31. Features

Features describe functionality which is not supported on all implementations. Features are properties of the physical device. Features are optional, and must be explicitly enabled before use. Support for features is reported and enabled on a per-feature basis.

Note

Features are reported via the basic VkPhysicalDeviceFeatures structure, as well as the extensible structure VkPhysicalDeviceFeatures2, which was added in the VK_KHR_get_physical_device_properties2 extension and included in Vulkan 1.1. When new features are added in future Vulkan versions or extensions, each extension should introduce one new feature structure, if needed. This structure can be added to the pNext chain of the VkPhysicalDeviceFeatures2 structure.

To query supported features, call:

void vkGetPhysicalDeviceFeatures(
    VkPhysicalDevice                            physicalDevice,
    VkPhysicalDeviceFeatures*                   pFeatures);
  • physicalDevice is the physical device from which to query the supported features.

  • pFeatures is a pointer to a VkPhysicalDeviceFeatures structure in which the physical device features are returned. For each feature, a value of VK_TRUE specifies that the feature is supported on this physical device, and VK_FALSE specifies that the feature is not supported.

Valid Usage (Implicit)
  • physicalDevice must be a valid VkPhysicalDevice handle

  • pFeatures must be a valid pointer to a VkPhysicalDeviceFeatures structure

Fine-grained features used by a logical device must be enabled at VkDevice creation time. If a feature is enabled that the physical device does not support, VkDevice creation will fail and return VK_ERROR_FEATURE_NOT_PRESENT.

The fine-grained features are enabled by passing a pointer to the VkPhysicalDeviceFeatures structure via the pEnabledFeatures member of the VkDeviceCreateInfo structure that is passed into the vkCreateDevice call. If a member of pEnabledFeatures is set to VK_TRUE or VK_FALSE, then the device will be created with the indicated feature enabled or disabled, respectively. Features can also be enabled by using the VkPhysicalDeviceFeatures2 structure.

If an application wishes to enable all features supported by a device, it can simply pass in the VkPhysicalDeviceFeatures structure that was previously returned by vkGetPhysicalDeviceFeatures. To disable an individual feature, the application can set the desired member to VK_FALSE in the same structure. Setting pEnabledFeatures to NULL and not including a VkPhysicalDeviceFeatures2 in the pNext member of VkDeviceCreateInfo is equivalent to setting all members of the structure to VK_FALSE.

Note

Some features, such as robustBufferAccess, may incur a run-time performance cost. Application writers should carefully consider the implications of enabling all supported features.

To query supported features defined by the core or extensions, call:

void vkGetPhysicalDeviceFeatures2KHR(
    VkPhysicalDevice                            physicalDevice,
    VkPhysicalDeviceFeatures2*                  pFeatures);
  • physicalDevice is the physical device from which to query the supported features.

  • pFeatures is a pointer to a VkPhysicalDeviceFeatures2 structure in which the physical device features are returned.

Each structure in pFeatures and its pNext chain contain members corresponding to fine-grained features. vkGetPhysicalDeviceFeatures2 writes each member to a boolean value indicating whether that feature is supported.

Valid Usage (Implicit)
  • physicalDevice must be a valid VkPhysicalDevice handle

  • pFeatures must be a valid pointer to a VkPhysicalDeviceFeatures2 structure

The VkPhysicalDeviceFeatures2 structure is defined as:

typedef struct VkPhysicalDeviceFeatures2 {
    VkStructureType             sType;
    void*                       pNext;
    VkPhysicalDeviceFeatures    features;
} VkPhysicalDeviceFeatures2;

or the equivalent

typedef VkPhysicalDeviceFeatures2 VkPhysicalDeviceFeatures2KHR;

The VkPhysicalDeviceFeatures2 structure is defined as:

  • sType is the type of this structure.

  • pNext is NULL or a pointer to an extension-specific structure.

  • features is a structure of type VkPhysicalDeviceFeatures describing the fine-grained features of the Vulkan 1.0 API.

The pNext chain of this structure is used to extend the structure with features defined by extensions. This structure can be used in vkGetPhysicalDeviceFeatures2 or can be in the pNext chain of a VkDeviceCreateInfo structure, in which case it controls which features are enabled in the device in lieu of pEnabledFeatures.

Valid Usage (Implicit)
  • sType must be VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2

The VkPhysicalDeviceFeatures structure is defined as:

typedef struct VkPhysicalDeviceFeatures {
    VkBool32    robustBufferAccess;
    VkBool32    fullDrawIndexUint32;
    VkBool32    imageCubeArray;
    VkBool32    independentBlend;
    VkBool32    geometryShader;
    VkBool32    tessellationShader;
    VkBool32    sampleRateShading;
    VkBool32    dualSrcBlend;
    VkBool32    logicOp;
    VkBool32    multiDrawIndirect;
    VkBool32    drawIndirectFirstInstance;
    VkBool32    depthClamp;
    VkBool32    depthBiasClamp;
    VkBool32    fillModeNonSolid;
    VkBool32    depthBounds;
    VkBool32    wideLines;
    VkBool32    largePoints;
    VkBool32    alphaToOne;
    VkBool32    multiViewport;
    VkBool32    samplerAnisotropy;
    VkBool32    textureCompressionETC2;
    VkBool32    textureCompressionASTC_LDR;
    VkBool32    textureCompressionBC;
    VkBool32    occlusionQueryPrecise;
    VkBool32    pipelineStatisticsQuery;
    VkBool32    vertexPipelineStoresAndAtomics;
    VkBool32    fragmentStoresAndAtomics;
    VkBool32    shaderTessellationAndGeometryPointSize;
    VkBool32    shaderImageGatherExtended;
    VkBool32    shaderStorageImageExtendedFormats;
    VkBool32    shaderStorageImageMultisample;
    VkBool32    shaderStorageImageReadWithoutFormat;
    VkBool32    shaderStorageImageWriteWithoutFormat;
    VkBool32    shaderUniformBufferArrayDynamicIndexing;
    VkBool32    shaderSampledImageArrayDynamicIndexing;
    VkBool32    shaderStorageBufferArrayDynamicIndexing;
    VkBool32    shaderStorageImageArrayDynamicIndexing;
    VkBool32    shaderClipDistance;
    VkBool32    shaderCullDistance;
    VkBool32    shaderFloat64;
    VkBool32    shaderInt64;
    VkBool32    shaderInt16;
    VkBool32    shaderResourceResidency;
    VkBool32    shaderResourceMinLod;
    VkBool32    sparseBinding;
    VkBool32    sparseResidencyBuffer;
    VkBool32    sparseResidencyImage2D;
    VkBool32    sparseResidencyImage3D;
    VkBool32    sparseResidency2Samples;
    VkBool32    sparseResidency4Samples;
    VkBool32    sparseResidency8Samples;
    VkBool32    sparseResidency16Samples;
    VkBool32    sparseResidencyAliased;
    VkBool32    variableMultisampleRate;
    VkBool32    inheritedQueries;
} VkPhysicalDeviceFeatures;

