These are found when building with -Wfloat-conversion.
Also use less stack space when printing single precision float. Addition of asinf and acosf addresses issue #851.