| Quellcodebibliothek Statistik Leitseite products/Sources/formale Sprachen/Java/Openjdk/src/hotspot/cpu/x86/ (Sun/Oracle ©) Datei vom 13.11.2022 mit Größe 120 kB |
|
Quelle assembler_x86.hpp Sprache: C |
|||||||||
|
/*
* Copyright (c) 1997, 2022, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it * under the terms of the GNU General Public License version 2 only, as * published by the Free Software Foundation. * * This code is distributed in the hope that it will be useful, but WITHOUT * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License * version 2 for more details (a copy is included in the LICENSE file that * accompanied this code). * * You should have received a copy of the GNU General Public License version * 2 along with this work; if not, write to the Free Software Foundation, * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. * * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA * or visit www.oracle.com if you need additional information or have any * questions. * */ #ifndef CPU_X86_ASSEMBLER_X86_HPP #define CPU_X86_ASSEMBLER_X86_HPP #include "asm/register.hpp" #include "utilities/powerOfTwo.hpp" // Contains all the definitions needed for x86 assembly code generation. // Calling convention class Argument { public: enum { #ifdef _LP64 #ifdef _WIN64 n_int_register_parameters_c = 4, // rcx, rdx, r8, r9 (c_rarg0, c_rarg1, ...) n_float_register_parameters_c = 4, // xmm0 - xmm3 (c_farg0, c_farg1, ... ) n_int_register_returns_c = 1, // rax n_float_register_returns_c = 1, // xmm0 #else n_int_register_parameters_c = 6, // rdi, rsi, rdx, rcx, r8, r9 (c_rarg0, c_rarg1, ...) n_float_register_parameters_c = 8, // xmm0 - xmm7 (c_farg0, c_farg1, ... ) n_int_register_returns_c = 2, // rax, rdx n_float_register_returns_c = 2, // xmm0, xmm1 #endif // _WIN64 n_int_register_parameters_j = 6, // j_rarg0, j_rarg1, ... n_float_register_parameters_j = 8 // j_farg0, j_farg1, ... #else n_register_parameters = 0, // 0 registers used to pass arguments n_int_register_parameters_j = 0, n_float_register_parameters_j = 0 #endif // _LP64 }; }; #ifdef _LP64 // Symbolically name the register arguments used by the c calling convention. // Windows is different from linux/solaris. So much for standards... #ifdef _WIN64 constexpr Register c_rarg0 = rcx; constexpr Register c_rarg1 = rdx; constexpr Register c_rarg2 = r8; constexpr Register c_rarg3 = r9; constexpr XMMRegister c_farg0 = xmm0; constexpr XMMRegister c_farg1 = xmm1; constexpr XMMRegister c_farg2 = xmm2; constexpr XMMRegister c_farg3 = xmm3; #else constexpr Register c_rarg0 = rdi; constexpr Register c_rarg1 = rsi; constexpr Register c_rarg2 = rdx; constexpr Register c_rarg3 = rcx; constexpr Register c_rarg4 = r8; constexpr Register c_rarg5 = r9; constexpr XMMRegister c_farg0 = xmm0; constexpr XMMRegister c_farg1 = xmm1; constexpr XMMRegister c_farg2 = xmm2; constexpr XMMRegister c_farg3 = xmm3; constexpr XMMRegister c_farg4 = xmm4; constexpr XMMRegister c_farg5 = xmm5; constexpr XMMRegister c_farg6 = xmm6; constexpr XMMRegister c_farg7 = xmm7; #endif // _WIN64 // Symbolically name the register arguments used by the Java calling convention. // We have control over the convention for java so we can do what we please. // What pleases us is to offset the java calling convention so that when // we call a suitable jni method the arguments are lined up and we don't // have to do little shuffling. A suitable jni method is non-static and a // small number of arguments (two fewer args on windows) // // |-------------------------------------------------------| // | c_rarg0 c_rarg1 c_rarg2 c_rarg3 c_rarg4 c_rarg5 | // |-------------------------------------------------------| // | rcx rdx r8 r9 rdi* rsi* | windows (* not a c_rarg) // | rdi rsi rdx rcx r8 r9 | solaris/linux // |-------------------------------------------------------| // | j_rarg5 j_rarg0 j_rarg1 j_rarg2 j_rarg3 j_rarg4 | // |-------------------------------------------------------| constexpr Register j_rarg0 = c_rarg1; constexpr Register j_rarg1 = c_rarg2; constexpr Register j_rarg2 = c_rarg3; // Windows runs out of register args here #ifdef _WIN64 constexpr Register j_rarg3 = rdi; constexpr Register j_rarg4 = rsi; #else constexpr Register j_rarg3 = c_rarg4; constexpr Register j_rarg4 = c_rarg5; #endif /* _WIN64 */ constexpr Register j_rarg5 = c_rarg0; constexpr XMMRegister j_farg0 = xmm0; constexpr XMMRegister j_farg1 = xmm1; constexpr XMMRegister j_farg2 = xmm2; constexpr XMMRegister j_farg3 = xmm3; constexpr XMMRegister j_farg4 = xmm4; constexpr XMMRegister j_farg5 = xmm5; constexpr XMMRegister j_farg6 = xmm6; constexpr XMMRegister j_farg7 = xmm7; constexpr Register rscratch1 = r10; // volatile constexpr Register rscratch2 = r11; // volatile constexpr Register r12_heapbase = r12; // callee-saved constexpr Register r15_thread = r15; // callee-saved #else // rscratch1 will appear in 32bit code that is dead but of course must compile // Using noreg ensures if the dead code is incorrectly live and executed it // will cause an assertion failure #define rscratch1 noreg #define rscratch2 noreg #endif // _LP64 // JSR 292 // On x86, the SP does not have to be saved when invoking method handle intrinsics // or compiled lambda forms. We indicate that by setting rbp_mh_SP_save to noreg. constexpr Register rbp_mh_SP_save = noreg; // Address is an abstraction used to represent a memory location // using any of the amd64 addressing modes with one object. // // Note: A register location is represented via a Register, not // via an address for efficiency & simplicity reasons. class ArrayAddress; class Address { public: enum ScaleFactor { no_scale = -1, times_1 = 0, times_2 = 1, times_4 = 2, times_8 = 3, times_ptr = LP64_ONLY(times_8) NOT_LP64(times_4) }; static ScaleFactor times(int size) { assert(size >= 1 && size <= 8 && is_power_of_2(size), "bad scale size"); if (size == 8) return times_8; if (size == 4) return times_4; if (size == 2) return times_2; return times_1; } static int scale_size(ScaleFactor scale) { assert(scale != no_scale, ""); assert(((1 << (int)times_1) == 1 && (1 << (int)times_2) == 2 && (1 << (int)times_4) == 4 && (1 << (int)times_8) == 8), ""); return (1 << (int)scale); } private: Register _base; Register _index; XMMRegister _xmmindex; ScaleFactor _scale; int _disp; bool _isxmmindex; RelocationHolder _rspec; // Easily misused constructors make them private // %%% can we make these go away? NOT_LP64(Address(address loc, RelocationHolder spec);) Address(int disp, address loc, relocInfo::relocType rtype); Address(int disp, address loc, RelocationHolder spec); public: int disp() { return _disp; } // creation Address() : _base(noreg), _index(noreg), _xmmindex(xnoreg), _scale(no_scale), _disp(0), _isxmmindex(false){ } // No default displacement otherwise Register can be implicitly // converted to 0(Register) which is quite a different animal. Address(Register base, int disp) : _base(base), _index(noreg), _xmmindex(xnoreg), _scale(no_scale), _disp(disp), _isxmmindex(false){ } Address(Register base, Register index, ScaleFactor scale, int disp = 0) : _base (base), _index(index), _xmmindex(xnoreg), _scale(scale), _disp (disp), _isxmmindex(false) { assert(!index->is_valid() == (scale == Address::no_scale), "inconsistent address"); } Address(Register base, RegisterOrConstant index, ScaleFactor scale = times_1, int disp = 0) : _base (base), _index(index.register_or_noreg()), _xmmindex(xnoreg), _scale(scale), _disp (disp + (index.constant_or_zero() * scale_size(scale))), _isxmmindex(false){ if (!index.is_register()) scale = Address::no_scale; assert(!