The members of the VkPhysicalDeviceFeatures structure describe the following features:

  • robustBufferAccess specifies that accesses to buffers are bounds-checked against the range of the buffer descriptor (as determined by VkDescriptorBufferInfo::range, VkBufferViewCreateInfo::range, or the size of the buffer). Out of bounds accesses must not cause application termination, and the effects of shader loads, stores, and atomics must conform to an implementation-dependent behavior as described below.

    • A buffer access is considered to be out of bounds if any of the following are true:

      • The pointer was formed by OpImageTexelPointer and the coordinate is less than zero or greater than or equal to the number of whole elements in the bound range.

      • The pointer was not formed by OpImageTexelPointer and the object pointed to is not wholly contained within the bound range. This includes accesses performed via variable pointers where the buffer descriptor being accessed cannot be statically determined. Uninitialized pointers and pointers equal to OpConstantNull are treated as pointing to a zero-sized object, so all accesses through such pointers are considered to be out of bounds.

        Note

        If a SPIR-V OpLoad instruction loads a structure and the tail end of the structure is out of bounds, then all members of the structure are considered out of bounds even if the members at the end are not statically used.

      • If any buffer access in a given SPIR-V block is determined to be out of bounds, then any other access of the same type (load, store, or atomic) in the same SPIR-V block that accesses an address less than 16 bytes away from the out of bounds address may also be considered out of bounds.

    • Out-of-bounds buffer loads will return any of the following values:

      • Values from anywhere within the memory range(s) bound to the buffer (possibly including bytes of memory past the end of the buffer, up to the end of the bound range).

      • Zero values, or (0,0,0,x) vectors for vector reads where x is a valid value represented in the type of the vector components and may be any of:

        • 0, 1, or the maximum representable positive integer value, for signed or unsigned integer components

        • 0.0 or 1.0, for floating-point components

    • Out-of-bounds writes may modify values within the memory range(s) bound to the buffer, but must not modify any other memory.

    • Out-of-bounds atomics may modify values within the memory range(s) bound to the buffer, but must not modify any other memory, and return an undefined value.

    • Vertex input attributes are considered out of bounds if the offset of the attribute in the bound vertex buffer range plus the size of the attribute is greater than either:

      • vertexBufferRangeSize, if bindingStride == 0; or

      • (vertexBufferRangeSize - (vertexBufferRangeSize % bindingStride))

      where vertexBufferRangeSize is the byte size of the memory range bound to the vertex buffer binding and bindingStride is the byte stride of the corresponding vertex input binding. Further, if any vertex input attribute using a specific vertex input binding is out of bounds, then all vertex input attributes using that vertex input binding for that vertex shader invocation are considered out of bounds.

      • If a vertex input attribute is out of bounds, it will be assigned one of the following values:

        • Values from anywhere within the memory range(s) bound to the buffer, converted according to the format of the attribute.

        • Zero values, format converted according to the format of the attribute.

        • Zero values, or (0,0,0,x) vectors, as described above.

    • If robustBufferAccess is not enabled, applications must not perform out of bounds accesses.

  • fullDrawIndexUint32 specifies the full 32-bit range of indices is supported for indexed draw calls when using a VkIndexType of VK_INDEX_TYPE_UINT32. maxDrawIndexedIndexValue is the maximum index value that may be used (aside from the primitive restart index, which is always 232-1 when the VkIndexType is VK_INDEX_TYPE_UINT32). If this feature is supported, maxDrawIndexedIndexValue must be 232-1; otherwise it must be no smaller than 224-1. See maxDrawIndexedIndexValue.

  • imageCubeArray specifies whether image views with a VkImageViewType of VK_IMAGE_VIEW_TYPE_CUBE_ARRAY can be created, and that the corresponding SampledCubeArray and ImageCubeArray SPIR-V capabilities can be used in shader code.

  • independentBlend specifies whether the VkPipelineColorBlendAttachmentState settings are controlled independently per-attachment. If this feature is not enabled, the VkPipelineColorBlendAttachmentState settings for all color attachments must be identical. Otherwise, a different VkPipelineColorBlendAttachmentState can be provided for each bound color attachment.

  • geometryShader specifies whether geometry shaders are supported. If this feature is not enabled, the VK_SHADER_STAGE_GEOMETRY_BIT and VK_PIPELINE_STAGE_GEOMETRY_SHADER_BIT enum values must not be used. This also specifies whether shader modules can declare the Geometry capability.

