Release v2026.3

• General
  • Support SPV_QCOM_multiple_wait_queues, SPV_KHR_opacity_micromap, and SPV_EXT_ocp_microscaling_types (#6671, #6670, #6772)
  • Add ArmExperimentalMLOperations and NonSemantic.Graph.DebugInfo instruction sets (#6692, #6666)
  • Fix unbounded memory usage in FriendlyNameMapper (#6763)
  • Improved hex-float compilation and float equality testing (#6778, #6783)
  • Various build system and dependency updates (Bazel, GN, Protobuf, Abseil)
• Optimizer
  • Stability fixes for DeadBranchElimPass, CodeSinkingPass, and OrderBlocks (#6764, #6703, #6676)
  • Improved handling of tensor constants, non-semantic instructions, and loop decorations (#6684, #6721, #6719, #6690)
  • Add folding rule for extract from copy logical and prevent word limit overflows (#6725, #6722)
• Validator
  • Support SPV_EXT_ocp_microscaling_types, SPV_EXT_split_barrier, and SPV_ARM_graph (#6779, #6732, #6694)
  • Support for SPV_NV_cooperative_matrix_decode_vector and Intel extensions (#6693, #6717, #6665)
  • Enhanced error reporting with source line printing and Shader Debug Info (#6648, #6698, #6691, #6686, #6685, #6681)
  • Correctness fixes for ray queries, descriptors, OpSpecConstantOp, and array validation (#6710, #6729, #6731, #6777, #6709)
  • Disallow duplicate targets for OpGroupDecorate and multiple graph entry points (#6720, #6754)
• Assembler/Disassembler
  • Support for SPV_QCOM_image_processing3 (#6776)
  • Performance and correctness improvements in the disassembler and tools (#6757, #6650)

Release v2026.2

• General
  • Fix build for NonSemanticShaderDebugInfo in GN (#6647)
  • Support SPV_KHR_abort and SPV_KHR_constant_data (#6625)
  • Support versioned NonSemanticShaderDebugInfo imports (#6555)
  • Fix C++20 compilation issue with large integer literals (#6557)
  • Remove SVA (Shader Validation Assistant) (#6550)
  • Improved Kokoro build configuration and storage usage (#6573, #6551)
  • Various dependency rolls and CI updates
• Optimizer
  • Fold OpCompositeExtract feeding from OpCopyLogical or OpLoad (#6614)
  • Fix DebugValue placements and initialization (#6599, #6626)
  • Support OpConstantCompositeReplicateEXT in constant manager (#6616)
  • Aggressive DCE support for untyped pointers and OpUntypedVariableKHR (#6602, #6607)
  • Optimize SPV_EXT_opacity_micromap and SPV_EXT_shader_invocation_reorder (#6571)
  • Add LegalizeMultidimArrayPass and improve access chain reordering (#6535, #6545)
  • Added various folding and reassociation rules (#6453, #6504, #6454, #6456)
  • Support for optimizing SPV_NV_push_constant_bank (#6547)
• Validator
  • Add validation for 4-bit integer types (#6644)
  • Support and validate SPV_KHR_constant_data (OpConstantDataKHR, OpSpecConstantDataKHR) (#6646, #6643, #6642)
  • Validate NonSemanticShaderDebugInfo 101 and versioned imports (#6635, #6555)
  • Improve validation for Intel SPIR-V extensions and various KHR/EXT extensions (#6232, #6589, #6583, etc.)
  • Extensive validation coverage for OpSpecConstantOp (#6596, #6585, #6582, #6579, #6578, #6576)
  • Check output location uniqueness for GeometryStreams (#6630)
  • Improved validation for variables, constants, and types (#6634, #6598, #6595, #6568)
  • Added validation for several new opcodes and capabilities (OpGroupAsyncCopy, OpLifetimeStart/Stop, etc.) (#6519, #6514, #6515)
• Assembler/Disassembler
  • Add --handle-unknown-opcodes flag to spirv-dis (#6604)
  • Handle leading + in hex float literals in assembler (#6565)
  • Validate bit width of float types with explicit encodings in assembler (#6562)

