Merge pull request #1144 from skirpichev/ulp

Add ulp(), see #1009

mpmath/__init__.py

@@ -222,6 +222,7 @@ mertens = mp.mertens
 
 ldexp = mp.ldexp
 frexp = mp.frexp
+ulp = mp.ulp
 
 fsum = mp.fsum
 fdot = mp.fdot

mpmath/ctx_fp.py

@@ -161,6 +161,7 @@ class FPContext(StandardBaseContext):
     ldexp = math.ldexp
     frexp = math.frexp
     hypot = math.hypot
+    ulp = math.ulp
 
     def mag(ctx, z):
         if z:

mpmath/ctx_mp.py

@@ -783,6 +783,19 @@ maxterms, or set zeroprec."""
         y, n = libmp.libmpf.mpf_frexp(x._mpf_)
         return ctx.make_mpf(y), n
 
+    def ulp(ctx, x):
+        """
+        Return the value of the least significant bit of the `x`.
+
+            >>> from mpmath import ulp
+            >>> ulp(1)
+            mpf('2.2204460492503131e-16')
+
+        """
+        x = ctx.convert(x)
+        *_, e, bc = x._mpf_
+        return ctx.make_mpf((0, 1, e + bc - ctx.prec, 1))
+
     def fneg(ctx, x, **kwargs):
         """
         Negates the number *x*, giving a floating-point result, optionally