_index->is_valid() == (scale == Address::no_scale), "inconsistent address"); } Address(Register base, XMMRegister index, ScaleFactor scale, int disp = 0) : _base (base), _index(noreg), _xmmindex(index), _scale(scale), _disp(disp), _isxmmindex(true) { assert(!index->is_valid() == (scale == Address::no_scale), "inconsistent address"); } // The following overloads are used in connection with the // ByteSize type (see sizes.hpp). They simplify the use of // ByteSize'd arguments in assembly code. Address(Register base, ByteSize disp) : Address(base, in_bytes(disp)) {} Address(Register base, Register index, ScaleFactor scale, ByteSize disp) : Address(base, index, scale, in_bytes(disp)) {} Address(Register base, RegisterOrConstant index, ScaleFactor scale, ByteSize disp) : Address(base, index, scale, in_bytes(disp)) {} Address plus_disp(int disp) const { Address a = (*this); a._disp += disp; return a; } Address plus_disp(RegisterOrConstant disp, ScaleFactor scale = times_1) const { Address a = (*this); a._disp += disp.constant_or_zero() * scale_size(scale); if (disp.is_register()) { assert(!a.index()->is_valid(), "competing indexes"); a._index = disp.as_register(); a._scale = scale; } return a; } bool is_same_address(Address a) const { // disregard _rspec return _base == a._base && _disp == a._disp && _index == a._index && _scale == a._scale; } // accessors bool uses(Register reg) const { return _base == reg || _index == reg; } Register base() const { return _base; } Register index() const { return _index; } XMMRegister xmmindex() const { return _xmmindex; } ScaleFactor scale() const { return _scale; } int disp() const { return _disp; } bool isxmmindex() const { return _isxmmindex; } // Convert the raw encoding form into the form expected by the constructor for // Address. An index of 4 (rsp) corresponds to having no index, so convert // that to noreg for the Address constructor. static Address make_raw(int base, int index, int scale, int disp, relocInfo::relocType disp static Address make_array(ArrayAddress); private: bool base_needs_rex() const { return _base->is_valid() && _base->encoding() >= 8; } bool index_needs_rex() const { return _index->is_valid() &&_index->encoding() >= 8; } bool xmmindex_needs_rex() const { return _xmmindex->is_valid() && _xmmindex->encoding() >= 8; } relocInfo::relocType reloc() const { return _rspec.type(); } friend class Assembler; friend class MacroAssembler; friend class LIR_Assembler; // base/index/scale/disp }; // // AddressLiteral has been split out from Address because operands of this type // need to be treated specially on 32bit vs. 64bit platforms. By splitting it out // the few instructions that need to deal with address literals are unique and the // MacroAssembler does not have to implement every instruction in the Assembler // in order to search for address literals that may need special handling depending // on the instruction and the platform. As small step on the way to merging i486/amd64 // directories. // class AddressLiteral { friend class ArrayAddress; RelocationHolder _rspec; // Typically we use AddressLiterals we want to use their rval // However in some situations we want the lval (effect address) of the item. // We provide a special factory for making those lvals. bool _is_lval; // If the target is far we'll need to load the ea of this to // a register to reach it. Otherwise if near we can do rip // relative addressing. address _target; protected: // creation AddressLiteral() : _is_lval(false), _target(NULL) {} public: AddressLiteral(address target, relocInfo::relocType rtype); AddressLiteral(address target, RelocationHolder const& rspec) : _rspec(rspec), _is_lval(false), _target(target) {} AddressLiteral addr() { AddressLiteral ret = *this; ret._is_lval = true; return ret; } private: address target() { return _target; } bool is_lval() const { return _is_lval; } relocInfo::relocType reloc() const { return _rspec.type(); } const RelocationHolder& rspec() const { return _rspec; } friend class Assembler; friend class MacroAssembler; friend class Address; friend class LIR_Assembler; }; // Convenience classes class RuntimeAddress: public AddressLiteral { public: RuntimeAddress(address target) : AddressLiteral(target, relocInfo::runtime_call_type }; class ExternalAddress: public AddressLiteral { private: static relocInfo::relocType reloc_for_target(address target) { // Sometimes ExternalAddress is used for values which aren't // exactly addresses, like the card table base. // external_word_type can't be used for values in the first page // so just skip the reloc in that case. return external_word_Relocation::can_be_relocated(target) ? relocInfo::external_wor } public: ExternalAddress(address target) : AddressLiteral(target, reloc_for_target(target)) {} }; class InternalAddress: public AddressLiteral { public: InternalAddress(address target) : AddressLiteral(target, relocInfo::internal_word_ty }; // x86 can do array addressing as a single operation since disp can be an absolute // address amd64 can't. We create a class that expresses the concept but does extra // magic on amd64 to get the final result class ArrayAddress { private: AddressLiteral _base; Address _index; public: ArrayAddress() {}; ArrayAddress(AddressLiteral base, Address index): _base(base), _index(index) {}; AddressLiteral base() { return _base; } Address index() { return _index; } }; class InstructionAttr; // 64-bit reflect the fxsave size which is 512 bytes and the new xsave area on EVEX which is anothe // See fxsave and xsave(EVEX enabled) documentation for layout const int FPUStateSizeInWords = NOT_LP64(27) LP64_ONLY(2688 / wordSize); // The Intel x86/Amd64 Assembler: Pure assembler doing NO optimizations on the instruction // level (e.g. mov rax, 0 is not translated into xor rax, rax!); i.e., what you write // is what you get. The Assembler is generating code into a CodeBuffer. class Assembler : public AbstractAssembler { friend class AbstractAssembler; // for the non-virtual hack friend class LIR_Assembler; // as_Address() friend class StubGenerator; public: enum Condition { // The x86 condition codes used for conditional jumps/moves. zero = 0x4, notZero = 0x5, equal = 0x4, notEqual = 0x5, less = 0xc, lessEqual = 0xe, greater = 0xf, greaterEqual = 0xd, below = 0x2, belowEqual = 0x6, above = 0x7, aboveEqual = 0x3, overflow = 0x0, noOverflow = 0x1, carrySet = 0x2, carryClear = 0x3, negative = 0x8, positive = 0x9, parity = 0xa, noParity = 0xb }; enum Prefix { // segment overrides CS_segment = 0x2e, SS_segment = 0x36, DS_segment = 0x3e, ES_segment = 0x26, FS_segment = 0x64, GS_segment = 0x65, REX = 0x40, REX_B = 0x41, REX_X = 0x42, REX_XB = 0x43, REX_R = 0x44, REX_RB = 0x45, REX_RX = 0x46, REX_RXB = 0x47, REX_W = 0x48, REX_WB = 0x49, REX_WX = 0x4A, REX_WXB = 0x4B, REX_WR = 0x4C, REX_WRB = 0x4D, REX_WRX = 0x4E, REX_WRXB = 0x4F, VEX_3bytes = 0xC4, VEX_2bytes = 0xC5, EVEX_4bytes = 0x62, Prefix_EMPTY = 0x0 }; enum VexPrefix { VEX_B = 0x20, VEX_X = 0x40, VEX_R = 0x80, VEX_W = 0x80 }; enum ExexPrefix { EVEX_F = 0x04, EVEX_V = 0x08, EVEX_Rb = 0x10, EVEX_X = 0x40, EVEX_Z = 0x80 }; enum VexSimdPrefix { VEX_SIMD_NONE = 0x0, VEX_SIMD_66 = 0x1, VEX_SIMD_F3 = 0x2, VEX_SIMD_F2 = 0x3 }; enum VexOpcode { VEX_OPCODE_NONE = 0x0, VEX_OPCODE_0F = 0x1, VEX_OPCODE_0F_38 = 0x2, VEX_OPCODE_0F_3A = 0x3, VEX_OPCODE_MASK = 0x1F }; enum AvxVectorLen { AVX_128bit = 0x0, AVX_256bit = 0x1, AVX_512bit = 0x2, AVX_NoVec = 0x4 }; enum EvexTupleType { EVEX_FV = 0, EVEX_HV = 4, EVEX_FVM = 6, EVEX_T1S = 7, EVEX_T1F = 11, EVEX_T2 = 13, EVEX_T4 = 15, EVEX_T8 = 17, EVEX_HVM = 18, EVEX_QVM = 19, EVEX_OVM = 20, EVEX_M128 = 21, EVEX_DUP = 22, EVEX_ETUP = 23 }; enum EvexInputSizeInBits { EVEX_8bit = 0, EVEX_16bit = 1, EVEX_32bit = 2, EVEX_64bit = 3, EVEX_NObit = 4 }; enum WhichOperand { // input to locate_operand, and format code for relocations imm_operand = 0, // embedded 32-bit|64-bit immediate operand disp32_operand = 1, // embedded 32-bit displacement or address call32_operand = 2, // embedded 32-bit self-relative displacement #ifndef _LP64 _WhichOperand_limit = 3 #else narrow_oop_operand = 3, // embedded 32-bit immediate narrow oop _WhichOperand_limit = 4 #endif }; // Comparison predicates for integral types & FP types when using SSE enum ComparisonPredicate { eq = 0, lt = 1, le = 2, _false = 3, neq = 4, nlt = 5, nle = 6, _true = 7 }; // Comparison predicates for FP types when using AVX // O means ordered. U is unordered. When using ordered, any NaN comparison is false. Otherwise, // S means signaling. Q means non-signaling. When signaling is true, instruction signals #IA o enum ComparisonPredicateFP { EQ_OQ = 0, LT_OS = 1, LE_OS = 2, UNORD_Q = 3, NEQ_UQ = 4, NLT_US = 5, NLE_US = 6, ORD_Q = 7, EQ_UQ = 8, NGE_US = 9, NGT_US = 0xA, FALSE_OQ = 0XB, NEQ_OQ = 0xC, GE_OS = 0xD, GT_OS = 0xE, TRUE_UQ = 0xF, EQ_OS = 0x10, LT_OQ = 0x11, LE_OQ = 0x12, UNORD_S = 0x13, NEQ_US = 0x14, NLT_UQ = 0x15, NLE_UQ = 0x16, ORD_S = 0x17, EQ_US = 0x18, NGE_UQ = 0x19, NGT_UQ = 0x1A, FALSE_OS = 0x1B, NEQ_OS = 0x1C, GE_OQ = 0x1D, GT_OQ = 0x1E, TRUE_US =0x1F }; enum Width { B = 0, W = 1, D = 2, Q = 3 }; //---< calculate length of instruction >--- // As instruction size can't be found out easily on x86/x64, // we just use '4' for len and maxlen. // instruction must start at passed address static unsigned int instr_len(unsigned char *instr) { return 4; } //---< longest instructions >--- // Max instruction length is not specified in architecture documentation. // We could use a "safe enough" estimate (15), but just default to // instruction length guess from above. static unsigned int instr_maxlen() { return 4; } // NOTE: The general philopsophy of the declarations here is that 64bit versions // of instructions are freely declared without the need for wrapping them an ifdef. // (Some dangerous instructions are ifdef's out of inappropriate jvm's.) // In the .cpp file the implementations are wrapped so that they are dropped out // of the resulting jvm. This is done mostly to keep the footprint of MINIMAL // to the size it was prior to merging up the 32bit and 64bit assemblers. // // This does mean you'll get a linker/runtime error if you use a 64bit only instruction // in a 32bit vm. This is somewhat unfortunate but keeps the ifdef noise down. private: bool _legacy_mode_bw; bool _legacy_mode_dq; bool _legacy_mode_vl; bool _legacy_mode_vlbw; NOT_LP64(bool _is_managed;) class InstructionAttr *_attributes; // 64bit prefixes void prefix(Register reg); void prefix(Register dst, Register src, Prefix p); void prefix(Register dst, Address adr, Prefix p); void prefix(Address adr); void prefix(Address adr, Register reg, bool byteinst = false); void prefix(Address adr, XMMRegister reg); int prefix_and_encode(int reg_enc, bool byteinst = false); int prefix_and_encode(int dst_enc, int src_enc) { return prefix_and_encode(dst_enc, false, src_enc, false); } int prefix_and_encode(int dst_enc, bool dst_is_byte, int src_enc, bool src_is_byte); // Some prefixq variants always emit exactly one prefix byte, so besides a // prefix-emitting method we provide a method to get the prefix byte to emit, // which can then be folded into a byte stream. int8_t get_prefixq(Address adr); int8_t get_prefixq(Address adr, Register reg); void prefixq(Address adr); void prefixq(Address adr, Register reg); void prefixq(Address adr, XMMRegister reg); int prefixq_and_encode(int reg_enc); int prefixq_and_encode(int dst_enc, int src_enc); void rex_prefix(Address adr, XMMRegister xreg, VexSimdPrefix pre, VexOpcode opc, bool rex_w); int rex_prefix_and_encode(int dst_enc, int src_enc, VexSimdPrefix pre, VexOpcode opc, bool rex_w); void vex_prefix(bool vex_r, bool vex_b, bool vex_x, int nds_enc, VexSimdPrefix pre, VexOpco void evex_prefix(bool vex_r, bool vex_b, bool vex_x, bool evex_r, bool evex_v, int nds_enc, VexSimdPrefix pre, VexOpcode opc); void vex_prefix(Address adr, int nds_enc, int xreg_enc, VexSimdPrefix pre, VexOpcode opc, InstructionAttr *attributes); int vex_prefix_and_encode(int dst_enc, int nds_enc, int src_enc, VexSimdPrefix pre, VexOpcode opc, InstructionAttr *attributes); void simd_prefix(XMMRegister xreg, XMMRegister nds, Address adr, VexSimdPrefix pre, VexOpcode opc, InstructionAttr *attributes); int simd_prefix_and_encode(XMMRegister dst, XMMRegister nds, XMMRegister src, VexSimdPr VexOpcode opc, InstructionAttr *attributes); // Helper functions for groups of instructions void emit_arith_b(int op1, int op2, Register dst, int imm8); void emit_arith(int op1, int op2, Register dst, int32_t imm32); // Force generation of a 4 byte immediate value even if it fits into 8bit void emit_arith_imm32(int op1, int op2, Register dst, int32_t imm32); void emit_arith(int op1, int op2, Register dst, Register src); bool emit_compressed_disp_byte(int &disp); void emit_modrm(int mod, int dst_enc, int src_enc); void emit_modrm_disp8(int mod, int dst_enc, int src_enc, int disp); void emit_modrm_sib(int mod, int dst_enc, int src_enc, Address::ScaleFactor scale, int index_enc, int base_enc); void emit_modrm_sib_disp8(int mod, int dst_enc, int src_enc, Address::ScaleFactor scale, int index_enc, int base_enc, int disp); void emit_operand_helper(int reg_enc, int base_enc, int index_enc, Address::ScaleFactor scale, int disp, RelocationHolder const& rspec, int post_addr_length); void emit_operand(Register reg, Register base, Register index, Address::ScaleFactor scale, int disp, RelocationHolder const& rspec, int post_addr_length); void emit_operand(Register reg, Register base, XMMRegister index, Address::ScaleFactor scale, int disp, RelocationHolder const& rspec, int post_addr_length); void emit_operand(XMMRegister xreg, Register base, XMMRegister xindex, Address::ScaleFactor scale, int disp, RelocationHolder const& rspec, int post_addr_length); void emit_operand(Register reg, Address adr, int post_addr_length); void emit_operand(XMMRegister reg, Register base, Register index, Address::ScaleFactor scale, int disp, RelocationHolder const& rspec, int post_addr_length); void emit_operand_helper(KRegister kreg, int base_enc, int index_enc, Address::ScaleFactor scale, int disp, RelocationHolder const& rspec, int post_addr_length); void emit_operand(KRegister kreg, Address adr, int post_addr_length); void emit_operand(KRegister kreg, Register base, Register index, Address::ScaleFactor scale, int disp, RelocationHolder const& rspec, int post_addr_length); void emit_operand(XMMRegister reg, Address adr, int post_addr_length); // Immediate-to-memory forms void emit_arith_operand(int op1, Register rm, Address adr, int32_t imm32); void emit_arith_operand_imm32(int op1, Register rm, Address adr, int32_t imm32); protected: #ifdef ASSERT void check_relocation(RelocationHolder const& rspec, int format); #endif void