gh-102021 : Allow multiple input files for interpreter loop generator (#102022)

The input files no longer use `-i`.

6 files changed, 105 insertions, 41 deletions

diff --git a/Makefile.pre.in b/Makefile.pre.in
index b12a1bc..1a1853b 100644
--- a/Makefile.pre.in
+++ b/Makefile.pre.in
@@ -1485,9 +1485,9 @@ regen-cases:
 	PYTHONPATH=$(srcdir)/Tools/cases_generator \
 	$(PYTHON_FOR_REGEN) \
 	    $(srcdir)/Tools/cases_generator/generate_cases.py \
-		-i $(srcdir)/Python/bytecodes.c \
 		-o $(srcdir)/Python/generated_cases.c.h.new \
-		-m $(srcdir)/Python/opcode_metadata.h.new
+		-m $(srcdir)/Python/opcode_metadata.h.new \
+		$(srcdir)/Python/bytecodes.c
 	$(UPDATE_FILE) $(srcdir)/Python/generated_cases.c.h $(srcdir)/Python/generated_cases.c.h.new
 	$(UPDATE_FILE) $(srcdir)/Python/opcode_metadata.h $(srcdir)/Python/opcode_metadata.h.new
 
diff --git a/Python/generated_cases.c.h b/Python/generated_cases.c.h
index f59f7c1..82e1850 100644
--- a/Python/generated_cases.c.h
+++ b/Python/generated_cases.c.h
@@ -1,5 +1,6 @@
 // This file is generated by Tools/cases_generator/generate_cases.py
-// from Python/bytecodes.c
+// from:
+//   Python/bytecodes.c
 // Do not edit!
 
         TARGET(NOP) {
diff --git a/Python/opcode_metadata.h b/Python/opcode_metadata.h
index f27906a..67cb008 100644
--- a/Python/opcode_metadata.h
+++ b/Python/opcode_metadata.h
@@ -1,5 +1,6 @@
-// This file is generated by Tools/cases_generator/generate_cases.py --metadata
-// from Python/bytecodes.c
+// This file is generated by Tools/cases_generator/generate_cases.py
+// from:
+//   Python/bytecodes.c
 // Do not edit!
 
 #ifndef NEED_OPCODE_TABLES
diff --git a/Tools/cases_generator/generate_cases.py b/Tools/cases_generator/generate_cases.py
index b760172..25cf75e 100644
--- a/Tools/cases_generator/generate_cases.py
+++ b/Tools/cases_generator/generate_cases.py
@@ -38,14 +38,14 @@ arg_parser = argparse.ArgumentParser(
     formatter_class=argparse.ArgumentDefaultsHelpFormatter,
 )
 arg_parser.add_argument(
-    "-i", "--input", type=str, help="Instruction definitions", default=DEFAULT_INPUT
-)
-arg_parser.add_argument(
     "-o", "--output", type=str, help="Generated code", default=DEFAULT_OUTPUT
 )
 arg_parser.add_argument(
     "-m", "--metadata", type=str, help="Generated metadata", default=DEFAULT_METADATA_OUTPUT
 )
+arg_parser.add_argument(
+    "input", nargs=argparse.REMAINDER, help="Instruction definition file(s)"
+)
 
 
 def effect_size(effect: StackEffect) -> tuple[int, str]:
@@ -485,6 +485,11 @@ class MacroInstruction(SuperOrMacroInstruction):
     parts: list[Component | parser.CacheEffect]
 
 
+@dataclasses.dataclass
+class OverriddenInstructionPlaceHolder:
+    name: str
+
+
 AnyInstruction = Instruction | SuperInstruction | MacroInstruction
 INSTR_FMT_PREFIX = "INSTR_FMT_"
 
@@ -492,32 +497,33 @@ INSTR_FMT_PREFIX = "INSTR_FMT_"
 class Analyzer:
     """Parse input, analyze it, and write to output."""
 