  • tessellationShader specifies whether tessellation control and evaluation shaders are supported. If this feature is not enabled, the VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT, VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT, VK_PIPELINE_STAGE_TESSELLATION_CONTROL_SHADER_BIT, VK_PIPELINE_STAGE_TESSELLATION_EVALUATION_SHADER_BIT, and VK_STRUCTURE_TYPE_PIPELINE_TESSELLATION_STATE_CREATE_INFO enum values must not be used. This also specifies whether shader modules can declare the Tessellation capability.

  • sampleRateShading specifies whether Sample Shading and multisample interpolation are supported. If this feature is not enabled, the sampleShadingEnable member of the VkPipelineMultisampleStateCreateInfo structure must be set to VK_FALSE and the minSampleShading member is ignored. This also specifies whether shader modules can declare the SampleRateShading capability.

  • dualSrcBlend specifies whether blend operations which take two sources are supported. If this feature is not enabled, the VK_BLEND_FACTOR_SRC1_COLOR, VK_BLEND_FACTOR_ONE_MINUS_SRC1_COLOR, VK_BLEND_FACTOR_SRC1_ALPHA, and VK_BLEND_FACTOR_ONE_MINUS_SRC1_ALPHA enum values must not be used as source or destination blending factors. See Dual-Source Blending.

  • logicOp specifies whether logic operations are supported. If this feature is not enabled, the logicOpEnable member of the VkPipelineColorBlendStateCreateInfo structure must be set to VK_FALSE, and the logicOp member is ignored.

  • multiDrawIndirect specifies whether multiple draw indirect is supported. If this feature is not enabled, the drawCount parameter to the vkCmdDrawIndirect and vkCmdDrawIndexedIndirect commands must be 0 or 1. The maxDrawIndirectCount member of the VkPhysicalDeviceLimits structure must also be 1 if this feature is not supported. See maxDrawIndirectCount.

  • drawIndirectFirstInstance specifies whether indirect draw calls support the firstInstance parameter. If this feature is not enabled, the firstInstance member of all VkDrawIndirectCommand and VkDrawIndexedIndirectCommand structures that are provided to the vkCmdDrawIndirect and vkCmdDrawIndexedIndirect commands must be 0.

  • depthClamp specifies whether depth clamping is supported. If this feature is not enabled, the depthClampEnable member of the VkPipelineRasterizationStateCreateInfo structure must be set to VK_FALSE. Otherwise, setting depthClampEnable to VK_TRUE will enable depth clamping.

  • depthBiasClamp specifies whether depth bias clamping is supported. If this feature is not enabled, the depthBiasClamp member of the VkPipelineRasterizationStateCreateInfo structure must be set to 0.0 unless the VK_DYNAMIC_STATE_DEPTH_BIAS dynamic state is enabled, and the depthBiasClamp parameter to vkCmdSetDepthBias must be set to 0.0.

  • fillModeNonSolid specifies whether point and wireframe fill modes are supported. If this feature is not enabled, the VK_POLYGON_MODE_POINT and VK_POLYGON_MODE_LINE enum values must not be used.

  • depthBounds specifies whether depth bounds tests are supported. If this feature is not enabled, the depthBoundsTestEnable member of the VkPipelineDepthStencilStateCreateInfo structure must be set to VK_FALSE. When depthBoundsTestEnable is set to VK_FALSE, the minDepthBounds and maxDepthBounds members of the VkPipelineDepthStencilStateCreateInfo structure are ignored.

  • wideLines specifies whether lines with width other than 1.0 are supported. If this feature is not enabled, the lineWidth member of the VkPipelineRasterizationStateCreateInfo structure must be set to 1.0 unless the VK_DYNAMIC_STATE_LINE_WIDTH dynamic state is enabled, and the lineWidth parameter to vkCmdSetLineWidth must be set to 1.0. When this feature is supported, the range and granularity of supported line widths are indicated by the lineWidthRange and lineWidthGranularity members of the VkPhysicalDeviceLimits structure, respectively.

  • largePoints specifies whether points with size greater than 1.0 are supported. If this feature is not enabled, only a point size of 1.0 written by a shader is supported. The range and granularity of supported point sizes are indicated by the pointSizeRange and pointSizeGranularity members of the VkPhysicalDeviceLimits structure, respectively.

  • alphaToOne specifies whether the implementation is able to replace the alpha value of the color fragment output from the fragment shader with the maximum representable alpha value for fixed-point colors or 1.0 for floating-point colors. If this feature is not enabled, then the alphaToOneEnable member of the VkPipelineMultisampleStateCreateInfo structure must be set to VK_FALSE. Otherwise setting alphaToOneEnable to VK_TRUE will enable alpha-to-one behavior.

  • multiViewport specifies whether more than one viewport is supported. If this feature is not enabled:

    • The viewportCount and scissorCount members of the VkPipelineViewportStateCreateInfo structure must be set to 1.

    • The firstViewport and viewportCount parameters to the vkCmdSetViewport command must be set to 0 and 1, respectively.

    • The firstScissor and scissorCount parameters to the vkCmdSetScissor command must be set to 0 and 1, respectively.

  • samplerAnisotropy specifies whether anisotropic filtering is supported. If this feature is not enabled, the anisotropyEnable member of the VkSamplerCreateInfo structure must be VK_FALSE.

  • textureCompressionETC2 specifies whether all of the ETC2 and EAC compressed texture formats are supported. If this feature is enabled, then the VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT, VK_FORMAT_FEATURE_BLIT_SRC_BIT and VK_FORMAT_FEATURE_SAMPLED_IMAGE_FILTER_LINEAR_BIT features must be supported in optimalTilingFeatures for the following formats:

    • VK_FORMAT_ETC2_R8G8B8_UNORM_BLOCK

    • VK_FORMAT_ETC2_R8G8B8_SRGB_BLOCK

    • VK_FORMAT_ETC2_R8G8B8A1_UNORM_BLOCK

    • VK_FORMAT_ETC2_R8G8B8A1_SRGB_BLOCK

    • VK_FORMAT_ETC2_R8G8B8A8_UNORM_BLOCK

    • VK_FORMAT_ETC2_R8G8B8A8_SRGB_BLOCK

    • VK_FORMAT_EAC_R11_UNORM_BLOCK

    • VK_FORMAT_EAC_R11_SNORM_BLOCK

    • VK_FORMAT_EAC_R11G11_UNORM_BLOCK

    • VK_FORMAT_EAC_R11G11_SNORM_BLOCK

    To query for additional properties, or if the feature is not enabled, vkGetPhysicalDeviceFormatProperties and vkGetPhysicalDeviceImageFormatProperties can be used to check for supported properties of individual formats as normal.