emit_data(jint data, relocInfo::relocType rtype, int format); void emit_data(jint data, RelocationHolder const& rspec, int format); void emit_data64(jlong data, relocInfo::relocType rtype, int format = 0); void emit_data64(jlong data, RelocationHolder const& rspec, int format = 0); bool always_reachable(AddressLiteral adr) NOT_LP64( { return true; } ); bool reachable(AddressLiteral adr) NOT_LP64( { return true; } ); // These are all easily abused and hence protected // 32BIT ONLY SECTION #ifndef _LP64 // Make these disappear in 64bit mode since they would never be correct void cmp_literal32(Register src1, int32_t imm32, RelocationHolder const& rspec); // 3 void cmp_literal32(Address src1, int32_t imm32, RelocationHolder const& rspec); // 32 void mov_literal32(Register dst, int32_t imm32, RelocationHolder const& rspec); // 32 void mov_literal32(Address dst, int32_t imm32, RelocationHolder const& rspec); // 32B void push_literal32(int32_t imm32, RelocationHolder const& rspec); // 32BIT ONLY #else // 64BIT ONLY SECTION void mov_literal64(Register dst, intptr_t imm64, RelocationHolder const& rspec); // 6 void cmp_narrow_oop(Register src1, int32_t imm32, RelocationHolder const& rspec); void cmp_narrow_oop(Address src1, int32_t imm32, RelocationHolder const& rspec); void mov_narrow_oop(Register dst, int32_t imm32, RelocationHolder const& rspec); void mov_narrow_oop(Address dst, int32_t imm32, RelocationHolder const& rspec); #endif // _LP64 // These are unique in that we are ensured by the caller that the 32bit // relative in these instructions will always be able to reach the potentially // 64bit address described by entry. Since they can take a 64bit address they // don't have the 32 suffix like the other instructions in this class. void call_literal(address entry, RelocationHolder const& rspec); void jmp_literal(address entry, RelocationHolder const& rspec); // Avoid using directly section // Instructions in this section are actually usable by anyone without danger // of failure but have performance issues that are addressed my enhanced // instructions which will do the proper thing base on the particular cpu. // We protect them because we don't trust you... // Don't use next inc() and dec() methods directly. INC & DEC instructions // could cause a partial flag stall since they don't set CF flag. // Use MacroAssembler::decrement() & MacroAssembler::increment() methods // which call inc() & dec() or add() & sub() in accordance with // the product flag UseIncDec value. void decl(Register dst); void decl(Address dst); void decq(Address dst); void incl(Register dst); void incl(Address dst); void incq(Register dst); void incq(Address dst); // New cpus require use of movsd and movss to avoid partial register stall // when loading from memory. But for old Opteron use movlpd instead of movsd. // The selection is done in MacroAssembler::movdbl() and movflt(). // Move Scalar Single-Precision Floating-Point Values void movss(XMMRegister dst, Address src); void movss(XMMRegister dst, XMMRegister src); void movss(Address dst, XMMRegister src); // Move Scalar Double-Precision Floating-Point Values void movsd(XMMRegister dst, Address src); void movsd(XMMRegister dst, XMMRegister src); void movsd(Address dst, XMMRegister src); void movlpd(XMMRegister dst, Address src); // New cpus require use of movaps and movapd to avoid partial register stall // when moving between registers. void movaps(XMMRegister dst, XMMRegister src); void movapd(XMMRegister dst, XMMRegister src); // End avoid using directly // Instruction prefixes void prefix(Prefix p); public: // Creation Assembler(CodeBuffer* code) : AbstractAssembler(code) { init_attributes(); } // Decoding static address locate_operand(address inst, WhichOperand which); static address locate_next_instruction(address inst); // Utilities static bool query_compressed_disp_byte(int disp, bool is_evex_inst, int vector_len, int cur_tuple_type, int in_size_in_bits, int cur_encoding); // Generic instructions // Does 32bit or 64bit as needed for the platform. In some sense these // belong in macro assembler but there is no need for both varieties to exist void init_attributes(void); void set_attributes(InstructionAttr *attributes) { _attributes = attributes; } void clear_attributes(void) { _attributes = NULL; } void set_managed(void) { NOT_LP64(_is_managed = true;) } void clear_managed(void) { NOT_LP64(_is_managed = false;) } bool is_managed(void) { NOT_LP64(return _is_managed;) LP64_ONLY(return false;) } void lea(Register dst, Address src); void mov(Register dst, Register src); #ifdef _LP64 // support caching the result of some routines // must be called before pusha(), popa(), vzeroupper() - checked with asserts static void precompute_instructions(); void pusha_uncached(); void popa_uncached(); #endif void vzeroupper_uncached(); void decq(Register dst); void pusha(); void popa(); void pushf(); void popf(); void push(int32_t imm32); void push(Register src); void pop(Register dst); // These do register sized moves/scans void rep_mov(); void rep_stos(); void rep_stosb(); void repne_scan(); #ifdef _LP64 void repne_scanl(); #endif // Vanilla instructions in lexical order void adcl(Address dst, int32_t imm32); void adcl(Address dst, Register src); void adcl(Register dst, int32_t imm32); void adcl(Register dst, Address src); void adcl(Register dst, Register src); void adcq(Register dst, int32_t imm32); void adcq(Register dst, Address src); void adcq(Register dst, Register src); void addb(Address dst, int imm8); void addw(Register dst, Register src); void addw(Address dst, int imm16); void addl(Address dst, int32_t imm32); void addl(Address dst, Register src); void addl(Register dst, int32_t imm32); void addl(Register dst, Address src); void addl(Register dst, Register src); void addq(Address dst, int32_t imm32); void addq(Address dst, Register src); void addq(Register dst, int32_t imm32); void addq(Register dst, Address src); void addq(Register dst, Register src); #ifdef _LP64 //Add Unsigned Integers with Carry Flag void adcxq(Register dst, Register src); //Add Unsigned Integers with Overflow Flag void adoxq(Register dst, Register src); #endif void addr_nop_4(); void addr_nop_5(); void addr_nop_7(); void addr_nop_8(); // Add Scalar Double-Precision Floating-Point Values void addsd(XMMRegister dst, Address src); void addsd(XMMRegister dst, XMMRegister src); // Add Scalar Single-Precision Floating-Point Values void addss(XMMRegister dst, Address src); void addss(XMMRegister dst, XMMRegister src); // AES instructions void aesdec(XMMRegister dst, Address src); void aesdec(XMMRegister dst, XMMRegister src); void aesdeclast(XMMRegister dst, Address src); void aesdeclast(XMMRegister dst, XMMRegister src); void aesenc(XMMRegister dst, Address src); void aesenc(XMMRegister dst, XMMRegister src); void aesenclast(XMMRegister dst, Address src); void aesenclast(XMMRegister dst, XMMRegister src); // Vector AES instructions void vaesenc(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); void vaesenclast(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); void vaesdec(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); void vaesdeclast(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); void andw(Register dst, Register src); void andb(Address dst, Register src); void andl(Address dst, int32_t imm32); void andl(Register dst, int32_t imm32); void andl(Register dst, Address src); void andl(Register