-    filename: str
+    input_filenames: list[str]
     output_filename: str
     metadata_filename: str
-    src: str
     errors: int = 0
 
-    def __init__(self, filename: str, output_filename: str, metadata_filename: str):
+    def __init__(self, input_filenames: list[str], output_filename: str, metadata_filename: str):
         """Read the input file."""
-        self.filename = filename
+        self.input_filenames = input_filenames
         self.output_filename = output_filename
         self.metadata_filename = metadata_filename
-        with open(filename) as f:
-            self.src = f.read()
 
     def error(self, msg: str, node: parser.Node) -> None:
         lineno = 0
+        filename = "<unknown file>"
         if context := node.context:
+            filename = context.owner.filename
             # Use line number of first non-comment in the node
             for token in context.owner.tokens[context.begin : context.end]:
                 lineno = token.line
                 if token.kind != "COMMENT":
                     break
-        print(f"{self.filename}:{lineno}: {msg}", file=sys.stderr)
+        print(f"{filename}:{lineno}: {msg}", file=sys.stderr)
         self.errors += 1
 
-    everything: list[parser.InstDef | parser.Super | parser.Macro]
+    everything: list[
+        parser.InstDef | parser.Super | parser.Macro | OverriddenInstructionPlaceHolder
+    ]
     instrs: dict[str, Instruction]  # Includes ops
     supers: dict[str, parser.Super]
     super_instrs: dict[str, SuperInstruction]
@@ -531,7 +537,31 @@ class Analyzer:
         We only want the parser to see the stuff between the
         begin and end markers.
         """
-        psr = parser.Parser(self.src, filename=self.filename)
+
+        self.everything = []
+        self.instrs = {}
+        self.supers = {}
+        self.macros = {}
+        self.families = {}
+
+        instrs_idx: dict[str, int] = dict()
+
+        for filename in self.input_filenames:
+            self.parse_file(filename, instrs_idx)
+
+        files = " + ".join(self.input_filenames)
+        print(
+            f"Read {len(self.instrs)} instructions/ops, "
+            f"{len(self.supers)} supers, {len(self.macros)} macros, "
+            f"and {len(self.families)} families from {files}",
+            file=sys.stderr,
+        )
+
+    def parse_file(self, filename: str, instrs_idx: dict[str, int]) -> None:
+        with open(filename) as file:
+            src = file.read()
+
+        psr = parser.Parser(src, filename=filename)
 
         # Skip until begin marker
         while tkn := psr.next(raw=True):
@@ -551,16 +581,27 @@ class Analyzer:
 
         # Parse from start
         psr.setpos(start)
-        self.everything = []
-        self.instrs = {}
-        self.supers = {}
-        self.macros = {}
-        self.families = {}
         thing: parser.InstDef | parser.Super | parser.Macro | parser.Family | None
+        thing_first_token = psr.peek()
         while thing := psr.definition():
             match thing:
                 case parser.InstDef(name=name):
+                    if name in self.instrs:
+                        if not thing.override:
+                            raise psr.make_syntax_error(
+                                f"Duplicate definition of '{name}' @ {thing.context} "
+                                f"previous definition @ {self.instrs[name].inst.context}",
+                                thing_first_token,
+                            )
+                        self.everything[instrs_idx[name]] = OverriddenInstructionPlaceHolder(name=name)
+                    if name not in self.instrs and thing.override:
+                        raise psr.make_syntax_error(
+                            f"Definition of '{name}' @ {thing.context} is supposed to be "
+                            "an override but no previous definition exists.",
+                            thing_first_token,
+                        )
                     self.instrs[name] = Instruction(thing)
+                    instrs_idx[name] = len(self.everything)
                     self.everything.append(thing)
                 case parser.Super(name):
                     self.supers[name] = thing
@@ -573,14 +614,7 @@ class Analyzer:
                 case _:
                     typing.assert_never(thing)
         if not psr.eof():
-            raise psr.make_syntax_error("Extra stuff at the end")
-
-        print(
-            f"Read {len(self.instrs)} instructions/ops, "
-            f"{len(self.supers)} supers, {len(self.macros)} macros, "
-            f"and {len(self.families)} families from {self.filename}",
-            file=sys.stderr,
-        )
+            raise psr.make_syntax_error(f"Extra stuff at the end of {filename}")
 
     def analyze(self) -> None:
         """Analyze the inputs.
@@ -879,6 +913,8 @@ class Analyzer:
         popped_data: list[tuple[AnyInstruction, str]] = []
         pushed_data: list[tuple[AnyInstruction, str]] = []
         for thing in self.everything:
+            if isinstance(thing, OverriddenInstructionPlaceHolder):
+                continue
             instr, popped, pushed = self.get_stack_effect_info(thing)
             if instr is not None:
                 popped_data.append((instr, popped))
@@ -907,6 +943,13 @@ class Analyzer:
         write_function("pushed", pushed_data)
         self.out.emit("")
 