  • textureCompressionASTC_LDR specifies whether all of the ASTC LDR compressed texture formats are supported. If this feature is enabled, then the VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT, VK_FORMAT_FEATURE_BLIT_SRC_BIT and VK_FORMAT_FEATURE_SAMPLED_IMAGE_FILTER_LINEAR_BIT features must be supported in optimalTilingFeatures for the following formats:

    • VK_FORMAT_ASTC_4x4_UNORM_BLOCK

    • VK_FORMAT_ASTC_4x4_SRGB_BLOCK

    • VK_FORMAT_ASTC_5x4_UNORM_BLOCK

    • VK_FORMAT_ASTC_5x4_SRGB_BLOCK

    • VK_FORMAT_ASTC_5x5_UNORM_BLOCK

    • VK_FORMAT_ASTC_5x5_SRGB_BLOCK

    • VK_FORMAT_ASTC_6x5_UNORM_BLOCK

    • VK_FORMAT_ASTC_6x5_SRGB_BLOCK

    • VK_FORMAT_ASTC_6x6_UNORM_BLOCK

    • VK_FORMAT_ASTC_6x6_SRGB_BLOCK

    • VK_FORMAT_ASTC_8x5_UNORM_BLOCK

    • VK_FORMAT_ASTC_8x5_SRGB_BLOCK

    • VK_FORMAT_ASTC_8x6_UNORM_BLOCK

    • VK_FORMAT_ASTC_8x6_SRGB_BLOCK

    • VK_FORMAT_ASTC_8x8_UNORM_BLOCK

    • VK_FORMAT_ASTC_8x8_SRGB_BLOCK

    • VK_FORMAT_ASTC_10x5_UNORM_BLOCK

    • VK_FORMAT_ASTC_10x5_SRGB_BLOCK

    • VK_FORMAT_ASTC_10x6_UNORM_BLOCK

    • VK_FORMAT_ASTC_10x6_SRGB_BLOCK

    • VK_FORMAT_ASTC_10x8_UNORM_BLOCK

    • VK_FORMAT_ASTC_10x8_SRGB_BLOCK

    • VK_FORMAT_ASTC_10x10_UNORM_BLOCK

    • VK_FORMAT_ASTC_10x10_SRGB_BLOCK

    • VK_FORMAT_ASTC_12x10_UNORM_BLOCK

    • VK_FORMAT_ASTC_12x10_SRGB_BLOCK

    • VK_FORMAT_ASTC_12x12_UNORM_BLOCK

    • VK_FORMAT_ASTC_12x12_SRGB_BLOCK

    To query for additional properties, or if the feature is not enabled, vkGetPhysicalDeviceFormatProperties and vkGetPhysicalDeviceImageFormatProperties can be used to check for supported properties of individual formats as normal.

  • textureCompressionBC specifies whether all of the BC compressed texture formats are supported. If this feature is enabled, then the VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT, VK_FORMAT_FEATURE_BLIT_SRC_BIT and VK_FORMAT_FEATURE_SAMPLED_IMAGE_FILTER_LINEAR_BIT features must be supported in optimalTilingFeatures for the following formats:

    • VK_FORMAT_BC1_RGB_UNORM_BLOCK

    • VK_FORMAT_BC1_RGB_SRGB_BLOCK

    • VK_FORMAT_BC1_RGBA_UNORM_BLOCK

    • VK_FORMAT_BC1_RGBA_SRGB_BLOCK

    • VK_FORMAT_BC2_UNORM_BLOCK

    • VK_FORMAT_BC2_SRGB_BLOCK

    • VK_FORMAT_BC3_UNORM_BLOCK

    • VK_FORMAT_BC3_SRGB_BLOCK

    • VK_FORMAT_BC4_UNORM_BLOCK

    • VK_FORMAT_BC4_SNORM_BLOCK

    • VK_FORMAT_BC5_UNORM_BLOCK

    • VK_FORMAT_BC5_SNORM_BLOCK

    • VK_FORMAT_BC6H_UFLOAT_BLOCK

    • VK_FORMAT_BC6H_SFLOAT_BLOCK

    • VK_FORMAT_BC7_UNORM_BLOCK

    • VK_FORMAT_BC7_SRGB_BLOCK

    To query for additional properties, or if the feature is not enabled, vkGetPhysicalDeviceFormatProperties and vkGetPhysicalDeviceImageFormatProperties can be used to check for supported properties of individual formats as normal.

  • occlusionQueryPrecise specifies whether occlusion queries returning actual sample counts are supported. Occlusion queries are created in a VkQueryPool by specifying the queryType of VK_QUERY_TYPE_OCCLUSION in the VkQueryPoolCreateInfo structure which is passed to vkCreateQueryPool. If this feature is enabled, queries of this type can enable VK_QUERY_CONTROL_PRECISE_BIT in the flags parameter to vkCmdBeginQuery. If this feature is not supported, the implementation supports only boolean occlusion queries. When any samples are passed, boolean queries will return a non-zero result value, otherwise a result value of zero is returned. When this feature is enabled and VK_QUERY_CONTROL_PRECISE_BIT is set, occlusion queries will report the actual number of samples passed.

