docs no supported

Author: Ehsan Totoni

docs/index.rst

@@ -12,3 +12,4 @@ HPAT documentation
 
    source/overview
    source/supported
+   source/notsupported

docs/source/notsupported.rst

@@ -0,0 +1,35 @@
+.. _notsupported:
+
+Not Supported Python Programs
+=============================
+
+HPAT statically compiles user codes to generate efficient parallel programs.
+Hence, the user code needs to be `statically compilable`.
+This means that HPAT should be able to infer all the variable types, and be able
+to analyze the computations.
+
+Type Stability
+--------------
+
+To enable type inference, the program should be `type stable`, which means every
+variable should have a single type. The example below is not type stable since
+variable ``a`` can be both a float and an array of floats::
+
+    if flag:
+        a = 1.0
+    else:
+        a = np.ones(10)
+
+The use of ``isinstance`` operator of Python often means type instabillity and
+is not supported.
+
+Similarly, function calls should also be deterministic. The below example is
+not supported since function ``f`` is not known in advance::
+
+    if flag:
+        f = np.zeros
+    else:
+        f = np.random.ranf
+    A = f(10)
+
+One can usually avoid these cases in numerical code without significant effort.

docs/source/supported.rst

@@ -1,7 +1,10 @@
 .. _supported:
 
-Supported Python Programs
-=========================
+User Guide
+==========
+
+HPAT supports a subset of Python that is commonly used for data analytics and
+machine learning. This section describes this subset.
 
 HPAT compiles and parallelizes the functions annotated with the `@hpat.jit`
 decorator. Therefore, file I/O and computations on large datasets should be