Skip to content

[mypyc] Do not preemptively diff & union op level dataflow sets #20897

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
hauntsaninja merged 9 commits into from
Feb 26, 2026
+49 −30
Merged

[mypyc] Do not preemptively diff & union op level dataflow sets #20897

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
3aed3b2
Fix: Do not preemptively diff & union op level dataflow sets
VaggelisD Feb 25, 2026
e90a04d
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Feb 25, 2026
0bb52f6
Do the same for block-level dataflow too
VaggelisD Feb 25, 2026
9e0bfc9
Do not copy for block-level dataflow, cache empty set creation
VaggelisD Feb 25, 2026
3492914
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Feb 25, 2026
6208e4f
Fix CI from EMPTY
VaggelisD Feb 25, 2026
5b6c1ac
Use frozenset for EMPTY
VaggelisD Feb 25, 2026
4ce4c07
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Feb 25, 2026
43799d3
Merge branch 'master' into dataflow_cur
hauntsaninja Feb 26, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
69 changes: 41 additions & 28 deletions mypyc/analysis/dataflow.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,8 +3,8 @@
from __future__ import annotations

from abc import abstractmethod
from collections.abc import Iterable, Iterator
from typing import Generic, TypeVar
from collections.abc import Iterable, Iterator, Set as AbstractSet
from typing import Any, Generic, TypeVar

from mypyc.ir.ops import (
Assign,
Expand Down Expand Up @@ -174,12 +174,14 @@ def __str__(self) -> str:
return f"before: {self.before}\nafter: {self.after}\n"


GenAndKill = tuple[set[T], set[T]]
GenAndKill = tuple[AbstractSet[T], AbstractSet[T]]

_EMPTY: tuple[frozenset[Any], frozenset[Any]] = (frozenset(), frozenset())


class BaseAnalysisVisitor(OpVisitor[GenAndKill[T]]):
def visit_goto(self, op: Goto) -> GenAndKill[T]:
return set(), set()
return _EMPTY

@abstractmethod
def visit_register_op(self, op: RegisterOp) -> GenAndKill[T]:
Expand Down Expand Up @@ -317,16 +319,16 @@ def __init__(self, strict_errors: bool = False) -> None:
self.strict_errors = strict_errors

def visit_branch(self, op: Branch) -> GenAndKill[Value]:
return set(), set()
return _EMPTY

def visit_return(self, op: Return) -> GenAndKill[Value]:
return set(), set()
return _EMPTY

def visit_unreachable(self, op: Unreachable) -> GenAndKill[Value]:
return set(), set()
return _EMPTY

def visit_register_op(self, op: RegisterOp) -> GenAndKill[Value]:
return set(), set()
return _EMPTY

def visit_assign(self, op: Assign) -> GenAndKill[Value]:
# Loading an error value may undefine the register.
Expand All @@ -337,10 +339,10 @@ def visit_assign(self, op: Assign) -> GenAndKill[Value]:

def visit_assign_multi(self, op: AssignMulti) -> GenAndKill[Value]:
# Array registers are special and we don't track the definedness of them.
return set(), set()
return _EMPTY

def visit_set_mem(self, op: SetMem) -> GenAndKill[Value]:
return set(), set()
return _EMPTY


def analyze_maybe_defined_regs(
Expand Down Expand Up @@ -392,27 +394,27 @@ def __init__(self, args: set[Value]) -> None:
self.args = args

def visit_branch(self, op: Branch) -> GenAndKill[Value]:
return set(), set()
return _EMPTY

def visit_return(self, op: Return) -> GenAndKill[Value]:
return set(), set()
return _EMPTY

def visit_unreachable(self, op: Unreachable) -> GenAndKill[Value]:
return set(), set()
return _EMPTY

def visit_register_op(self, op: RegisterOp) -> GenAndKill[Value]:
return set(), set()
return _EMPTY

def visit_assign(self, op: Assign) -> GenAndKill[Value]:
if op.dest in self.args:
return set(), {op.dest}
return set(), set()
return _EMPTY

def visit_assign_multi(self, op: AssignMulti) -> GenAndKill[Value]:
return set(), set()
return _EMPTY

