Fix ARM bits/mathdef.h

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Fix ARM bits/mathdef.h

Joseph Myers
No version of ARM hard or soft float has excess precision; all operate
on float values with float arithmetic, and on double values with
double arithmetic.  Thus glibc's default <bits/mathdef.h>, which
defines float_t to be double (so describing a generic system with
excess precision), is incorrect for ARM; in fact
sysdeps/arm/fpu/bits/mathdef.h is correct for all ARM cases, minus a
comment about FPA.

2006-11-09  Joseph Myers  <[hidden email]>

        * sysdeps/arm/fpu/bits/mathdef.h: Move to
        sysdeps/arm/bits/mathdef.h.  Remove comment about FPA.

diff -rupN ports.orig/sysdeps/arm/bits/mathdef.h ports/sysdeps/arm/bits/mathdef.h
--- ports.orig/sysdeps/arm/bits/mathdef.h 1970-01-01 00:00:00.000000000 +0000
+++ ports/sysdeps/arm/bits/mathdef.h 2006-11-09 22:26:19.000000000 +0000
@@ -0,0 +1,42 @@
+/* Copyright (C) 1999, 2000, 2004, 2006 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#if !defined _MATH_H && !defined _COMPLEX_H
+# error "Never use <bits/mathdef.h> directly; include <math.h> instead"
+#endif
+
+#if defined  __USE_ISOC99 && defined _MATH_H && !defined _MATH_H_MATHDEF
+# define _MATH_H_MATHDEF 1
+
+/* GCC does not promote `float' values to `double'.  */
+typedef float float_t; /* `float' expressions are evaluated as
+   `float'.  */
+typedef double double_t; /* `double' expressions are evaluated as
+   `double'.  */
+
+/* The values returned by `ilogb' for 0 and NaN respectively.  */
+# define FP_ILOGB0 (-2147483647)
+# define FP_ILOGBNAN (2147483647)
+
+#endif /* ISO C99 */
+
+#ifndef __NO_LONG_DOUBLE_MATH
+/* Signal that we do not really have a `long double'.  This disables the
+   declaration of all the `long double' function variants.  */
+# define __NO_LONG_DOUBLE_MATH 1
+#endif
diff -rupN ports.orig/sysdeps/arm/fpu/bits/mathdef.h ports/sysdeps/arm/fpu/bits/mathdef.h
--- ports.orig/sysdeps/arm/fpu/bits/mathdef.h 2004-03-15 21:44:08.000000000 +0000
+++ ports/sysdeps/arm/fpu/bits/mathdef.h 1970-01-01 00:00:00.000000000 +0000
@@ -1,44 +0,0 @@
-/* Copyright (C) 1999, 2000, 2004 Free Software Foundation, Inc.
-   This file is part of the GNU C Library.
-
-   The GNU C Library is free software; you can redistribute it and/or
-   modify it under the terms of the GNU Lesser General Public
-   License as published by the Free Software Foundation; either
-   version 2.1 of the License, or (at your option) any later version.
-
-   The GNU C Library is distributed in the hope that it will be useful,
-   but WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Lesser General Public License for more details.
-
-   You should have received a copy of the GNU Lesser General Public
-   License along with the GNU C Library; if not, write to the Free
-   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
-   02111-1307 USA.  */
-
-#if !defined _MATH_H && !defined _COMPLEX_H
-# error "Never use <bits/mathdef.h> directly; include <math.h> instead"
-#endif
-
-#if defined  __USE_ISOC99 && defined _MATH_H && !defined _MATH_H_MATHDEF
-# define _MATH_H_MATHDEF 1
-
-/* GCC does not promote `float' values to `double'.  */
-typedef float float_t; /* `float' expressions are evaluated as
-   `float'.  */
-typedef double double_t; /* `double' expressions are evaluated as
-   `double'.  */
-
-/* The values returned by `ilogb' for 0 and NaN respectively.  */
-# define FP_ILOGB0 (-2147483647)
-# define FP_ILOGBNAN (2147483647)
-
-#endif /* ISO C99 */
-
-#ifndef __NO_LONG_DOUBLE_MATH
-/* Signal that we do not really have a `long double'.  This disables the
-   declaration of all the `long double' function variants.  */
-/* XXX The FPA does support this but the patterns in GCC are currently
-   turned off.  */
-# define __NO_LONG_DOUBLE_MATH 1
-#endif

--
Joseph S. Myers
[hidden email]
Reply | Threaded
Open this post in threaded view
|

Re: Fix ARM bits/mathdef.h

Daniel Jacobowitz-2
On Thu, Nov 09, 2006 at 10:28:26PM +0000, Joseph S. Myers wrote:
> 2006-11-09  Joseph Myers  <[hidden email]>
>
> * sysdeps/arm/fpu/bits/mathdef.h: Move to
> sysdeps/arm/bits/mathdef.h.  Remove comment about FPA.

Thanks, you're right - I checked this in.

--
Daniel Jacobowitz
CodeSourcery