Skip to content

dataflow

BasicBlock

Linear sequence of statements with a single entry and exit.

Successor/predecessor lists encode the CFG edges. Statements execute in order; no internal branching occurs within a BasicBlock.

Source code in src/finch/symbolic/dataflow.py 
 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
class BasicBlock:
    """Linear sequence of statements with a single entry and exit.

    Successor/predecessor lists encode the CFG edges. Statements execute in
    order; no internal branching occurs within a BasicBlock.
    """

    def __init__(self, id: str) -> None:
        self.id = id
        self.statements: list = []
        self.successors: list[BasicBlock] = []
        self.predecessors: list[BasicBlock] = []

    def add_statement(self, statement) -> None:
        self.statements.append(statement)

    def add_successor(self, successor: "BasicBlock") -> None:
        if successor not in self.successors:
            self.successors.append(successor)

        if self not in successor.predecessors:
            successor.predecessors.append(self)

    def __str__(self) -> str:
        """String representation of BasicBlock in LLVM style."""
        lines = []

        if self.successors:
            succ_names = [succ.id for succ in self.successors]
            succ_str = f" #succs=[{', '.join(succ_names)}]"
        else:
            succ_str = " #succs=[]"

        # Block header
        lines.append(f"{self.id}:{succ_str}")

        # Block statements
        lines.extend(f"    {stmt}" for stmt in self.statements)

        return "\n".join(lines)

__str__()

String representation of BasicBlock in LLVM style.

Source code in src/finch/symbolic/dataflow.py 
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
def __str__(self) -> str:
    """String representation of BasicBlock in LLVM style."""
    lines = []

    if self.successors:
        succ_names = [succ.id for succ in self.successors]
        succ_str = f" #succs=[{', '.join(succ_names)}]"
    else:
        succ_str = " #succs=[]"

    # Block header
    lines.append(f"{self.id}:{succ_str}")

    # Block statements
    lines.extend(f"    {stmt}" for stmt in self.statements)

    return "\n".join(lines)

ControlFlowGraph

Collection of BasicBlocks plus explicit ENTRY and EXIT nodes.

Provides helpers to allocate uniquely named blocks and holds the global block mapping used by dataflow analyses.

Source code in src/finch/symbolic/dataflow.py 
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
class ControlFlowGraph:
    """Collection of BasicBlocks plus explicit ENTRY and EXIT nodes.

    Provides helpers to allocate uniquely named blocks and holds the global
    block mapping used by dataflow analyses.
    """

    def __init__(self) -> None:
        self.block_counter = 0
        self.block_name = ""
        self.blocks: dict[str, BasicBlock] = {}

        self.entry_block = self.new_block_custom("ENTRY")
        self.exit_block = self.new_block_custom("EXIT")

    def new_block(self) -> BasicBlock:
        bid = f"{self.block_name}_{self.block_counter}"
        self.block_counter += 1
        block = BasicBlock(bid)
        self.blocks[bid] = block
        return block

    def new_block_custom(self, name: str) -> BasicBlock:
        bid = f"{name}_{self.block_counter}"
        self.block_counter += 1
        block = BasicBlock(bid)
        self.blocks[bid] = block
        return block

    def __str__(self) -> str:
        """Print the CFG in LLVM style format."""
        blocks = list(self.blocks.values())

        # Use list comprehension with join for better performance
        block_strings = [str(block) for block in blocks]
        return "\n\n".join(block_strings)

__str__()

Print the CFG in LLVM style format.

Source code in src/finch/symbolic/dataflow.py 
76
77
78
79
80
81
82
def __str__(self) -> str:
    """Print the CFG in LLVM style format."""
    blocks = list(self.blocks.values())

    # Use list comprehension with join for better performance
    block_strings = [str(block) for block in blocks]
    return "\n\n".join(block_strings)

DataFlowAnalysis

Bases: ABC

Base class for performing data flow analyses over a ControlFlowGraph.

This abstract class implements a generic work-list algorithm that can run either in a "forward" or "backward" direction. Subclasses must provide IR-specific semantics by implementing the abstract methods: transfer, join, stmt_str and direction.

Key attributes - cfg: the ControlFlowGraph to analyze. - input_states: a mapping from basic block id to the lattice state that holds at the entry of the block. - output_states: a mapping from basic block id to the lattice state that holds at the exit of the block.

Source code in src/finch/symbolic/dataflow.py 
 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
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
class DataFlowAnalysis(ABC):
    """
    Base class for performing data flow analyses over a ControlFlowGraph.

    This abstract class implements a generic work-list algorithm that can run either
    in a "forward" or "backward" direction. Subclasses must provide IR-specific
    semantics by implementing the abstract methods: transfer, join, stmt_str and
    direction.