+    def from_source_files(self) -> str:
+        paths = "\n//   ".join(
+            os.path.relpath(filename, ROOT).replace(os.path.sep, posixpath.sep)
+            for filename in self.input_filenames
+        )
+        return f"// from:\n//   {paths}\n"
+
     def write_metadata(self) -> None:
         """Write instruction metadata to output file."""
 
@@ -914,6 +957,8 @@ class Analyzer:
         all_formats: set[str] = set()
         for thing in self.everything:
             match thing:
+                case OverriddenInstructionPlaceHolder():
+                    continue
                 case parser.InstDef():
                     format = self.instrs[thing.name].instr_fmt
                 case parser.Super():
@@ -928,8 +973,8 @@ class Analyzer:
 
         with open(self.metadata_filename, "w") as f:
             # Write provenance header
-            f.write(f"// This file is generated by {THIS} --metadata\n")
-            f.write(f"// from {os.path.relpath(self.filename, ROOT).replace(os.path.sep, posixpath.sep)}\n")
+            f.write(f"// This file is generated by {THIS}\n")
+            f.write(self.from_source_files())
             f.write(f"// Do not edit!\n")
 
             # Create formatter; the rest of the code uses this
@@ -959,6 +1004,8 @@ class Analyzer:
             # Write metadata for each instruction
             for thing in self.everything:
                 match thing:
+                    case OverriddenInstructionPlaceHolder():
+                        continue
                     case parser.InstDef():
                         if thing.kind != "op":
                             self.write_metadata_for_inst(self.instrs[thing.name])
@@ -1008,7 +1055,7 @@ class Analyzer:
         with open(self.output_filename, "w") as f:
             # Write provenance header
             f.write(f"// This file is generated by {THIS}\n")
-            f.write(f"// from {os.path.relpath(self.filename, ROOT).replace(os.path.sep, posixpath.sep)}\n")
+            f.write(self.from_source_files())
             f.write(f"// Do not edit!\n")
 
             # Create formatter; the rest of the code uses this
@@ -1020,6 +1067,8 @@ class Analyzer:
             n_macros = 0
             for thing in self.everything:
                 match thing:
+                    case OverriddenInstructionPlaceHolder():
+                        self.write_overridden_instr_place_holder(thing)
                     case parser.InstDef():
                         if thing.kind != "op":
                             n_instrs += 1
@@ -1039,9 +1088,17 @@ class Analyzer:
             file=sys.stderr,
         )
 
