"""Parser for LLVM--.""" from collections import namedtuple import ply.lex as lex import ply.yacc as yacc from enum import Enum class SimpleType(Enum): Void = 1 I1 = 2 I8 = 3 I32 = 4 I64 = 5 # Namedtuples for storing AST Program = namedtuple('Program', ['tdecls', 'gdecls', 'fdecls']) TypeDec = namedtuple('TypeDec', ['name', 'body']) PointerType = namedtuple('PointerType', ['inner_ty']) StructType = namedtuple('StructType', ['fields']) ArrayType = namedtuple('ArrayType', ['length', 'inner_ty']) FunctionType = namedtuple('FunctionType', ['return_ty', 'parameters']) NamedType = namedtuple('NamedType', ['other_name']) GlobalDec = namedtuple('GlobalDec', ['name', 'ty', 'body']) GNull = namedtuple('GNull', []) GGid = namedtuple('GGid', ['val']) GInt = namedtuple('GInt', ['val']) GString = namedtuple('GString', ['val']) GArray = namedtuple('GArray', ['entries']) GStruct = namedtuple('GStruct', ['fields']) FunctionDec = namedtuple('FunctionDec', ['return_type', 'name', 'parameters', 'body']) FunctionBody = namedtuple('FunctionBody', ['first_block', 'named_blocks']) Block = namedtuple('Block', ['insns', 'terminator']) Binop = namedtuple('Binop', ['bop', 'ty', 'left', 'right']) Alloca = namedtuple('Alloca', ['ty']) Load = namedtuple('Load', ['ty', 'oper']) Store = namedtuple('Store', ['ty', 'value', 'location']) Icmp = namedtuple('Icmp', ['cnd', 'ty', 'left', 'right']) Call = namedtuple('Call', ['return_ty', 'callee', 'arguments']) Bitcast = namedtuple('Bitcast', ['from_ty', 'oper', 'to_ty']) Gep = namedtuple('Gep', ['base_ty', 'oper_ty', 'oper', 'steps']) Zext = namedtuple('Zext', ['from_ty', 'oper', 'to_ty']) Ptrtoint = namedtuple('Ptrtoint', ['pointer_ty', 'oper', 'to_ty']) Ret = namedtuple('Ret', ['ty', 'oper']) Br = namedtuple('Br', ['label']) Cbr = namedtuple('Cbr', ['ty', 'oper', 'then_label', 'else_label']) class LLVMParser(object): # Lexer current_string = "" line_begin = 0 states = [ ('string', 'exclusive') ] reserved = { 'add': 'ADD', 'sub': 'SUB', 'mul': 'MUL', 'shl': 'SHL', 'lshr': 'LSHR', 'ashr': 'ASHR', 'and': 'AND', 'or': 'OR', 'xor': 'XOR', 'sdiv': 'SDIV', 'eq': 'EQ', 'ne': 'NE', 'slt': 'SLT', 'sle': 'SLE', 'sgt': 'SGT', 'sge': 'SGE', 'alloca': 'ALLOCA', 'load': 'LOAD', 'store': 'STORE', 'icmp': 'ICMP', 'call': 'CALL', 'bitcast': 'BITCAST', 'getelementptr': 'GETELEMENTPTR', 'zext': 'ZEXT', 'ptrtoint': 'PTRTOINT', 'ret': 'RET', 'br': 'BR', 'label': 'LABEL', 'define': 'DEFINE', 'null': 'NULL', 'global': 'GLOBAL', 'type': 'TYPE', 'to': 'TO', 'void': 'VOID', 'i1': 'I1', 'i8': 'I8', 'i32': 'I32', 'i64': 'I64', } tokens = [ 'INT', 'STRING', 'ASTERIX', 'LPAREN', 'RPAREN', 'LBRACK', 'RBRACK', 'LBRACE', 'RBRACE', 'ASSIGN', 'COLON', 'COMMA', 'PercentID', 'AtID', 'ID' ] t_ignore = ' \t' t_string_ignore = '' t_ASTERIX = r'\*' t_LPAREN = r'\(' t_RPAREN = r'\)' t_LBRACK = r'\[' t_RBRACK = r'\]' t_LBRACE = r'{' t_RBRACE = r'}' t_ASSIGN = r'=' t_COLON = r':' t_COMMA = r',' def t_COMMENT(self, t): r'(;|declare|target).