On Wed, 2017-01-25 at 07:53 -0700, Keith Bierman wrote: > > Date: Tue, 24 Jan 2017 00:55:20 -0800 > > From: Anton Shterenlikht <[log in to unmask]> > > Subject: f2008, minus zero, complex intrinsics on branch cuts and > > points > > > > Some time ago I was hit by incorrect > > implementation of some complex intrinsic functions > I would suggest trying it with fdlibm (http://www.netlib.org/fdlibm/) > in place of the system one. It's helpful to have the same interfaces for procedures that claim to calculate elementary functions and a few special functions for Fortran and for C, and the same set of procedure functionality. But that doesn't require the intrinsic functions provided with Fortran processors to reproduce all the mistakes in libm bit for bit.