new file mode 100644
--- /dev/null
+++ b/lib/float+.h
@@ -0,0 +1,150 @@
+/* Supplemental information about the floating-point formats.
+   Copyright (C) 2007 Free Software Foundation, Inc.
+   Written by Bruno Haible <bruno@clisp.org>, 2007.
+
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 2, or (at your option)
+   any later version.
+
+   This program 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 General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software Foundation,
+   Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.  */
+
+#ifndef _FLOATPLUS_H
+#define _FLOATPLUS_H
+
+#include <float.h>
+#include <limits.h>
+
+/* Number of bits in the mantissa of a floating-point number, including the
+   "hidden bit".  */
+#if FLT_RADIX == 2
+# define FLT_MANT_BIT FLT_MANT_DIG
+# define DBL_MANT_BIT DBL_MANT_DIG
+# define LDBL_MANT_BIT LDBL_MANT_DIG
+#elif FLT_RADIX == 4
+# define FLT_MANT_BIT (FLT_MANT_DIG * 2)
+# define DBL_MANT_BIT (DBL_MANT_DIG * 2)
+# define LDBL_MANT_BIT (LDBL_MANT_DIG * 2)
+#elif FLT_RADIX == 16
+# define FLT_MANT_BIT (FLT_MANT_DIG * 4)
+# define DBL_MANT_BIT (DBL_MANT_DIG * 4)
+# define LDBL_MANT_BIT (LDBL_MANT_DIG * 4)
+#endif
+
+/* Bit mask that can be used to mask the exponent, as an unsigned number.  */
+#define FLT_EXP_MASK ((FLT_MAX_EXP - FLT_MIN_EXP) | 7)
+#define DBL_EXP_MASK ((DBL_MAX_EXP - DBL_MIN_EXP) | 7)
+#define LDBL_EXP_MASK ((LDBL_MAX_EXP - LDBL_MIN_EXP) | 7)
+
+/* Number of bits used for the exponent of a floating-point number, including
+   the exponent's sign.  */
+#define FLT_EXP_BIT \
+  (FLT_EXP_MASK < 0x100 ? 8 : \
+   FLT_EXP_MASK < 0x200 ? 9 : \
+   FLT_EXP_MASK < 0x400 ? 10 : \
+   FLT_EXP_MASK < 0x800 ? 11 : \
+   FLT_EXP_MASK < 0x1000 ? 12 : \
+   FLT_EXP_MASK < 0x2000 ? 13 : \
+   FLT_EXP_MASK < 0x4000 ? 14 : \
+   FLT_EXP_MASK < 0x8000 ? 15 : \
+   FLT_EXP_MASK < 0x10000 ? 16 : \
+   FLT_EXP_MASK < 0x20000 ? 17 : \
+   FLT_EXP_MASK < 0x40000 ? 18 : \
+   FLT_EXP_MASK < 0x80000 ? 19 : \
+   FLT_EXP_MASK < 0x100000 ? 20 : \
+   FLT_EXP_MASK < 0x200000 ? 21 : \
+   FLT_EXP_MASK < 0x400000 ? 22 : \
+   FLT_EXP_MASK < 0x800000 ? 23 : \
+   FLT_EXP_MASK < 0x1000000 ? 24 : \
+   FLT_EXP_MASK < 0x2000000 ? 25 : \
+   FLT_EXP_MASK < 0x4000000 ? 26 : \
+   FLT_EXP_MASK < 0x8000000 ? 27 : \
+   FLT_EXP_MASK < 0x10000000 ? 28 : \
+   FLT_EXP_MASK < 0x20000000 ? 29 : \
+   FLT_EXP_MASK < 0x40000000 ? 30 : \
+   FLT_EXP_MASK <= 0x7fffffff ? 31 : \
+   32)
+#define DBL_EXP_BIT \
+  (DBL_EXP_MASK < 0x100 ? 8 : \
+   DBL_EXP_MASK < 0x200 ? 9 : \
+   DBL_EXP_MASK < 0x400 ? 10 : \
+   DBL_EXP_MASK < 0x800 ? 11 : \
+   DBL_EXP_MASK < 0x1000 ? 12 : \
+   DBL_EXP_MASK < 0x2000 ? 13 : \
+   DBL_EXP_MASK < 0x4000 ? 14 : \
+   DBL_EXP_MASK < 0x8000 ? 15 : \
