Chris Lattner | 27aa7d2 | 2009-06-21 20:16:42 +0000 | [diff] [blame] | 1 | //===- AsmParser.cpp - Parser for Assembly Files --------------------------===// |
| 2 | // |
| 3 | // The LLVM Compiler Infrastructure |
| 4 | // |
| 5 | // This file is distributed under the University of Illinois Open Source |
| 6 | // License. See LICENSE.TXT for details. |
| 7 | // |
| 8 | //===----------------------------------------------------------------------===// |
| 9 | // |
| 10 | // This class implements the parser for assembly files. |
| 11 | // |
| 12 | //===----------------------------------------------------------------------===// |
| 13 | |
| 14 | #include "AsmParser.h" |
Chris Lattner | b0789ed | 2009-06-21 20:54:55 +0000 | [diff] [blame] | 15 | #include "llvm/Support/SourceMgr.h" |
| 16 | #include "llvm/Support/raw_ostream.h" |
Chris Lattner | 27aa7d2 | 2009-06-21 20:16:42 +0000 | [diff] [blame] | 17 | using namespace llvm; |
| 18 | |
Chris Lattner | 14ee48a | 2009-06-21 21:22:11 +0000 | [diff] [blame] | 19 | bool AsmParser::Error(SMLoc L, const char *Msg) { |
| 20 | Lexer.PrintMessage(L, Msg); |
| 21 | return true; |
| 22 | } |
| 23 | |
| 24 | bool AsmParser::TokError(const char *Msg) { |
| 25 | Lexer.PrintMessage(Lexer.getLoc(), Msg); |
| 26 | return true; |
| 27 | } |
| 28 | |
Chris Lattner | 27aa7d2 | 2009-06-21 20:16:42 +0000 | [diff] [blame] | 29 | bool AsmParser::Run() { |
Chris Lattner | b0789ed | 2009-06-21 20:54:55 +0000 | [diff] [blame] | 30 | // Prime the lexer. |
| 31 | Lexer.Lex(); |
| 32 | |
| 33 | while (Lexer.isNot(asmtok::Eof)) |
| 34 | if (ParseStatement()) |
| 35 | return true; |
| 36 | |
| 37 | return false; |
| 38 | } |
| 39 | |
Chris Lattner | 2cf5f14 | 2009-06-22 01:29:09 +0000 | [diff] [blame^] | 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)) |
| 44 | Lexer.Lex(); |
| 45 | |
| 46 | // Eat EOL. |
| 47 | if (Lexer.is(asmtok::EndOfStatement)) |
| 48 | Lexer.Lex(); |
| 49 | } |
| 50 | |
| 51 | struct AsmParser::X86Operand { |
| 52 | enum { |
| 53 | Register, |
| 54 | Immediate |
| 55 | } Kind; |
| 56 | |
| 57 | union { |
| 58 | struct { |
| 59 | unsigned RegNo; |
| 60 | } Reg; |
| 61 | |
| 62 | struct { |
| 63 | // FIXME: Should be a general expression. |
| 64 | int64_t Val; |
| 65 | } Imm; |
| 66 | }; |
| 67 | |
| 68 | static X86Operand CreateReg(unsigned RegNo) { |
| 69 | X86Operand Res; |
| 70 | Res.Kind = Register; |
| 71 | Res.Reg.RegNo = RegNo; |
| 72 | return Res; |
| 73 | } |
| 74 | static X86Operand CreateImm(int64_t Val) { |
| 75 | X86Operand Res; |
| 76 | Res.Kind = Immediate; |
| 77 | Res.Imm.Val = Val; |
| 78 | return Res; |
| 79 | } |
| 80 | }; |
| 81 | |
| 82 | bool AsmParser::ParseX86Operand(X86Operand &Op) { |
| 83 | switch (Lexer.getKind()) { |
| 84 | default: |
| 85 | return TokError("unknown token at start of instruction operand"); |
| 86 | case asmtok::Register: |
| 87 | // FIXME: Decode reg #. |
| 88 | Op = X86Operand::CreateReg(0); |
| 89 | Lexer.Lex(); // Eat register. |
| 90 | return false; |
| 91 | case asmtok::Dollar: |
| 92 | // $42 -> immediate. |
| 93 | Lexer.Lex(); |
| 94 | // FIXME: Parse an arbitrary expression here, like $(4+5) |
| 95 | if (Lexer.isNot(asmtok::IntVal)) |
| 96 | return TokError("expected integer constant"); |
| 97 | |
| 98 | Op = X86Operand::CreateReg(Lexer.getCurIntVal()); |
| 99 | Lexer.Lex(); // Eat register. |
| 100 | return false; |
| 101 | case asmtok::Identifier: |
| 102 | // This is a label, this should be parsed as part of an expression, to |
| 103 | // handle things like LFOO+4 |
| 104 | Op = X86Operand::CreateImm(0); // FIXME. |
