// Copyright (c) 1994-2006 Sun Microsystems Inc. // All Rights Reserved. // // Redistribution and use in source and binary forms, with or without // modification, are permitted provided that the following conditions are // met: // // - Redistributions of source code must retain the above copyright notice, // this list of conditions and the following disclaimer. // // - Redistribution in binary form must reproduce the above copyright // notice, this list of conditions and the following disclaimer in the // documentation and/or other materials provided with the distribution. // // - Neither the name of Sun Microsystems or the names of contributors may // be used to endorse or promote products derived from this software without // specific prior written permission. // // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS // IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, // THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR // PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR // CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, // EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, // PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR // PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF // LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING // NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS // SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. // The original source code covered by the above license above has been // modified significantly by Google Inc. // Copyright 2006-2008 the V8 project authors. All rights reserved. // A light-weight IA32 Assembler. #ifndef V8_IA32_ASSEMBLER_IA32_H_ #define V8_IA32_ASSEMBLER_IA32_H_ namespace v8 { namespace internal { // CPU Registers. // // 1) We would prefer to use an enum, but enum values are assignment- // compatible with int, which has caused code-generation bugs. // // 2) We would prefer to use a class instead of a struct but we don't like // the register initialization to depend on the particular initialization // order (which appears to be different on OS X, Linux, and Windows for the // installed versions of C++ we tried). Using a struct permits C-style // "initialization". Also, the Register objects cannot be const as this // forces initialization stubs in MSVC, making us dependent on initialization // order. // // 3) By not using an enum, we are possibly preventing the compiler from // doing certain constant folds, which may significantly reduce the // code generated for some assembly instructions (because they boil down // to a few constants). If this is a problem, we could change the code // such that we use an enum in optimized mode, and the struct in debug // mode. This way we get the compile-time error checking in debug mode // and best performance in optimized code. // struct Register { bool is_valid() const { return 0 <= code_ && code_ < 8; } bool is(Register reg) const { return code_ == reg.code_; } // eax, ebx, ecx and edx are byte registers, the rest are not. bool is_byte_register() const { return code_ <= 3; } int code() const { ASSERT(is_valid()); return code_; } int bit() const { ASSERT(is_valid()); return 1 << code_; } // (unfortunately we can't make this private in a struct) int code_; }; const Register eax = { 0 }; const Register ecx = { 1 }; const Register edx = { 2 }; const Register ebx = { 3 }; const Register esp = { 4 }; const Register ebp = { 5 }; const Register esi = { 6 }; const Register edi = { 7 }; const Register no_reg = { -1 }; struct XMMRegister { bool is_valid() const { return 0 <= code_ && code_ < 2; } // currently int code() const { ASSERT(is_valid()); return code_; } int code_; }; const XMMRegister xmm0 = { 0 }; const XMMRegister xmm1 = { 1 }; const XMMRegister xmm2 = { 2 }; const XMMRegister xmm3 = { 3 }; const XMMRegister xmm4 = { 4 }; const XMMRegister xmm5 = { 5 }; const XMMRegister xmm6 = { 6 }; const XMMRegister xmm7 = { 7 }; enum Condition { // any value < 0 is considered no_condition no_condition = -1, overflow = 0, no_overflow = 1, below = 2, above_equal = 3, equal = 4, not_equal = 5, below_equal = 6, above = 7, negative = 8, positive = 9, parity_even = 10, parity_odd = 11, less = 12, greater_equal = 13, less_equal = 14, greater = 15, // aliases carry = below, not_carry = above_equal, zero = equal, not_zero = not_equal, sign = negative, not_sign = positive }; // Returns the equivalent of !cc. // Negation of the default no_condition (-1) results in a non-default // no_condition value (-2). As long as tests for no_condition check // for condition < 0, this will work as expected. inline Condition NegateCondition(Condition cc); // Corresponds to transposing the operands of a comparison. inline Condition ReverseCondition(Condition cc) { switch (cc) { case below: return above; case above: return below; case above_equal: return below_equal; case below_equal: return above_equal; case less: return greater; case greater: return less; case greater_equal: return less_equal; case less_equal: return greater_equal; default: return cc; }; } enum Hint { no_hint = 0, not_taken = 0x2e, taken = 0x3e }; // The result of negating a hint is as if the corresponding condition // were negated by NegateCondition. That is, no_hint is mapped to // itself and not_taken and taken are mapped to each other. inline Hint NegateHint(Hint hint) { return (hint == no_hint) ? no_hint : ((hint == not_taken) ? taken : not_taken); } // ----------------------------------------------------------------------------- // Machine instruction Immediates class Immediate BASE_EMBEDDED { public: inline explicit Immediate(int x); inline explicit Immediate(const char* s); inline explicit Immediate(const ExternalReference& ext); inline explicit Immediate(Handle handle); inline explicit Immediate(Smi* value); static Immediate CodeRelativeOffset(Label* label) { return Immediate(label); } bool is_zero() const { return x_ == 0 && rmode_ == RelocInfo::NONE; } bool is_int8() const { return -128 <= x_ && x_ < 128 && rmode_ == RelocInfo::NONE; } bool is_int16() const { return -32768 <= x_ && x_ < 32768 && rmode_ == RelocInfo::NONE; } private: inline explicit Immediate(Label* value); int x_; RelocInfo::Mode rmode_; friend class Assembler; }; // ----------------------------------------------------------------------------- // Machine instruction Operands enum ScaleFactor { times_1 = 0, times_2 = 1, times_4 = 2, times_8 = 3, times_pointer_size = times_4, times_half_pointer_size = times_2 }; class Operand BASE_EMBEDDED { public: // reg INLINE(explicit Operand(Register reg)); // [disp/r] INLINE(explicit Operand(int32_t disp, RelocInfo::Mode rmode)); // disp only must always be relocated // [base + disp/r] explicit Operand(Register base, int32_t disp, RelocInfo::Mode rmode = RelocInfo::NONE); // [base + index*scale + disp/r] explicit Operand(Register base, Register index, ScaleFactor scale, int32_t disp, RelocInfo::Mode rmode = RelocInfo::NONE); // [index*scale + disp/r] explicit Operand(Register index, ScaleFactor scale, int32_t disp, RelocInfo::Mode rmode = RelocInfo::NONE); static Operand StaticVariable(const ExternalReference& ext) { return Operand(reinterpret_cast(ext.address()), RelocInfo::EXTERNAL_REFERENCE); } static Operand StaticArray(Register index, ScaleFactor scale, const ExternalReference& arr) { return Operand(index, scale, reinterpret_cast(arr.address()), RelocInfo::EXTERNAL_REFERENCE); } // Returns true if this Operand is a wrapper for the specified register. bool is_reg(Register reg) const; private: byte buf_[6]; // The number of bytes in buf_. unsigned int len_; // Only valid if len_ > 4. RelocInfo::Mode rmode_; // Set the ModRM byte without an encoded 'reg' register. The // register is encoded later as part of the emit_operand operation. inline void set_modrm(int mod, Register rm); inline void set_sib(ScaleFactor scale, Register index, Register base); inline void set_disp8(int8_t disp); inline void set_dispr(int32_t disp, RelocInfo::Mode rmode); friend class Assembler; }; // ----------------------------------------------------------------------------- // A Displacement describes the 32bit immediate field of an instruction which // may be used together with a Label in order to refer to a yet unknown code // position. Displacements stored in the instruction stream are used to describe // the instruction and to chain a list of instructions using the same Label. // A Displacement contains 2 different fields: // // next field: position of next displacement in the chain (0 = end of list) // type field: instruction type // // A next value of null (0) indicates the end of a chain (note that there can // be no displacement at position zero, because there is always at least one // instruction byte before the displacement). // // Displacement _data field layout // // |31.....2|1......0| // [ next | type | class Displacement BASE_EMBEDDED { public: enum Type { UNCONDITIONAL_JUMP, CODE_RELATIVE, OTHER }; int data() const { return data_; } Type type() const { return TypeField::decode(data_); } void next(Label* L) const { int n = NextField::decode(data_); n > 0 ? L->link_to(n) : L->Unuse(); } void link_to(Label* L) { init(L, type()); } explicit Displacement(int data) { data_ = data; } Displacement(Label* L, Type type) { init(L, type); } void print() { PrintF("%s (%x) ", (type() == UNCONDITIONAL_JUMP ? "jmp" : "[other]"), NextField::decode(data_)); } private: int data_; class TypeField: public BitField {}; class NextField: public BitField {}; void init(Label* L, Type type); }; // CpuFeatures keeps track of which features are supported by the target CPU. // Supported features must be enabled by a Scope before use. // Example: // if (CpuFeatures::IsSupported(SSE2)) { // CpuFeatures::Scope fscope(SSE2); // // Generate SSE2 floating point code. // } else { // // Generate standard x87 floating point code. // } class CpuFeatures : public AllStatic { public: // Feature flags bit positions. They are mostly based on the CPUID spec. // (We assign CPUID itself to one of the currently reserved bits -- // feel free to change this if needed.) enum Feature { SSE3 = 32, SSE2 = 26, CMOV = 15, RDTSC = 4, CPUID = 10 }; // Detect features of the target CPU. Set safe defaults if the serializer // is enabled (snapshots must be portable). static void Probe(); // Check whether a feature is supported by the target CPU. static bool IsSupported(Feature f) { if (f == SSE2 && !FLAG_enable_sse2) return false; if (f == SSE3 && !FLAG_enable_sse3) return false; if (f == CMOV && !FLAG_enable_cmov) return false; if (f == RDTSC && !FLAG_enable_rdtsc) return false; return (supported_ & (static_cast(1) << f)) != 0; } // Check whether a feature is currently enabled. static bool IsEnabled(Feature f) { return (enabled_ & (static_cast(1) << f)) != 0; } // Enable a specified feature within a scope. class Scope BASE_EMBEDDED { #ifdef DEBUG public: explicit Scope(Feature f) { ASSERT(CpuFeatures::IsSupported(f)); old_enabled_ = CpuFeatures::enabled_; CpuFeatures::enabled_ |= (static_cast(1) << f); } ~Scope() { CpuFeatures::enabled_ = old_enabled_; } private: uint64_t old_enabled_; #else public: explicit Scope(Feature f) {} #endif }; private: static uint64_t supported_; static uint64_t enabled_; }; class Assembler : public Malloced { private: // We check before assembling an instruction that there is sufficient // space to write an instruction and its relocation information. // The relocation writer's position must be kGap bytes above the end of // the generated instructions. This leaves enough space for the // longest possible ia32 instruction, 15 bytes, and the longest possible // relocation information encoding, RelocInfoWriter::kMaxLength == 16. // (There is a 15 byte limit on ia32 instruction length that rules out some // otherwise valid instructions.) // This allows for a single, fast space check per instruction. static const int kGap = 32; public: // Create an assembler. Instructions and relocation information are emitted // into a buffer, with the instructions starting from the beginning and the // relocation information starting from the end of the buffer. See CodeDesc // for a detailed comment on the layout (globals.h). // // If the provided buffer is NULL, the assembler allocates and grows its own // buffer, and buffer_size determines the initial buffer size. The buffer is // owned by the assembler and deallocated upon destruction of the assembler. // // If the provided buffer is not NULL, the assembler uses the provided buffer // for code generation and assumes its size to be buffer_size. If the buffer // is too small, a fatal error occurs. No deallocation of the buffer is done // upon destruction of the assembler. Assembler(void* buffer, int buffer_size); ~Assembler(); // GetCode emits any pending (non-emitted) code and fills the descriptor // desc. GetCode() is idempotent; it returns the same result if no other // Assembler functions are invoked in between GetCode() calls. void GetCode(CodeDesc* desc); // Read/Modify the code target in the branch/call instruction at pc. inline static Address target_address_at(Address pc); inline static void set_target_address_at(Address pc, Address target); // Distance between the address of the code target in the call instruction // and the return address static const int kCallTargetAddressOffset = kPointerSize; // Distance between start of patched return sequence and the emitted address // to jump to. static const int kPatchReturnSequenceAddressOffset = 1; // JMP imm32. // --------------------------------------------------------------------------- // Code generation // // - function names correspond one-to-one to ia32 instruction mnemonics // - unless specified otherwise, instructions operate on 32bit operands // - instructions on 8bit (byte) operands/registers have a trailing '_b' // - instructions on 16bit (word) operands/registers have a trailing '_w' // - naming conflicts with C++ keywords are resolved via a trailing '_' // NOTE ON INTERFACE: Currently, the interface is not very consistent // in the sense that some operations (e.g. mov()) can be called in more // the one way to generate the same instruction: The Register argument // can in some cases be replaced with an Operand(Register) argument. // This should be cleaned up and made more orthogonal. The questions // is: should we always use Operands instead of Registers where an // Operand is possible, or should we have a Register (overloaded) form // instead? We must be careful to make sure that the selected instruction // is obvious from the parameters to avoid hard-to-find code generation // bugs. // Insert the smallest number of nop instructions // possible to align the pc offset to a multiple // of m. m must be a power of 2. void Align(int m); // Stack void pushad(); void popad(); void pushfd(); void popfd(); void push(const Immediate& x); void push(Register src); void push(const Operand& src); void push(Label* label, RelocInfo::Mode relocation_mode); void pop(Register dst); void pop(const Operand& dst); void enter(const Immediate& size); void leave(); // Moves void mov_b(Register dst, const Operand& src); void mov_b(const Operand& dst, int8_t imm8); void mov_b(const Operand& dst, Register src); void mov_w(Register dst, const Operand& src); void mov_w(const Operand& dst, Register src); void mov(Register dst, int32_t imm32); void mov(Register dst, const Immediate& x); void mov(Register dst, Handle handle); void mov(Register dst, const Operand& src); void mov(Register dst, Register src); void mov(const Operand& dst, const Immediate& x); void mov(const Operand& dst, Handle handle); void mov(const Operand& dst, Register src); void movsx_b(Register dst, const Operand& src); void movsx_w(Register dst, const Operand& src); void movzx_b(Register dst, const Operand& src); void movzx_w(Register dst, const Operand& src); // Conditional moves void cmov(Condition cc, Register dst, int32_t imm32); void cmov(Condition cc, Register dst, Handle handle); void cmov(Condition cc, Register dst, const Operand& src); // Exchange two registers void xchg(Register dst, Register src); // Arithmetics void adc(Register dst, int32_t imm32); void adc(Register dst, const Operand& src); void add(Register dst, const Operand& src); void add(const Operand& dst, const Immediate& x); void and_(Register dst, int32_t imm32); void and_(Register dst, const Operand& src); void and_(const Operand& src, Register dst); void and_(const Operand& dst, const Immediate& x); void cmpb(const Operand& op, int8_t imm8); void cmpb_al(const Operand& op); void cmpw_ax(const Operand& op); void cmpw(const Operand& op, Immediate imm16); void cmp(Register reg, int32_t imm32); void cmp(Register reg, Handle handle); void cmp(Register reg, const Operand& op); void cmp(const Operand& op, const Immediate& imm); void cmp(const Operand& op, Handle handle); void dec_b(Register dst); void dec(Register dst); void dec(const Operand& dst); void cdq(); void idiv(Register src); // Signed multiply instructions. void imul(Register src); // edx:eax = eax * src. void imul(Register dst, const Operand& src); // dst = dst * src. void imul(Register dst, Register src, int32_t imm32); // dst = src * imm32. void