gh-112962: in dis module, put cache information in the Instruction instead of creating fake Instructions to represent it (#113016)

This commit is contained in:
Irit Katriel 2023-12-13 12:00:21 +00:00 committed by GitHub
parent 3531ea441b
commit 428c9812cb
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
7 changed files with 115 additions and 47 deletions

View file

@ -328,13 +328,17 @@ operation is being performed, so the intermediate analysis object isn't useful:
source line information (if any) is taken directly from the disassembled code
object.
The *show_caches* and *adaptive* parameters work as they do in :func:`dis`.
The *adaptive* parameter works as it does in :func:`dis`.
.. versionadded:: 3.4
.. versionchanged:: 3.11
Added the *show_caches* and *adaptive* parameters.
.. versionchanged:: 3.13
The *show_caches* parameter is deprecated and has no effect. The *cache_info*
field of each instruction is populated regardless of its value.
.. function:: findlinestarts(code)
@ -482,6 +486,14 @@ details of bytecode instructions as :class:`Instruction` instances:
:class:`dis.Positions` object holding the
start and end locations that are covered by this instruction.
.. data::cache_info
Information about the cache entries of this instruction, as
triplets of the form ``(name, size, data)``, where the ``name``
and ``size`` describe the cache format and data is the contents
of the cache. ``cache_info`` is ``None`` if the instruction does not have
caches.
.. versionadded:: 3.4
.. versionchanged:: 3.11
@ -493,8 +505,8 @@ details of bytecode instructions as :class:`Instruction` instances:
Changed field ``starts_line``.
Added fields ``start_offset``, ``cache_offset``, ``end_offset``,
``baseopname``, ``baseopcode``, ``jump_target``, ``oparg``, and
``line_number``.
``baseopname``, ``baseopcode``, ``jump_target``, ``oparg``,
``line_number`` and ``cache_info``.
.. class:: Positions