*' pass def t_newline(self, t): r'\n+' t.lexer.lineno += len(t.value) self.line_begin = t.lexpos def t_string_newline(self, t): r'\n' print("{}:{}: Newline is not allowed inside strings." .format(t.lineno, t.lexpos - self.line_begin)) t.lexer.lineno += 1 self.line_begin = t.lexpos def t_INT(self, t): r'\d+' t.value = int(t.value) return t def t_begin_string(self, t): r'c\"' t.lexer.begin('string') self.current_string = "" def t_string_doublequote(self, t): r'\\"' self.current_string += '"' def t_string_backslash(self, t): r'\\\\' self.current_string += '\\' def t_string_hex(self, t): r'\\[0-9a-f][0-9a-f]' code = int(t.value[1:], 16) self.current_string += chr(code) def t_string_singlebackslash(self, t): r'\\' print("{}:{}: Single backslash is not allowed inside strings." .format(t.lineno, t.lexpos - self.line_begin)) def t_string_end(self, t): r'"' t.value = self.current_string self.current_string = "" t.type = "STRING" t.lexer.begin('INITIAL') return t def t_string_meat(self, t): r'.' self.current_string += t.value def t_ID(self, t): r'[a-zA-Z0-9_-]+' t.type = self.reserved.get(t.value, 'ID') return t def t_PercentID(self, t): r'%[a-zA-Z0-9_-]+' t.value = t.value[1:] return t def t_AtID(self, t): r'@[a-zA-Z0-9_-]+' t.value = t.value[1:] return t def t_ANY_error(self, t): print("{}:{}: Illegal character '{}'" .format(t.lineno, t.lexpos - self.line_begin, t.value[0])) t.lexer.skip(1) # Parser def handle_top_decs(self, smth): tdecs = {} gdecs = {} fdecs = {} for top_dec in smth: name = top_dec.name if isinstance(top_dec, FunctionDec): if name in fdecs.keys(): print('ERROR: Function {} is declared more than once' .format(name)) fdecs[name] = top_dec elif isinstance(top_dec, TypeDec): if name in tdecs.keys(): print('ERROR: Type {} is declared more than once' .format(name)) tdecs[name] = top_dec elif isinstance(top_dec, GlobalDec): if name in gdecs.keys(): print('ERROR: Global {} is declared more than once' .format(name)) gdecs[name] = top_dec else: print('Parser Error: Unknown top level declaration {}' .format(top_dec)) return tdecs, gdecs, fdecs def p_program(self, p): 'program : top_decs' tdecls, gdecls, fdecls = self.handle_top_decs(p[1]) p[0] = Program(tdecls, gdecls, fdecls) def p_topdecs_some(self, p): '''top_decs : tdec top_decs | gdec top_decs | fdec top_decs''' p[0] = [p[1]] + p[2] def p_topdecs_empty(self, p): 'top_decs : ' p[0] = [] def p_tdec(self, p): 'tdec : PercentID ASSIGN TYPE ty' p[0] = TypeDec(p[1], p[4]) def p_ty_void(self, p): 'ty : VOID' p[0] = SimpleType.Void def p_ty_i1(self, p): 'ty : I1' p[0] = SimpleType.I1 def p_ty_i8(self, p): 'ty : I8' p[0] = SimpleType.I8 def p_ty_i32(self, p): 