[PATCH] frexpl: Support smaller long double.
Takashi Yano
takashi.yano@nifty.ne.jp
Wed Dec 1 19:34:34 GMT 2021
- Currently, frexpl() supports only the following cases.
1) LDBL_MANT_DIG == 64 or 113
2) 'long double' is equivalent to 'double'
This patch add support for smaller long double, LDBL_MANT_DIG is
24 or 53.
---
newlib/libm/common/frexpl.c | 78 +++++++++++++++++++++++++++++++------
1 file changed, 66 insertions(+), 12 deletions(-)
diff --git a/newlib/libm/common/frexpl.c b/newlib/libm/common/frexpl.c
index 22c811689..8fc6b4a9c 100644
--- a/newlib/libm/common/frexpl.c
+++ b/newlib/libm/common/frexpl.c
@@ -33,16 +33,66 @@ POSSIBILITY OF SUCH DAMAGE.
#include "local.h"
/* On platforms where long double is as wide as double. */
-#if defined(_LDBL_EQ_DBL) || (LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024)
+#if defined(_LDBL_EQ_DBL)
long double
frexpl (long double x, int *eptr)
{
return frexp(x, eptr);
}
-#endif
+#else /* !_DBL_EQ_DBL */
+# if (LDBL_MANT_DIG == 24) /* 32-bit long double */
+static const double scale = 0x1p25;
+
+union ldbl {
+ long double x;
+ struct {
+# ifdef __IEEE_LITTLE_ENDIAN /* for Intel CPU */
+ __uint32_t frac:23;
+ __uint32_t exp:8;
+ __uint32_t sign:1;
+# endif
+# ifdef __IEEE_BIG_ENDIAN
+# ifndef ___IEEE_BYTES_LITTLE_ENDIAN
+ __uint32_t sign:1;
+ __uint32_t exp:8;
+ __uint32_t frac:23;
+# else /* ARMEL without __VFP_FP__ */
+ __uint32_t frac:23;
+ __uint32_t exp:8;
+ __uint32_t sign:1;
+# endif
+# endif
+ } u32;
+};
+# elif (LDBL_MANT_DIG == 53) /* 64-bit long double */
+static const double scale = 0x1p54;
+
+union ldbl {
+ long double x;
+ struct {
+# ifdef __IEEE_LITTLE_ENDIAN /* for Intel CPU */
+ __uint32_t fracl;
+ __uint32_t frach:20;
+ __uint32_t exp:11;
+ __uint32_t sign:1;
+# endif
+# ifdef __IEEE_BIG_ENDIAN
+ __uint32_t sign:1;
+ __uint32_t exp:11;
+ __uint32_t frach:20;
+# ifndef ___IEEE_BYTES_LITTLE_ENDIAN
+# else /* ARMEL without __VFP_FP__ */
+ __uint32_t frach:20;
+ __uint32_t exp:11;
+ __uint32_t sign:1;
+# endif
+ __uint32_t fracl;
+# endif
+ } u32;
+};
+# elif (LDBL_MANT_DIG == 64) /* 80-bit long double */
+static const double scale = 0x1p65;
-#if (LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 113) && LDBL_MAX_EXP == 16384
-# if (LDBL_MANT_DIG == 64) /* 80-bit long double */
union ldbl {
long double x;
struct {
@@ -68,7 +118,9 @@ union ldbl {
# endif
} u32;
};
-# else /* LDBL_MANT_DIG == 113, 128-bit long double */
+# elif (LDBL_MANT_DIG == 113) /* 128-bit long double */
+static const double scale = 0x1p114;
+
union ldbl {
long double x;
struct {
@@ -96,9 +148,11 @@ union ldbl {
# endif
} u32;
};
+# else
+# error Unsupported long double format.
# endif
-static const double two114 = 0x1p114;
+static const int scale_exp = LDBL_MANT_DIG + 1;
long double
frexpl (long double x, int *eptr)
@@ -107,16 +161,16 @@ frexpl (long double x, int *eptr)
u.x = x;
int e = u.u32.exp;
*eptr = 0;
- if (e == 0x7fff || x == 0)
+ if (e == (LDBL_MAX_EXP*2 - 1) || x == 0)
return x; /* inf,nan,0 */
if (e == 0) /* subnormal */
{
- u.x *= two114;
+ u.x *= scale;
e = u.u32.exp;
- *eptr -= 114;
+ *eptr -= scale_exp;
}
- *eptr += e - 16382;
- u.u32.exp = 0x3ffe; /* 0 */
+ *eptr += e - (LDBL_MAX_EXP - 2);
+ u.u32.exp = LDBL_MAX_EXP - 2; /* -1 */
return u.x;
}
-#endif /* End of 80-bit or 128-bit long double */
+#endif /* !_LDBL_EQ_DBL */
--
2.33.0
More information about the Newlib
mailing list