2002-08-17 Jeff Johnston <jjohnstn@redhat.com>

        * configure.host: Add powerpc*-*-eabispe* configuration.
        * libc/machine/powerpc/atosfix16.c: New fixed-point conversion file.
        * libc/machine/powerpc/atosfix32.c: Ditto.
        * libc/machine/powerpc/atosfix64.c: Ditto.
        * libc/machine/powerpc/atoufix16.c: Ditto.
        * libc/machine/powerpc/atoufix32.c: Ditto.
        * libc/machine/powerpc/atoufix64.c: Ditto.
        * libc/machine/powerpc/fix64.h: Ditto.
        * libc/machine/powerpc/simdldtoa.c: Ditto.
        * libc/machine/powerpc/strtosfix16.c: Ditto.
        * libc/machine/powerpc/strtosfix32.c: Ditto.
        * libc/machine/powerpc/strtosfix64.c: Ditto.
        * libc/machine/powerpc/strtoufix16.c: Ditto.
        * libc/machine/powerpc/strtoufix32.c: Ditto.
        * libc/machine/powerpc/strtoufix64.c: Ditto.
        * libc/machine/powerpc/ufix64toa.c: Ditto.
        * libc/machine/powerpc/configure.in: Add check for
        powerpc*-eabispe and add fixed-point conversion functions.
        * libc/machine/powerpc/configure: Regenerated.
        * libc/machine/powerpc/vfprintf.c[__SPE__]: Add support for
        %r and %R format specifiers which handle fixed-point data.
        * libc/machine/powerpc/vfscanf.c[__SPE__]: Ditto.
        * libc/machine/powerpc/machine/stdlib.h[__SPE__]: Add fixed-point
        function prototypes.

