Lines Matching defs:ulp
79 * On the definition of ULP
81 * The GLSL spec does not define ULP. However, it refers to IEEE 754, which
84 * ULP(x) is the distance between the closest floating point numbers
87 * Note that this means that when x = 2^n, ULP(x) = 2^(n-p-1), i.e. it is the
90 * Furthermore, it is assumed that ULP is calculated relative to the exact
94 * For details, see "On the definition of ulp(x)" by Jean-Michel Muller
98 double FloatFormat::ulp (double x, double count) const
109 // Harrison's ULP: choose distance to closest (i.e. next lower) at binade
116 // ULP cannot be lower than the smallest quantum.
335 oss << "ulp(" << arg << ")";
336 check(oss.str(), m_fmt->ulp(arg), ref);