'ty : I32' p[0] = SimpleType.I32 def p_ty_i64(self, p): 'ty : I64' p[0] = SimpleType.I64 def p_ty_ptr(self, p): 'ty : ty ASTERIX' p[0] = PointerType(p[1]) def p_ty_struct(self, p): 'ty : LBRACE ty_list RBRACE' p[0] = StructType(p[2]) def p_ty_array(self, p): 'ty : LBRACK INT ID ty RBRACK' if p[3] == 'x': p[0] = ArrayType(p[2], p[4]) else: print('Invalid name in array definition: {}\n It should have been an x.' .format(p[3])) raise SyntaxError def p_ty_fun(self, p): 'ty : ty LPAREN ty_list RPAREN' p[0] = FunctionType(p[1], p[3]) def p_ty_id(self, p): 'ty : PercentID' p[0] = NamedType(p[1]) def p_ty_list_single(self, p): 'ty_list : ty' p[0] = [p[1]] def p_ty_list_multiple(self, p): 'ty_list : ty COMMA ty_list' p[0] = [p[1]] + p[3] def p_ty_list_empty(self, p): 'ty_list : ' p[0] = [] def p_gdec(self, p): 'gdec : AtID ASSIGN GLOBAL ty ginit' p[0] = GlobalDec(p[1], p[4], p[5]) def p_ginit_null(self, p): 'ginit : NULL' p[0] = GNull() def p_ginit_id(self, p): 'ginit : AtID' p[0] = GGid(p[1]) def p_ginit_int(self, p): 'ginit : INT' p[0] = GInt(p[1]) def p_ginit_string(self, p): 'ginit : STRING' p[0] = GString(p[1]) def p_ginit_array(self, p): 'ginit : LBRACK ty_ginit_list RBRACK' # TODO This syntax seems weird p[0] = GArray(p[2]) def p_ginit_struct(self, p): 'ginit : LBRACE ty_ginit_list RBRACE' p[0] = GStruct(p[2]) def p_ty_ginit_list_single(self, p): 'ty_ginit_list : ty ginit' p[0] = [(p[1], p[2])] def p_ty_ginit_list_multiple(self, p): 'ty_ginit_list : ty ginit COMMA ty_ginit_list' p[0] = [(p[1], p[2])] + p[4] def p_ty_ginit_list_empty(self, p): 'ty_ginit_list : ' p[0] = [] def p_fdec(self, p): 'fdec : DEFINE ty AtID LPAREN ty_id_list RPAREN LBRACE fbody RBRACE' p[0] = FunctionDec(p[2], p[3], p[5], p[8]) def p_ty_id_list_single(self, p): 'ty_id_list : ty PercentID' p[0] = [(p[1], p[2])] def p_ty_id_list_multiple(self, p): 'ty_id_list : ty PercentID COMMA ty_id_list' p[0] = [(p[1], p[2])] + p[4] def p_ty_id_list_empty(self, p): 'ty_id_list : ' p[0] = [] def p_fbody_multiple_blocks(self, p): 'fbody : block named_block_list' p[0] = FunctionBody(p[1], p[2]) def p_fbody_one_block(self, p): 'fbody : block' p[0] = FunctionBody(p[1], []) def p_block_insns_terminator(self, p): 'block : insns terminator' p[0] = Block(p[1], p[2]) def p_block_terminator(self, p): 'block : terminator' p[0] = Block([], p[1]) def p_insns_single(self, p): 'insns : optionally_named_insn' p[0] = [p[1]] def p_insns_multiple(self, p): 'insns : optionally_named_insn insns' p[0] = [p[1]] + p[2] def p_optionally_named_insn_some(self, p): 'optionally_named_insn : PercentID ASSIGN insn' p[0] = (p[1], p[3]) def p_optionally_named_insn_none(self, p): 'optionally_named_insn : insn' p[0] = (None, p[1]) def p_insn_bop(self, p): 'insn : bop ty operand COMMA operand' p[0] = Binop(p[1], p[2], p[3], p[5]) def