On sound relative error bounds for floating-point arithmetic | IEEE Conference Publication | IEEE Xplore