def visit_set_mem(self, op: SetMem) -> GenAndKill[Value]:
return set(), set()
return _EMPTY


def analyze_borrowed_arguments(
Expand All @@ -435,13 +437,13 @@ def analyze_borrowed_arguments(

class UndefinedVisitor(BaseAnalysisVisitor[Value]):
def visit_branch(self, op: Branch) -> GenAndKill[Value]:
return set(), set()
return _EMPTY

def visit_return(self, op: Return) -> GenAndKill[Value]:
return set(), set()
return _EMPTY

def visit_unreachable(self, op: Unreachable) -> GenAndKill[Value]:
return set(), set()
return _EMPTY

def visit_register_op(self, op: RegisterOp) -> GenAndKill[Value]:
return set(), {op} if not op.is_void else set()
Expand All @@ -453,7 +455,7 @@ def visit_assign_multi(self, op: AssignMulti) -> GenAndKill[Value]:
return set(), {op.dest}

def visit_set_mem(self, op: SetMem) -> GenAndKill[Value]:
return set(), set()
return _EMPTY


def non_trivial_sources(op: Op) -> set[Value]:
Expand All @@ -472,10 +474,10 @@ def visit_return(self, op: Return) -> GenAndKill[Value]:
if not isinstance(op.value, (Integer, Float)):
return {op.value}, set()
else:
return set(), set()
return _EMPTY

def visit_unreachable(self, op: Unreachable) -> GenAndKill[Value]:
return set(), set()
return _EMPTY

def visit_register_op(self, op: RegisterOp) -> GenAndKill[Value]:
gen = non_trivial_sources(op)
Expand All @@ -494,10 +496,10 @@ def visit_set_mem(self, op: SetMem) -> GenAndKill[Value]:
return non_trivial_sources(op), set()

def visit_inc_ref(self, op: IncRef) -> GenAndKill[Value]:
return set(), set()
return _EMPTY

def visit_dec_ref(self, op: DecRef) -> GenAndKill[Value]:
return set(), set()
return _EMPTY


def analyze_live_regs(blocks: list[BasicBlock], cfg: CFG) -> AnalysisResult[Value]:
Expand Down Expand Up @@ -559,8 +561,16 @@ def run_analysis(
ops = list(reversed(ops))
for op in ops:
opgen, opkill = op.accept(gen_and_kill)
Copy link
Contributor Author

@VaggelisD VaggelisD Feb 25, 2026 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think we could even cache the visitor result per op so we don't have to accept() twice, here and in L632, right? Not sure how worth it it is though, the visitor work is not that intensive.

gen = (gen - opkill) | opgen
kill = (kill - opgen) | opkill
if opkill:
gen -= opkill

if opgen:
gen |= opgen
kill -= opgen

if opkill:
kill |= opkill

block_gen[block] = gen
block_kill[block] = kill

Expand Down Expand Up @@ -624,7 +634,10 @@ def run_analysis(
for idx, op in ops_enum:
op_before[label, idx] = cur
opgen, opkill = op.accept(gen_and_kill)
cur = (cur - opkill) | opgen
if opkill:
cur = cur - opkill
if opgen:
cur = cur | opgen
op_after[label, idx] = cur
if backward:
op_after, op_before = op_before, op_after
Expand Down
10 changes: 8 additions & 2 deletions mypyc/analysis/selfleaks.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,12 @@
from __future__ import annotations

from mypyc.analysis.dataflow import CFG, MAYBE_ANALYSIS, AnalysisResult, run_analysis
from mypyc.analysis.dataflow import (
CFG,
MAYBE_ANALYSIS,
AnalysisResult,
GenAndKill as _DataflowGenAndKill,
run_analysis,
)
from mypyc.ir.ops import (
Assign,
AssignMulti,
Expand Down Expand Up @@ -49,7 +55,7 @@
)
from mypyc.ir.rtypes import RInstance

GenAndKill = tuple[set[None], set[None]]
GenAndKill = _DataflowGenAndKill[None]

CLEAN: GenAndKill = (set(), set())
DIRTY: GenAndKill = ({None}, {None})
Expand Down