1 //===- AsmParser.cpp - Parser for Assembly Files --------------------------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This class implements the parser for assembly files.
12 //===----------------------------------------------------------------------===//
14 #include "AsmParser.h"
15 #include "llvm/Support/SourceMgr.h"
16 #include "llvm/Support/raw_ostream.h"
19 bool AsmParser::Error(SMLoc L, const char *Msg) {
20 Lexer.PrintMessage(L, Msg);
24 bool AsmParser::TokError(const char *Msg) {
25 Lexer.PrintMessage(Lexer.getLoc(), Msg);
29 bool AsmParser::Run() {
33 while (Lexer.isNot(asmtok::Eof))
40 /// EatToEndOfStatement - Throw away the rest of the line for testing purposes.
41 void AsmParser::EatToEndOfStatement() {
42 while (Lexer.isNot(asmtok::EndOfStatement) &&
43 Lexer.isNot(asmtok::Eof))
47 if (Lexer.is(asmtok::EndOfStatement))
52 struct AsmParser::X86Operand {
65 // FIXME: Should be a general expression.
71 int64_t Disp; // FIXME: Should be a general expression.
78 static X86Operand CreateReg(unsigned RegNo) {
81 Res.Reg.RegNo = RegNo;
84 static X86Operand CreateImm(int64_t Val) {
90 static X86Operand CreateMem(unsigned SegReg, int64_t Disp, unsigned BaseReg,
91 unsigned Scale, unsigned ScaleReg) {
94 Res.Mem.SegReg = SegReg;
96 Res.Mem.BaseReg = BaseReg;
97 Res.Mem.Scale = Scale;
98 Res.Mem.ScaleReg = ScaleReg;
103 bool AsmParser::ParseX86Operand(X86Operand &Op) {
104 switch (Lexer.getKind()) {
106 return ParseX86MemOperand(Op);
107 case asmtok::Register:
108 // FIXME: Decode reg #.
109 // FIXME: if a segment register, this could either be just the seg reg, or
110 // the start of a memory operand.
111 Op = X86Operand::CreateReg(123);
112 Lexer.Lex(); // Eat register.
114 case asmtok::Dollar: {
118 if (ParseExpression(Val))
119 return TokError("expected integer constant");
120 Op = X86Operand::CreateReg(Val);
123 Lexer.Lex(); // Eat the star.
125 if (Lexer.is(asmtok::Register)) {
126 Op = X86Operand::CreateReg(123);
127 Lexer.Lex(); // Eat register.
128 } else if (ParseX86MemOperand(Op))
131 // FIXME: Note that these are 'dereferenced' so that clients know the '*' is
138 /// ParseX86MemOperand: segment: disp(basereg, indexreg, scale)
139 bool AsmParser::ParseX86MemOperand(X86Operand &Op) {
140 // FIXME: If SegReg ':' (e.g. %gs:), eat and remember.
144 // We have to disambiguate a parenthesized expression "(4+5)" from the start
145 // of a memory operand with a missing displacement "(%ebx)" or "(,%eax)". The
146 // only way to do this without lookahead is to eat the ( and see what is after
149 if (Lexer.isNot(asmtok::LParen)) {
150 if (ParseExpression(Disp)) return true;
152 // After parsing the base expression we could either have a parenthesized
153 // memory address or not. If not, return now. If so, eat the (.
154 if (Lexer.isNot(asmtok::LParen)) {
155 Op = X86Operand::CreateMem(SegReg, Disp, 0, 0, 0);
162 // Okay, we have a '('. We don't know if this is an expression or not, but
163 // so we have to eat the ( to see beyond it.
164 Lexer.Lex(); // Eat the '('.
166 if (Lexer.is(asmtok::Register) || Lexer.is(asmtok::Comma)) {
167 // Nothing to do here, fall into the code below with the '(' part of the
168 // memory operand consumed.
170 // It must be an parenthesized expression, parse it now.
171 if (ParseParenExpr(Disp)) return true;
173 // After parsing the base expression we could either have a parenthesized
174 // memory address or not. If not, return now. If so, eat the (.
175 if (Lexer.isNot(asmtok::LParen)) {
176 Op = X86Operand::CreateMem(SegReg, Disp, 0, 0, 0);
185 // If we reached here, then we just ate the ( of the memory operand. Process
186 // the rest of the memory operand.
187 unsigned BaseReg = 0, ScaleReg = 0, Scale = 0;
189 if (Lexer.is(asmtok::Register)) {
190 BaseReg = 123; // FIXME: decode reg #
191 Lexer.Lex(); // eat the register.
194 if (Lexer.is(asmtok::Comma)) {
195 Lexer.Lex(); // eat the comma.
197 if (Lexer.is(asmtok::Register)) {
198 ScaleReg = 123; // FIXME: decode reg #
199 Lexer.Lex(); // eat the register.
200 Scale = 1; // If not specified, the scale defaults to 1.
203 if (Lexer.is(asmtok::Comma)) {
204 Lexer.Lex(); // eat the comma.
206 // If present, get and validate scale amount.
