Analysis of high-performance floating-point arithmetic on FPGAs