dst, Register src); void andl(Address dst, Register src); void andq(Address dst, int32_t imm32); void andq(Register dst, int32_t imm32); void andq(Register dst, Address src); void andq(Register dst, Register src); void andq(Address dst, Register src); // BMI instructions void andnl(Register dst, Register src1, Register src2); void andnl(Register dst, Register src1, Address src2); void andnq(Register dst, Register src1, Register src2); void andnq(Register dst, Register src1, Address src2); void blsil(Register dst, Register src); void blsil(Register dst, Address src); void blsiq(Register dst, Register src); void blsiq(Register dst, Address src); void blsmskl(Register dst, Register src); void blsmskl(Register dst, Address src); void blsmskq(Register dst, Register src); void blsmskq(Register dst, Address src); void blsrl(Register dst, Register src); void blsrl(Register dst, Address src); void blsrq(Register dst, Register src); void blsrq(Register dst, Address src); void bsfl(Register dst, Register src); void bsrl(Register dst, Register src); #ifdef _LP64 void bsfq(Register dst, Register src); void bsrq(Register dst, Register src); #endif void bswapl(Register reg); void bswapq(Register reg); void call(Label& L, relocInfo::relocType rtype); void call(Register reg); // push pc; pc <- reg void call(Address adr); // push pc; pc <- adr void cdql(); void cdqq(); void cld(); void clflush(Address adr); void clflushopt(Address adr); void clwb(Address adr); void cmovl(Condition cc, Register dst, Register src); void cmovl(Condition cc, Register dst, Address src); void cmovq(Condition cc, Register dst, Register src); void cmovq(Condition cc, Register dst, Address src); void cmpb(Address dst, int imm8); void cmpl(Address dst, int32_t imm32); void cmpl(Register dst, int32_t imm32); void cmpl(Register dst, Register src); void cmpl(Register dst, Address src); void cmpl_imm32(Address dst, int32_t imm32); void cmpq(Address dst, int32_t imm32); void cmpq(Address dst, Register src); void cmpq(Register dst, int32_t imm32); void cmpq(Register dst, Register src); void cmpq(Register dst, Address src); void cmpw(Address dst, int imm16); void cmpxchg8 (Address adr); void cmpxchgb(Register reg, Address adr); void cmpxchgl(Register reg, Address adr); void cmpxchgq(Register reg, Address adr); void cmpxchgw(Register reg, Address adr); // Ordered Compare Scalar Double-Precision Floating-Point Values and set EFLAGS void comisd(XMMRegister dst, Address src); void comisd(XMMRegister dst, XMMRegister src); // Ordered Compare Scalar Single-Precision Floating-Point Values and set EFLAGS void comiss(XMMRegister dst, Address src); void comiss(XMMRegister dst, XMMRegister src); // Identify processor type and features void cpuid(); // CRC32C void crc32(Register crc, Register v, int8_t sizeInBytes); void crc32(Register crc, Address adr, int8_t sizeInBytes); // Convert Scalar Double-Precision Floating-Point Value to Scalar Single-Precision Floati void cvtsd2ss(XMMRegister dst, XMMRegister src); void cvtsd2ss(XMMRegister dst, Address src); // Convert Doubleword Integer to Scalar Double-Precision Floating-Point Value void cvtsi2sdl(XMMRegister dst, Register src); void cvtsi2sdl(XMMRegister dst, Address src); void cvtsi2sdq(XMMRegister dst, Register src); void cvtsi2sdq(XMMRegister dst, Address src); // Convert Doubleword Integer to Scalar Single-Precision Floating-Point Value void cvtsi2ssl(XMMRegister dst, Register src); void cvtsi2ssl(XMMRegister dst, Address src); void cvtsi2ssq(XMMRegister dst, Register src); void cvtsi2ssq(XMMRegister dst, Address src); // Convert Packed Signed Doubleword Integers to Packed Double-Precision Floating-Point Val void cvtdq2pd(XMMRegister dst, XMMRegister src); void vcvtdq2pd(XMMRegister dst, XMMRegister src, int vector_len); // Convert Halffloat to Single Precision Floating-Point value void vcvtps2ph(XMMRegister dst, XMMRegister src, int imm8, int vector_len); void vcvtph2ps(XMMRegister dst, XMMRegister src, int vector_len); void evcvtps2ph(Address dst, KRegister mask, XMMRegister src, int imm8, int vector_len); void vcvtps2ph(Address dst, XMMRegister src, int imm8, int vector_len); void vcvtph2ps(XMMRegister dst, Address src, int vector_len); // Convert Packed Signed Doubleword Integers to Packed Single-Precision Floating-Point Val void cvtdq2ps(XMMRegister dst, XMMRegister src); void vcvtdq2ps(XMMRegister dst, XMMRegister src, int vector_len); // Convert Scalar Single-Precision Floating-Point Value to Scalar Double-Precision Floati void cvtss2sd(XMMRegister dst, XMMRegister src); void cvtss2sd(XMMRegister dst, Address src); // Convert with Truncation Scalar Double-Precision Floating-Point Value to Doubleword Inte void cvtsd2siq(Register dst, XMMRegister src); void cvttsd2sil(Register dst, Address src); void cvttsd2sil(Register dst, XMMRegister src); void cvttsd2siq(Register dst, Address src); void cvttsd2siq(Register dst, XMMRegister src); // Convert with Truncation Scalar Single-Precision Floating-Point Value to Doubleword Inte void cvttss2sil(Register dst, XMMRegister src); void cvttss2siq(Register dst, XMMRegister src); void cvtss2sil(Register dst, XMMRegister src); // Convert vector double to int void cvttpd2dq(XMMRegister dst, XMMRegister src); // Convert vector float and double void vcvtps2pd(XMMRegister dst, XMMRegister src, int vector_len); void vcvtpd2ps(XMMRegister dst, XMMRegister src, int vector_len); // Convert vector float to int/long void vcvtps2dq(XMMRegister dst, XMMRegister src, int vector_len); void vcvttps2dq(XMMRegister dst, XMMRegister src, int vector_len); void evcvttps2qq(XMMRegister dst, XMMRegister src, int vector_len); // Convert vector long to vector FP void evcvtqq2ps(XMMRegister dst, XMMRegister src, int vector_len); void evcvtqq2pd(XMMRegister dst, XMMRegister src, int vector_len); // Convert vector double to long void evcvtpd2qq(XMMRegister dst, XMMRegister src, int vector_len); void evcvttpd2qq(XMMRegister dst, XMMRegister src, int vector_len); // Convert vector double to int void vcvttpd2dq(XMMRegister dst, XMMRegister src, int vector_len); // Evex casts with truncation void evpmovwb(XMMRegister dst, XMMRegister src, int vector_len); void evpmovdw(XMMRegister dst, XMMRegister src, int vector_len); void evpmovdb(XMMRegister dst, XMMRegister src, int vector_len); void evpmovqd(XMMRegister dst, XMMRegister src, int vector_len); void evpmovqb(XMMRegister dst, XMMRegister src, int vector_len); void evpmovqw(XMMRegister dst, XMMRegister src, int vector_len); // Evex casts with signed saturation void evpmovsqd(XMMRegister dst, XMMRegister src, int vector_len); //Abs of packed Integer values void pabsb(XMMRegister dst, XMMRegister src); void pabsw(XMMRegister dst, XMMRegister src); void pabsd(XMMRegister dst, XMMRegister src); void vpabsb(XMMRegister dst, XMMRegister src, int vector_len); void vpabsw(XMMRegister dst, XMMRegister src, int vector_len); void vpabsd(XMMRegister dst, XMMRegister src, int vector_len); void evpabsq(XMMRegister dst, XMMRegister src, int vector_len); // Divide Scalar Double-Precision Floating-Point Values void divsd(XMMRegister dst, Address src); void divsd(XMMRegister dst, XMMRegister src); // Divide Scalar Single-Precision Floating-Point Values void divss(XMMRegister dst, Address src); void divss(XMMRegister dst, XMMRegister src); #ifndef _LP64 private: void emit_farith(int b1, int b2, int i); public: void emms(); void fabs(); void fadd(int i); void fadd_d(Address src); void