From: Kbwms AT aol DOT com Message-ID: Date: Sun, 13 Jul 2003 18:26:02 EDT Subject: Re: Testing the Math Functions To: djgpp-workers AT delorie DOT com MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="part1_f3.2dfdd493.2c43367a_boundary" X-Mailer: 8.0 for Windows sub 6011 Reply-To: djgpp-workers AT delorie DOT com --part1_f3.2dfdd493.2c43367a_boundary Content-Type: text/plain; charset="ISO-8859-1" Content-Transfer-Encoding: quoted-printable News from the C99 library development front: The existing assembler version of modfl() has at least two bugs: y =3D modfl(x,&K) When x =3D -N, return should be -0 and -N stored in K. =A0 Currently, +0 is=20 returned. When x =3D Inf, return should be 0. =A0 Currently, NaN is returned. My new library of long double functions contains a C version of modfl(). Th= e=20 library will be delivered in the near future. Currently, my attention is=20 fixed on errors of various kinds in my documentation. I have written several new functions for the long double library including=20 hypotl(). There is a version of hypotl() extant in libc but it isn't workin= g=20 properly. The new version is fast and very reliable.=20 Also, I'm writing new test-vector generators for the test suite that resides= =20 in archive djtst203.zip. These generators will produce data for tests of th= e=20 long double precision functions. KB Williams --part1_f3.2dfdd493.2c43367a_boundary Content-Type: text/html; charset="ISO-8859-1" Content-Transfer-Encoding: quoted-printable News from the C99 library development front:


The existing assembler version of modfl() has at least two bugs:

y =3D modfl(x,&K)

When x =3D -N, return should be -0 and -N stored in K. =A0 Currently, +0 is=20= returned.

When x =3D Inf, return should be 0. =A0 Currently, NaN is returned.


My new library of long double functions contains a C version of modfl().&nbs= p; The library will be delivered in the near future.  Currently, my att= ention is fixed on errors of various kinds in my documentation.

I have written several new functions for the long double library including h= ypotl().  There is a version of hypotl() extant in libc but it isn't wo= rking properly.  The new version is fast and very reliable.

Also, I'm writing new test-vector generators for the test suite that resides= in archive djtst203.zip.  These generators will produce data for tests= of the long double precision functions.


KB Williams --part1_f3.2dfdd493.2c43367a_boundary--