The case for a redundant format in floating point arithmetic | IEEE Conference Publication | IEEE Xplore