Release v2026.1

• General
  • spirv-tools: Fix infinite recursion in SmallVector::operator== on MSVC (#6470)
  • Add support for SPV_NV_push_constant_bank (#6507)
  • Enable UseMultiToolTask for msbuild performance (#6472)
  • Fix asan error in stack usage (#6434)
  • Fix crash in array length validation (#6500)
  • Fix infinite loop in logical pointer validation (#6450)
  • Implement SPV_EXT_descriptor_heap (#6503)
• Optimizer
  • Adding constant folding for NMin, NMax and NClamp (#6433)
  • Also remap DebugScope instructions (#6501)
  • Change ValueNumber to use spivOpcodeIsCommutativeBinaryOperator (#6431)
  • Extending factoring folding rules to support subtraction (#6432)
  • Extending negation folding rules (#6444)
  • Folding Rules For Common Comparisons (#6437)
  • Folding rules for negation and cancellation (#6452)
  • Harden CFGCleanupPass against ID overflow (#6439)
  • Harden CombineAccessChainsPass against ID overflow (#6440)
  • Harden CopyPropagateArraysPass against ID overflow (#6441)
  • Remove lingering debug-scopes. (#6416)
  • add adce test for debug declare (#6462)
  • fix adce when ns debug info present (#6217) (#6420)
• Validator
  • Handle OpTypeVectorIdEXT in validate_interfaces.cpp (#6492)
  • Validate SPV_EXT_long_vector (#6471)
  • Validate floats/int Image Operands are 32-bit (#6477)
  • Add Mesh Shading Read Output check (#6426)
  • Add Mesh/Task to check for LocalSize (#6459)
  • Add missing LOD non-zero MS (#6496)
  • Add width to IsFloatScalarType (#6489)
  • Better atomic storage class message (#6425)
  • Better error message when types are the same (#6427)
  • Check Mesh Topology builtin by entry point (#6457)
  • Check for CoopMat scope (#6442)
  • Combine Image Coordinate checks (#6494)
  • Fix crash with bad OpUntypedArrayLengthKHR (#6493)
  • Label VUID 10823 (#6488)
  • Provide RuntimeDescriptorArray hint in message (#6448)
  • Use new StorageClassToString util (#6460)
  • Use validation_state.cpp helpers (#6478)
  • Validate ImageQuerySizeLod is 32-bit (#6479)
  • Validate Mesh Shader output are an array (#6449)
  • add missing relational ops to InvalidTypePass (#6495)
• Diff
  • diff: Match OpExtInst in debug and types/values sections (#6430)

Release v2025.5

• General
• Validator
• Optimizer
  • spirv-opt: Add a folding rule for OpBitReverse and fix OpBitCast to support lower to higher bit conversions. (#6321)
  • spirv-opt: Add basic support for graphs (#6351)
  • spirv-opt: Adding folding rules for bitwise and (#6361)
  • spirv-opt: Adding normal form to ValueNumber calculation (#6371)
  • spirv-opt: Cache IsReadOnlyLoad in ValueNumberTable (#6385)
  • spirv-opt: Clone decorations when unrolling loops (#6373)
  • spirv-opt: Correct ADCE bug keeping dead DebugValue Value operands (#6368)
  • spirv-opt: Fix constant folding for OpBitcast on signed integers. (#6378)
  • spirv-opt: Fix id overflow in ConvertToSampledImage (#6339)
  • spirv-opt: Handle ID overflow in AggressiveDCEPass (#6366)
  • spirv-opt: Handle ID overflow in SplitCombinedImageSamplerPass (#6406)
  • spirv-opt: Handle id overflow in CCP and constant folding (#6367)
  • spirv-opt: Handle id overflow in MergeReturnPass (#6340)
  • spirv-opt: Handle id overflow in amd_ext_to_khr (#6403)
  • Add SPV_KHR_maximal_reconvergence to the allowlist (#6372)
  • Adding folding rules for commutive bitwise operations (xor/or/and). (#6358)
  • [OPT] Move definition of Pass::Status to Utils (#6402)
• Validator:
  • spirv-val: Add LocalInvocationIndex checks (#6395)
  • spirv-val: Add validation for SPV_KHR_fma (#6388)
  • spirv-val: Break up the 3000 line ValidateExtInst function (#6364)
  • spirv-val: Fix VUID churn for 10685 (#6389)
  • spirv-val: Improve Explicit Layout message (#6331)
  • spirv-val: Use spvOpcodeString consistently (#6393)
  • Relax validation rule for DebugGlobalVariable variable (#6382)
  • Validate SPV_EXT_shader_64bit_indexing (#6376)
  • Validate logical pointer restrictions (#6240)
  • [tools/val] fix u8str for real this time (#6341)
  • Add VK_EXT_shader_uniform_buffer_unsized_array (#6386)
• Assembler
  • Add as/dis support for Arm.MotionEngine.100 extended instruction set (#6284)
  • spirv-as: fix command line for stdin as input (#6398)

v2025.4

Prepare release v2025.4 (#6335)

v2025.3

Prepare release v2025.3 (#6192)