1 files changed, 206 insertions, 0 deletions

diff --git a/newlib/libc/machine/powerpc/strtosfix16.c b/newlib/libc/machine/powerpc/strtosfix16.c
new file mode 100644
index 000000000..ad629978c
--- /dev/null
+++ b/newlib/libc/machine/powerpc/strtosfix16.c
@@ -0,0 +1,206 @@
+/*
+FUNCTION
+        <<strtosfix16>>, <<strtosfix32>>, <<strtosfix64>>---string to signed fixed point
+
+INDEX
+	strtosfix16
+INDEX
+	strtosfix32
+INDEX
+	strtosfix64
+INDEX
+	_strtosfix16_r
+INDEX
+	_strtosfix32_r
+INDEX
+	_strtosfix64_r
+
+ANSI_SYNOPSIS
+	#include <stdlib.h>
+        __int16 strtosfix16 (const char *<[s]>, char **<[ptr]>);
+
+        __int32 strtosfix32 (const char *<[s]>, char **<[ptr]>);
+
+        __int64 strtosfix64 (const char *<[s]>, char **<[ptr]>);
+
+        __int16 _strtosfix16_r (void *<[reent]>, 
+                       const char *<[s]>, char **<[ptr]>);
+
+        __int32 _strtosfix32_r (void *<[reent]>, 
+                       const char *<[s]>, char **<[ptr]>);
+
+        __int64 _strtosfix64_r (void *<[reent]>, 
+                       const char *<[s]>, char **<[ptr]>);
+
+TRAD_SYNOPSIS
+	#include <stdlib.h>
+	__int16 strtosfix16 (<[s]>, <[ptr]>)
+        char *<[s]>;
+        char **<[ptr]>;
+
+	__int32 strtosfix32 (<[s]>, <[ptr]>)
+        char *<[s]>;
+        char **<[ptr]>;
+
+	__int64 strtosfix64 (<[s]>, <[ptr]>)
+        char *<[s]>;
+        char **<[ptr]>;
+
+	__int16 _strtosfix16_r (<[reent]>, <[s]>, <[ptr]>)
+	char *<[reent]>;
+        char *<[s]>;
+        char **<[ptr]>;
+
+	__int32 _strtosfix32_r (<[reent]>, <[s]>, <[ptr]>)
+	char *<[reent]>;
+        char *<[s]>;
+        char **<[ptr]>;
+
+	__int64 _strtosfix64_r (<[reent]>, <[s]>, <[ptr]>)
+	char *<[reent]>;
+        char *<[s]>;
+        char **<[ptr]>;
+
+DESCRIPTION
+        The function <<strtosfix16>> converts the string <<*<[s]>>> to
+	a fixed-point sign + 15-bits fraction representation.  The function 
+	follows the same rules as <<strtod>>.
+
+	The substring converted is the longest initial
+	subsequence of <[s]>, beginning with the first
+	non-whitespace character, that has the format:
+	.[+|-]<[digits]>[.][<[digits]>][(e|E)[+|-]<[digits]>] 
+	The substring contains no characters if <[s]> is empty, consists
+	entirely of whitespace, or if the first non-whitespace
+	character is something other than <<+>>, <<->>, <<.>>, or a
+	digit. If the substring is empty, no conversion is done, and
+	the value of <[s]> is stored in <<*<[ptr]>>>.  Otherwise,
+	the substring is converted, and a pointer to the final string
+	(which will contain at least the terminating null character of
+	<[s]>) is stored in <<*<[ptr]>>>.  If you want no
+	assignment to <<*<[ptr]>>>, pass a null pointer as <[ptr]>.
+
+	<<strtosfix32>> is identical to <<strtosfix16>> except that it 
+	converts to fixed-point sign + 31-bits fraction representation.
+	<<strtosfix64>> is also similar, except that it converts
+	to fixed-point sign + 63-bit fraction format.
+
+	The alternate functions <<_strtosfix16_r>>, <<_strtosfix32_r>>,
+	and <<_strtosfix64_r>> are reentrant versions.
+	The extra argument <[reent]> is a pointer to a reentrancy structure.
+
+RETURNS
+	The functions return the converted substring value, if any.  If
+	no conversion can be performed, then 0 is returned.  If the converted
+	value is a NaN, 0 is returned and errno is set to <<EDOM>>.
+	If the converted value exceeds the maximum positive fixed-point value, 
+	the output value is saturated to the maximum value and <<ERANGE>> is stored in 
+	errno.  If the converted value is less than the minimum fixed-point negative
+	value, then the output is saturated to the minimum value  and <<ERANGE>> is stored
+	in errno.  Otherwise, the converted value is returned in the
+	specified fixed-point format.
+
+PORTABILITY
+        <<strtosfix16>>, <<strtosfix32>>, and <<strtosfix64>> are non-standard.
+
+        The OS subroutines of <<strtod>> are required.
+*/
+
+#include <_ansi.h>
+#include <limits.h>
+#include <errno.h>
+#include <stdlib.h>
+#include <reent.h>
+#include "vfieeefp.h"
+
+/*
+ * Convert a string to a fixed-point (sign + 15-bits) value.
+ *
+ * Ignores `locale' stuff.
+ */
+__int16_t
+_DEFUN (_strtosfix16_r, (rptr, nptr, endptr),
+	struct _reent *rptr _AND
+	_CONST char *nptr _AND
+	char **endptr)
+{
+  union double_union dbl;
+  unsigned long tmp, tmp2;
+  int exp, negexp, sign;
+  __int16_t result;
+
+  dbl.d = _strtod_r (rptr, nptr, endptr);
+
+  /* treat NAN as domain error, +/- infinity as saturation */
+  if (!finite(dbl.d))
+    {
+      if (isnan (dbl.d))
+	{
+	  rptr->_errno = EDOM;
+	  return 0;
+	}
+      rptr->_errno = ERANGE;
+      if (word0(dbl) & Sign_bit)
+	return SHRT_MIN;
+      return SHRT_MAX;
+    }
+
+  /* check for normal saturation */
+  if (dbl.d >= 1.0)
+    {
+      rptr->_errno = ERANGE;
+      return SHRT_MAX;
+    }
+  else if (dbl.d < -1.0)
+    {
+      rptr->_errno = ERANGE;
+      return SHRT_MIN;
+    }
+
+  /* otherwise we have normal number in range */
+
+  /* strip off sign and exponent */
+  sign = word0(dbl) & Sign_bit;
+  exp = ((word0(dbl) & Exp_mask) >> Exp_shift) - Bias;
+  negexp = -exp;
+  if (negexp > 15)
+    return 0;
+  /* add in implicit normalized bit */
+  tmp = word0(dbl) | Exp_msk1;
+  /* remove exponent and sign */
+  tmp <<= Ebits;
+  if (negexp != 0)
+    {
+      /* perform rounding */
+      tmp2 = tmp + (1 << (negexp - 1));
+      result = (short)(tmp2 >> (negexp + 16));
+      /* check if rounding caused carry bit which must be added into result */
+      if (tmp2 < tmp)
+	result |= (1 << (16 - negexp));
+      /* check if positive saturation has occurred because of rounding */
+      if (!sign && result < 0)
+	{
+	  rptr->_errno = ERANGE;
+	  return SHRT_MAX;
+	}
+    }
+  else
+    {
+      /* we have -1.0, no rounding necessary */
+      return SHRT_MIN;
+    }
+
+  return  sign ? -result : result;
+}
+
+#ifndef _REENT_ONLY
+
+__int16_t
+_DEFUN (strtosfix16, (s, ptr, base),
+	_CONST char *s _AND
+	char **ptr)
+{
+  return _strtosfix16_r (_REENT, s, ptr);
+}
+
+#endif