memoized: further fixes; hash objects whre possible

This is further to commit 6d854fd4dc59936142fff811e575cd034a308c43

Due to the way certain objects are handled, getting the frist few bytes
of an object may not be enough (e.g. strings do not store their values
inline). To further mitigate the issue, hash the object where possible.

devlib/utils/misc.py

@@ -33,6 +33,7 @@ import ctypes
 from operator import itemgetter
 from itertools import groupby
 from functools import partial
+from collections import Hashable
 
 import wrapt
 
@@ -557,19 +558,22 @@ def __get_memo_id(obj):
     ID string.
     """
     obj_id = id(obj)
-    obj_pyobj = ctypes.cast(obj_id, ctypes.py_object)
+    if isinstance(obj, Hashable):
-    # TODO: Note: there is still a possibility of a clash here. If Two
+        return '{}/{}'.format(obj_id, hash(obj))
-    # different objects get assigned the same ID, an are large and are
+    else:
-    # identical in the first thirty two bytes. This shouldn't be much of an
+        obj_pyobj = ctypes.cast(obj_id, ctypes.py_object)
-    # issue in the current application of memoizing Target calls, as it's very
+        # TODO: Note: there is still a possibility of a clash here. If Two
-    # unlikely that a target will get passed large params; but may cause
+        # different objects get assigned the same ID, an are large and are
-    # problems in other applications, e.g. when memoizing results of operations
+        # identical in the first thirty two bytes. This shouldn't be much of an
-    # on large arrays. I can't really think of a good way around that apart
+        # issue in the current application of memoizing Target calls, as it's very
-    # form, e.g., md5 hashing the entire raw object, which will have an
+        # unlikely that a target will get passed large params; but may cause
-    # undesirable impact on performance.
+        # problems in other applications, e.g. when memoizing results of operations
-    num_bytes = min(ctypes.sizeof(obj_pyobj), 32)
+        # on large arrays. I can't really think of a good way around that apart
-    obj_bytes = ctypes.string_at(ctypes.addressof(obj_pyobj), num_bytes)
+        # form, e.g., md5 hashing the entire raw object, which will have an
-    return '{}/{}'.format(obj_id, obj_bytes)
+        # undesirable impact on performance.
+        num_bytes = min(ctypes.sizeof(obj_pyobj), 32)
+        obj_bytes = ctypes.string_at(ctypes.addressof(obj_pyobj), num_bytes)
+        return '{}/{}'.format(obj_id, obj_bytes)
 
 
 @wrapt.decorator