Customising floating-point designs | IEEE Conference Publication | IEEE Xplore