uint64_t low;
uint16_t high;
} floatx80;
+#define make_floatx80(exp, mant) ((floatx80) { mant, exp })
#endif
#ifdef FLOAT128
typedef struct {
return ((a.high & 0x7fff) == 0x7fff) && (a.low<<1);
}
+#define floatx80_zero make_floatx80(0x0000, 0x0000000000000000LL)
+#define floatx80_one make_floatx80(0x3fff, 0x8000000000000000LL)
+#define floatx80_ln2 make_floatx80(0x3ffe, 0xb17217f7d1cf79acLL)
+#define floatx80_half make_floatx80(0x3ffe, 0x8000000000000000LL)
+#define floatx80_infinity make_floatx80(0x7fff, 0x8000000000000000LL)
+
/*----------------------------------------------------------------------------
| The pattern for a default generated extended double-precision NaN. The
| `high' and `low' values hold the most- and least-significant bits,