+    def write_overridden_instr_place_holder(self,
+            place_holder: OverriddenInstructionPlaceHolder) -> None:
+        self.out.emit("")
+        self.out.emit(
+            f"// TARGET({place_holder.name}) overridden by later definition")
+
     def write_instr(self, instr: Instruction) -> None:
         name = instr.name
         self.out.emit("")
+        if instr.inst.override:
+            self.out.emit("// Override")
         with self.out.block(f"TARGET({name})"):
             if instr.predicted:
                 self.out.emit(f"PREDICTED({name});")
@@ -1190,6 +1247,8 @@ def variable_used(node: parser.Node, name: str) -> bool:
 def main():
     """Parse command line, parse input, analyze, write output."""
     args = arg_parser.parse_args()  # Prints message and sys.exit(2) on error
+    if len(args.input) == 0:
+        args.input.append(DEFAULT_INPUT)
     a = Analyzer(args.input, args.output, args.metadata)  # Raises OSError if input unreadable
     a.parse()  # Raises SyntaxError on failure
     a.analyze()  # Prints messages and sets a.errors on failure
diff --git a/Tools/cases_generator/lexer.py b/Tools/cases_generator/lexer.py
index 39b6a21..1c70d1c 100644
--- a/Tools/cases_generator/lexer.py
+++ b/Tools/cases_generator/lexer.py
@@ -119,7 +119,7 @@ letter = re.compile(r'[a-zA-Z_]')
 kwds = (
     'AUTO', 'BREAK', 'CASE', 'CHAR', 'CONST',
     'CONTINUE', 'DEFAULT', 'DO', 'DOUBLE', 'ELSE', 'ENUM', 'EXTERN',
-    'FLOAT', 'FOR', 'GOTO', 'IF', 'INLINE', 'INT', 'LONG',
+    'FLOAT', 'FOR', 'GOTO', 'IF', 'INLINE', 'INT', 'LONG', 'OVERRIDE',
     'REGISTER', 'OFFSETOF',
     'RESTRICT', 'RETURN', 'SHORT', 'SIGNED', 'SIZEOF', 'STATIC', 'STRUCT',
     'SWITCH', 'TYPEDEF', 'UNION', 'UNSIGNED', 'VOID',
diff --git a/Tools/cases_generator/parser.py b/Tools/cases_generator/parser.py
index c7c8d8a..7bf45a3 100644
--- a/Tools/cases_generator/parser.py
+++ b/Tools/cases_generator/parser.py
@@ -33,7 +33,7 @@ class Context(NamedTuple):
     owner: PLexer
 
     def __repr__(self):
-        return f"<{self.begin}-{self.end}>"
+        return f"<{self.owner.filename}: {self.begin}-{self.end}>"
 
 
 @dataclass
@@ -99,6 +99,7 @@ UOp = OpName | CacheEffect
 
 @dataclass
 class InstHeader(Node):
+    override: bool
     register: bool
     kind: Literal["inst", "op", "legacy"]  # Legacy means no (inputs -- outputs)
     name: str
@@ -108,6 +109,7 @@ class InstHeader(Node):
 
 @dataclass
 class InstDef(Node):
+    override: bool
     register: bool
     kind: Literal["inst", "op", "legacy"]
     name: str
@@ -152,17 +154,18 @@ class Parser(PLexer):
         if hdr := self.inst_header():
             if block := self.block():
                 return InstDef(
-                    hdr.register, hdr.kind, hdr.name, hdr.inputs, hdr.outputs, block
+                    hdr.override, hdr.register, hdr.kind, hdr.name, hdr.inputs, hdr.outputs, block
                 )
             raise self.make_syntax_error("Expected block")
         return None
 
     @contextual
     def inst_header(self) -> InstHeader | None:
-        # inst(NAME)
-        #   | [register] inst(NAME, (inputs -- outputs))
-        #   | [register] op(NAME, (inputs -- outputs))
+        # [override] inst(NAME)
+        #   | [override] [register] inst(NAME, (inputs -- outputs))
+        #   | [override] [register] op(NAME, (inputs -- outputs))
         # TODO: Make INST a keyword in the lexer.
+        override = bool(self.expect(lx.OVERRIDE))
         register = bool(self.expect(lx.REGISTER))
         if (tkn := self.expect(lx.IDENTIFIER)) and (kind := tkn.text) in ("inst", "op"):
             if self.expect(lx.LPAREN) and (tkn := self.expect(lx.IDENTIFIER)):
@@ -171,10 +174,10 @@ class Parser(PLexer):
                     inp, outp = self.io_effect()
                     if self.expect(lx.RPAREN):
                         if (tkn := self.peek()) and tkn.kind == lx.LBRACE:
-                            return InstHeader(register, kind, name, inp, outp)
+                            return InstHeader(override, register, kind, name, inp, outp)
                 elif self.expect(lx.RPAREN) and kind == "inst":
                     # No legacy stack effect if kind is "op".
-                    return InstHeader(register, "legacy", name, [], [])
+                    return InstHeader(override, register, "legacy", name, [], [])
         return None
 
     def io_effect(self) -> tuple[list[InputEffect], list[OutputEffect]]: