Mercurial > octave-nkf
changeset 9837:7c70084b125e
improve comment for 9835
| author | Jaroslav Hajek <highegg@gmail.com> |
|---|---|
| date | Thu, 19 Nov 2009 12:22:08 +0100 |
| parents | 804c21f3659b |
| children | 55219e65c7cd |
| files | liboctave/lo-specfun.cc |
| diffstat | 1 files changed, 3 insertions(+), 3 deletions(-) [+] |
line wrap: on
line diff
--- a/liboctave/lo-specfun.cc Thu Nov 19 11:15:23 2009 +0100 +++ b/liboctave/lo-specfun.cc Thu Nov 19 12:22:08 2009 +0100 @@ -3101,9 +3101,9 @@ return x < -1.0f ? FloatComplex (logf (-(1.0f + x)), pi) : FloatComplex (log1pf (x)); } -// This algorithm is due to P. Jacklam -// See http:// -// The rational approximation has relative accuracy 1e-9 in the whole region. +// This algorithm is due to P. Jacklam. +// See http://home.online.no/~pjacklam/notes/invnorm/ +// The rational approximation has relative accuracy 1.15e-9 in the whole region. // For doubles, it is refined by a single step of Higham's 3rd order method. // For single precision, the accuracy is already OK, so we skip it to get // faster evaluation.