iframe-proxy | Sunbelt Computer Software

History

177 lines (136 loc) · 4.78 KB

Raw

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

"""

Generate Lib/_opcode_metadata.py for RustPython bytecode.

This file generates opcode metadata that is compatible with CPython 3.13.

"""

import itertools

import pathlib

import re

import typing

ROOT = pathlib.Path(__file__).parents[1]

BYTECODE_FILE = (

ROOT / "crates" / "compiler-core" / "src" / "bytecode" / "instruction.rs"

)

OPCODE_METADATA_FILE = ROOT / "Lib" / "_opcode_metadata.py"

# Opcodes that needs to be first, regardless of their opcode ID.

PRIORITY_OPMAP = {

"CACHE",

"RESERVED",

"RESUME",

"INSTRUMENTED_LINE",

"ENTER_EXECUTOR",

}

def to_pascal_case(s: str) -> str:

res = re.sub(r"(?<=[a-z0-9])([A-Z])", r"_\1", s)

return re.sub(r"(\D)(\d+)$", r"\1_\2", res).upper()

class Opcode(typing.NamedTuple):

rust_name: str

id: int

have_oparg: bool

@property

def cpython_name(self) -> str:

return to_pascal_case(self.rust_name)

@property

def is_instrumented(self):

return self.cpython_name.startswith("INSTRUMENTED_")

@classmethod

def from_str(cls, body: str):

raw_variants = re.split(r"(\d+),", body.strip())

raw_variants.remove("")

for raw_name, raw_id in itertools.batched(raw_variants, 2, strict=True):

have_oparg = "Arg<" in raw_name # Hacky but works

name = re.findall(r"\b[A-Z][A-Za-z]*\d*\b(?=\s*[\({=])", raw_name)[0]

yield cls(rust_name=name.strip(), id=int(raw_id), have_oparg=have_oparg)

def __lt__(self, other: typing.Self) -> bool:

sprio, oprio = (

opcode.cpython_name not in PRIORITY_OPMAP for opcode in (self, other)

)

return (sprio, self.id) < (oprio, other.id)

def extract_enum_body(contents: str, enum_name: str) -> str:

res = re.search(f"pub enum {enum_name} " + r"\{(.+?)\n\}", contents, re.DOTALL)

if not res:

raise ValueError(f"Could not find {enum_name} enum")

return "\n".join(

line.split("//")[0].strip() # Remove any comment. i.e. "foo // some comment"

for line in res.group(1).splitlines()

if not line.strip().startswith("//") # Ignore comment lines

)

def build_deopts(contents: str) -> dict[str, list[str]]:

raw_body = re.search(

r"fn deopt$self$ -> Option<Self>(.*)", contents, re.DOTALL

).group(1)

body = "\n".join(

itertools.takewhile(

lambda l: not l.startswith("_ =>"), # Take until reaching fallback

filter(

lambda l: (

not l.startswith(

("//", "Some(match")

) # Skip comments or start of match

map(str.strip, raw_body.splitlines()),

)

).removeprefix("{")

depth = 0

arms = []

buf = []

for char in body:

if char == "{":

depth += 1

elif char == "}":

depth -= 1

if depth == 0 and (char in ("}", ",")):

arm = "".join(buf).strip()

arms.append(arm)

buf = []

else:

buf.append(char)

# last arm

arms.append("".join(buf))

arms = [arm for arm in arms if arm]

deopts = {}

for arm in arms:

*specialized, deopt = map(to_pascal_case, re.findall(r"Self::(\w*)\b", arm))

deopts[deopt] = specialized

return deopts

contents = BYTECODE_FILE.read_text(encoding="utf-8")

deopts = build_deopts(contents)

enum_body = "\n".join(

extract_enum_body(contents, enum_name)

for enum_name in ("Instruction", "PseudoInstruction")

)

opcodes = list(Opcode.from_str(enum_body))

have_oparg = min(opcode.id for opcode in opcodes if opcode.have_oparg) - 1

min_instrumented = min(opcode.id for opcode in opcodes if opcode.is_instrumented)

# Generate the output file

output = """# This file is generated by scripts/generate_opcode_metadata.py

# for RustPython bytecode format (CPython 3.14 compatible opcode numbers).

# Do not edit!

"""

output += "\n_specializations = {\n"

for key, lst in deopts.items():

output += f' "{key}": [\n'

for item in lst:

output += f' "{item}",\n'

output += " ],\n"

output += "}\n"

specialized = set(itertools.chain.from_iterable(deopts.values()))

output += "\n_specialized_opmap = {\n"

for opcode in sorted(opcodes, key=lambda op: op.cpython_name):

cpython_name = opcode.cpython_name

if cpython_name not in specialized:

continue

output += f" '{cpython_name}': {opcode.id},\n"

output += "}\n"

output += "\nopmap = {\n"

for opcode in sorted(opcodes):

cpython_name = opcode.cpython_name

if cpython_name in specialized:

continue

output += f" '{cpython_name}': {opcode.id},\n"

output += "}\n"

output += f"""

HAVE_ARGUMENT = {have_oparg}

MIN_INSTRUMENTED_OPCODE = {min_instrumented}

"""

OPCODE_METADATA_FILE.write_text(output, encoding="utf-8")

Sunbelt Computer Software

PL/B Language Development and Support

Navigation Menu

Uh oh!

Sunbelt Computer Software

PL/B Language Development and Support

Uh oh!

FilesExpand file tree

generate_opcode_metadata.py

Latest commit

History

generate_opcode_metadata.py

File metadata and controls