Skip to content

GH-135904: Improve the JIT's performance on macOS #136528

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
merged 8 commits into from
Jul 14, 2025
Merged

GH-135904: Improve the JIT's performance on macOS #136528

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
aa2b0df
Make jump targets local instead of extern
brandtbucher Jul 7, 2025
3a6c20d
Add __attribute__((visibility("hidden")))
brandtbucher Jul 8, 2025
3ff9ee7
Rework AArch64 alignment
brandtbucher Jul 8, 2025
d66cc85
fixup
brandtbucher Jul 8, 2025
1f6bdfc
fixup
brandtbucher Jul 8, 2025
cff0f09
Detect ";" comments
brandtbucher Jul 8, 2025
997a858
Handle symbol and label prefixes in a principled way
brandtbucher Jul 9, 2025
9544d5d
Fix mypy
brandtbucher Jul 11, 2025
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
Prev Previous commit
Next Next commit
Handle symbol and label prefixes in a principled way
  • Loading branch information
@brandtbucher
brandtbucher committed Jul 9, 2025
commit 997a858eed3067dac0e09593d51ebc15e7c95c76
15 changes: 10 additions & 5 deletions Tools/jit/_optimizers.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -70,8 +70,9 @@ class Optimizer:

path: pathlib.Path
_: dataclasses.KW_ONLY
# prefix used to mangle symbols on some platforms:
prefix: str = ""
# Prefixes used to mangle local labels and symbols:
label_prefix: str
symbol_prefix: str
# The first block in the linked list:
_root: _Block = dataclasses.field(init=False, default_factory=_Block)
_labels: dict[str, _Block] = dataclasses.field(init=False, default_factory=dict)
Expand Down Expand Up @@ -130,8 +131,12 @@ def __post_init__(self) -> None:
block.fallthrough = False

def _preprocess(self, text: str) -> str:
# Override this method to do preprocessing of the textual assembly:
return text
# Override this method to do preprocessing of the textual assembly.
# In all cases, replace references to the _JIT_CONTINUE symbol with
# references to a local _JIT_CONTINUE label (which we will add later):
continue_symbol = rf"\b{re.escape(self.symbol_prefix)}_JIT_CONTINUE\b"
continue_label = f"{self.label_prefix}_JIT_CONTINUE"
return re.sub(continue_symbol, continue_label, text)

@classmethod
def _invert_branch(cls, line: str, target: str) -> str | None:
Expand Down Expand Up @@ -198,7 +203,7 @@ def _insert_continue_label(self) -> None:
# jmp FOO
# _JIT_CONTINUE:
# This lets the assembler encode _JIT_CONTINUE jumps at build time!
continuation = self._lookup_label(f"{self.prefix}_JIT_CONTINUE")
continuation = self._lookup_label(f"{self.label_prefix}_JIT_CONTINUE")
assert continuation.label
continuation.noninstructions.append(f"{continuation.label}:")
end.link, continuation.link = continuation, end.link
Expand Down
63 changes: 42 additions & 21 deletions Tools/jit/_targets.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,7 +44,8 @@ class _Target(typing.Generic[_S, _R]):
_: dataclasses.KW_ONLY
args: typing.Sequence[str] = ()
optimizer: type[_optimizers.Optimizer] = _optimizers.Optimizer
prefix: str = ""
label_prefix: typing.ClassVar[str]
symbol_prefix: typing.ClassVar[str]
stable: bool = False
debug: bool = False
verbose: bool = False
Expand Down Expand Up @@ -164,7 +165,9 @@ async def _compile(
*shlex.split(self.cflags),
]
await _llvm.run("clang", args_s, echo=self.verbose)
self.optimizer(s, prefix=self.prefix).run()
self.optimizer(
s, label_prefix=self.label_prefix, symbol_prefix=self.symbol_prefix
).run()
args_o = [f"--target={self.triple}", "-c", "-o", f"{o}", f"{s}"]
await _llvm.run("clang", args_o, echo=self.verbose)
return await self._parse(o)
Expand Down Expand Up @@ -266,7 +269,7 @@ def _handle_section(
symbol = wrapped_symbol["Symbol"]
offset = base + symbol["Value"]
name = symbol["Name"]
name = name.removeprefix(self.prefix)
name = name.removeprefix(self.symbol_prefix)
if name not in group.symbols:
group.symbols[name] = value, offset
for wrapped_relocation in section["Relocations"]:
Expand All @@ -277,9 +280,9 @@ def _handle_section(
def _unwrap_dllimport(self, name: str) -> tuple[_stencils.HoleValue, str | None]:
if name.startswith("__imp_"):
name = name.removeprefix("__imp_")
name = name.removeprefix(self.prefix)
name = name.removeprefix(self.symbol_prefix)
return _stencils.HoleValue.GOT, name
name = name.removeprefix(self.prefix)
name = name.removeprefix(self.symbol_prefix)
return _stencils.symbol_to_value(name)

