// Copyright (c) 2015-2016 The Khronos Group Inc. // Copyright notice at https://www.khronos.org/registry/speccopyright.html [appendix] [[spirvenv]] = Vulkan Environment for SPIR-V Shaders for Vulkan are defined by the <> as well as the <>. This appendix defines additional SPIR-V requirements applying to Vulkan shaders. == Required Versions and Formats A Vulkan 1.0 implementation must: support the 1.0 version of SPIR-V and the 1.0 version of the SPIR-V Extended Instructions for GLSL. A SPIR-V module passed into flink:vkCreateShaderModule is interpreted as a series of 32-bit words in host endianness, with literal strings packed as described in section 2.2 of the SPIR-V Specification. The first few words of the SPIR-V module must: be a magic number and a SPIR-V version number, as described in section 2.3 of the SPIR-V Specification. [[spirvenv-capabilities]] == Capabilities Implementations must: support the following capability operands declared by code:OpCapability: - code:Matrix - code:Shader - code:InputAttachment - code:Sampled1D - code:Image1D - code:SampledBuffer - code:ImageBuffer - code:ImageQuery - code:DerivativeControl Implementations may: support features that are not required: by the Specification, as described in the <> chapter. If such a feature is supported, then any capability operand(s) corresponding to that feature must: also be supported. [[spirvenv-capabilities-table]] .SPIR-V Capabilities which are not required:, and corresponding feature names [options="header"] |==== | SPIR-V OpCapability | Vulkan feature name | code:Geometry | <> | code:Tessellation | <> | code:Float64 | <> | code:Int64 | <> | code:Int16 | <> | code:TessellationPointSize | <> | code:GeometryPointSize | <> | code:ImageGatherExtended | <> | code:StorageImageMultisample | <> | code:UniformBufferArrayDynamicIndexing | <> | code:SampledImageArrayDynamicIndexing | <> | code:StorageBufferArrayDynamicIndexing | <> | code:StorageImageArrayDynamicIndexing | <> | code:ClipDistance | <> | code:CullDistance | <> | code:ImageCubeArray | <> | code:SampleRateShading | <> | code:SparseResidency | <> | code:MinLod | <> | code:SampledCubeArray | <> | code:ImageMSArray | <> | code:StorageImageExtendedFormats | <> | code:InterpolationFunction | <> | code:StorageImageReadWithoutFormat | <> | code:StorageImageWriteWithoutFormat | <> | code:MultiViewport | <> |==== The application mustnot: pass a SPIR-V module containing any of the following to flink:vkCreateShaderModule: - any OpCapability not listed above, - an unsupported capability, or - a capability which corresponds to a Vulkan feature which has not been enabled. [[spirvenv-module-validation]] == Validation Rules within a Module A SPIR-V module passed to flink:vkCreateShaderModule must: conform to the following rules: * Every entry point must: have no return value and accept no arguments. * Recursion: The static function-call graph for an entry point mustnot: contain cycles. * The *Logical* addressing model must: be selected. * *Scope* for execution must: be limited to: ** *Workgroup* ** *Subgroup* * *Scope* for memory must: be limited to: ** *Device* ** *Workgroup* ** *Invocation* * The code:OriginLowerLeft execution mode mustnot: be used; fragment entry points must: declare code:OriginUpperLeft. * The code:PixelCenterInteger execution mode mustnot: be used. Pixels are always centered at half-integer coordinates. * Images ** code:OpTypeImage must: declare a scalar 32-bit float or 32-bit integer type for the ``Sampled Type''. (code:RelaxedPrecision can: be applied to a sampling instruction and to the variable holding the result of a sampling instruction.) ** code:OpSampledImage must: only consume an ``Image'' operand whose type has its ``Sampled'' operand set to 1. ** The ``(u, v)'' coordinates used for a code:SubpassData must: be the of a constant vector (0, 0), or if a layer coordinate is used, must: be a vector that was formed with constant 0 for the ``u'' and ``v'' components. ** The ``Depth'' operand of code:OpTypeImage is ignored. * Decorations ** The code:GLSLShared and code:GLSLPacked decorations mustnot: be used. ** The code:Flat, code:NoPerspective, code:Sample, and code:Centroid decorations mustnot: be used on variables with storage class other than code:Input or on variables used in the interface of non-fragment shader entry points. ** The code:Patch decoration mustnot: be used on variables in the interface of a vertex, geometry, or fragment shader stage's entry point. * code:OpTypeRuntimeArray must: only be used for the last member of an code:OpTypeStruct in the code:Uniform storage class. * Linkage: See <> for additional linking and validation rules. * Compute Shaders ** For each compute shader entry point, either a code:LocalSize execution mode or an