  • pipelineStatisticsQuery specifies whether the pipeline statistics queries are supported. If this feature is not enabled, queries of type VK_QUERY_TYPE_PIPELINE_STATISTICS cannot be created, and none of the VkQueryPipelineStatisticFlagBits bits can be set in the pipelineStatistics member of the VkQueryPoolCreateInfo structure.

  • vertexPipelineStoresAndAtomics specifies whether storage buffers and images support stores and atomic operations in the vertex, tessellation, and geometry shader stages. If this feature is not enabled, all storage image, storage texel buffers, and storage buffer variables used by these stages in shader modules must be decorated with the NonWritable decoration (or the readonly memory qualifier in GLSL).

  • fragmentStoresAndAtomics specifies whether storage buffers and images support stores and atomic operations in the fragment shader stage. If this feature is not enabled, all storage image, storage texel buffers, and storage buffer variables used by the fragment stage in shader modules must be decorated with the NonWritable decoration (or the readonly memory qualifier in GLSL).

  • shaderTessellationAndGeometryPointSize specifies whether the PointSize built-in decoration is available in the tessellation control, tessellation evaluation, and geometry shader stages. If this feature is not enabled, members decorated with the PointSize built-in decoration must not be read from or written to and all points written from a tessellation or geometry shader will have a size of 1.0. This also specifies whether shader modules can declare the TessellationPointSize capability for tessellation control and evaluation shaders, or if the shader modules can declare the GeometryPointSize capability for geometry shaders. An implementation supporting this feature must also support one or both of the tessellationShader or geometryShader features.

  • shaderImageGatherExtended specifies whether the extended set of image gather instructions are available in shader code. If this feature is not enabled, the OpImage*Gather instructions do not support the Offset and ConstOffsets operands. This also specifies whether shader modules can declare the ImageGatherExtended capability.

  • shaderStorageImageExtendedFormats specifies whether all the extended storage image formats are available in shader code. If this feature is enabled then the VK_FORMAT_FEATURE_STORAGE_IMAGE_BIT feature must be supported in optimalTilingFeatures for all of the extended formats. To query for additional properties, or if the feature is not enabled, vkGetPhysicalDeviceFormatProperties and vkGetPhysicalDeviceImageFormatProperties can be used to check for supported properties of individual formats as normal.

  • shaderStorageImageMultisample specifies whether multisampled storage images are supported. If this feature is not enabled, images that are created with a usage that includes VK_IMAGE_USAGE_STORAGE_BIT must be created with samples equal to VK_SAMPLE_COUNT_1_BIT. This also specifies whether shader modules can declare the StorageImageMultisample capability.

  • shaderStorageImageReadWithoutFormat specifies whether storage images require a format qualifier to be specified when reading from storage images. If this feature is not enabled, the OpImageRead instruction must not have an OpTypeImage of Unknown. This also specifies whether shader modules can declare the StorageImageReadWithoutFormat capability.

  • shaderStorageImageWriteWithoutFormat specifies whether storage images require a format qualifier to be specified when writing to storage images. If this feature is not enabled, the OpImageWrite instruction must not have an OpTypeImage of Unknown. This also specifies whether shader modules can declare the StorageImageWriteWithoutFormat capability.

  • shaderUniformBufferArrayDynamicIndexing specifies whether arrays of uniform buffers can be indexed by dynamically uniform integer expressions in shader code. If this feature is not enabled, resources with a descriptor type of VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER or VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC must be indexed only by constant integral expressions when aggregated into arrays in shader code. This also specifies whether shader modules can declare the UniformBufferArrayDynamicIndexing capability.

  • shaderSampledImageArrayDynamicIndexing specifies whether arrays of samplers or sampled images can be indexed by dynamically uniform integer expressions in shader code. If this feature is not enabled, resources with a descriptor type of VK_DESCRIPTOR_TYPE_SAMPLER, VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, or VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE must be indexed only by constant integral expressions when aggregated into arrays in shader code. This also specifies whether shader modules can declare the SampledImageArrayDynamicIndexing capability.

  • shaderStorageBufferArrayDynamicIndexing specifies whether arrays of storage buffers can be indexed by dynamically uniform integer expressions in shader code. If this feature is not enabled, resources with a descriptor type of VK_DESCRIPTOR_TYPE_STORAGE_BUFFER or VK_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC must be indexed only by constant integral expressions when aggregated into arrays in shader code. This also specifies whether shader modules can declare the StorageBufferArrayDynamicIndexing capability.

  • shaderStorageImageArrayDynamicIndexing specifies whether arrays of storage images can be indexed by dynamically uniform integer expressions in shader code. If this feature is not enabled, resources with a descriptor type of VK_DESCRIPTOR_TYPE_STORAGE_IMAGE must be indexed only by constant integral expressions when aggregated into arrays in shader code. This also specifies whether shader modules can declare the StorageImageArrayDynamicIndexing capability.

  • shaderClipDistance specifies whether clip distances are supported in shader code. If this feature is not enabled, any members decorated with the ClipDistance built-in decoration must not be read from or written to in shader modules. This also specifies whether shader modules can declare the ClipDistance capability.

  • shaderCullDistance specifies whether cull distances are supported in shader code. If this feature is not enabled, any members decorated with the CullDistance built-in decoration must not be read from or written to in shader modules. This also specifies whether shader modules can declare the CullDistance capability.

  • shaderFloat64 specifies whether 64-bit floats (doubles) are supported in shader code. If this feature is not enabled, 64-bit floating-point types must not be used in shader code. This also specifies whether shader modules can declare the Float64 capability.

  • shaderInt64 specifies whether 64-bit integers (signed and unsigned) are supported in shader code. If this feature is not enabled, 64-bit integer types must not be used in shader code. This also specifies whether shader modules can declare the Int64 capability.