+   DBL_EXP_MASK < 0x10000 ? 16 : \
+   DBL_EXP_MASK < 0x20000 ? 17 : \
+   DBL_EXP_MASK < 0x40000 ? 18 : \
+   DBL_EXP_MASK < 0x80000 ? 19 : \
+   DBL_EXP_MASK < 0x100000 ? 20 : \
+   DBL_EXP_MASK < 0x200000 ? 21 : \
+   DBL_EXP_MASK < 0x400000 ? 22 : \
+   DBL_EXP_MASK < 0x800000 ? 23 : \
+   DBL_EXP_MASK < 0x1000000 ? 24 : \
+   DBL_EXP_MASK < 0x2000000 ? 25 : \
+   DBL_EXP_MASK < 0x4000000 ? 26 : \
+   DBL_EXP_MASK < 0x8000000 ? 27 : \
+   DBL_EXP_MASK < 0x10000000 ? 28 : \
+   DBL_EXP_MASK < 0x20000000 ? 29 : \
+   DBL_EXP_MASK < 0x40000000 ? 30 : \
+   DBL_EXP_MASK <= 0x7fffffff ? 31 : \
+   32)
+#define LDBL_EXP_BIT \
+  (LDBL_EXP_MASK < 0x100 ? 8 : \
+   LDBL_EXP_MASK < 0x200 ? 9 : \
+   LDBL_EXP_MASK < 0x400 ? 10 : \
+   LDBL_EXP_MASK < 0x800 ? 11 : \
+   LDBL_EXP_MASK < 0x1000 ? 12 : \
+   LDBL_EXP_MASK < 0x2000 ? 13 : \
+   LDBL_EXP_MASK < 0x4000 ? 14 : \
+   LDBL_EXP_MASK < 0x8000 ? 15 : \
+   LDBL_EXP_MASK < 0x10000 ? 16 : \
+   LDBL_EXP_MASK < 0x20000 ? 17 : \
+   LDBL_EXP_MASK < 0x40000 ? 18 : \
+   LDBL_EXP_MASK < 0x80000 ? 19 : \
+   LDBL_EXP_MASK < 0x100000 ? 20 : \
+   LDBL_EXP_MASK < 0x200000 ? 21 : \
+   LDBL_EXP_MASK < 0x400000 ? 22 : \
+   LDBL_EXP_MASK < 0x800000 ? 23 : \
+   LDBL_EXP_MASK < 0x1000000 ? 24 : \
+   LDBL_EXP_MASK < 0x2000000 ? 25 : \
+   LDBL_EXP_MASK < 0x4000000 ? 26 : \
+   LDBL_EXP_MASK < 0x8000000 ? 27 : \
+   LDBL_EXP_MASK < 0x10000000 ? 28 : \
+   LDBL_EXP_MASK < 0x20000000 ? 29 : \
+   LDBL_EXP_MASK < 0x40000000 ? 30 : \
+   LDBL_EXP_MASK <= 0x7fffffff ? 31 : \
+   32)
+
+/* Number of bits used for a floating-point number: the mantissa (not
+   counting the "hidden bit", since it may or may not be explicit), the
+   exponent, and the sign.  */
+#define FLT_TOTAL_BIT ((FLT_MANT_BIT - 1) + FLT_EXP_BIT + 1)
+#define DBL_TOTAL_BIT ((DBL_MANT_BIT - 1) + DBL_EXP_BIT + 1)
+#define LDBL_TOTAL_BIT ((LDBL_MANT_BIT - 1) + LDBL_EXP_BIT + 1)
+
+/* Number of bytes used for a floating-point number.
+   This can be smaller than the 'sizeof'.  For example, on i386 systems,
+   'long double' most often have LDBL_MANT_BIT = 64, LDBL_EXP_BIT = 16, hence
+   LDBL_TOTAL_BIT = 80 bits, i.e. 10 bytes of consecutive memory, but
+   sizeof (long double) = 12 or = 16.  */
+#define SIZEOF_FLT ((FLT_TOTAL_BIT + CHAR_BIT - 1) / CHAR_BIT)
+#define SIZEOF_DBL ((DBL_TOTAL_BIT + CHAR_BIT - 1) / CHAR_BIT)
+#define SIZEOF_LDBL ((LDBL_TOTAL_BIT + CHAR_BIT - 1) / CHAR_BIT)
+
+/* Verify that SIZEOF_FLT <= sizeof (float) etc.  */
+typedef int verify_sizeof_flt[2 * (SIZEOF_FLT <= sizeof (float)) - 1];
+typedef int verify_sizeof_dbl[2 * (SIZEOF_DBL <= sizeof (double)) - 1];
+#if HAVE_LONG_DOUBLE
+typedef int verify_sizeof_ldbl[2 * (SIZEOF_LDBL <= sizeof (long double)) - 1];
+#endif
+
+#endif /* _FLOATPLUS_H */