inc(Register dst); void inc(const Operand& dst); void lea(Register dst, const Operand& src); // Unsigned multiply instruction. void mul(Register src); // edx:eax = eax * reg. void neg(Register dst); void not_(Register dst); void or_(Register dst, int32_t imm32); void or_(Register dst, const Operand& src); void or_(const Operand& dst, Register src); void or_(const Operand& dst, const Immediate& x); void rcl(Register dst, uint8_t imm8); void sar(Register dst, uint8_t imm8); void sar(Register dst); void sbb(Register dst, const Operand& src); void shld(Register dst, const Operand& src); void shl(Register dst, uint8_t imm8); void shl(Register dst); void shrd(Register dst, const Operand& src); void shr(Register dst, uint8_t imm8); void shr(Register dst); void shr_cl(Register dst); void subb(const Operand& dst, int8_t imm8); void sub(const Operand& dst, const Immediate& x); void sub(Register dst, const Operand& src); void sub(const Operand& dst, Register src); void test(Register reg, const Immediate& imm); void test(Register reg, const Operand& op); void test(const Operand& op, const Immediate& imm); void xor_(Register dst, int32_t imm32); void xor_(Register dst, const Operand& src); void xor_(const Operand& src, Register dst); void xor_(const Operand& dst, const Immediate& x); // Bit operations. void bt(const Operand& dst, Register src); void bts(const Operand& dst, Register src); // Miscellaneous void hlt(); void int3(); void nop(); void rdtsc(); void ret(int imm16); // Label operations & relative jumps (PPUM Appendix D) // // Takes a branch opcode (cc) and a label (L) and generates // either a backward branch or a forward branch and links it // to the label fixup chain. Usage: // // Label L; // unbound label // j(cc, &L); // forward branch to unbound label // bind(&L); // bind label to the current pc // j(cc, &L); // backward branch to bound label // bind(&L); // illegal: a label may be bound only once // // Note: The same Label can be used for forward and backward branches // but it may be bound only once. void bind(Label* L); // binds an unbound label L to the current code position // Calls void call(Label* L); void call(byte* entry, RelocInfo::Mode rmode); void call(const Operand& adr); void call(Handle code, RelocInfo::Mode rmode); // Jumps void jmp(Label* L); // unconditional jump to L void jmp(byte* entry, RelocInfo::Mode rmode); void jmp(const Operand& adr); void jmp(Handle code, RelocInfo::Mode rmode); // Conditional jumps void j(Condition cc, Label* L, Hint hint = no_hint); void j(Condition cc, byte* entry, RelocInfo::Mode rmode, Hint hint = no_hint); void j(Condition cc, Handle code, Hint hint = no_hint); // Floating-point operations void fld(int i); void fld1(); void fldz(); void fld_s(const Operand& adr); void fld_d(const Operand& adr); void fstp_s(const Operand& adr); void fstp_d(const Operand& adr); void fild_s(const Operand& adr); void fild_d(const Operand& adr); void fist_s(const Operand& adr); void fistp_s(const Operand& adr); void fistp_d(const Operand& adr); void fisttp_s(const Operand& adr); void fabs(); void fchs(); void fcos(); void fsin(); void fadd(int i); void fsub(int i); void fmul(int i); void fdiv(int i); void fisub_s(const Operand& adr); void faddp(int i = 1); void fsubp(int i = 1); void fsubrp(int i = 1); void fmulp(int i = 1); void fdivp(int i = 1); void fprem(); void fprem1(); void fxch(int i = 1); void fincstp(); void ffree(int i = 0); void ftst(); void fucomp(int i); void fucompp(); void fucomip(); void fcompp(); void fnstsw_ax(); void fwait(); void fnclex(); void frndint(); void sahf(); void setcc(Condition cc, Register reg); void cpuid(); // SSE2 instructions void cvttss2si(Register dst, const Operand& src); void cvttsd2si(Register dst, const Operand& src); void cvtsi2sd(XMMRegister dst, const Operand& src); void addsd(XMMRegister dst, XMMRegister src); void subsd(XMMRegister dst, XMMRegister src); void mulsd(XMMRegister dst, XMMRegister src); void divsd(XMMRegister dst, XMMRegister src); void comisd(XMMRegister dst, XMMRegister src); // Use either movsd or movlpd. void movdbl(XMMRegister dst, const Operand& src); void movdbl(const Operand& dst, XMMRegister src); // Debugging void Print(); // Check the code size generated from label to here. int SizeOfCodeGeneratedSince(Label* l) { return