  • shaderInt16 specifies whether 16-bit integers (signed and unsigned) are supported in shader code. If this feature is not enabled, 16-bit integer types must not be used in shader code. This also specifies whether shader modules can declare the Int16 capability.

  • shaderResourceResidency specifies whether image operations that return resource residency information are supported in shader code. If this feature is not enabled, the OpImageSparse* instructions must not be used in shader code. This also specifies whether shader modules can declare the SparseResidency capability. The feature requires at least one of the sparseResidency* features to be supported.

  • shaderResourceMinLod specifies whether image operations that specify the minimum resource LOD are supported in shader code. If this feature is not enabled, the MinLod image operand must not be used in shader code. This also specifies whether shader modules can declare the MinLod capability.

  • sparseBinding specifies whether resource memory can be managed at opaque sparse block level instead of at the object level. If this feature is not enabled, resource memory must be bound only on a per-object basis using the vkBindBufferMemory and vkBindImageMemory commands. In this case, buffers and images must not be created with VK_BUFFER_CREATE_SPARSE_BINDING_BIT and VK_IMAGE_CREATE_SPARSE_BINDING_BIT set in the flags member of the VkBufferCreateInfo and VkImageCreateInfo structures, respectively. Otherwise resource memory can be managed as described in Sparse Resource Features.

  • sparseResidencyBuffer specifies whether the device can access partially resident buffers. If this feature is not enabled, buffers must not be created with VK_BUFFER_CREATE_SPARSE_RESIDENCY_BIT set in the flags member of the VkBufferCreateInfo structure.

  • sparseResidencyImage2D specifies whether the device can access partially resident 2D images with 1 sample per pixel. If this feature is not enabled, images with an imageType of VK_IMAGE_TYPE_2D and samples set to VK_SAMPLE_COUNT_1_BIT must not be created with VK_IMAGE_CREATE_SPARSE_RESIDENCY_BIT set in the flags member of the VkImageCreateInfo structure.

  • sparseResidencyImage3D specifies whether the device can access partially resident 3D images. If this feature is not enabled, images with an imageType of VK_IMAGE_TYPE_3D must not be created with VK_IMAGE_CREATE_SPARSE_RESIDENCY_BIT set in the flags member of the VkImageCreateInfo structure.

  • sparseResidency2Samples specifies whether the physical device can access partially resident 2D images with 2 samples per pixel. If this feature is not enabled, images with an imageType of VK_IMAGE_TYPE_2D and samples set to VK_SAMPLE_COUNT_2_BIT must not be created with VK_IMAGE_CREATE_SPARSE_RESIDENCY_BIT set in the flags member of the VkImageCreateInfo structure.

  • sparseResidency4Samples specifies whether the physical device can access partially resident 2D images with 4 samples per pixel. If this feature is not enabled, images with an imageType of VK_IMAGE_TYPE_2D and samples set to VK_SAMPLE_COUNT_4_BIT must not be created with VK_IMAGE_CREATE_SPARSE_RESIDENCY_BIT set in the flags member of the VkImageCreateInfo structure.

  • sparseResidency8Samples specifies whether the physical device can access partially resident 2D images with 8 samples per pixel. If this feature is not enabled, images with an imageType of VK_IMAGE_TYPE_2D and samples set to VK_SAMPLE_COUNT_8_BIT must not be created with VK_IMAGE_CREATE_SPARSE_RESIDENCY_BIT set in the flags member of the VkImageCreateInfo structure.

  • sparseResidency16Samples specifies whether the physical device can access partially resident 2D images with 16 samples per pixel. If this feature is not enabled, images with an imageType of VK_IMAGE_TYPE_2D and samples set to VK_SAMPLE_COUNT_16_BIT must not be created with VK_IMAGE_CREATE_SPARSE_RESIDENCY_BIT set in the flags member of the VkImageCreateInfo structure.

  • sparseResidencyAliased specifies whether the physical device can correctly access data aliased into multiple locations. If this feature is not enabled, the VK_BUFFER_CREATE_SPARSE_ALIASED_BIT and VK_IMAGE_CREATE_SPARSE_ALIASED_BIT enum values must not be used in flags members of the VkBufferCreateInfo and VkImageCreateInfo structures, respectively.

  • variableMultisampleRate specifies whether all pipelines that will be bound to a command buffer during a subpass with no attachments must have the same value for VkPipelineMultisampleStateCreateInfo::rasterizationSamples. If set to VK_TRUE, the implementation supports variable multisample rates in a subpass with no attachments. If set to VK_FALSE, then all pipelines bound in such a subpass must have the same multisample rate. This has no effect in situations where a subpass uses any attachments.

  • inheritedQueries specifies whether a secondary command buffer may be executed while a query is active.

The VkPhysicalDeviceVariablePointersFeatures structure is defined as:

typedef struct VkPhysicalDeviceVariablePointersFeatures {
    VkStructureType    sType;
    void*              pNext;
    VkBool32           variablePointersStorageBuffer;
    VkBool32           variablePointers;
} VkPhysicalDeviceVariablePointersFeatures;

or the equivalent

typedef VkPhysicalDeviceVariablePointersFeatures VkPhysicalDeviceVariablePointersFeaturesKHR;

The members of the VkPhysicalDeviceVariablePointersFeatures structure describe the following features:

  • variablePointersStorageBuffer specifies whether the implementation supports the SPIR-V VariablePointersStorageBuffer capability. When this feature is not enabled, shader modules must not declare the SPV_KHR_variable_pointers extension or the VariablePointersStorageBuffer capability.

  • variablePointers specifies whether the implementation supports the SPIR-V VariablePointers capability. When this feature is not enabled, shader modules must not declare the VariablePointers capability.

If the VkPhysicalDeviceVariablePointersFeatures structure is included in the pNext chain of VkPhysicalDeviceFeatures2, it is filled with values indicating whether each feature is supported. VkPhysicalDeviceVariablePointersFeatures can also be used in the pNext chain of VkDeviceCreateInfo to enable the features.

Valid Usage
  • If variablePointers is enabled then variablePointersStorageBuffer must also be enabled.

Valid Usage (Implicit)
  • sType must be VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VARIABLE_POINTERS_FEATURES

The VkPhysicalDeviceMultiviewFeatures structure is defined as:

typedef struct VkPhysicalDeviceMultiviewFeatures {
    VkStructureType    sType;
    void*              pNext;
    VkBool32           multiview;
    VkBool32           multiviewGeometryShader;
    VkBool32           multiviewTessellationShader;
} VkPhysicalDeviceMultiviewFeatures;

or the equivalent

typedef VkPhysicalDeviceMultiviewFeatures VkPhysicalDeviceMultiviewFeaturesKHR;

The members of the VkPhysicalDeviceMultiviewFeatures structure describe the following features:

  • multiview specifies whether the implementation supports multiview rendering within a render pass. If this feature is not enabled, the view mask of each subpass must always be zero.

  • multiviewGeometryShader specifies whether the implementation supports multiview rendering within a render pass, with geometry shaders. If this feature is not enabled, then a pipeline compiled against a subpass with a non-zero view mask must not include a geometry shader.

  • multiviewTessellationShader specifies whether the implementation supports multiview rendering within a render pass, with tessellation shaders. If this feature is not enabled, then a pipeline compiled against a subpass with a non-zero view mask must not include any tessellation shaders.

If the VkPhysicalDeviceMultiviewFeatures structure is included in the pNext chain of VkPhysicalDeviceFeatures2, it is filled with values indicating whether each feature is supported. VkPhysicalDeviceMultiviewFeatures can also be used in the pNext chain of VkDeviceCreateInfo to enable the features.

Valid Usage
  • If multiviewGeometryShader is enabled then multiview must also be enabled.

  • If multiviewTessellationShader is enabled then multiview must also be enabled.

Valid Usage (Implicit)
  • sType must be VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MULTIVIEW_FEATURES

To query 64-bit atomic support for signed and unsigned integers call vkGetPhysicalDeviceFeatures2 with a VkPhysicalDeviceShaderAtomicInt64FeaturesKHR structure included in the pNext chain of its pFeatures parameter.

The VkPhysicalDeviceShaderAtomicInt64FeaturesKHR structure is defined as:

typedef struct VkPhysicalDeviceShaderAtomicInt64FeaturesKHR {
    VkStructureType    sType;
    void*              pNext;
    VkBool32           shaderBufferInt64Atomics;
    VkBool32           shaderSharedInt64Atomics;
} VkPhysicalDeviceShaderAtomicInt64FeaturesKHR;
  • sType is the type of this structure.

  • pNext is NULL or a pointer to an extension-specific structure.

  • shaderBufferInt64Atomics indicates whether shaders can support 64-bit unsigned and signed integer atomic operations on buffers.

  • shaderSharedInt64Atomics indicates whether shaders can support 64-bit unsigned and signed integer atomic operations on shared memory.

Valid Usage (Implicit)
  • sType must be VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_ATOMIC_INT64_FEATURES_KHR

To query 8-bit storage features additionally supported call vkGetPhysicalDeviceFeatures2 with a VkPhysicalDevice8BitStorageFeaturesKHR structure included in the pNext chain of its pFeatures parameter. The VkPhysicalDevice8BitStorageFeaturesKHR structure can also be in the pNext chain of a VkDeviceCreateInfo structure, in which case it controls which additional features are enabled in the device.

The VkPhysicalDevice8BitStorageFeaturesKHR structure is defined as:

typedef struct VkPhysicalDevice8BitStorageFeaturesKHR {
    VkStructureType    sType;
    void*              pNext;
    VkBool32           storageBuffer8BitAccess;
    VkBool32           uniformAndStorageBuffer8BitAccess;
    VkBool32           storagePushConstant8;
} VkPhysicalDevice8BitStorageFeaturesKHR;
  • sType is the type of this structure.

  • pNext is NULL or a pointer to an extension-specific structure.

  • storageBuffer8BitAccess indicates whether objects in the StorageBuffer storage class with the Block decoration can have 8-bit integer members. If this feature is not enabled, 8-bit integer members must not be used in such objects. This also indicates whether shader modules can declare the StorageBuffer8BitAccess capability.

  • uniformAndStorageBuffer8BitAccess indicates whether objects in the Uniform storage class with the Block decoration and in the StorageBuffer storage class with the same decoration can have 8-bit integer members. If this feature is not enabled, 8-bit integer members must not be used in such objects. This also indicates whether shader modules can declare the UniformAndStorageBuffer8BitAccess capability.

  • storagePushConstant8 indicates whether objects in the PushConstant storage class can have 8-bit integer members. If this feature is not enabled, 8-bit integer members must not be used in such objects. This also indicates whether shader modules can declare the StoragePushConstant8 capability.

Valid Usage (Implicit)
  • sType must be VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_8BIT_STORAGE_FEATURES_KHR

To query 16-bit storage features additionally supported call vkGetPhysicalDeviceFeatures2 with a VkPhysicalDevice16BitStorageFeatures structure included in the pNext chain of its pFeatures parameter. The VkPhysicalDevice16BitStorageFeatures structure can also be in the pNext chain of a VkDeviceCreateInfo structure, in which case it controls which additional features are enabled in the device.

The VkPhysicalDevice16BitStorageFeatures structure is defined as:

typedef struct VkPhysicalDevice16BitStorageFeatures {
    VkStructureType    sType;
    void*              pNext;
    VkBool32           storageBuffer16BitAccess;
    VkBool32           uniformAndStorageBuffer16BitAccess;
    VkBool32           storagePushConstant16;
    VkBool32           storageInputOutput16;
} VkPhysicalDevice16BitStorageFeatures;

or the equivalent

typedef VkPhysicalDevice16BitStorageFeatures VkPhysicalDevice16BitStorageFeaturesKHR;
  • sType is the type of this structure.

  • pNext is NULL or a pointer to an extension-specific structure.

  • storageBuffer16BitAccess specifies whether objects in the StorageBuffer storage class with the Block decoration can have 16-bit integer and 16-bit floating-point members. If this feature is not enabled, 16-bit integer or 16-bit floating-point members must not be used in such objects. This also specifies whether shader modules can declare the StorageBuffer16BitAccess capability.

  • uniformAndStorageBuffer16BitAccess specifies whether objects in the Uniform storage class with the Block decoration and in the StorageBuffer storage class with the same decoration can have 16-bit integer and 16-bit floating-point members. If this feature is not enabled, 16-bit integer or 16-bit floating-point members must not be used in such objects. This also specifies whether shader modules can declare the UniformAndStorageBuffer16BitAccess capability.

  • storagePushConstant16 specifies whether objects in the PushConstant storage class can have 16-bit integer and 16-bit floating-point members. If this feature is not enabled, 16-bit integer or floating-point members must not be used in such objects. This also specifies whether shader modules can declare the StoragePushConstant16 capability.

  • storageInputOutput16 specifies whether objects in the Input and Output storage classes can have 16-bit integer and 16-bit floating-point members. If this feature is not enabled, 16-bit integer or 16-bit floating-point members must not be used in such objects. This also specifies whether shader modules can declare the StorageInputOutput16 capability.

Valid Usage (Implicit)
  • sType must be VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_16BIT_STORAGE_FEATURES

To query features additionally supported by the VK_KHR_shader_float16_int8 extension, call vkGetPhysicalDeviceFeatures2KHR with a VkPhysicalDeviceFloat16Int8FeaturesKHR structure in the pNext chain. The VkPhysicalDeviceFloat16Int8FeaturesKHR structure can also be in the pNext chain of a VkDeviceCreateInfo structure, in which case it controls which additional features are enabled in the device.

The VkPhysicalDeviceFloat16Int8FeaturesKHR structure is defined as:

typedef struct VkPhysicalDeviceFloat16Int8FeaturesKHR {
    VkStructureType    sType;
    void*              pNext;
    VkBool32           shaderFloat16;
    VkBool32           shaderInt8;
} VkPhysicalDeviceFloat16Int8FeaturesKHR;
  • sType is the type of this structure.

  • pNext is NULL or a pointer to an extension-specific structure.

  • shaderFloat16 indicates whether 16-bit floats (halfs) are supported in shader code. This also indicates whether shader modules can declare the Float16 capability.

  • shaderInt8 indicates whether 8-bit integers (signed and unsigned) are supported in shader code. This also indicates whether shader modules can declare the Int8 capability.

Valid Usage (Implicit)
  • sType must be VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FLOAT16_INT8_FEATURES_KHR

The VkPhysicalDeviceSamplerYcbcrConversionFeatures structure is defined as:

typedef struct VkPhysicalDeviceSamplerYcbcrConversionFeatures {
    VkStructureType    sType;
    void*              pNext;
    VkBool32           samplerYcbcrConversion;
} VkPhysicalDeviceSamplerYcbcrConversionFeatures;

or the equivalent

typedef VkPhysicalDeviceSamplerYcbcrConversionFeatures VkPhysicalDeviceSamplerYcbcrConversionFeaturesKHR;

The members of the VkPhysicalDeviceSamplerYcbcrConversionFeatures structure describe the following feature:

  • samplerYcbcrConversion specifies whether the implementation supports sampler Y’CBCR conversion. If samplerYcbcrConversion is VK_FALSE, sampler Y’CBCR conversion is not supported, and samplers using sampler Y’CBCR conversion must not be used.

Valid Usage (Implicit)
  • sType must be VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SAMPLER_YCBCR_CONVERSION_FEATURES

To query memory model features additionally supported call vkGetPhysicalDeviceFeatures2 with a VkPhysicalDeviceVulkanMemoryModelFeaturesKHR structure included in the pNext chain of its pFeatures parameter. The VkPhysicalDeviceVulkanMemoryModelFeaturesKHR structure can also be in the pNext chain of a VkDeviceCreateInfo structure, in which case it controls which additional features are enabled in the device.

The VkPhysicalDeviceVulkanMemoryModelFeaturesKHR structure is defined as:

typedef struct VkPhysicalDeviceVulkanMemoryModelFeaturesKHR {
    VkStructureType    sType;
    void*              pNext;
    VkBool32           vulkanMemoryModel;
    VkBool32           vulkanMemoryModelDeviceScope;
    VkBool32           vulkanMemoryModelAvailabilityVisibilityChains;
} VkPhysicalDeviceVulkanMemoryModelFeaturesKHR;
  • sType is the type of this structure.

  • pNext is NULL or a pointer to an extension-specific structure.

  • vulkanMemoryModel indicates whether the Vulkan Memory Model is supported, as defined in Vulkan Memory Model. This also indicates whether shader modules can declare the VulkanMemoryModelKHR capability.

  • vulkanMemoryModelDeviceScope indicates whether the Vulkan Memory Model can use Device scope synchronization. This also indicates whether shader modules can declare the VulkanMemoryModelDeviceScopeKHR capability.

  • vulkanMemoryModelAvailabilityVisibilityChains indicates whether the Vulkan Memory Model can use availability and visibility chains with more than one element.

Valid Usage (Implicit)
  • sType must be VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VULKAN_MEMORY_MODEL_FEATURES_KHR

31.1. Feature Requirements

All Vulkan graphics implementations must support the following features:

All other features defined in the Specification are optional.