fadd_s(Address src); // "Alternate" versions of x87 instructions place result down in FPU // stack instead of on TOS void fadda(int i); // "alternate" fadd void faddp(int i = 1); void fchs(); void fcom(int i); void fcomp(int i = 1); void fcomp_d(Address src); void fcomp_s(Address src); void fcompp(); void fcos(); void fdecstp(); void fdiv(int i); void fdiv_d(Address src); void fdivr_s(Address src); void fdiva(int i); // "alternate" fdiv void fdivp(int i = 1); void fdivr(int i); void fdivr_d(Address src); void fdiv_s(Address src); void fdivra(int i); // "alternate" reversed fdiv void fdivrp(int i = 1); void ffree(int i = 0); void fild_d(Address adr); void fild_s(Address adr); void fincstp(); void finit(); void fist_s (Address adr); void fistp_d(Address adr); void fistp_s(Address adr); void fld1(); void fld_d(Address adr); void fld_s(Address adr); void fld_s(int index); void fldcw(Address src); void fldenv(Address src); void fldlg2(); void fldln2(); void fldz(); void flog(); void flog10(); void fmul(int i); void fmul_d(Address src); void fmul_s(Address src); void fmula(int i); // "alternate" fmul void fmulp(int i = 1); void fnsave(Address dst); void fnstcw(Address src); void fnstsw_ax(); void fprem(); void fprem1(); void frstor(Address src); void fsin(); void fsqrt(); void fst_d(Address adr); void fst_s(Address adr); void fstp_d(Address adr); void fstp_d(int index); void fstp_s(Address adr); void fsub(int i); void fsub_d(Address src); void fsub_s(Address src); void fsuba(int i); // "alternate" fsub void fsubp(int i = 1); void fsubr(int i); void fsubr_d(Address src); void fsubr_s(Address src); void fsubra(int i); // "alternate" reversed fsub void fsubrp(int i = 1); void ftan(); void ftst(); void fucomi(int i = 1); void fucomip(int i = 1); void fwait(); void fxch(int i = 1); void fyl2x(); void frndint(); void f2xm1(); void fldl2e(); #endif // !_LP64 // operands that only take the original 32bit registers void emit_operand32(Register reg, Address adr, int post_addr_length); void fld_x(Address adr); // extended-precision (80-bit) format void fstp_x(Address adr); // extended-precision (80-bit) format void fxrstor(Address src); void xrstor(Address src); void fxsave(Address dst); void xsave(Address dst); void hlt(); void idivl(Register src); void divl(Register src); // Unsigned division #ifdef _LP64 void idivq(Register src); void divq(Register src); // Unsigned division #endif void imull(Register src); void imull(Register dst, Register src); void imull(Register dst, Register src, int value); void imull(Register dst, Address src, int value); void imull(Register dst, Address src); #ifdef _LP64 void imulq(Register dst, Register src); void imulq(Register dst, Register src, int value); void imulq(Register dst, Address src, int value); void imulq(Register dst, Address src); void imulq(Register dst); #endif // jcc is the generic conditional branch generator to run- // time routines, jcc is used for branches to labels. jcc // 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 // jcc(cc, L); // forward branch to unbound label // bind(L); // bind label to the current pc // jcc(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 jcc(Condition cc, Label& L, bool maybe_short = true); // Conditional jump to a 8-bit offset to L. // WARNING: be very careful using this for forward jumps. If the label is // not bound within an 8-bit offset of this instruction, a run-time error // will occur. // Use macro to record file and line number. #define jccb(cc, L) jccb_0(cc, L, __FILE__, __LINE__) void jccb_0(Condition cc, Label& L, const char* file, int line); void jmp(Address entry); // pc <- entry // Label operations & relative jumps (PPUM Appendix D) void jmp(Label& L, bool maybe_short = true); // unconditional jump to L void jmp(Register entry); // pc <- entry // Unconditional 8-bit offset jump to L. // WARNING: be very careful using this for forward jumps. If the label is // not bound within an 8-bit offset of this instruction, a run-time error // will occur. // Use macro to record file and line number. #define jmpb(L) jmpb_0(L, __FILE__, __LINE__) void jmpb_0(Label& L, const char* file, int line); void ldmxcsr( Address src ); void leal(Register dst, Address src); void leaq(Register dst, Address src); void lfence(); void lock(); void size_prefix(); void lzcntl(Register dst, Register src); void lzcntl(Register dst, Address src); #ifdef _LP64 void lzcntq(Register dst, Register src); void lzcntq(Register dst, Address src); #endif enum Membar_mask_bits { StoreStore = 1 << 3, LoadStore = 1 << 2, StoreLoad = 1 << 1, LoadLoad = 1 << 0 }; // Serializes memory and blows flags void membar(Membar_mask_bits order_constraint); void mfence(); void sfence(); // Moves void mov64(Register dst, int64_t imm64); void mov64(Register dst, int64_t imm64, relocInfo::relocType rtype, int format); void movb(Address dst, Register src); void movb(Address dst, int imm8); void movb(Register dst, Address src); void movddup(XMMRegister dst, XMMRegister src); void movddup(XMMRegister dst, Address src); void vmovddup(XMMRegister dst, Address src, int vector_len); void kandbl(KRegister dst, KRegister src1, KRegister src2); void kandwl(KRegister dst, KRegister src1, KRegister src2); void kanddl(KRegister dst, KRegister src1, KRegister src2); void kandql(KRegister dst, KRegister src1, KRegister src2); void korbl(KRegister dst, KRegister src1, KRegister src2); void korwl(KRegister dst, KRegister src1, KRegister src2); void kordl(KRegister dst, KRegister src1, KRegister src2); void korql(KRegister dst, KRegister src1, KRegister src2); void kxorbl(KRegister dst, KRegister src1, KRegister src2); void kxorwl(KRegister dst, KRegister src1, KRegister src2); void kxordl(KRegister dst, KRegister src1, KRegister src2); void kxorql(KRegister dst, KRegister src1, KRegister src2); void kmovbl(KRegister dst, Register src); void kmovbl(Register dst, KRegister src); void kmovbl(KRegister dst, KRegister src); void kmovwl(KRegister dst, Register src); void kmovwl(KRegister dst, Address src); void kmovwl(Register dst, KRegister src); void kmovwl(Address dst, KRegister src); void kmovwl(KRegister dst, KRegister src); void kmovdl(KRegister dst, Register src); void kmovdl(Register dst, KRegister src); void kmovql(KRegister dst, KRegister src); void kmovql(Address dst, KRegister src); void kmovql(KRegister dst, Address src); void kmovql(KRegister dst, Register src); void kmovql(Register dst, KRegister src); void knotbl(KRegister dst, KRegister src); void knotwl(KRegister dst, KRegister src); void knotdl(KRegister dst, KRegister src); void knotql(KRegister dst, KRegister src); void kortestbl(KRegister dst, KRegister src); void kortestwl(KRegister dst, KRegister src); void kortestdl(KRegister dst, KRegister src); void kortestql(KRegister dst, KRegister src); void kxnorbl(KRegister dst, KRegister src1, KRegister src2); void kshiftlbl(KRegister dst, KRegister src, int imm8); void kshiftlql(KRegister dst, KRegister src, int imm8); void kshiftrbl(KRegister dst, KRegister src, int imm8); void kshiftrwl(KRegister dst, KRegister src, int imm8); void kshiftrdl(KRegister dst, KRegister src, int imm8); void kshiftrql(KRegister dst, KRegister src, int imm8); void ktestq(KRegister src1, KRegister src2); void ktestd(KRegister src1, KRegister src2); void kunpckdql(KRegister dst, KRegister src1, KRegister src2); void ktestql(KRegister dst, KRegister src); void ktestdl(KRegister dst, KRegister src); void