high performance computing on graphics processing units: hgpu.org

Galois Field arithmetic has been used very frequently in popular security and error-correction applications. Montgomery multiplication is among the suitable methods used for accelerating modular multiplication, which is the most time consuming basic arithmetic operation. Montgomery multiplication is also suitable to be implemented in parallel. OpenCL, which is a portable, heterogeneous and parallel programming framework, is recently supported by a major FPGA vendor, Altera. Therefore it is now possible to exploit the advantages of using both FPGA and C based OpenCL language simulataneously. In this thesis, Montgomery multiplication algorithm is implemented on FPGA using OpenCL programming language. Performance of the proposed FPGA implementation is evaluated and compared with CPU and GPU platforms. Using different OpenCL specific directives, several FPGA configurations corresponding to different parallel architectures are implemented for different multiplication sizes.