1
2
3
4
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
45
46
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
83
84
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
|
"""Generate the main interpreter switch.
Reads the instruction definitions from bytecodes.c.
Writes the cases to generated_cases.c.h, which is #included in ceval.c.
"""
import argparse
import os.path
import sys
from analyzer import (
Analysis,
Instruction,
Uop,
Part,
analyze_files,
Skip,
StackItem,
analysis_error,
)
from generators_common import (
DEFAULT_INPUT,
ROOT,
write_header,
emit_tokens,
)
from cwriter import CWriter
from typing import TextIO, Iterator
from lexer import Token
from stack import StackOffset, Stack, SizeMismatch
DEFAULT_OUTPUT = ROOT / "Python/generated_cases.c.h"
FOOTER = "#undef TIER_ONE\n"
def declare_variables(inst: Instruction, out: CWriter) -> None:
variables = {"unused"}
for uop in inst.parts:
if isinstance(uop, Uop):
for var in reversed(uop.stack.inputs):
if var.name not in variables:
type = var.type if var.type else "PyObject *"
variables.add(var.name)
if var.condition:
out.emit(f"{type}{var.name} = NULL;\n")
else:
out.emit(f"{type}{var.name};\n")
for var in uop.stack.outputs:
if var.name not in variables:
variables.add(var.name)
type = var.type if var.type else "PyObject *"
if var.condition:
out.emit(f"{type}{var.name} = NULL;\n")
else:
out.emit(f"{type}{var.name};\n")
def write_uop(
uop: Part, out: CWriter, offset: int, stack: Stack, inst: Instruction, braces: bool
) -> int:
# out.emit(stack.as_comment() + "\n")
if isinstance(uop, Skip):
entries = "entries" if uop.size > 1 else "entry"
out.emit(f"/* Skip {uop.size} cache {entries} */\n")
return offset + uop.size
try:
out.start_line()
if braces:
out.emit(f"// {uop.name}\n")
for var in reversed(uop.stack.inputs):
out.emit(stack.pop(var))
if braces:
out.emit("{\n")
if not uop.properties.stores_sp:
for i, var in enumerate(uop.stack.outputs):
out.emit(stack.push(var))
for cache in uop.caches:
if cache.name != "unused":
if cache.size == 4:
type = "PyObject *"
reader = "read_obj"
else:
type = f"uint{cache.size*16}_t "
reader = f"read_u{cache.size*16}"
out.emit(
f"{type}{cache.name} = {reader}(&this_instr[{offset}].cache);\n"
)
if inst.family is None:
out.emit(f"(void){cache.name};\n")
offset += cache.size
emit_tokens(out, uop, stack, inst)
if uop.properties.stores_sp:
for i, var in enumerate(uop.stack.outputs):
out.emit(stack.push(var))
if braces:
out.start_line()
out.emit("}\n")
# out.emit(stack.as_comment() + "\n")
return offset
except SizeMismatch as ex:
raise analysis_error(ex.args[0], uop.body[0])
def uses_this(inst: Instruction) -> bool:
if inst.properties.needs_this:
return True
for uop in inst.parts:
if isinstance(uop, Skip):
continue
for cache in uop.caches:
if cache.name != "unused":
return True
return False
def generate_tier1(
filenames: list[str], analysis: Analysis, outfile: TextIO, lines: bool
) -> None:
write_header(__file__, filenames, outfile)
outfile.write(
"""
#ifdef TIER_TWO
#error "This file is for Tier 1 only"
#endif
#define TIER_ONE 1
"""
)
out = CWriter(outfile, 2, lines)
out.emit("\n")
for name, inst in sorted(analysis.instructions.items()):
needs_this = uses_this(inst)
out.emit("\n")
out.emit(f"TARGET({name}) {{\n")
unused_guard = "(void)this_instr;\n" if inst.family is None else ""
if needs_this and not inst.is_target:
out.emit(f"_Py_CODEUNIT *this_instr = frame->instr_ptr = next_instr;\n")
out.emit(unused_guard)
else:
out.emit(f"frame->instr_ptr = next_instr;\n")
out.emit(f"next_instr += {inst.size};\n")
out.emit(f"INSTRUCTION_STATS({name});\n")
if inst.is_target:
out.emit(f"PREDICTED({name});\n")
if needs_this:
out.emit(f"_Py_CODEUNIT *this_instr = next_instr - {inst.size};\n")
out.emit(unused_guard)
if inst.family is not None:
out.emit(
f"static_assert({inst.family.size} == {inst.size-1}"
', "incorrect cache size");\n'
)
declare_variables(inst, out)
offset = 1 # The instruction itself
stack = Stack()
for part in inst.parts:
# Only emit braces if more than one uop
insert_braces = len([p for p in inst.parts if isinstance(p, Uop)]) > 1
offset = write_uop(part, out, offset, stack, inst, insert_braces)
out.start_line()
if not inst.parts[-1].properties.always_exits:
stack.flush(out)
if inst.parts[-1].properties.ends_with_eval_breaker:
out.emit("CHECK_EVAL_BREAKER();\n")
out.emit("DISPATCH();\n")
out.start_line()
out.emit("}")
out.emit("\n")
outfile.write(FOOTER)
arg_parser = argparse.ArgumentParser(
description="Generate the code for the interpreter switch.",
formatter_class=argparse.ArgumentDefaultsHelpFormatter,
)
arg_parser.add_argument(
"-o", "--output", type=str, help="Generated code", default=DEFAULT_OUTPUT
)
arg_parser.add_argument(
"-l", "--emit-line-directives", help="Emit #line directives", action="store_true"
)
arg_parser.add_argument(
"input", nargs=argparse.REMAINDER, help="Instruction definition file(s)"
)
def generate_tier1_from_files(
filenames: list[str], outfilename: str, lines: bool
) -> None:
data = analyze_files(filenames)
with open(outfilename, "w") as outfile:
generate_tier1(filenames, data, outfile, lines)
if __name__ == "__main__":
args = arg_parser.parse_args()
if len(args.input) == 0:
args.input.append(DEFAULT_INPUT)
data = analyze_files(args.input)
with open(args.output, "w") as outfile:
generate_tier1(args.input, data, outfile, args.emit_line_directives)
|
