package mesos // fixed point scalar math from mesos:src/common/values.cpp // -- // We manipulate scalar values by converting them from floating point to a // fixed point representation, doing a calculation, and then converting // the result back to floating point. We deliberately only preserve three // decimal digits of precision in the fixed point representation. This // ensures that client applications see predictable numerical behavior, at // the expense of sacrificing some precision. import "math" func convertToFloat64(f int64) float64 { // NOTE: We do the conversion from fixed point via integer division // and then modulus, rather than a single floating point division. // This ensures that we only apply floating point division to inputs // in the range [0,999], which is easier to check for correctness. var ( quotient = float64(f / 1000) remainder = float64(f%1000) / 1000.0 ) return quotient + remainder } func convertToFixed64(f float64) int64 { return round64(f * 1000) } func round64(f float64) int64 { if math.Abs(f) < 0.5 { return 0 } return int64(f + math.Copysign(0.5, f)) }