207 if (Lexer.is(asmtok::IntVal)) {
208 int64_t ScaleVal = Lexer.getCurIntVal();
209 if (ScaleVal != 1 && ScaleVal != 2 && ScaleVal != 4 && ScaleVal != 8)
210 return TokError("scale factor in address must be 1, 2, 4 or 8");
211 Lexer.Lex(); // eat the scale.
212 Scale = (unsigned)ScaleVal;
217 // Ok, we've eaten the memory operand, verify we have a ')' and eat it too.
218 if (Lexer.isNot(asmtok::RParen))
219 return TokError("unexpected token in memory operand");
220 Lexer.Lex(); // Eat the ')'.
222 Op = X86Operand::CreateMem(SegReg, Disp, BaseReg, Scale, ScaleReg);
226 /// ParseParenExpr - Parse a paren expression and return it.
227 /// NOTE: This assumes the leading '(' has already been consumed.
229 /// parenexpr ::= expr)
231 bool AsmParser::ParseParenExpr(int64_t &Res) {
232 if (ParseExpression(Res)) return true;
233 if (Lexer.isNot(asmtok::RParen))
234 return TokError("expected ')' in parentheses expression");
239 /// ParsePrimaryExpr - Parse a primary expression and return it.
240 /// primaryexpr ::= (parenexpr
241 /// primaryexpr ::= symbol
242 /// primaryexpr ::= number
243 /// primaryexpr ::= ~,+,- primaryexpr
244 bool AsmParser::ParsePrimaryExpr(int64_t &Res) {
245 switch (Lexer.getKind()) {
247 return TokError("unknown token in expression");
248 case asmtok::Identifier:
249 // This is a label, this should be parsed as part of an expression, to
250 // handle things like LFOO+4
252 Lexer.Lex(); // Eat identifier.
255 Res = Lexer.getCurIntVal();
256 Lexer.Lex(); // Eat identifier.
259 Lexer.Lex(); // Eat the '('.
260 return ParseParenExpr(Res);
264 Lexer.Lex(); // Eat the operator.
265 return ParsePrimaryExpr(Res);
269 /// ParseExpression - Parse an expression and return it.
271 /// expr ::= expr +,- expr -> lowest.
272 /// expr ::= expr |,^,&,! expr -> middle.
273 /// expr ::= expr *,/,%,<<,>> expr -> highest.
274 /// expr ::= primaryexpr
276 bool AsmParser::ParseExpression(int64_t &Res) {
277 return ParsePrimaryExpr(Res);
284 /// ::= EndOfStatement
285 /// ::= Label* Directive ...Operands... EndOfStatement
286 /// ::= Label* Identifier OperandList* EndOfStatement
287 bool AsmParser::ParseStatement() {
288 switch (Lexer.getKind()) {
290 return TokError("unexpected token at start of statement");
291 case asmtok::EndOfStatement:
294 case asmtok::Identifier:
296 // TODO: Recurse on local labels etc.
299 // If we have an identifier, handle it as the key symbol.
300 SMLoc IDLoc = Lexer.getLoc();
301 std::string IDVal = Lexer.getCurStrVal();
303 // Consume the identifier, see what is after it.
304 if (Lexer.Lex() == asmtok::Colon) {
305 // identifier ':' -> Label.
307 return ParseStatement();
310 // Otherwise, we have a normal instruction or directive.
311 if (IDVal[0] == '.') {
312 Lexer.PrintMessage(IDLoc, "warning: ignoring directive for now");
313 EatToEndOfStatement();
317 // If it's an instruction, parse an operand list.
318 std::vector<X86Operand> Operands;
320 // Read the first operand, if present. Note that we require a newline at the
321 // end of file, so we don't have to worry about Eof here.
322 if (Lexer.isNot(asmtok::EndOfStatement)) {
324 if (ParseX86Operand(Op))
326 Operands.push_back(Op);
329 while (Lexer.is(asmtok::Comma)) {
330 Lexer.Lex(); // Eat the comma.
332 // Parse and remember the operand.
334 if (ParseX86Operand(Op))
336 Operands.push_back(Op);
339 if (Lexer.isNot(asmtok::EndOfStatement))
340 return TokError("unexpected token in operand list");
342 // Eat the end of statement marker.
345 // Instruction is good, process it.
346 outs() << "Found instruction: " << IDVal << " with " << Operands.size()
349 // Skip to end of line for now.