伟. 高效可配置浮点FFT处理器设计[J]. 微电子学与计算机, 2012, 29(4): 36-40.
SANG Hong-shi, GAO Wei. An Efficient Architecture Design of Reconfigurable Float-point FFT Processor[J]. Microelectronics & Computer, 2012, 29(4): 36-40.
Citation: SANG Hong-shi, GAO Wei. An Efficient Architecture Design of Reconfigurable Float-point FFT Processor[J]. Microelectronics & Computer, 2012, 29(4): 36-40.

高效可配置浮点FFT处理器设计

An Efficient Architecture Design of Reconfigurable Float-point FFT Processor

  • 摘要: 为了克服高精度浮点FFT处理器具有较大资源开销的设计瓶颈, 采用基于单口存储器的FIFO构建共享蝶形结构的R2/22SDF流水可配置结构.采用适合浮点设计的基2/22算法实现流水结构, 不仅有利于可配置电路的实现, 还能够有效减少复数乘法次数, 提高复数乘法器的计算效率.采用双倍数据位宽的单口存储器实现FIFO存储器, 有效避免了双口存储器面积和功耗较大的问题.改进的蝶形共享结构实现两级蝶形的合并, 解决了单路径延迟反馈流水线结构蝶形单元利用率低的问题.与传统流水线结构FFT处理器设计相比, 有效降低了浮点设计中的资源开销, 提高了计算单元的利用效率.

     

    Abstract: Large resource cost is the design bottleneck of high-precision float-point FFT processor, a novel R2/22SDF reconfigurable architecture using shared-butterfly which employs single-port-based FIFO.Radix 2/22algorithm and pipeline architecture, which is suitable for float-point design, can reduce the multiplicative complexity and improve the multiplication efficiency.The FIFO memory using double-width single-port ram can avoid the larger area and power coat of dual-port ram.Two butterfly units can be merged by the proposed shared-butterfly architecture, which solves the low utilization factor problem of traditional single-path-delay-feedback architecture.The float-point design cost is efficiently reduced and the calculator utilization factor is improved, compared with the traditional pipeline method.

     

/

返回文章
返回