Comparison of Scalable Montgomery Modular Multiplication Implementations Embedded in Reconfigurable Hardware
Résumé
This paper presents a comparison of possible approaches for an efficient implementation of Multiple-word radix-2 Montgomery Modular Multiplication (MM) on modern Field Programmable Gate Arrays (FPGAs). The hardware implementation of MM coprocessor is fully scalable what means that it can be reused in order to generate long-precision results independently on the word length of the originally proposed coprocessor. The first of analyzed implementations uses a data path based on traditionally used redundant carry-save adders, the second one exploits, in scalable designs not yet applied, standard carry-propagate adders with fast carry chain logic. As a control unit and a platform for purely software implementation an embedded soft-core processor Altera NIOS is employed. All implementations use large embedded memory blocks available in recent FPGAs. Speed and logic requirements comparisons are performed on the optimized software and combined hardware-software designs in Altera FPGAs. The issues of targeting a design specifically for a FPGA are considered taking into account the underlying architecture imposed by the target FPGA technology. It is shown that the coprocessors based on carry-save adders and carry-propagate adders provide comparable results in constrained FPGA implementations but in case of carry-propagate logic, the solution requires less embedded memory and provides some additional implementation advantages presented in the paper.
Origine : Fichiers produits par l'(les) auteur(s)