Lines Matching full:gamma
2 * Implementation of the true gamma function (as opposed to lgamma)
10 * This module implements the float128 gamma function under the name
46 * relates gamma(-x) and gamma(x).
76 * In this range we compute gamma(x) as x^(x-1/2) * e^-x * K, in tgamma_large()
88 * but gamma(x) does not, we split x^(x-0.5) in half and in tgamma_large()
105 * turn a case where gamma(+x) would overflow into a case in tgamma_large()
106 * where gamma(-x) doesn't underflow. Not only that, but the in tgamma_large()
122 * g = 1/(x*gamma(x)), and then return 1/(g*x). in tgamma_tiny()
135 /* On this interval, gamma can't even be distinguished from 1/x, in tgamma_ultratiny()
150 * difference between x and the point where gamma has a minimum, in tgamma_central()
197 return x-x; /* gamma(-inf) has indeterminate sign, so provoke an in tgamma128()
200 return x; /* but gamma(+inf) is just +inf with no error */ in tgamma128()
217 * gamma(1-x) gamma(x) = pi/sin(pi*x) in tgamma128()
220 * => gamma(x) = -------------------- in tgamma128()
221 * gamma(1-x) sin(pi*x) in tgamma128()
225 * gamma(t+1)=t gamma(t). Setting t=-x, this gives us in tgamma128()
226 * gamma(1-x) = -x gamma(-x), so we now have in tgamma128()
229 * gamma(x) = ---------------------- in tgamma128()
230 * -x gamma(-x) sin(pi*x) in tgamma128()
232 * which relates gamma(x) to gamma(-x), which is much nicer, in tgamma128()
242 * 1 / (gamma(x) * x * negadjust) in tgamma128()
246 * For some of the cases below, we'll compute gamma(x) in tgamma128()
276 * gamma of any positive value this large overflows, and gamma in tgamma128()
301 * gamma(x) = gamma(x+1)/x. in tgamma128()