    Key attributes
    - cfg: the ControlFlowGraph to analyze.
    - input_states: a mapping from basic block id to the lattice state that holds
        at the entry of the block.
    - output_states: a mapping from basic block id to the lattice state that holds
        at the exit of the block.
    """

    def __init__(self, cfg: ControlFlowGraph):
        self.cfg: ControlFlowGraph = cfg
        self.input_states: dict[str, dict] = {
            block.id: {} for block in cfg.blocks.values()
        }
        self.output_states: dict[str, dict] = {
            block.id: {} for block in cfg.blocks.values()
        }

    def __str__(self) -> str:
        """
        Return a string representation describing the current dataflow
        analysis state over the control-flow graph (CFG).

        The output lists each basic block in the order
        produced by self.cfg.blocks.values().
        For every block the printed form is:

            <block.id>: #succs=[<succ_id>, ...]
                <stmt_representation_for_first_stmt>
                <stmt_representation_for_second_stmt>
                ...
        """
        lines: list[str] = []
        blocks = list(self.cfg.blocks.values())

        for block in blocks:
            # get successors of the block
            if block.successors:
                succ_names = [succ.id for succ in block.successors]
                succ_str = f"#succs=[{', '.join(succ_names)}]"
            else:
                succ_str = "#succs=[]"

            lines.append(f"{block.id}: {succ_str}")

            # get the input state for this block
            input_state = self.input_states.get(block.id, {})

            if self.direction() == "forward":
                # deep copy the input_state, otherwise original state
                # in the input_states will be changed,
                # which can lead to errors in future
                state = copy.deepcopy(input_state)

                # print each statement using subclasses stmt_str method
                for stmt in block.statements:
                    lines.append(f"    {self.stmt_str(stmt, state)}")
                    # advance state with current statement
                    state = self.transfer([stmt], state)
            else:
                # TODO: implement "backward" direction printing
                lines.extend(
                    f"    {self.stmt_str(stmt, input_state)}"
                    for stmt in block.statements
                )

            lines.append("")

        return "\n".join(lines)

    @abstractmethod
    def stmt_str(self, stmt, state: dict) -> str:
        """
        Format a single statement given the current lattice state.
        Implement in subclasses or IR-specific abstract bases.
        """
        ...

    @abstractmethod
    def transfer(self, stmts, state: dict) -> dict:
        """
        Transfer function for the data flow analysis.
        This should be implemented by subclasses.
        """
        ...

    @abstractmethod
    def join(self, state_1: dict, state_2: dict) -> dict:
        """
        Join function for the data flow analysis.
        This should be implemented by subclasses.
        """
        ...

    @abstractmethod
    def direction(self) -> str:
        """
        Return the direction of the data flow analysis, either "forward" or "backward".
        This should be implemented by subclasses.
        """
        ...

    def analyze(self) -> None:
        """
        Perform the data flow analysis on the control flow graph.
        This method initializes the work list and processes each block.
        """
        work_list: list[BasicBlock] = list(self.cfg.blocks.values())
        if self.direction() == "forward":
            while work_list:
                block = work_list.pop(0)

                # get current input state based on the predecessors output states
                if not block.predecessors:
                    input_state = {}
                else:
                    pred_outputs = [
                        self.output_states.get(pred.id, {})
                        for pred in block.predecessors
                    ]
                    input_state = pred_outputs[0].copy()
                    for pred_output in pred_outputs[1:]:
                        input_state = self.join(input_state, pred_output)

                self.input_states[block.id] = input_state

                # perform transfer based on the statements in the current basic block
                output_state = self.transfer(block.statements, input_state)

                # check if output_state changed
                if output_state != self.output_states.get(block.id, {}):
                    self.output_states[block.id] = output_state

                    for successor in block.successors:
                        if successor not in work_list:
                            work_list.append(successor)
        else:
            while work_list:
                block = work_list.pop(0)

                # get current input state based on the successors output states
                if not block.successors:
                    input_state = {}
                else:
                    succ_outputs = [
                        self.output_states.get(succ.id, {}) for succ in block.successors
                    ]
                    input_state = succ_outputs[0].copy()
                    for succ_output in succ_outputs[1:]:
                        input_state = self.join(input_state, succ_output)

                self.input_states[block.id] = input_state

                # perform trasnfer based on the statements in the current basic block
                output_state = self.transfer(block.statements, input_state)

                # check if output state changed
                if output_state != self.output_states.get(block.id, {}):
                    self.output_states[block.id] = output_state

                    for predecessor in block.predecessors:
                        if predecessor not in work_list:
                            work_list.append(predecessor)

__str__()

Return a string representation describing the current dataflow analysis state over the control-flow graph (CFG).

The output lists each basic block in the order produced by self.cfg.blocks.values(). For every block the printed form is:

<block.id>: #succs=[<succ_id>, ...]
    <stmt_representation_for_first_stmt>
    <stmt_representation_for_second_stmt>
    ...
Source code in src/finch/symbolic/dataflow.py 
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
def __str__(self) -> str:
    """
    Return a string representation describing the current dataflow
    analysis state over the control-flow graph (CFG).

