1 //===-EDDisassembler.h - LLVM Enhanced Disassembler -------------*- C++ -*-===//
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 file defines the interface for the Enhanced Disassembly library's
11 // disassembler class. The disassembler is responsible for vending individual
12 // instructions according to a given architecture and disassembly syntax.
14 //===----------------------------------------------------------------------===//
16 #ifndef EDDisassembler_
17 #define EDDisassembler_
21 #include "llvm-c/EnhancedDisassembly.h"
23 #include "llvm/ADT/OwningPtr.h"
24 #include "llvm/ADT/Triple.h"
25 #include "llvm/Support/raw_ostream.h"
26 #include "llvm/System/Mutex.h"
41 class TargetAsmParser;
45 class MCParsedAsmOperand;
47 template <typename T> class SmallVectorImpl;
50 class TargetRegisterInfo;
55 /// EDDisassembler - Encapsulates a disassembler for a single architecture and
56 /// disassembly syntax. Also manages the static disassembler registry.
57 struct EDDisassembler {
62 /// CPUKey - Encapsulates the descriptor of an architecture/disassembly-syntax
65 /// The architecture type
66 llvm::Triple::ArchType Arch;
68 /// The assembly syntax
69 EDAssemblySyntax_t Syntax;
71 /// operator== - Equality operator
72 bool operator==(const CPUKey &key) const {
73 return (Arch == key.Arch &&
74 Syntax == key.Syntax);
77 /// operator< - Less-than operator
78 bool operator<(const CPUKey &key) const {
81 if(Syntax >= key.Syntax)
87 typedef std::map<CPUKey, EDDisassembler*> DisassemblerMap_t;
89 /// True if the disassembler registry has been initialized; false if not
90 static bool sInitialized;
91 /// A map from disassembler specifications to disassemblers. Populated
93 static DisassemblerMap_t sDisassemblers;
95 /// getDisassembler - Returns the specified disassemble, or NULL on failure
97 /// @arg arch - The desired architecture
98 /// @arg syntax - The desired disassembly syntax
99 static EDDisassembler *getDisassembler(llvm::Triple::ArchType arch,
100 EDAssemblySyntax_t syntax);
102 /// getDisassembler - Returns the disassembler for a given combination of
103 /// CPU type, CPU subtype, and assembly syntax, or NULL on failure
105 /// @arg str - The string representation of the architecture triple, e.g.,
106 /// "x86_64-apple-darwin"
107 /// @arg syntax - The disassembly syntax for the required disassembler
108 static EDDisassembler *getDisassembler(llvm::StringRef str,
109 EDAssemblySyntax_t syntax);
111 /// initialize - Initializes the disassembler registry and the LLVM backend
112 static void initialize();
114 ////////////////////////
115 // Per-object members //
116 ////////////////////////
118 /// True only if the object has been successfully initialized
120 /// True if the disassembler can provide semantic information
123 /// The stream to write errors to
124 llvm::raw_ostream &ErrorStream;
126 /// The architecture/syntax pair for the current architecture
128 /// The LLVM target corresponding to the disassembler
129 const llvm::Target *Tgt;
130 /// The assembly information for the target architecture
131 llvm::OwningPtr<const llvm::MCAsmInfo> AsmInfo;
132 /// The disassembler for the target architecture
133 llvm::OwningPtr<const llvm::MCDisassembler> Disassembler;
134 /// The output string for the instruction printer; must be guarded with
136 llvm::OwningPtr<std::string> InstString;
137 /// The output stream for the disassembler; must be guarded with
139 llvm::OwningPtr<llvm::raw_string_ostream> InstStream;
140 /// The instruction printer for the target architecture; must be guarded with
141 /// PrinterMutex when printing
142 llvm::OwningPtr<llvm::MCInstPrinter> InstPrinter;
143 /// The mutex that guards the instruction printer's printing functions, which
144 /// use a shared stream
145 llvm::sys::Mutex PrinterMutex;
146 /// The array of instruction information provided by the TableGen backend for