pc_offset() - l->pos(); } // Mark address of the ExitJSFrame code. void RecordJSReturn(); // Record a comment relocation entry that can be used by a disassembler. // Use --debug_code to enable. void RecordComment(const char* msg); void RecordPosition(int pos); void RecordStatementPosition(int pos); void WriteRecordedPositions(); // Writes a single word of data in the code stream. // Used for inline tables, e.g., jump-tables. void dd(uint32_t data, RelocInfo::Mode reloc_info); int pc_offset() const { return pc_ - buffer_; } int current_statement_position() const { return current_statement_position_; } int current_position() const { return current_position_; } // Check if there is less than kGap bytes available in the buffer. // If this is the case, we need to grow the buffer before emitting // an instruction or relocation information. inline bool overflow() const { return pc_ >= reloc_info_writer.pos() - kGap; } // Get the number of bytes available in the buffer. inline int available_space() const { return reloc_info_writer.pos() - pc_; } // Avoid overflows for displacements etc. static const int kMaximalBufferSize = 512*MB; static const int kMinimalBufferSize = 4*KB; protected: void movsd(XMMRegister dst, const Operand& src); void movsd(const Operand& dst, XMMRegister src); void emit_sse_operand(XMMRegister reg, const Operand& adr); void emit_sse_operand(XMMRegister dst, XMMRegister src); private: byte* addr_at(int pos) { return buffer_ + pos; } byte byte_at(int pos) { return buffer_[pos]; } uint32_t long_at(int pos) { return *reinterpret_cast(addr_at(pos)); } void long_at_put(int pos, uint32_t x) { *reinterpret_cast(addr_at(pos)) = x; } // code emission void GrowBuffer(); inline void emit(uint32_t x); inline void emit(Handle handle); inline void emit(uint32_t x, RelocInfo::Mode rmode); inline void emit(const Immediate& x); inline void emit_w(const Immediate& x); // Emit the code-object-relative offset of the label's position inline void emit_code_relative_offset(Label* label); // instruction generation void emit_arith_b(int op1, int op2, Register dst, int imm8); // Emit a basic arithmetic instruction (i.e. first byte of the family is 0x81) // with a given destination expression and an immediate operand. It attempts // to use the shortest encoding possible. // sel specifies the /n in the modrm byte (see the Intel PRM). void emit_arith(int sel, Operand dst, const Immediate& x); void emit_operand(Register reg, const Operand& adr); void emit_farith(int b1, int b2, int i); // labels void print(Label* L); void bind_to(Label* L, int pos); void link_to(Label* L, Label* appendix); // displacements inline Displacement disp_at(Label* L); inline void disp_at_put(Label* L, Displacement disp); inline void emit_disp(Label* L, Displacement::Type type); // record reloc info for current pc_ void RecordRelocInfo(RelocInfo::Mode rmode, intptr_t data = 0); friend class CodePatcher; friend class EnsureSpace; // Code buffer: // The buffer into which code and relocation info are generated. byte* buffer_; int buffer_size_; // True if the assembler owns the buffer, false if buffer is external. bool own_buffer_; // A previously allocated buffer of kMinimalBufferSize bytes, or NULL. static byte* spare_buffer_; // code generation byte* pc_; // the program counter; moves forward RelocInfoWriter reloc_info_writer; // push-pop elimination byte* last_pc_; // source position information int current_statement_position_; int current_position_; int written_statement_position_; int written_position_; }; // Helper class that ensures that there is enough space for generating // instructions and relocation information. The constructor makes // sure that there is enough space and (in debug mode) the destructor // checks that we did not generate too much. class EnsureSpace BASE_EMBEDDED { public: explicit EnsureSpace(Assembler* assembler) : assembler_(assembler) { if (assembler_->overflow()) assembler_->GrowBuffer(); #ifdef DEBUG space_before_ = assembler_->available_space(); #endif } #ifdef DEBUG ~EnsureSpace() { int bytes_generated = space_before_ - assembler_->available_space(); ASSERT(bytes_generated < assembler_->kGap); } #endif private: Assembler* assembler_; #ifdef DEBUG int space_before_; #endif }; } } // namespace v8::internal #endif // V8_IA32_ASSEMBLER_IA32_H_