ktestwl(KRegister dst, KRegister src); void ktestbl(KRegister dst, KRegister src); void movdl(XMMRegister dst, Register src); void movdl(Register dst, XMMRegister src); void movdl(XMMRegister dst, Address src); void movdl(Address dst, XMMRegister src); // Move Double Quadword void movdq(XMMRegister dst, Register src); void movdq(Register dst, XMMRegister src); // Move Aligned Double Quadword void movdqa(XMMRegister dst, XMMRegister src); void movdqa(XMMRegister dst, Address src); // Move Unaligned Double Quadword void movdqu(Address dst, XMMRegister src); void movdqu(XMMRegister dst, Address src); void movdqu(XMMRegister dst, XMMRegister src); // Move Unaligned 256bit Vector void vmovdqu(Address dst, XMMRegister src); void vmovdqu(XMMRegister dst, Address src); void vmovdqu(XMMRegister dst, XMMRegister src); // Move Unaligned 512bit Vector void evmovdqub(XMMRegister dst, XMMRegister src, int vector_len); void evmovdqub(XMMRegister dst, Address src, int vector_len); void evmovdqub(XMMRegister dst, KRegister mask, XMMRegister src, bool merge, int vector_le void evmovdqub(XMMRegister dst, KRegister mask, Address src, bool merge, int vector_len); void evmovdqub(Address dst, KRegister mask, XMMRegister src, bool merge, int vector_len); void evmovdquw(XMMRegister dst, Address src, int vector_len); void evmovdquw(Address dst, XMMRegister src, int vector_len); void evmovdquw(XMMRegister dst, KRegister mask, XMMRegister src, bool merge, int vector_le void evmovdquw(XMMRegister dst, KRegister mask, Address src, bool merge, int vector_len); void evmovdquw(Address dst, KRegister mask, XMMRegister src, bool merge, int vector_len); void evmovdqul(XMMRegister dst, XMMRegister src, int vector_len); void evmovdqul(XMMRegister dst, Address src, int vector_len); void evmovdqul(Address dst, XMMRegister src, int vector_len); void evmovdqul(XMMRegister dst, KRegister mask, XMMRegister src, bool merge, int vector_le void evmovdqul(XMMRegister dst, KRegister mask, Address src, bool merge, int vector_len); void evmovdqul(Address dst, KRegister mask, XMMRegister src, bool merge, int vector_len); void evmovdquq(Address dst, XMMRegister src, int vector_len); void evmovdquq(XMMRegister dst, Address src, int vector_len); void evmovdquq(XMMRegister dst, XMMRegister src, int vector_len); void evmovdquq(XMMRegister dst, KRegister mask, XMMRegister src, bool merge, int vector_le void evmovdquq(XMMRegister dst, KRegister mask, Address src, bool merge, int vector_len); void evmovdquq(Address dst, KRegister mask, XMMRegister src, bool merge, int vector_len); // Move lower 64bit to high 64bit in 128bit register void movlhps(XMMRegister dst, XMMRegister src); void movl(Register dst, int32_t imm32); void movl(Address dst, int32_t imm32); void movl(Register dst, Register src); void movl(Register dst, Address src); void movl(Address dst, Register src); #ifdef _LP64 void movq(Register dst, Register src); void movq(Register dst, Address src); void movq(Address dst, Register src); void movq(Address dst, int32_t imm32); void movq(Register dst, int32_t imm32); #endif // Move Quadword void movq(Address dst, XMMRegister src); void movq(XMMRegister dst, Address src); void movq(XMMRegister dst, XMMRegister src); void movq(Register dst, XMMRegister src); void movq(XMMRegister dst, Register src); void movsbl(Register dst, Address src); void movsbl(Register dst, Register src); #ifdef _LP64 void movsbq(Register dst, Address src); void movsbq(Register dst, Register src); // Move signed 32bit immediate to 64bit extending sign void movslq(Address dst, int32_t imm64); void movslq(Register dst, int32_t imm64); void movslq(Register dst, Address src); void movslq(Register dst, Register src); #endif void movswl(Register dst, Address src); void movswl(Register dst, Register src); #ifdef _LP64 void movswq(Register dst, Address src); void movswq(Register dst, Register src); #endif void movups(XMMRegister dst, Address src); void vmovups(XMMRegister dst, Address src, int vector_len); void movups(Address dst, XMMRegister src); void vmovups(Address dst, XMMRegister src, int vector_len); void movw(Address dst, int imm16); void movw(Register dst, Address src); void movw(Address dst, Register src); void movzbl(Register dst, Address src); void movzbl(Register dst, Register src); #ifdef _LP64 void movzbq(Register dst, Address src); void movzbq(Register dst, Register src); #endif void movzwl(Register dst, Address src); void movzwl(Register dst, Register src); #ifdef _LP64 void movzwq(Register dst, Address src); void movzwq(Register dst, Register src); #endif // Unsigned multiply with RAX destination register void mull(Address src); void mull(Register src); #ifdef _LP64 void mulq(Address src); void mulq(Register src); void mulxq(Register dst1, Register dst2, Register src); #endif // Multiply Scalar Double-Precision Floating-Point Values void mulsd(XMMRegister dst, Address src); void mulsd(XMMRegister dst, XMMRegister src); // Multiply Scalar Single-Precision Floating-Point Values void mulss(XMMRegister dst, Address src); void mulss(XMMRegister dst, XMMRegister src); void negl(Register dst); void negl(Address dst); #ifdef _LP64 void negq(Register dst); void negq(Address dst); #endif void nop(int i = 1); void notl(Register dst); #ifdef _LP64 void notq(Register dst); void btsq(Address dst, int imm8); void btrq(Address dst, int imm8); #endif void orw(Register dst, Register src); void orl(Address dst, int32_t imm32); void orl(Register dst, int32_t imm32); void orl(Register dst, Address src); void orl(Register dst, Register src); void orl(Address dst, Register src); void orb(Address dst, int imm8); void orb(Address dst, Register src); void orq(Address dst, int32_t imm32); void orq(Address dst, Register src); void orq(Register dst, int32_t imm32); void orq(Register dst, Address src); void orq(Register dst, Register src); // Pack with signed saturation void packsswb(XMMRegister dst, XMMRegister src); void vpacksswb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); void packssdw(XMMRegister dst, XMMRegister src); void vpackssdw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); // Pack with unsigned saturation void packuswb(XMMRegister dst, XMMRegister src); void packuswb(XMMRegister dst, Address src); void packusdw(XMMRegister dst, XMMRegister src); void vpackuswb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); void vpackusdw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); // Permutations void vpermq(XMMRegister dst, XMMRegister src, int imm8, int vector_len); void vpermq(XMMRegister dst, XMMRegister src, int imm8); void vpermq(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); void vpermb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); void vpermb(XMMRegister dst, XMMRegister nds, Address src, int vector_len); void vpermw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); void vpermd(XMMRegister dst, XMMRegister nds, Address src, int vector_len); void vpermd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); void vperm2i128(XMMRegister dst, XMMRegister nds, XMMRegister src, int imm8); void vperm2f128(XMMRegister dst, XMMRegister nds, XMMRegister src, int imm8); void vpermilps(XMMRegister dst, XMMRegister src, int imm8, int vector_len); void vpermilpd(XMMRegister dst, XMMRegister src, int imm8, int vector_len); void vpermpd(XMMRegister dst, XMMRegister src, int imm8, int vector_len); void evpermi2q(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); void