Accuracy of Mathematical Functions in Single, Double, and Quadruple Precision

       Hi,

I have published a new version of my note, including results for double and
quadruple precision, and also for bivariate functions (atan2, hypot, pow):

https://homepages.loria.fr/PZimmermann/papers/#accuracy

Apart from the Bessel functions for which Newlib has large errors (like other
libraries except the Intel Math Library), this note shows large errors in
Newlib:

* for the binary32 lgamma (7.50e6 ulps), tgamma (239 ulps)
  and pow (169 ulps) functions

* for the binary64 lgamma and tgamma functions

Best regards,
Paul Zimmermann