object decorated with the code:WorkgroupSize decoration must: be specified. [[spirvenv-precision-operation]] == Precision and Operation of SPIR-V Instructions The following rules apply to both single and double-precision floating point instructions: - Positive and negative infinities and positive and negative zeros are generated as dictated by <>, but subject to the precisions allowed in the following table. - Dividing a non-zero by a zero results in the appropriately signed <> infinity. - Any denormalized value input into a shader or potentially generated by any instruction in a shader may: be flushed to 0. - The rounding mode cannot: be set and is undefined. - NaNs may: not be generated. Instructions that operate on a NaN may: not result in a NaN. - Support for signaling NaNs is optional: and exceptions are never raised. The precision of double-precision instructions is at least that of single precision. For single precision (32 bit) instructions, precisions are required: to be at least as follows, unless decorated with RelaxedPrecision: .Precision of core SPIR-V Instructions [options="header"] |==== | Instruction | Precision | code:OpFAdd | Correctly rounded. | code:OpFSub | Correctly rounded. | code:OpFMul | Correctly rounded. | code:OpFOrdEqual, code:OpFUnordEqual | Correct result. | code:OpFOrdLessThan, code:OpFUnordLessThan | Correct result. | code:OpFOrdGreaterThan, code:OpFUnordGreaterThan | Correct result. | code:OpFOrdLessThanEqual, code:OpFUnordLessThanEqual | Correct result. | code:OpFOrdGreaterThanEqual, code:OpFUnordGreaterThanEqual| Correct result. | code:OpFDiv | 2.5 ULP for b in the range [2^-126^, 2^126^]. | conversions between types | Correctly rounded. |==== .Precision of GLSL.std.450 Instructions [options="header"] |==== |Instruction | Precision | code:fma() | Inherited from code:OpFMul followed by code:OpFAdd. | code:exp(x), code:exp2(x) | latexmath:[$(3 + 2 \times \|x\|)$] ULP. | code:log(), code:log2() | 3 ULP outside the range [0.5, 2.0]. Absolute error < 2^-21^ inside the range [0.5, 2.0]. | code:pow(x, y) | Inherited from code:exp2(y × code:log2(x)). | code:sqrt() | Inherited from 1.0 / code:inversesqrt(). | code:inversesqrt() | 2 ULP. |==== GLSL.std.450 extended instructions specifically defined in terms of the above instructions inherit the above errors. GLSL.std.450 extended instructions not listed above and not defined in terms of the above have undefined precision. These include, for example, the trigonometric functions and determinant. For the code:OpSRem and code:OpSMod instructions, if either operand is negative the result is undefined. [NOTE] .Note ==== While the code:OpSRem and code:OpSMod instructions are supported by the Vulkan environment, they require non-negative values and thus do not enable additional functionality beyond what code:OpUMod provides. ==== [[spirvenv-image-formats]] Compatibility Between SPIR-V Image Formats And Vulkan Formats ------------------------------------------------------------- [cols="2*", options="header"] |=== |SPIR-V Image Format |Vulkan Format |code:Rgba32f |ename:VK_FORMAT_R32G32B32A32_SFLOAT |code:Rgba16f |ename:VK_FORMAT_R16G16B16A16_SFLOAT |code:R32f |ename:VK_FORMAT_R32_SFLOAT |code:Rgba8 |ename:VK_FORMAT_R8G8B8A8_UNORM |code:Rgba8Snorm |ename:VK_FORMAT_R8G8B8A8_SNORM |code:Rg32f |ename:VK_FORMAT_R32G32_SFLOAT |code:Rg16f |ename:VK_FORMAT_R16G16_SFLOAT |code:R11fG11fB10f |ename:VK_FORMAT_B10G11R11_UFLOAT_PACK32 |code:R16f |ename:VK_FORMAT_R16_SFLOAT |code:Rgba16 |ename:VK_FORMAT_R16G16B16A16_UNORM |code:Rgb10A2 |ename:VK_FORMAT_A2B10G10R10_UNORM_PACK32 |code:Rg16 |ename:VK_FORMAT_R16G16_UNORM |code:Rg8 |ename:VK_FORMAT_R8G8_UNORM |code:R16 |ename:VK_FORMAT_R16_UNORM |code:R8 |ename:VK_FORMAT_R8_UNORM |code:Rgba16Snorm |ename:VK_FORMAT_R16G16B16A16_SNORM |code:Rg16Snorm |ename:VK_FORMAT_R16G16_SNORM |code:Rg8Snorm |ename:VK_FORMAT_R8G8_SNORM |code:R16Snorm |ename:VK_FORMAT_R16_SNORM |code:R8Snorm |ename:VK_FORMAT_R8_SNORM |code:Rgba32i |ename:VK_FORMAT_R32G32B32A32_SINT |code:Rgba16i |ename:VK_FORMAT_R16G16B16A16_SINT |code:Rgba8i |ename:VK_FORMAT_R8G8B8A8_SINT |code:R32i |ename:VK_FORMAT_R32_SINT |code:Rg32i |ename:VK_FORMAT_R32G32_SINT |code:Rg16i |ename:VK_FORMAT_R16G16_SINT |code:Rg8i |ename:VK_FORMAT_R8G8_SINT |code:R16i |ename:VK_FORMAT_R16_SINT |code:R8i |ename:VK_FORMAT_R8_SINT |code:Rgba32ui |ename:VK_FORMAT_R32G32B32A32_UINT |code:Rgba16ui |ename:VK_FORMAT_R16G16B16A16_UINT |code:Rgba8ui |ename:VK_FORMAT_R8G8B8A8_UINT |code:R32ui |ename:VK_FORMAT_R32_UINT |code:Rgb10a2ui |ename:VK_FORMAT_A2B10G10R10_UINT_PACK32 |code:Rg32ui |ename:VK_FORMAT_R32G32_UINT |code:Rg16ui |ename:VK_FORMAT_R16G16_UINT |code:Rg8ui |ename:VK_FORMAT_R8G8_UINT |code:R16ui |ename:VK_FORMAT_R16_UINT |code:R8ui |ename:VK_FORMAT_R8_UINT |===