def _handle_relocation(
Expand Down Expand Up @@ -327,9 +330,24 @@ def _handle_relocation(
return _stencils.Hole(offset, kind, value, symbol, addend)


class _COFF32(_COFF):
# These mangle like Mach-O and other "older" formats:
label_prefix: str = "L"
symbol_prefix: str = "_"


class _COFF64(_COFF):
# These mangle like ELF and other "newer" formats:
label_prefix: str = ".L"
symbol_prefix: str = ""


class _ELF(
_Target[_schema.ELFSection, _schema.ELFRelocation]
): # pylint: disable = too-few-public-methods
label_prefix = ".L"
symbol_prefix = ""

def _handle_section(
self, section: _schema.ELFSection, group: _stencils.StencilGroup
) -> None:
Expand Down Expand Up @@ -366,7 +384,7 @@ def _handle_section(
symbol = wrapped_symbol["Symbol"]
offset = len(stencil.body) + symbol["Value"]
name = symbol["Name"]["Name"]
name = name.removeprefix(self.prefix)
name = name.removeprefix(self.symbol_prefix)
group.symbols[name] = value, offset
stencil.body.extend(section["SectionData"]["Bytes"])
assert not section["Relocations"]
Expand Down Expand Up @@ -401,7 +419,7 @@ def _handle_relocation(
},
}:
offset += base
s = s.removeprefix(self.prefix)
s = s.removeprefix(self.symbol_prefix)
value, symbol = _stencils.HoleValue.GOT, s
case {
"Addend": addend,
Expand All @@ -410,7 +428,7 @@ def _handle_relocation(
"Type": {"Name": kind},
}:
offset += base
s = s.removeprefix(self.prefix)
s = s.removeprefix(self.symbol_prefix)
value, symbol = _stencils.symbol_to_value(s)
case _:
raise NotImplementedError(relocation)
Expand All @@ -420,14 +438,17 @@ def _handle_relocation(
class _MachO(
_Target[_schema.MachOSection, _schema.MachORelocation]
): # pylint: disable = too-few-public-methods
label_prefix = "L"
symbol_prefix = "_"

