NewYour coding agent can read the release notes before it upgrades.Set up the MCP server β
NuGet
NuGet, #745 most downloaded on NuGet
iced is a blazing fast and correct x86 (16/32/64-bit) instruction decoder, disassembler and assembler written in C#. - π Supports all Intel and AMD instructions - π Correct: All instructions are tested and iced has been tested against other disassemblers/assemblers (xed, gas, objdump, masm, dumpbin, nasm, ndisasm) and fuzzed - π 100% C# code - π The formatter supports masm, nasm, gas (AT&T), Intel (XED) and there are many options to customize the output - π The decoder decodes >130 MB/s - π Small decoded instructions, only 40 bytes and the decoder doesn't allocate any memory - π Create instructions with code assembler, eg. `asm.mov(eax, edx)` - π The encoder can be used to re-encode decoded instructions at any address - π API to get instruction info, eg. read/written registers, memory and rflags bits; CPUID feature flag, control flow info, etc - π Supports `.NET Standard 2.0/2.1+` and `.NET Framework 4.5+` - π License: MIT
Latest stable
1.22.0
Last release 21 days ago
31 Oct 2026
Ships unpredictably
gaps range from 3 weeks to 2.8 years
Most releases are documented
notes for 22 of 30 stable releases
Nothing withdrawn
no release was ever pulled
8 years old
30 releases Β· first in 2018
Drag across the ruler to compare, or tap a release to open it.
Rust: Fixed an integer underflow panic in the decoder (debug builds / overflow-checks) when the decoded code crosses a 4 GiB address boundary #696 (Cr
overflow-checks) when the decoded code crosses a 4 GiB address boundary #696 (Credit @plzin)no_std now works even if another crate in the dependency graph uses std #666 (Credit @gkgoat1)musllinux_1_2 (was musllinux_1_1) and 32-bit Linux (i686) wheels are no longer published (use the sdist)always_show_scale now documents when the scale is always shown #743Fixed BlockEncoder returning an error when trying to encode a 32-bit branch from 0 -> FFFFFF00 #511
BlockEncoder returning an error when trying to encode a 32-bit branch from 0 -> FFFFFF00 #511Instruction::try_immediate() to the API docs #438 (Credit @mbbutler)Your coding agent can read these notes before it upgrades. Set up the MCP server β
Added new Intel instructions #428
AVX-VNNI-INT16SHA512SM3SM4TSEAdded new Intel instructions #400
AMX-COMPLEXCMPccXADD instructions #374hashbrown dependency that was used if no_std + serde has now been removed.Added new AMD instructions #369
RMPQUERYAMX-FP16AVX-IFMAAVX-NE-CONVERTAVX-VNNI-INT8CMPCCXADDMSRLISTPREFETCHITIRAO-INTWRMSRNSImproved BlockEncoder speed #262, #265
BlockEncoder speed #262, #265
Decoder speed (100MB/s -> 130MB/s, .NET 6 x64) #266, #267
XSHA512 alias #261a.zero_bytes()) that can be used as an assembler label #272CodeAssembler::assemble_options() and an API to get the address of labels after the code has been assembled #273Removed deprecated Instruction::with_* methods (use with{1,2,3,4,5}() instead)
PADLOCK instructions, added missing XSHA512 instruction #260 (VIA info from @tremalrik)IsJcxShort/IsLoopcc/IsLoop methods/props #227 (Credit @am0nsec), #259CodeAssembler traits public #236 (Credit @Kixiron)Instruction/Register fns const #240 (Credit @i509VCB)HashMap (hashbrown crate if no_std) dependency from block_encoder feature #248
no_std on Windows)Instruction::with_* methods (use with{1,2,3,4,5}() instead)BigInt is supported by default now #258Added Intel Knights Corner (KNC) instructions
DecoderOptions::KNC to the Decoder constructormvex feature must also be enabled to decode KNC instructionsmypy reporting errors, see hereserde feature): json deserialization (Instruction) is faster (json is still inefficient and slow, though)Jcc, SETcc, CMOVcc, REP{,N}E aliases (eg. a.jz(lbl))a.vex()/a.evex() fns to select VEX or EVEX encodingRust: Added Instruction::with{1,2,3,4,5}() methods to create instructions. The older methods with longer names have been deprecated. ``rust // old: leβ¦
let mut a = CodeAssembler::new(64)?;
a.xor(ecx, ecx)?;
a.add(byte_ptr(rdx + r15 * 8 + 7), 0x10)?;
assert_eq!(a.instructions().len(), 2);
let bytes = a.assemble(0x1234_5678)?;
assert_eq!(bytes, b"\x31\xC9\x42\x80\x44\xFA\x07\x10");
All instructions are supported, including VEX/EVEX instructions.
It requires the code_asm feature which is disabled by default.
See the Rust README for a longer example.Instruction::with{1,2,3,4,5}() methods to create instructions. The older methods with longer names have been deprecated.// old:
let _ = Instruction::try_with_reg_reg_u32(Code::Imul_r16_rm16_imm16, Register::CX, Register::DX, 0x5AA5)?;
// new: (the '3' suffix means '3 operands')
let _ = Instruction::with3(Code::Imul_r16_rm16_imm16, Register::CX, Register::DX, 0x5AA5)?;
// A number suffix (`u32`, `u64`, `i64`) is sometimes needed to help the compiler:
let _ = Instruction::with2(Code::Mov_r64_imm64, Register::RAX, 0x1234_5678_9ABC_DEF0u64)?;
Instruction can be serialized/deserialized with pickle (Credit: @paulfariello)Instruction::is_string_instruction() (Credit: @woodruffw) which returns true if it's eg. SCASB, MOVSQ, or any other string instruction.serde feature added to serialize/deserialize InstructionAdded Intel AVX512-FP16 instructions
AVX512-FP16 instructions0F 0E and 0F 0F instructions (enable with DecoderOptions.Udbg/::UDBG), see https://github.com/chip-red-pill/udbgInstrMemoryOperand.ctor_u64() to create an instance with an unsigned displRust: Decoder perf improvements
Decoder perf improvements
VPCMP{U,}{B,W,D,Q} pseudo ops, eg. VPCMPLEBgas formatter updated to add SYSEXIT L/Q suffixes in 64-bit mode instead of a REX.W prefixImprove BlockEncoder forward branch encoding perf (Credit: @susitsm)
BlockEncoder forward branch encoding perf (Credit: @susitsm)Rust: Decoder now implements Send + Sync (Credit: @h33p)
Decoder now implements Send + Sync (Credit: @h33p)AssemblerRegister* ctors are now public (Credit: @HoLLy-HaCKeR)Updated instruction info: LOADIWKEY may cause a VM exit
LOADIWKEY may cause a VM exitFastFormatter is now type FastFormatter = SpecializedFormatter (no breaking changes)
FRED and LKGS instructionsInstruction's memory displacement is now 64 bits (was 32 bits). This solves a couple of issues and will make creating RIP-relative memory operands much easier.
MemoryDisplacement{32,64}/memory_displacement{32,64}() (deprecated old 32-bit members: MemoryDisplacement/memory_displacement())EIP/RIP relative memory operand, MemoryDisplacement64/memory_displacement64() is now the absolute address (used to be an RIP-relative displacement)OpKind.Memory64 is deprecated since it's not used by iced (use OpKind.Memory instead)MemoryOperand's displacement argument is now an i64 (used to be an i32)Decoder::with_ip()/Decoder.Create() with an extra ip argument (Rust/.NET/Python)FastFormatter perf improvements:
SpecializedFormatter<TraitOptions> is 2x faster than the old FastFormatter code. see hereFastFormatter is now type FastFormatter = SpecializedFormatter<T> (no breaking changes)try_*() methods that return a Result<T, E>Register::values() (Credit: @tnballo) that returns an iterator that returns all enum values. Same associated method was also added to the other public enums.TryFrom<usize>Instruction::op_kinds() which returns all (0-5) OpKind valuesRust: deprecated some methods that can panic and added try_*() methods that return a Result
C0, C1, C2, C3) to RflagsBitsInstruction.GetFpuStackIncrementInfo() / Instruction.fpu_stack_increment_info() to get the value added to FPU TOP and whether it's written by the instructionUsedMemory.{try_,}virtual_address() (Credit: @woodruffw){UsedMemory,Instruction}.TryGetVirtualAddress()try_*() methods that return a Result<T, E>Added new Intel instructions: TDX, Key Locker, UINTR, HRESET, AVX-VNNI
