Mixed abstractions for floating-point arithmetic | IEEE Conference Publication | IEEE Xplore