V2025.2

[OPT] Don't fold n % 1.0 (#6100)

If values are not whole number, then the result is not 0. For example,
`(1.5 % 1.0) == 0.5`.

Fixes #6097

Release v2025.1

• General
  • diff: Fix crash in OpString matching (#5988)
  • Add SPV_AMDX_shader_enqueue version 2 support (#5838)
  • add support for SPV_INTEL_subgroup_matrix_multiply_accumulate (#5928)
  • update cmake_minimum_required to 3.22.1 (#5925)
  • Add OpImageSampleFootprintNV to IsAllowedSampledImageOperand (#5914)
  • assembler: ensure progress when seeking the version string (#5910)
• Optimizer
  • opt: keep all OpSource instructions (#5901)
  • [opt] Fix bug opt::InstructionBuilder::AddVariable (#6007)
  • [OPT] Add SPV_KHR_ray_tracing to allow list (#5941)
  • opt: keep all OpSource instructions (#5901)
• Validator
  • spirv-val: Add AllowVulkan32BitBitwise option (#6001)
  • Fix untyped pointer comparison validation (#6004)
  • spirv-val: Update VUIDs for 308 header (#5990)
  • spirv-val: fix env parsing for vk1.1spv1.4 (#5985)
  • Add validation for SPV_NV_linear_swept_spheres. (#5975)
  • Add validation SPV_NV_cluster_acceleration_structure. (#5974)
  • Improve the instruction diagnostic for some access chain errors (#5978)
  • Update physical storage buffer validation to match SPIR-V 1.6.5 (#5977)
  • Validate SPV_NV_cooperative_vector (#5972)
  • Fix layout checks with untyped pointers (#5970)
  • spirv-val: Update mislabeled VUIDs (#5964)
  • More explicit layout validation (#5958)
  • spirv-val: Add VK_KHR_maintenance8 support (#5951)
  • Add EXT_mesh_shader validation support (#5640)
  • spirv-val: Remove OpenCL ivec3 req (#5940)
  • spirv-val: Validate zero product workgroup size (#5407)
  • Relax DebugLine validation (#5916)
• Linker
  • linker: remove LinkOnceODR decorations when linking executables (#5979)
  • fix: handle LinkOnceODR correctly (#5938)

Release v2024.4

• General
  • Add FPEncoding operand type. (#5726)
  • Support SPV_KHR_untyped_pointers (#5736)
  • add support for SPV_INTEL_global_variable_host_access (#5786)
  • Add support for SPV_KHR_compute_shader_derivative (#5817)
  • Accept hex representation as binary input (#5870)
  • Vulkan 1.4 support (#5899)
• Optimizer
  • Add knowledge of cooperative matrices (#5720)
  • Add struct-packing pass and unit test. (#5778)
• Validator
  • Validate presence of Stride operand to OpCooperativeMatrix{Load,Store}KHR (#5777)
  • Update sampled image validation (#5789)
  • Disallow stores according to VUID 06924 (#5368)
  • Add validation for SPV_NV_tensor_addressing and SPV_NV_cooperative_matrix2 (#5865)
• Linker
  • allow linking functions with different pointer arguments (#5534)

Release v2024.3

• General
• Optimizer
  • Do not fold mul and adds to generate fmas (#5682)
  • add OpExtInst forward ref fixup pass (#5708)
• Validator
  • Separate Location check for tess patch (#5654)
  • Validate MemoryAccessMask of OpCooperativeMatrixStoreKHR (#5668)
  • OpSampledImage extra validation (#5695)
  • add support for OpExtInstWithForwardRefs (#5698)A
• Disassembler
  • add decorations to comments (#5675)
  • Add --nested-indent and --reorder-blocks (#5671)