1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
|
""" ArgumentEffects computes write effect on arguments. """
from pythran.analyses.aliases import Aliases
from pythran.analyses.intrinsics import Intrinsics
from pythran.analyses.global_declarations import GlobalDeclarations
from pythran.passmanager import ModuleAnalysis
from pythran.tables import MODULES
from pythran.graph import DiGraph
# FIXME: investigate why we need to import it that way
from pythran import intrinsic
import gast as ast
from functools import reduce
class FunctionEffects(object):
def __init__(self, node):
self.func = node
if isinstance(node, ast.FunctionDef):
self.update_effects = [False] * len(node.args.args)
elif isinstance(node, intrinsic.Intrinsic):
self.update_effects = [isinstance(x, intrinsic.UpdateEffect)
for x in node.argument_effects]
elif isinstance(node, ast.alias):
self.update_effects = []
elif isinstance(node, intrinsic.Class):
self.update_effects = []
else:
raise NotImplementedError
# Compute the intrinsic effects only once
IntrinsicArgumentEffects = {}
def save_function_effect(module):
""" Recursively save function effect for pythonic functions. """
for intr in module.values():
if isinstance(intr, dict): # Submodule case
save_function_effect(intr)
else:
fe = FunctionEffects(intr)
IntrinsicArgumentEffects[intr] = fe
if isinstance(intr, intrinsic.Class):
save_function_effect(intr.fields)
for module in MODULES.values():
save_function_effect(module)
class ArgumentEffectsHelper(ModuleAnalysis[Aliases, GlobalDeclarations, Intrinsics]):
"""Gathers inter-procedural effects on function arguments."""
ResultType = DiGraph
def __init__(self):
# There's an edge between src and dest if a parameter of dest is
# modified by src
super().__init__()
self.node_to_functioneffect = {}
def prepare(self, node):
"""
Initialise arguments effects as this analyse is inter-procedural.
Initialisation done for Pythonic functions and default value set for
user defined functions.
"""
super().prepare(node)
for i in self.intrinsics:
fe = IntrinsicArgumentEffects[i]
self.node_to_functioneffect[i] = fe
self.result.add_node(fe)
for n in self.global_declarations.values():
fe = FunctionEffects(n)
self.node_to_functioneffect[n] = fe
self.result.add_node(fe)
def argument_index(self, node):
while isinstance(node, ast.Subscript):
node = node.value
for node_alias in self.aliases[node]:
while isinstance(node_alias, ast.Subscript):
node_alias = node_alias.value
if node_alias in self.current_arguments:
return self.current_arguments[node_alias]
if node_alias in self.current_subscripted_arguments:
return self.current_subscripted_arguments[node_alias]
return -1
def visit_FunctionDef(self, node):
self.current_function = self.node_to_functioneffect[node]
self.current_arguments = {arg: i
for i, arg
in enumerate(node.args.args)}
self.current_subscripted_arguments = dict()
assert self.current_function in self.result
self.generic_visit(node)
def visit_For(self, node):
ai = self.argument_index(node.iter)
if ai >= 0:
self.current_subscripted_arguments[node.target] = ai
self.generic_visit(node)
def visit_AugAssign(self, node):
n = self.argument_index(node.target)
if n >= 0:
self.current_function.update_effects[n] = True
self.generic_visit(node)
def visit_Assign(self, node):
targets = node.targets if isinstance(node, ast.Assign) else (node.target,)
for t in targets:
if isinstance(t, ast.Subscript):
n = self.argument_index(t)
if n >= 0:
self.current_function.update_effects[n] = True
self.generic_visit(node)
visit_AnnAssign = visit_Assign
def visit_Call(self, node):
for i, arg in enumerate(node.args):
n = self.argument_index(arg)
if n >= 0:
func_aliases = self.aliases[node.func]
# pessimistic case: no alias found
if func_aliases is None:
self.current_function.update_effects[n] = True
continue
# expand argument if any
func_aliases = reduce(
lambda x, y: x + (
# all functions
list(self.node_to_functioneffect.keys())
if (isinstance(y, ast.Name) and
self.argument_index(y) >= 0)
else [y]),
func_aliases,
list())
for func_alias in func_aliases:
# special hook for binded functions
if isinstance(func_alias, ast.Call):
bound_name = func_alias.args[0].id
func_alias = self.global_declarations[bound_name]
if func_alias is intrinsic.UnboundValue:
continue
if func_alias not in self.node_to_functioneffect:
continue
if func_alias is MODULES['functools']['partial']:
base_func_aliases = self.aliases[node.args[0]]
fe = self.node_to_functioneffect[func_alias]
if len(base_func_aliases) == 1:
base_func_alias = next(iter(base_func_aliases))
fe = self.node_to_functioneffect.get(
base_func_alias,
fe)
else:
fe = self.node_to_functioneffect[func_alias]
if not self.result.has_edge(fe, self.current_function):
self.result.add_edge(
fe,
self.current_function,
effective_parameters=[],
formal_parameters=[])
edge = self.result.edges[fe, self.current_function]
edge["effective_parameters"].append(n)
edge["formal_parameters"].append(i)
self.generic_visit(node)
class ArgumentEffects(ModuleAnalysis[ArgumentEffectsHelper]):
"""Gathers inter-procedural effects on function arguments."""
ResultType = dict
def visit_Module(self, node):
result = self.argument_effects_helper
candidates = set(result)
while candidates:
function = candidates.pop()
for ue in enumerate(function.update_effects):
update_effect_idx, update_effect = ue
if not update_effect:
continue
for pred in result.successors(function):
edge = result.edges[function, pred]
for fp in enumerate(edge["formal_parameters"]):
i, formal_parameter_idx = fp
# propagate the impurity backward if needed.
# Afterward we may need another graph iteration
ith_effectiv = edge["effective_parameters"][i]
if(formal_parameter_idx == update_effect_idx and
not pred.update_effects[ith_effectiv]):
pred.update_effects[ith_effectiv] = True
candidates.add(pred)
self.result = {f.func: f.update_effects for f in result}
|
