.SPV File Format – Enumerants

Many SPIR-V operands are numbers taken from a fixed list of named values: a storage class, a capability, a decoration, a built-in variable and so on. This page lists every such enumeration in Khronos’ spirv.h header for SPIR-V 1.6 (SPV_VERSION 0x10600), with every value, including the values that extensions add.

Names are shown without the prefix the header puts on them, so SpvStorageClassUniform is Uniform. Where several names share a value, they are in the same row. The header ends most enumerations with a Max entry of 0x7fffffff; those entries are left out.

Some enumerations are bit masks, whose values may be combined. The header names each of their values with a Mask ending and writes them in hexadecimal. Here the values keep that form and the names drop the ending, so SpvImageOperandsBiasMask is Bias and SpvImageOperandsMaskNone is None. The header also has a matching Shift enumeration for each bit mask, which this page leaves out.

In all, the 58 enumerations below hold 1,299 names for 1,120 values. They are in the order the header gives them:

• SourceLanguage, ExecutionModel, AddressingModel, MemoryModel, ExecutionMode, StorageClass, Dim
• SamplerAddressingMode, SamplerFilterMode, ImageFormat, ImageChannelOrder, ImageChannelDataType, ImageOperands
• FPFastMathMode, FPRoundingMode, LinkageType, AccessQualifier, FunctionParameterAttribute, Decoration, BuiltIn
• SelectionControl, LoopControl, FunctionControl, MemorySemantics, MemoryAccess, Scope, GroupOperation, KernelEnqueueFlags, KernelProfilingInfo, Capability
• RayFlags, RayQueryIntersection, RayQueryCommittedIntersectionType, RayQueryCandidateIntersectionType, FragmentShadingRate
• FPDenormMode, FPOperationMode, QuantizationModes, OverflowModes, PackedVectorFormat
• CooperativeMatrixOperands, CooperativeMatrixLayout, CooperativeMatrixUse, CooperativeMatrixReduce
• TensorClampMode, TensorAddressingOperands, TensorOperands
• InitializationModeQualifier, HostAccessQualifier, LoadCacheControl, StoreCacheControl, NamedMaximumNumberOfRegisters, MatrixMultiplyAccumulateOperands
• RawAccessChainOperands, FPEncoding, CooperativeVectorMatrixLayout, ComponentType, GatherModes