View file

@ -267,9 +267,10 @@ _Instruction = collections.namedtuple(
'starts_line',
'line_number',
'label',
'positions'
'positions',
'cache_info',
],
defaults=[None, None]
defaults=[None, None, None]
)
_Instruction.opname.__doc__ = "Human readable name for operation"
@ -286,6 +287,7 @@ _Instruction.starts_line.__doc__ = "True if this opcode starts a source line, ot
_Instruction.line_number.__doc__ = "source line number associated with this opcode (if any), otherwise None"
_Instruction.label.__doc__ = "A label (int > 0) if this instruction is a jump target, otherwise None"
_Instruction.positions.__doc__ = "dis.Positions object holding the span of source code covered by this instruction"
_Instruction.cache_info.__doc__ = "list of (name, size, data), one for each cache entry of the instruction"
_ExceptionTableEntryBase = collections.namedtuple("_ExceptionTableEntryBase",
"start end target depth lasti")
@ -334,6 +336,8 @@ class Instruction(_Instruction):
label - A label if this instruction is a jump target, otherwise None
positions - Optional dis.Positions object holding the span of source code
covered by this instruction
cache_info - information about the format and content of the instruction's cache
entries (if any)
"""
@property
@ -570,7 +574,6 @@ def get_instructions(x, *, first_line=None, show_caches=False, adaptive=False):
linestarts=linestarts,
line_offset=line_offset,
co_positions=co.co_positions(),
show_caches=show_caches,
original_code=original_code,
arg_resolver=arg_resolver)
@ -645,8 +648,7 @@ def _is_backward_jump(op):
'ENTER_EXECUTOR')
def _get_instructions_bytes(code, linestarts=None, line_offset=0, co_positions=None,
show_caches=False, original_code=None, labels_map=None,
arg_resolver=None):
original_code=None, labels_map=None, arg_resolver=None):
"""Iterate over the instructions in a bytecode string.
Generates a sequence of Instruction namedtuples giving the details of each
@ -682,32 +684,28 @@ def _get_instructions_bytes(code, linestarts=None, line_offset=0, co_positions=N
else:
argval, argrepr = arg, repr(arg)
yield Instruction(_all_opname[op], op, arg, argval, argrepr,
offset, start_offset, starts_line, line_number,
labels_map.get(offset, None), positions)
instr = Instruction(_all_opname[op], op, arg, argval, argrepr,
offset, start_offset, starts_line, line_number,
labels_map.get(offset, None), positions)
caches = _get_cache_size(_all_opname[deop])
if not caches:
continue
if not show_caches:
# We still need to advance the co_positions iterator:
for _ in range(caches):
next(co_positions, ())
continue
for name, size in _cache_format[opname[deop]].items():
for i in range(size):
offset += 2
# Only show the fancy argrepr for a CACHE instruction when it's
# the first entry for a particular cache value:
if i == 0:
data = code[offset: offset + 2 * size]
argrepr = f"{name}: {int.from_bytes(data, sys.byteorder)}"
else:
argrepr = ""
yield Instruction(
"CACHE", CACHE, 0, None, argrepr, offset, offset, False, None, None,
Positions(*next(co_positions, ()))
)
# Advance the co_positions iterator:
for _ in range(caches):
next(co_positions, ())
if caches:
cache_info = []
for name, size in _cache_format[opname[deop]].items():
data = code[offset + 2: offset + 2 + 2 * size]
cache_info.append((name, size, data))
else:
cache_info = None
yield Instruction(_all_opname[op], op, arg, argval, argrepr,
offset, start_offset, starts_line, line_number,
labels_map.get(offset, None), positions, cache_info)
def disassemble(co, lasti=-1, *, file=None, show_caches=False, adaptive=False,
show_offsets=False):
@ -787,7 +785,6 @@ def _disassemble_bytes(code, lasti=-1, varname_from_oparg=None,
instrs = _get_instructions_bytes(code, linestarts=linestarts,
line_offset=line_offset,
co_positions=co_positions,
show_caches=show_caches,
original_code=original_code,
labels_map=labels_map,
arg_resolver=arg_resolver)
@ -805,6 +802,23 @@ def print_instructions(instrs, exception_entries, formatter, show_caches=False,
is_current_instr = instr.offset <= lasti \
<= instr.offset + 2 * _get_cache_size(_all_opname[_deoptop(instr.opcode)])
formatter.print_instruction(instr, is_current_instr)
deop = _deoptop(instr.opcode)
if show_caches and instr.cache_info:
offset = instr.offset
for name, size, data in instr.cache_info:
for i in range(size):
offset += 2
# Only show the fancy argrepr for a CACHE instruction when it's
# the first entry for a particular cache value:
if i == 0:
argrepr = f"{name}: {int.from_bytes(data, sys.byteorder)}"
else:
argrepr = ""
formatter.print_instruction(
Instruction("CACHE", CACHE, 0, None, argrepr, offset, offset,
False, None, None, instr.positions),
is_current_instr)
formatter.print_exception_table(exception_entries)
def _disassemble_str(source, **kwargs):
@ -952,7 +966,6 @@ class Bytecode:
linestarts=self._linestarts,
line_offset=self._line_offset,
co_positions=co.co_positions(),
show_caches=self.show_caches,
original_code=original_code,
labels_map=labels_map,
arg_resolver=arg_resolver)

View file

@ -7,6 +7,18 @@ from _testinternalcapi import compiler_codegen, optimize_cfg, assemble_code_obje
_UNSPECIFIED = object()
def instructions_with_positions(instrs, co_positions):
# Return (instr, positions) pairs from the instrs list and co_positions
# iterator. The latter contains items for cache lines and the former
# doesn't, so those need to be skipped.
co_positions = co_positions or iter(())
for instr in instrs:
yield instr, next(co_positions, ())
for _, size, _ in (instr.cache_info or ()):
for i in range(size):
next(co_positions, ())
class BytecodeTestCase(unittest.TestCase):
"""Custom assertion methods for inspecting bytecode."""

View file

@ -144,6 +144,8 @@ from test.support import (cpython_only,
gc_collect)
from test.support.script_helper import assert_python_ok
from test.support import threading_helper
from test.support.bytecode_helper import (BytecodeTestCase,
instructions_with_positions)
from opcode import opmap, opname
COPY_FREE_VARS = opmap['COPY_FREE_VARS']
@ -384,10 +386,8 @@ class CodeTest(unittest.TestCase):
code = traceback.tb_frame.f_code
artificial_instructions = []
for instr, positions in zip(
dis.get_instructions(code, show_caches=True),
code.co_positions(),
strict=True
for instr, positions in instructions_with_positions(
dis.get_instructions(code), code.co_positions()
):
# If any of the positions is None, then all have to
# be None as well for the case above. There are still

