Design of linear algebra hardware accelerators dedicated to implementation in FPGA devices

In this paper, methods for hardware linear algebra accelerators dedicated to implementation in FPGA devices, are presented. The two design methods with the use an evolutionary algorithm for allocation mapping are described. The first method is dedicated to creation a parallel architectures which process data on systolic type, dedicated to implementation into multicontext FPGA devices. The second method is used for design of processor array implemented in classic FPGA devices. Parameters of designed accelerators for exemplary linear algebra algorithms are presented and compared with parameters described parallel architectures obtained with the use of another design methods. (Design of linear algebra hardware accelerators dedicated to implementation in FPGA devices.)