| 105 | Lexer.Lex(); // Eat identifier. |
| 106 | return false; |
| 107 | |
| 108 | //case asmtok::Star: |
| 109 | // * %eax |
| 110 | // * <memaddress> |
| 111 | // Note that these are both "dereferenced". |
| 112 | } |
| 113 | } |
| 114 | |
Chris Lattner | b0789ed | 2009-06-21 20:54:55 +0000 | [diff] [blame] | 115 | |
| 116 | /// ParseStatement: |
| 117 | /// ::= EndOfStatement |
Chris Lattner | 2cf5f14 | 2009-06-22 01:29:09 +0000 | [diff] [blame^] | 118 | /// ::= Label* Directive ...Operands... EndOfStatement |
| 119 | /// ::= Label* Identifier OperandList* EndOfStatement |
Chris Lattner | b0789ed | 2009-06-21 20:54:55 +0000 | [diff] [blame] | 120 | bool AsmParser::ParseStatement() { |
| 121 | switch (Lexer.getKind()) { |
| 122 | default: |
Chris Lattner | 14ee48a | 2009-06-21 21:22:11 +0000 | [diff] [blame] | 123 | return TokError("unexpected token at start of statement"); |
Chris Lattner | b0789ed | 2009-06-21 20:54:55 +0000 | [diff] [blame] | 124 | case asmtok::EndOfStatement: |
| 125 | Lexer.Lex(); |
| 126 | return false; |
| 127 | case asmtok::Identifier: |
| 128 | break; |
| 129 | // TODO: Recurse on local labels etc. |
| 130 | } |
| 131 | |
| 132 | // If we have an identifier, handle it as the key symbol. |
Chris Lattner | 2cf5f14 | 2009-06-22 01:29:09 +0000 | [diff] [blame^] | 133 | SMLoc IDLoc = Lexer.getLoc(); |
Chris Lattner | b0789ed | 2009-06-21 20:54:55 +0000 | [diff] [blame] | 134 | std::string IDVal = Lexer.getCurStrVal(); |
| 135 | |
| 136 | // Consume the identifier, see what is after it. |
| 137 | if (Lexer.Lex() == asmtok::Colon) { |
| 138 | // identifier ':' -> Label. |
| 139 | Lexer.Lex(); |
| 140 | return ParseStatement(); |
| 141 | } |
| 142 | |
| 143 | // Otherwise, we have a normal instruction or directive. |
Chris Lattner | 2cf5f14 | 2009-06-22 01:29:09 +0000 | [diff] [blame^] | 144 | if (IDVal[0] == '.') { |
| 145 | Lexer.PrintMessage(IDLoc, "warning: ignoring directive for now"); |
| 146 | EatToEndOfStatement(); |
| 147 | return false; |
| 148 | } |
Chris Lattner | b0789ed | 2009-06-21 20:54:55 +0000 | [diff] [blame] | 149 | |
Chris Lattner | 2cf5f14 | 2009-06-22 01:29:09 +0000 | [diff] [blame^] | 150 | // If it's an instruction, parse an operand list. |
| 151 | std::vector<X86Operand> Operands; |
Chris Lattner | b0789ed | 2009-06-21 20:54:55 +0000 | [diff] [blame] | 152 | |
Chris Lattner | 2cf5f14 | 2009-06-22 01:29:09 +0000 | [diff] [blame^] | 153 | // Read the first operand, if present. Note that we require a newline at the |
| 154 | // end of file, so we don't have to worry about Eof here. |
| 155 | if (Lexer.isNot(asmtok::EndOfStatement)) { |
| 156 | X86Operand Op; |
| 157 | if (ParseX86Operand(Op)) |
| 158 | return true; |
| 159 | Operands.push_back(Op); |
| 160 | } |
| 161 | |
| 162 | while (Lexer.is(asmtok::Comma)) { |
| 163 | Lexer.Lex(); // Eat the comma. |
| 164 | |
| 165 | // Parse and remember the operand. |
| 166 | X86Operand Op; |
| 167 | if (ParseX86Operand(Op)) |
| 168 | return true; |
| 169 | Operands.push_back(Op); |
| 170 | } |
| 171 | |
| 172 | if (Lexer.isNot(asmtok::EndOfStatement)) |
| 173 | return TokError("unexpected token in operand list"); |
| 174 | |
| 175 | // Eat the end of statement marker. |
| 176 | Lexer.Lex(); |
| 177 | |
| 178 | // Instruction is good, process it. |
| 179 | outs() << "Found instruction: " << IDVal << " with " << Operands.size() |
| 180 | << " operands.\n"; |
| 181 | |
| 182 | // Skip to end of line for now. |
Chris Lattner | 27aa7d2 | 2009-06-21 20:16:42 +0000 | [diff] [blame] | 183 | return false; |
| 184 | } |