p_insn_alloca(self, p): 'insn : ALLOCA ty' p[0] = Alloca(p[2]) def p_insn_load(self, p): 'insn : LOAD ty COMMA ty operand' p[0] = Load(p[2], p[5]) def p_insn_store(self, p): 'insn : STORE ty operand COMMA ty operand' p[0] = Store(p[2], p[3], p[6]) def p_insn_icmp(self, p): 'insn : ICMP cnd ty operand COMMA operand' p[0] = Icmp(p[2], p[3], p[4], p[6]) def p_insn_call(self, p): 'insn : CALL ty operand LPAREN ty_operand_list RPAREN' p[0] = Call(p[2], p[3], p[5]) def p_insn_call_empty(self, p): 'insn : CALL ty operand LPAREN RPAREN' p[0] = Call(p[2], p[3], []) def p_insn_bitcast(self, p): 'insn : BITCAST ty operand TO ty' p[0] = Bitcast(p[2], p[3], p[5]) def p_insn_gep(self, p): 'insn : GETELEMENTPTR ty COMMA ty operand COMMA ty_operand_list' p[0] = Gep(p[2], p[4], p[5], p[7]) def p_insn_gep_empty(self, p): 'insn : GETELEMENTPTR ty COMMA ty operand' p[0] = Gep(p[2], p[4], p[5], []) def p_insn_zext(self, p): 'insn : ZEXT ty operand TO ty' p[0] = Zext(p[2], p[3], p[4]) def p_insn_ptrtoint(self, p): 'insn : PTRTOINT ty ASTERIX operand TO ty' p[0] = Ptrtoint(p[2], p[4], p[6]) def p_bop(self, p): '''bop : ADD | SUB | MUL | SHL | LSHR | ASHR | AND | OR | XOR | SDIV''' p[0] = p[1] def p_cnd(self, p): '''cnd : EQ | NE | SLT | SLE | SGT | SGE''' p[0] = p[1] def p_ty_operand_list_single(self, p): 'ty_operand_list : ty operand' p[0] = [(p[1], p[2])] def p_ty_operand_list_multiple(self, p): 'ty_operand_list : ty operand COMMA ty_operand_list' p[0] = [(p[1], p[2])] + p[4] def p_terminator_ret_void(self, p): 'terminator : RET VOID' p[0] = Ret(SimpleType.Void, None) def p_terminator_ret_oper(self, p): 'terminator : RET ty operand' p[0] = Ret(p[2], p[3]) def p_terminator_branch(self, p): 'terminator : BR LABEL PercentID' p[0] = Br(p[3]) def p_terminator_conditional_branch(self, p): 'terminator : BR ty operand COMMA LABEL PercentID COMMA LABEL PercentID' p[0] = Cbr(p[2], p[3], p[6], p[9]) def p_operand(self, p): '''operand : NULL | INT | AtID | PercentID''' # TODO: distinguish ids p[0] = p[1] def p_named_block_list_single(self, p): 'named_block_list : ID COLON block' p[0] = [(p[1], p[3])] def p_named_block_list_multiple(self, p): 'named_block_list : ID COLON block named_block_list' p[0] = [(p[1], p[3])] + p[4] def p_error(self, t): if t is None: print('Syntax error at end of file') else: print('Syntax error at token {}' .format(t)) def __init__(self): self.tokens += self.reserved.values() def build(self, **kwargs): self.lexer = lex.lex(module=self, **kwargs) self.parser = yacc.yacc(module=self, **kwargs) def test(self, data): result = self.parser.parse(data, lexer=self.lexer) print(result) if __name__ == '__main__': p = LLVMParser() p.build() data = r''' define void @tigermain (i64 %U_mainSL_8, i64 %U_mainDummy_9) { %a = add i64 3, 5 ; please be 8 ret i64 %a } ''' p.test(data)