Implemented NAN method for Spin2

Author: totalspectrum

Changelog.txt

@@ -1,4 +1,5 @@
 Version 6.5.3
+- Implemented NAN() method for Spin2
 - Improved error message for objects that are not found
 - Reduced likelihood of duplicate function definitions in sub-objects
 

frontends/common.c

@@ -218,6 +218,8 @@ Aliases spin2alias[] = {
 
     { "strcopy", "__builtin_strncpy" },
 
+    { "nan", "__builtin_isnanf" },
+    
     /* obsolete aliases */
     { "outl_", "_outl" },
     { "outh_", "_outh" },

sys/float.spin

@@ -321,6 +321,7 @@ pri __builtin_copysign(x=float, y=float) : r=float
   r := (y>>31) << 31
   r |= x
 
+' calculate the log base 2 of a float
 pri __builtin_ilogb(a=float) : r=long | s, x, m
   (s,x,m) := _float_Unpack(a)
   if m == 0
@@ -331,10 +332,19 @@ pri __builtin_ilogb(a=float) : r=long | s, x, m
     return $8000_0000  ' NaN
   return x
 
+' determine whether a float is NaN or not
+pri __builtin_isnanf(a=float) : r | s, x, m
+  (s,x,m) := _float_Unpack(a)
+  if x < 128 || m == _float_one
+    return false
+  return true
+
+' calculate log2 of a fixed point number
 pri __builtin_ilogb_fixed(a) : r=long | n
   n := >|a
   r := n - 17
 
+' copy the sign of a fixed point number
 pri __builtin_copysign_fixed(x, s) : r=long
   if s < 0
     if x < 0