PowerPC floating point little-endian [6 of 15]

http://sourceware.org/ml/libc-alpha/2013-07/msg00197.html

A rewrite to make this code correct for little-endian.

	* sysdeps/ieee754/ldbl-128ibm/e_sqrtl.c (mynumber): Replace
	union 32-bit int array member with 64-bit int array.
	(t515, tm256): Double rather than long double.
	(__ieee754_sqrtl): Rewrite using 64-bit arithmetic.

1 files changed, 23 insertions, 28 deletions

diff --git a/sysdeps/ieee754/ldbl-128ibm/e_sqrtl.c b/sysdeps/ieee754/ldbl-128ibm/e_sqrtl.c
index 2b0f7c62e4..61feb367f7 100644
--- a/sysdeps/ieee754/ldbl-128ibm/e_sqrtl.c
+++ b/sysdeps/ieee754/ldbl-128ibm/e_sqrtl.c
@@ -34,15 +34,13 @@
 
 #include <math_private.h>
 
-typedef unsigned int int4;
-typedef union {int4 i[4]; long double x; double d[2]; } mynumber;
+typedef union {int64_t i[2]; long double x; double d[2]; } mynumber;
 
-static const  mynumber
-  t512 = {{0x5ff00000, 0x00000000, 0x00000000, 0x00000000 }},  /* 2^512  */
-  tm256 = {{0x2ff00000, 0x00000000, 0x00000000, 0x00000000 }};  /* 2^-256 */
 static const double
-two54 = 1.80143985094819840000e+16, /* 0x4350000000000000 */
-twom54 = 5.55111512312578270212e-17; /* 0x3C90000000000000 */
+  t512 = 0x1p512,
+  tm256 = 0x1p-256,
+  two54 = 0x1p54,	/* 0x4350000000000000 */
+  twom54 = 0x1p-54;	/* 0x3C90000000000000 */
 
 /*********************************************************************/
 /* An ultimate sqrt routine. Given an IEEE double machine number x   */
@@ -54,56 +52,53 @@ long double __ieee754_sqrtl(long double x)
   static const long double big = 134217728.0, big1 = 134217729.0;
   long double t,s,i;
   mynumber a,c;
-  int4 k, l, m;
-  int n;
+  uint64_t k, l;
+  int64_t m, n;
   double d;
 
   a.x=x;
-  k=a.i[0] & 0x7fffffff;
+  k=a.i[0] & INT64_C(0x7fffffffffffffff);
   /*----------------- 2^-1022  <= | x |< 2^1024  -----------------*/
-  if (k>0x000fffff && k<0x7ff00000) {
+  if (k>INT64_C(0x000fffff00000000) && k<INT64_C(0x7ff0000000000000)) {
     if (x < 0) return (big1-big1)/(big-big);
-    l = (k&0x001fffff)|0x3fe00000;
-    if (((a.i[2] & 0x7fffffff) | a.i[3]) != 0) {
-      n = (int) ((l - k) * 2) >> 21;
-      m = (a.i[2] >> 20) & 0x7ff;
+    l = (k&INT64_C(0x001fffffffffffff))|INT64_C(0x3fe0000000000000);
+    if ((a.i[1] & INT64_C(0x7fffffffffffffff)) != 0) {
+      n = (int64_t) ((l - k) * 2) >> 53;
+      m = (a.i[1] >> 52) & 0x7ff;
       if (m == 0) {
 	a.d[1] *= two54;
-	m = ((a.i[2] >> 20) & 0x7ff) - 54;
+	m = ((a.i[1] >> 52) & 0x7ff) - 54;
       }
       m += n;
-      if ((int) m > 0)
-	a.i[2] = (a.i[2] & 0x800fffff) | (m << 20);
-      else if ((int) m <= -54) {
-	a.i[2] &= 0x80000000;
-	a.i[3] = 0;
+      if (m > 0)
+	a.i[1] = (a.i[1] & INT64_C(0x800fffffffffffff)) | (m << 52);
+      else if (m <= -54) {
+	a.i[1] &= INT64_C(0x8000000000000000);
       } else {
 	m += 54;
-	a.i[2] = (a.i[2] & 0x800fffff) | (m << 20);
+	a.i[1] = (a.i[1] & INT64_C(0x800fffffffffffff)) | (m << 52);
 	a.d[1] *= twom54;
       }
     }
     a.i[0] = l;
     s = a.x;
     d = __ieee754_sqrt (a.d[0]);
-    c.i[0] = 0x20000000+((k&0x7fe00000)>>1);
+    c.i[0] = INT64_C(0x2000000000000000)+((k&INT64_C(0x7fe0000000000000))>>1);
     c.i[1] = 0;
-    c.i[2] = 0;
-    c.i[3] = 0;
     i = d;
     t = 0.5L * (i + s / i);
     i = 0.5L * (t + s / t);
     return c.x * i;
   }
   else {
-    if (k>=0x7ff00000) {
-      if (a.i[0] == 0xfff00000 && a.i[1] == 0)
+    if (k>=INT64_C(0x7ff0000000000000)) {
+      if (a.i[0] == INT64_C(0xfff0000000000000))
 	return (big1-big1)/(big-big); /* sqrt (-Inf) = NaN.  */
       return x; /* sqrt (NaN) = NaN, sqrt (+Inf) = +Inf.  */
     }
     if (x == 0) return x;
     if (x < 0) return (big1-big1)/(big-big);
-    return tm256.x*__ieee754_sqrtl(x*t512.x);
+    return tm256*__ieee754_sqrtl(x*t512);
   }
 }
 strong_alias (__ieee754_sqrtl, __sqrtl_finite)