Lines Matching refs:error
1185 # If both INF, same sign => same as both, opposite => error.
2549 Similar to self._rescale(exp._exp) but with error checking.
4039 """Handles an error
4046 error = _condition_map.get(condition, condition)
4047 if error in self._ignored_flags:
4049 return error().handle(self, *args)
4051 self.flags[error] = 1
4052 if not self.traps[error]:
4058 raise error(explanation)
4097 # if an error occurs in the middle.
5247 an error condition, the exponent of the result of a quantize is always
5739 """Integer approximation to M*log(x/M), with absolute error boundable
5746 both cases these are upper bounds on the error; it will usually be
5788 approximation to 10**p * log10(c*10**e), with an absolute error of
5810 log_d = _ilog(c, M) # error < 5 + 22 = 27
5811 log_10 = _log10_digits(p) # error < 1
5815 log_d = 0 # error < 2.31
5816 log_tenpower = _div_nearest(f, 10**-p) # error < 0.5
5822 approximation to 10**p * log(c*10**e), with an absolute error of
5835 # compute approximation to 10**p*log(d), with error < 27
5841 c = _div_nearest(c, 10**-k) # error of <= 0.5 in c
5843 # _ilog magnifies existing error in c by a factor of at most 10
5844 log_d = _ilog(c, 10**p) # error < 5 + 22 = 27
5846 # p <= 0: just approximate the whole thing by 0; error < 2.31
5849 # compute approximation to f*10**p*log(10), with error < 11.
5853 # error in f * _log10_digits(p+extra) < |f| * 1 = |f|
5854 # after division, error < |f|/10**extra + 0.5 < 10 + 0.5 < 11
5861 # error in sum < 11+27 = 38; error after division < 0.38 + 0.5 < 1
5904 x/M <= 2.4, the absolute error in the result is bounded by 60 (and
5948 digits of precision, and with an error in d of at most 1. This is
5949 almost, but not quite, the same as the error being < 1ulp: when d
5950 = 10**(p-1) the error could be up to 10 ulp."""
5971 # error in result of _iexp < 120; error after division < 0.62
5982 of precision, and with an error in c of at most 1. (This is
5983 almost, but not quite, the same as the error being < 1ulp: when c
5984 == 10**(p-1) we can only guarantee error < 10ulp.)