The Rise of Mixed Precision Arithmetic Nick Higham, 20 Oct 2015 For the last 30 years, most floating point calculations in scientific computing have been carried out in 64-bit IEEE double…