1 Low-Level Instructions and Macro Instructions
2 =============================================
3
4 Migrate macro instructions such as the *Index instructions to library code implemented
5 using low-level instructions.
6
7 Consider introducing classic machine level instructions (word addition, subtraction, and
8 so on) in order to implement all current RSVP instructions.
9
10 Class and Module Attribute Assignment
11 =====================================
12
13 Verify that the context information is correctly set, particularly for the unoptimised
14 cases.
15
16 Update docs/assignment.txt.
17
18 Consider attribute assignment observations, along with the possibility of class and module
19 attribute assignment.
20
21 (Note direct assignments as usual, indirect assignments via the attribute usage
22 mechanism. During attribute collection and inference, add assigned values to all
23 inferred targets.)
24
25 (Since class attributes can be assigned, StoreAttrIndex would no longer need to reject
26 static attributes, although this might still be necessary where attribute usage analysis
27 has not been performed.)
28
29 Potentially consider changing static attribute details to use object-relative offsets in
30 order to simplify the instruction implementations. This might allow us to eliminate the
31 static attribute flag for attributes in the object table, at least at run-time.
32
33 Dynamic Attribute Access
34 ========================
35
36 Consider explicit accessor initialisation:
37
38 attr = accessor("attr")
39 getattr(C, attr)
40
41 Attribute Usage
42 ===============
43
44 Consider attribute usage observations being suspended inside blocks where AttributeError
45 may be caught (although this doesn't anticipate such exceptions being caught outside a
46 function altogether).
47
48 Consider type deduction and its consequences where types belong to the same hierarchy
49 and where a guard could be generated for the most general type.
50
51 Consider permitting multiple class alternatives where the attributes are all identical.
52
53 Support class attribute positioning similar to instance attribute positioning, potentially
54 (for both) based on usage observations. For example, if __iter__ is used on two classes,
55 the class attribute could be exposed at a similar relative position to the class (and
56 potentially accessible using a LoadAttr-style instruction).
57
58 **** Constant attribute users need not maintain usage since they are already resolved. ****
59
60 Loop entry points should capture usage to update later assignments in the loop.
61 The continue and break statements should affect usage propagation.
62
63 Consider handling CallFunc in micropython.inspect in order to produce instances of specific classes.
64 Then, consider adding support for guard removal/verification where known instances are involved.
65 Consider handling branches of values within namespaces in order to support more precise value usage.
66
67 Frame Optimisations
68 ===================
69
70 Stack frame replacement where a local frame is unused after a call, such as in a tail call
71 situation.
72
73 Local assignment detection plus frame re-use. Example: slice.__init__ calls
74 xrange.__init__ with the same arguments which are unchanged in xrange.__init__. There is
75 therefore no need to build a new frame for this call, although in some cases the locals
76 frame might need expanding.
77
78 Inlining
79 ========
80
81 Where a function or method call can always be determined, the body of the target could be
82 inlined - copied into place - within the caller. If the target is only ever called by a
83 single caller it could be moved into place.
84
85 Function Specialisation
86 =======================
87
88 Specialisation of certain functions, such as isinstance(x, cls) where cls is a known
89 constant.
90
91 Structure and Object Table Optimisations
92 ========================================
93
94 Fix object table entries for attributes not provided by any known object, or provide an
95 error, potentially overridden by options. For example, the augmented assignment methods
96 are not supported by the built-in objects and thus the operator module functions cause
97 the compilation to fail. Alternatively, just supply the methods since something has to do
98 so in the builtins.
99
100 Consider attribute merging where many attributes are just aliases for the same underlying
101 definition.
102
103 Consider references to defaults as occurring only within the context of a particular
104 function, thus eliminating default value classes if such functions are not themselves
105 invoked.
106
107 Scope Handling
108 ==============
109
110 Consider merging the InspectedModule.store tests with the scope conflict handling.
111
112 Consider labelling _scope on assignments and dealing with the assignment of removed
113 attributes, possibly removing the entire assignment, and distinguishing between such cases
114 and unknown names.
115
116 Check name origin where multiple branches could yield multiple scope interpretations:
117
118 ----
119 try:
120 set # built-in name
121 except NameError:
122 from sets import Set as set # local definition of name
123
124 set # could be confused by the local definition at run-time
125 ----
126
127 Object Coverage
128 ===============
129
130 Incorporate constants into the coverage, eliminating unused constants.
131
132 Support __init__ traversal (and other implicit names) more effectively.
133
134 Other
135 =====
136
137 Support tuple as a function returning any input tuple uncopied.
138
139 Check context_value initialisation (avoiding or handling None effectively).
140
141 __getitem__ could be written in Python, using a native method only to access fragments.
142
143 Consider better "macro" support where new expressions need to be generated and processed.
144
145 Detect TestIdentity results involving constants, potentially optimising status-affected
146 instructions:
147
148 TestIdentity(x, y) # where x is always y
149 JumpIfFalse(...) # would be removed (never false)
150 JumpIfTrue(...) # changed to Jump(...)
151
152 Status-affected blocks could be optimised away for such constant-related results.