    The output lists each basic block in the order
    produced by self.cfg.blocks.values().
    For every block the printed form is:

        <block.id>: #succs=[<succ_id>, ...]
            <stmt_representation_for_first_stmt>
            <stmt_representation_for_second_stmt>
            ...
    """
    lines: list[str] = []
    blocks = list(self.cfg.blocks.values())

    for block in blocks:
        # get successors of the block
        if block.successors:
            succ_names = [succ.id for succ in block.successors]
            succ_str = f"#succs=[{', '.join(succ_names)}]"
        else:
            succ_str = "#succs=[]"

        lines.append(f"{block.id}: {succ_str}")

        # get the input state for this block
        input_state = self.input_states.get(block.id, {})

        if self.direction() == "forward":
            # deep copy the input_state, otherwise original state
            # in the input_states will be changed,
            # which can lead to errors in future
            state = copy.deepcopy(input_state)

            # print each statement using subclasses stmt_str method
            for stmt in block.statements:
                lines.append(f"    {self.stmt_str(stmt, state)}")
                # advance state with current statement
                state = self.transfer([stmt], state)
        else:
            # TODO: implement "backward" direction printing
            lines.extend(
                f"    {self.stmt_str(stmt, input_state)}"
                for stmt in block.statements
            )

        lines.append("")

    return "\n".join(lines)

analyze()

Perform the data flow analysis on the control flow graph. This method initializes the work list and processes each block.

Source code in src/finch/symbolic/dataflow.py 
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
def analyze(self) -> None:
    """
    Perform the data flow analysis on the control flow graph.
    This method initializes the work list and processes each block.
    """
    work_list: list[BasicBlock] = list(self.cfg.blocks.values())
    if self.direction() == "forward":
        while work_list:
            block = work_list.pop(0)

            # get current input state based on the predecessors output states
            if not block.predecessors:
                input_state = {}
            else:
                pred_outputs = [
                    self.output_states.get(pred.id, {})
                    for pred in block.predecessors
                ]
                input_state = pred_outputs[0].copy()
                for pred_output in pred_outputs[1:]:
                    input_state = self.join(input_state, pred_output)

            self.input_states[block.id] = input_state

            # perform transfer based on the statements in the current basic block
            output_state = self.transfer(block.statements, input_state)

            # check if output_state changed
            if output_state != self.output_states.get(block.id, {}):
                self.output_states[block.id] = output_state

                for successor in block.successors:
                    if successor not in work_list:
                        work_list.append(successor)
    else:
        while work_list:
            block = work_list.pop(0)

            # get current input state based on the successors output states
            if not block.successors:
                input_state = {}
            else:
                succ_outputs = [
                    self.output_states.get(succ.id, {}) for succ in block.successors
                ]
                input_state = succ_outputs[0].copy()
                for succ_output in succ_outputs[1:]:
                    input_state = self.join(input_state, succ_output)

            self.input_states[block.id] = input_state

            # perform trasnfer based on the statements in the current basic block
            output_state = self.transfer(block.statements, input_state)

            # check if output state changed
            if output_state != self.output_states.get(block.id, {}):
                self.output_states[block.id] = output_state

                for predecessor in block.predecessors:
                    if predecessor not in work_list:
                        work_list.append(predecessor)

direction() abstractmethod

Return the direction of the data flow analysis, either "forward" or "backward". This should be implemented by subclasses.

Source code in src/finch/symbolic/dataflow.py 
187
188
189
190
191
192
193
@abstractmethod
def direction(self) -> str:
    """
    Return the direction of the data flow analysis, either "forward" or "backward".
    This should be implemented by subclasses.
    """
    ...

join(state_1, state_2) abstractmethod

Join function for the data flow analysis. This should be implemented by subclasses.

Source code in src/finch/symbolic/dataflow.py 
179
180
181
182
183
184
185
@abstractmethod
def join(self, state_1: dict, state_2: dict) -> dict:
    """
    Join function for the data flow analysis.
    This should be implemented by subclasses.
    """
    ...

stmt_str(stmt, state) abstractmethod

Format a single statement given the current lattice state. Implement in subclasses or IR-specific abstract bases.

Source code in src/finch/symbolic/dataflow.py 
163
164
165
166
167
168
169
@abstractmethod
def stmt_str(self, stmt, state: dict) -> str:
    """
    Format a single statement given the current lattice state.
    Implement in subclasses or IR-specific abstract bases.
    """
    ...

transfer(stmts, state) abstractmethod

Transfer function for the data flow analysis. This should be implemented by subclasses.

Source code in src/finch/symbolic/dataflow.py 
171
172
173
174
175
176
177
@abstractmethod
def transfer(self, stmts, state: dict) -> dict:
    """
    Transfer function for the data flow analysis.
    This should be implemented by subclasses.
    """
    ...