evpermt2b(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); void evpmultishiftqb(XMMRegister dst, XMMRegister ctl, XMMRegister src, int vector_len); void pause(); // Undefined Instruction void ud2(); // SSE4.2 string instructions void pcmpestri(XMMRegister xmm1, XMMRegister xmm2, int imm8); void pcmpestri(XMMRegister xmm1, Address src, int imm8); void pcmpeqb(XMMRegister dst, XMMRegister src); void vpcmpCCbwd(XMMRegister dst, XMMRegister nds, XMMRegister src, int cond_encoding, int void vpcmpeqb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); void evpcmpeqb(KRegister kdst, XMMRegister nds, XMMRegister src, int vector_len); void evpcmpeqb(KRegister kdst, XMMRegister nds, Address src, int vector_len); void evpcmpeqb(KRegister kdst, KRegister mask, XMMRegister nds, Address src, int vector_le void vpcmpgtb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); void evpcmpgtb(KRegister kdst, XMMRegister nds, Address src, int vector_len); void evpcmpgtb(KRegister kdst, KRegister mask, XMMRegister nds, Address src, int vector_le void evpcmpuw(KRegister kdst, XMMRegister nds, XMMRegister src, ComparisonPredicate vcc, void evpcmpuw(KRegister kdst, XMMRegister nds, Address src, ComparisonPredicate vcc, int v void pcmpeqw(XMMRegister dst, XMMRegister src); void vpcmpeqw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); void evpcmpeqw(KRegister kdst, XMMRegister nds, XMMRegister src, int vector_len); void evpcmpeqw(KRegister kdst, XMMRegister nds, Address src, int vector_len); void vpcmpgtw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); void pcmpeqd(XMMRegister dst, XMMRegister src); void vpcmpeqd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); void evpcmpeqd(KRegister kdst, KRegister mask, XMMRegister nds, XMMRegister src, int vecto void evpcmpeqd(KRegister kdst, KRegister mask, XMMRegister nds, Address src, int vector_le void pcmpeqq(XMMRegister dst, XMMRegister src); void evpcmpeqq(KRegister kdst, KRegister mask, XMMRegister nds, XMMRegister src, int vecto void vpcmpCCq(XMMRegister dst, XMMRegister nds, XMMRegister src, int cond_encoding, int ve void vpcmpeqq(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); void evpcmpeqq(KRegister kdst, XMMRegister nds, XMMRegister src, int vector_len); void evpcmpeqq(KRegister kdst, XMMRegister nds, Address src, int vector_len); void pcmpgtq(XMMRegister dst, XMMRegister src); void vpcmpgtq(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); void pmovmskb(Register dst, XMMRegister src); void vpmovmskb(Register dst, XMMRegister src, int vec_enc); void vmovmskps(Register dst, XMMRegister src, int vec_enc); void vmovmskpd(Register dst, XMMRegister src, int vec_enc); void vpmaskmovd(XMMRegister dst, XMMRegister nds, Address src, int vector_len); void vpmaskmovq(XMMRegister dst, XMMRegister mask, Address src, int vector_len); void vmaskmovps(XMMRegister dst, Address src, XMMRegister mask, int vector_len); void vmaskmovpd(XMMRegister dst, Address src, XMMRegister mask, int vector_len); void vmaskmovps(Address dst, XMMRegister src, XMMRegister mask, int vector_len); void vmaskmovpd(Address dst, XMMRegister src, XMMRegister mask, int vector_len); // SSE 4.1 extract void pextrd(Register dst, XMMRegister src, int imm8); void pextrq(Register dst, XMMRegister src, int imm8); void pextrd(Address dst, XMMRegister src, int imm8); void pextrq(Address dst, XMMRegister src, int imm8); void pextrb(Register dst, XMMRegister src, int imm8); void pextrb(Address dst, XMMRegister src, int imm8); // SSE 2 extract void pextrw(Register dst, XMMRegister src, int imm8); void pextrw(Address dst, XMMRegister src, int imm8); // SSE 4.1 insert void pinsrd(XMMRegister dst, Register src, int imm8); void pinsrq(XMMRegister dst, Register src, int imm8); void pinsrb(XMMRegister dst, Register src, int imm8); void pinsrd(XMMRegister dst, Address src, int imm8); void pinsrq(XMMRegister dst, Address src, int imm8); void pinsrb(XMMRegister dst, Address src, int imm8); void insertps(XMMRegister dst, XMMRegister src, int imm8); // SSE 2 insert void pinsrw(XMMRegister dst, Register src, int imm8); void pinsrw(XMMRegister dst, Address src, int imm8); // AVX insert void vpinsrd(XMMRegister dst, XMMRegister nds, Register src, int imm8); void vpinsrb(XMMRegister dst, XMMRegister nds, Register src, int imm8); void vpinsrq(XMMRegister dst, XMMRegister nds, Register src, int imm8); void vpinsrw(XMMRegister dst, XMMRegister nds, Register src, int imm8); void vinsertps(XMMRegister dst, XMMRegister nds, XMMRegister src, int imm8); // Zero extend moves void pmovzxbw(XMMRegister dst, XMMRegister src); void pmovzxbw(XMMRegister dst, Address src); void pmovzxbd(XMMRegister dst, XMMRegister src); void vpmovzxbw(XMMRegister dst, Address src, int vector_len); void vpmovzxbw(XMMRegister dst, XMMRegister src, int vector_len); void vpmovzxbd(XMMRegister dst, XMMRegister src, int vector_len); void vpmovzxbq(XMMRegister dst, XMMRegister src, int vector_len); void vpmovzxwd(XMMRegister dst, XMMRegister src, int vector_len); void vpmovzxwq(XMMRegister dst, XMMRegister src, int vector_len); void pmovzxdq(XMMRegister dst, XMMRegister src); void vpmovzxdq(XMMRegister dst, XMMRegister src, int vector_len); void evpmovzxbw(XMMRegister dst, KRegister mask, Address src, int vector_len); // Sign extend moves void pmovsxbd(XMMRegister dst, XMMRegister src); void pmovsxbq(XMMRegister dst, XMMRegister src); void pmovsxbw(XMMRegister dst, XMMRegister src); void pmovsxwd(XMMRegister dst, XMMRegister src); void vpmovsxbd(XMMRegister dst, XMMRegister src, int vector_len); void vpmovsxbq(XMMRegister dst, XMMRegister src, int vector_len); void vpmovsxbw(XMMRegister dst, XMMRegister src, int vector_len); void vpmovsxwd(XMMRegister dst, XMMRegister src, int vector_len); void vpmovsxwq(XMMRegister dst, XMMRegister src, int vector_len); void vpmovsxdq(XMMRegister dst, XMMRegister src, int vector_len); void evpmovwb(Address dst, XMMRegister src, int vector_len); void evpmovwb(Address dst, KRegister mask, XMMRegister src, int vector_len); void evpmovdb(Address dst, XMMRegister src, int vector_len); // Multiply add void pmaddwd(XMMRegister dst, XMMRegister src); void vpmaddwd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); void vpmaddubsw(XMMRegister dst, XMMRegister src1, XMMRegister src2, int vector_len); void evpmadd52luq(XMMRegister dst, XMMRegister src1, XMMRegister src2, int vector_len); void evpmadd52luq(XMMRegister dst, KRegister mask, XMMRegister src1, XMMRegister src2, bo void evpmadd52huq(XMMRegister dst, XMMRegister src1, XMMRegister src2, int vector_len); void evpmadd52huq(XMMRegister dst, KRegister mask, XMMRegister src1, XMMRegister src2, bo // Multiply add accumulate void evpdpwssd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len); #ifndef _LP64 // no 32bit push/pop on amd64 void popl(Address dst); #endif #ifdef _LP64 void popq(Address dst); void popq(Register dst); #endif void popcntl(Register dst, Address src); void popcntl(Register dst, Register src); void evpopcntb(XMMRegister dst, KRegister mask, XMMRegister src, bool merge, int vector_le void evpopcntw(XMMRegister dst, KRegister mask, XMMRegister src, bool merge, int vector_le --> -------------------- --> maximum size reached --> -------------------- ¤ Diese beiden folgenden Angebotsgruppen bietet das Unternehmen0.23Angebot Wie Sie bei der Firma Beratungs- und Dienstleistungen beauftragen können ¤*Eine klare Vorstellung vom Zielzustand |
|
2026-03-28