147 /// the target architecture
148 const llvm::EDInstInfo *InstInfos;
149 /// The target-specific lexer for use in tokenizing strings, in
150 /// target-independent and target-specific portions
151 llvm::OwningPtr<llvm::AsmLexer> GenericAsmLexer;
152 llvm::OwningPtr<llvm::TargetAsmLexer> SpecificAsmLexer;
153 /// The guard for the above
154 llvm::sys::Mutex ParserMutex;
155 /// The LLVM number used for the target disassembly syntax variant
156 int LLVMSyntaxVariant;
158 typedef std::vector<std::string> regvec_t;
159 typedef std::map<std::string, unsigned> regrmap_t;
161 /// A vector of registers for quick mapping from LLVM register IDs to names
163 /// A map of registers for quick mapping from register names to LLVM IDs
166 /// A set of register IDs for aliases of the stack pointer for the current
168 std::set<unsigned> stackPointers;
169 /// A set of register IDs for aliases of the program counter for the current
171 std::set<unsigned> programCounters;
173 /// Constructor - initializes a disassembler with all the necessary objects,
174 /// which come pre-allocated from the registry accessor function
176 /// @arg key - the architecture and disassembly syntax for the
178 EDDisassembler(CPUKey& key);
180 /// valid - reports whether there was a failure in the constructor.
185 /// hasSemantics - reports whether the disassembler can provide operands and
187 bool hasSemantics() {
193 /// createInst - creates and returns an instruction given a callback and
194 /// memory address, or NULL on failure
196 /// @arg byteReader - A callback function that provides machine code bytes
197 /// @arg address - The address of the first byte of the instruction,
198 /// suitable for passing to byteReader
199 /// @arg arg - An opaque argument for byteReader
200 EDInst *createInst(EDByteReaderCallback byteReader,
204 /// initMaps - initializes regVec and regRMap using the provided register
207 /// @arg registerInfo - the register information to use as a source
208 void initMaps(const llvm::TargetRegisterInfo ®isterInfo);
209 /// nameWithRegisterID - Returns the name (owned by the EDDisassembler) of a
210 /// register for a given register ID, or NULL on failure
212 /// @arg registerID - the ID of the register to be queried
213 const char *nameWithRegisterID(unsigned registerID) const;
214 /// registerIDWithName - Returns the ID of a register for a given register
215 /// name, or (unsigned)-1 on failure
217 /// @arg name - The name of the register
218 unsigned registerIDWithName(const char *name) const;
220 /// registerIsStackPointer - reports whether a register ID is an alias for the
221 /// stack pointer register
223 /// @arg registerID - The LLVM register ID
224 bool registerIsStackPointer(unsigned registerID);
225 /// registerIsStackPointer - reports whether a register ID is an alias for the
226 /// stack pointer register
228 /// @arg registerID - The LLVM register ID
229 bool registerIsProgramCounter(unsigned registerID);
231 /// printInst - prints an MCInst to a string, returning 0 on success, or -1
234 /// @arg str - A reference to a string which is filled in with the string
235 /// representation of the instruction
236 /// @arg inst - A reference to the MCInst to be printed
237 int printInst(std::string& str,
240 /// parseInst - extracts operands and tokens from a string for use in
241 /// tokenizing the string. Returns 0 on success, or -1 otherwise.
243 /// @arg operands - A reference to a vector that will be filled in with the
245 /// @arg tokens - A reference to a vector that will be filled in with the
247 /// @arg str - The string representation of the instruction
248 int parseInst(llvm::SmallVectorImpl<llvm::MCParsedAsmOperand*> &operands,
249 llvm::SmallVectorImpl<llvm::AsmToken> &tokens,
250 const std::string &str);
252 /// llvmSyntaxVariant - returns the LLVM syntax variant for this disassembler
253 int llvmSyntaxVariant() const;