Exploiting the Performance of 32 bit Floating Point Arithmetic in Obtaining 64 bit Accuracy