Accurate arithmetic results for decimal data on non-decimal computers (Q1060810)

Recently, techniques have been devised and implemented which permit the computation of smallest enclosing machine number interval for the exact results of a good number of highly composite operations. These exact results refer, however, to the data as they are represented in the computer. This note shows how the conversion of decimal data into non- decimal representations may be joined with the mathematical operation on the data into one high-accuracy algorithm. Such an algorithm is explicitly presented for the solution of systems of linear equations.