TDX, Key Locker, UINTR, HRESET, AVX-VNNIALTINSTOpCodeInfo properties: cpl, mode (rm, v86, vmx, etc), tsx, serializing, vm exit and more.FastFormatter with a masm-like syntax that can be used if formatting speed is more important than being able to re-assemble formatted instructions. It also uses less code (good if you're targeting wasm).Instruction.HasOpKind() (Credit: @Symbai)Added Decoder.LastError property, deprecated InvalidNoMoreBytes
AMX instructions, AMD VMGEXIT and REX.W GETSECREX.W CALL/JMP FAR [mem], REX.W LSS/LFS/LGS [mem]), add UD0 (without a modrm byte). Use the AMD decoder option, the older AmdBranches is now obsolete but does the same thing.MPX ISA in 2019. MPX instructions are now disabled by default but can be enabled with the MPX decoder option.Decoder.LastError property, deprecated InvalidNoMoreBytesAssembler.db(byte[]) (Credit: @Symbai)try_virtual_address() (Credit: @Ralith)--{in,ex}clude-cpuid <name> command line optionsSupports new AMD instructions: INVLPGB, TLBSYNC
INVLPGB, TLBSYNCNegateConditionCode()/ConditionCode() now also support LOOPE/LOOPNEJcc/SETcc/CMOVcc/LOOPcc mnemonics (eg. JE or JZ)ShowUselessPrefixes formatter optionIEnumerable<Instruction>Decoder.Create(bitness, byte[], options) overloadSupports new Intel instructions: SERIALIZE, XSUSLDTRK, XRESLDTRK
SERIALIZE, XSUSLDTRK, XRESLDTRKPSMASH, PVALIDATE, RMPADJUST, RMPUPDATEPreferST0 formatter option (fadd st,st(3) vs fadd st(0),st(3))Decoder.InvalidNoMoreBytes which is true if the decoded instruction is invalid because there are no more bytes left.Assembler class was added (thanks to @xoofx for adding this feature!) which lets you easily add instructions with a syntax that looks like assembler,
Assembler class was added (thanks to @xoofx for adding this feature!) which lets you easily add instructions with a syntax that looks like assembler, eg. c.imul(rcx, __[rdx*8+0x10], -10). It supports all instructions and all modes (16/32/64-bit). See the example in the README for more info.net35 tfm, added netstandard2.1 tfmInstruction.HasSegmentPrefix propertyEncoder.WriteByte()OpCodeOperandKind.r{16,32,64}_reg_mem for movdir64b/enqcmd{s,}StreamCodeWriter and StreamCodeReaderVZERO{UPPER,ALL} don't touch regs 16-31BlockEncoder.TryEncode() signature got updated with a BlockEncoderResultInstruction.GetInfo()/GetUsedRegisters()/GetUsedMemory() were removed, and the docs never recommended using them. Use InstructionInfoFactory instead.FormatterOperandOptions is now a struct, used to be a flags enum.Instruction.Create() methods now throw if the immediate value is invalidStringBuilderFormatterOutput, use StringOutput instead (shorter name!)FormatterOutputTextKind -> FormatterTextKindSupports new Intel instructions: VP2INTERSECTD, VP2INTERSECTQ, ENQCMD, ENQCMDS
VP2INTERSECTD, VP2INTERSECTQ, ENQCMD, ENQCMDSRDPRU, MCOMMITstruct Instruction)OpCodeInfo (Code.ToOpCode() / instruction.OpCode)DecoderOptions.NoInvalidCheck so it's possible to decode some invalid instructions (eg. lock add eax,ecx, mov cs,eax)Instruction methods: NegateConditionCode(), ToShortBranch(), ToNearBranch(), ConditionCode and similar Code extension methodsref Instruction parameter now have an in Instruction parameter if they don't modify itFormatterOutput has more virtual methods, eg. WriteMnemonic() that can be overridden. They call Write() by default.Encoder properties: VEX_WIG, VEX_LIG, EVEX_WIG, EVEX_LIGdr0, dr1, etc register names instead of db0, db1, etcNothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
One column per quarter.