def _handle_section(
self, section: _schema.MachOSection, group: _stencils.StencilGroup
) -> None:
assert section["Address"] >= len(group.code.body)
assert "SectionData" in section
flags = {flag["Name"] for flag in section["Attributes"]["Flags"]}
name = section["Name"]["Value"]
name = name.removeprefix(self.prefix)
name = name.removeprefix(self.symbol_prefix)
if "Debug" in flags:
return
if "PureInstructions" in flags:
Expand All @@ -451,7 +472,7 @@ def _handle_section(
symbol = wrapped_symbol["Symbol"]
offset = symbol["Value"] - start_address
name = symbol["Name"]["Name"]
name = name.removeprefix(self.prefix)
name = name.removeprefix(self.symbol_prefix)
group.symbols[name] = value, offset
assert "Relocations" in section
for wrapped_relocation in section["Relocations"]:
Expand All @@ -476,7 +497,7 @@ def _handle_relocation(
},
}:
offset += base
s = s.removeprefix(self.prefix)
s = s.removeprefix(self.symbol_prefix)
value, symbol = _stencils.HoleValue.GOT, s
addend = 0
case {
Expand All @@ -485,7 +506,7 @@ def _handle_relocation(
"Type": {"Name": "X86_64_RELOC_GOT" | "X86_64_RELOC_GOT_LOAD" as kind},
}:
offset += base
s = s.removeprefix(self.prefix)
s = s.removeprefix(self.symbol_prefix)
value, symbol = _stencils.HoleValue.GOT, s
addend = (
int.from_bytes(raw[offset : offset + 4], "little", signed=True) - 4
Expand All @@ -500,7 +521,7 @@ def _handle_relocation(
"Type": {"Name": "X86_64_RELOC_BRANCH" | "X86_64_RELOC_SIGNED" as kind},
}:
offset += base
s = s.removeprefix(self.prefix)
s = s.removeprefix(self.symbol_prefix)
value, symbol = _stencils.symbol_to_value(s)
addend = (
int.from_bytes(raw[offset : offset + 4], "little", signed=True) - 4
Expand All @@ -515,27 +536,27 @@ def _handle_relocation(
"Type": {"Name": kind},
}:
offset += base
s = s.removeprefix(self.prefix)
s = s.removeprefix(self.symbol_prefix)
value, symbol = _stencils.symbol_to_value(s)
addend = 0
case _:
raise NotImplementedError(relocation)
return _stencils.Hole(offset, kind, value, symbol, addend)


def get_target(host: str) -> _COFF | _ELF | _MachO:
def get_target(host: str) -> _COFF32 | _COFF64 | _ELF | _MachO:
"""Build a _Target for the given host "triple" and options."""
optimizer: type[_optimizers.Optimizer]
target: _COFF | _ELF | _MachO
target: _COFF32 | _COFF64 | _ELF | _MachO
if re.fullmatch(r"aarch64-apple-darwin.*", host):
condition = "defined(__aarch64__) && defined(__APPLE__)"
optimizer = _optimizers.OptimizerAArch64
target = _MachO(host, condition, optimizer=optimizer, prefix="_")
target = _MachO(host, condition, optimizer=optimizer)
elif re.fullmatch(r"aarch64-pc-windows-msvc", host):
args = ["-fms-runtime-lib=dll", "-fplt"]
condition = "defined(_M_ARM64)"
optimizer = _optimizers.OptimizerAArch64
target = _COFF(host, condition, args=args, optimizer=optimizer)
target = _COFF64(host, condition, args=args, optimizer=optimizer)
elif re.fullmatch(r"aarch64-.*-linux-gnu", host):
# -mno-outline-atomics: Keep intrinsics from being emitted.
args = ["-fpic", "-mno-outline-atomics"]
Expand All @@ -547,16 +568,16 @@ def get_target(host: str) -> _COFF | _ELF | _MachO:
args = ["-DPy_NO_ENABLE_SHARED", "-Wno-ignored-attributes"]
optimizer = _optimizers.OptimizerX86
condition = "defined(_M_IX86)"
target = _COFF(host, condition, args=args, optimizer=optimizer, prefix="_")
target = _COFF32(host, condition, args=args, optimizer=optimizer)
elif re.fullmatch(r"x86_64-apple-darwin.*", host):
condition = "defined(__x86_64__) && defined(__APPLE__)"
optimizer = _optimizers.OptimizerX86
target = _MachO(host, condition, optimizer=optimizer, prefix="_")
target = _MachO(host, condition, optimizer=optimizer)
elif re.fullmatch(r"x86_64-pc-windows-msvc", host):
args = ["-fms-runtime-lib=dll"]
condition = "defined(_M_X64)"
optimizer = _optimizers.OptimizerX86
target = _COFF(host, condition, args=args, optimizer=optimizer)
target = _COFF64(host, condition, args=args, optimizer=optimizer)
elif re.fullmatch(r"x86_64-.*-linux-gnu", host):
args = ["-fno-pic", "-mcmodel=medium", "-mlarge-data-threshold=0"]
condition = "defined(__x86_64__) && defined(__linux__)"
Expand Down
Loading
pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy