diff --git a/Testing/float.c b/Testing/float.c
index 9df23ac97..6f803bfc3 100644
--- a/Testing/float.c
+++ b/Testing/float.c
@@ -1,24 +1,73 @@
 #include <math.h>
 
+#define FACTOR (1 << 30)
+/* #define FACTOR 1e8 */
+
 int main()
 {
-  int i = 0;
-  double d = 0.123456789;
+  int i1, i2, i3;
+  unsigned u2, u3 = 0;
+  volatile double d = 0.123456789;
+  volatile double di;
 
+#if 0
   printf ("%f %d\n", d, i);
   d = frexp (d, &i);
+  printf ("%f %d\n", d, i);
 
   d = ldexp (d, i);
   printf("%f\n", d);
 
-  printf ("%f %d\n", d, i);
-  d = frexp (d, &i);
+#elif 1
 
-  printf ("%f %d\n", d, i);
-  d = frexp (d, &i);
+  i1 = i2 = i3 = 0;
+  d = -0.123456789;
+  printf ("encoded value = %.15g\n", d);
+  d = frexp (d, &i1);
+  printf ("%g %d %d %d\n", d, i1, i2, i3);
+  d *= FACTOR;
+  i2 = d;
+  d -= i2;
+  printf ("%g %d %d %d\n", d, i1, i2, i3);
+  d *= FACTOR;
+  i3 = d;
+  d -= i3;
+  printf ("%g %d %d %d\n", d, i1, i2, i3);
 
-  printf ("%f %d\n", d, i);
-  d = frexp (d, &i);
+  d = 0;
+  d = i3;
+  d /= FACTOR;
+  d += i2;
+  d /= FACTOR;
+  d = ldexp (d, i1);
+  printf ("decoded value = %.15g\n", d);
+
+#else
+
+  d = 0.123456789;
+  printf ("original value = %g\n", d);
+
+  d = frexp (d, &i1);
+  d *= FACTOR;
+  d = modf (d, &di);
+  u2 = di;
+  if (d != 0)
+    {
+      d *= FACTOR;
+      d = modf (d, &di);
+      u3 = di;
+    }
+  printf ("%d %u %u\n", i1, u2, u3);
+  
+  d = 0;
+  d = u3;
+  d /= FACTOR;
+  d += u2;
+  d /= FACTOR;
+  d = ldexp (d, i1);
+  printf ("decoded value = %g\n", d);
+
+#endif
 
   exit (0);
 }