View file

@ -12,6 +12,7 @@ import warnings
from test import support
from test.support import (script_helper, requires_debug_ranges,
requires_specialization, Py_C_RECURSION_LIMIT)
from test.support.bytecode_helper import instructions_with_positions
from test.support.os_helper import FakePath
class TestSpecifics(unittest.TestCase):
@ -1346,8 +1347,8 @@ class TestSourcePositions(unittest.TestCase):
def assertOpcodeSourcePositionIs(self, code, opcode,
line, end_line, column, end_column, occurrence=1):
for instr, position in zip(
dis.Bytecode(code, show_caches=True), code.co_positions(), strict=True
for instr, position in instructions_with_positions(
dis.Bytecode(code), code.co_positions()
):
if instr.opname == opcode:
occurrence -= 1

View file

@ -13,6 +13,7 @@ from test.support.bytecode_helper import BytecodeTestCase
import opcode
CACHE = dis.opmap["CACHE"]
def get_tb():
def _error():
@ -1227,9 +1228,9 @@ class DisTests(DisTestBase):
else:
# "copy" the code to un-quicken it:
f.__code__ = f.__code__.replace()
for instruction in dis.get_instructions(
for instruction in _unroll_caches_as_Instructions(dis.get_instructions(
f, show_caches=True, adaptive=adaptive
):
), show_caches=True):
if instruction.opname == "CACHE":
yield instruction.argrepr
@ -1262,7 +1263,8 @@ class DisTests(DisTestBase):
# However, this might change in the future. So we explicitly try to find
# a CACHE entry in the instructions. If we can't do that, fail the test
for inst in dis.get_instructions(f, show_caches=True):
for inst in _unroll_caches_as_Instructions(
dis.get_instructions(f, show_caches=True), show_caches=True):
if inst.opname == "CACHE":
op_offset = inst.offset - 2
cache_offset = inst.offset
@ -1775,8 +1777,8 @@ expected_opinfo_simple = [
class InstructionTestCase(BytecodeTestCase):
def assertInstructionsEqual(self, instrs_1, instrs_2, /):
instrs_1 = [instr_1._replace(positions=None) for instr_1 in instrs_1]
instrs_2 = [instr_2._replace(positions=None) for instr_2 in instrs_2]
instrs_1 = [instr_1._replace(positions=None, cache_info=None) for instr_1 in instrs_1]
instrs_2 = [instr_2._replace(positions=None, cache_info=None) for instr_2 in instrs_2]
self.assertEqual(instrs_1, instrs_2)
class InstructionTests(InstructionTestCase):
@ -1890,9 +1892,9 @@ class InstructionTests(InstructionTestCase):
instruction.positions.col_offset,
instruction.positions.end_col_offset,
)
for instruction in dis.get_instructions(
for instruction in _unroll_caches_as_Instructions(dis.get_instructions(
code, adaptive=adaptive, show_caches=show_caches
)
), show_caches=show_caches)
]
self.assertEqual(co_positions, dis_positions)
@ -2233,6 +2235,31 @@ class TestDisTracebackWithFile(TestDisTraceback):
dis.distb(tb, file=output)
return output.getvalue()
def _unroll_caches_as_Instructions(instrs, show_caches=False):
# Cache entries are no longer reported by dis as fake instructions,
# but some tests assume that do. We should rewrite the tests to assume
# the new API, but it will be clearer to keep the tests working as
# before and do that in a separate PR.
for instr in instrs:
yield instr
if not show_caches:
continue
offset = instr.offset
for name, size, data in (instr.cache_info or ()):
for i in range(size):
offset += 2
# Only show the fancy argrepr for a CACHE instruction when it's
# the first entry for a particular cache value:
if i == 0:
argrepr = f"{name}: {int.from_bytes(data, sys.byteorder)}"
else:
argrepr = ""
yield Instruction("CACHE", CACHE, 0, None, argrepr, offset, offset,
False, None, None, instr.positions)
if __name__ == "__main__":
unittest.main()

View file

@ -0,0 +1,3 @@
:mod:`dis` module functions add cache information to the
:class:`~dis.Instruction` instance rather than creating fake
:class:`~dis.Instruction` instances to represent the cache entries.