Vulkan^® 1.1.130 - A Specification (with all registered Vulkan extensions)

The VK_KHR_8bit_storage extension allows use of 8-bit types in uniform and storage buffers, and push constant blocks. This extension introduces several new optional features which map to SPIR-V capabilities and allow access to 8-bit data in Block-decorated objects in the Uniform and the StorageBuffer storage classes, and objects in the PushConstant storage class.

The StorageBuffer8BitAccess capability must be supported by all implementations of this extension. The other capabilities are optional.

The VK_KHR_android_surface extension is an instance extension. It provides a mechanism to create a VkSurfaceKHR object (defined by the VK_KHR_surface extension) that refers to an ANativeWindow, Android’s native surface type. The ANativeWindow represents the producer endpoint of any buffer queue, regardless of consumer endpoint. Common consumer endpoints for ANativeWindows are the system window compositor, video encoders, and application-specific compositors importing the images through a SurfaceTexture.

None

None

1) Does Android need a way to query for compatibility between a particular physical device (and queue family?) and a specific Android display?

RESOLVED: No. Currently on Android, any physical device is expected to be able to present to the system compositor, and all queue families must support the necessary image layout transitions and synchronization operations.

This extension allows the application to query a 64-bit buffer device address value for a buffer, which can be used to access the buffer memory via the PhysicalStorageBuffer storage class in the GL_EXT_buffer_reference GLSL extension and SPV_KHR_physical_storage_buffer SPIR-V extension.

This extension also allows opaque addresses for buffers and memory objects to be queried and later supplied by a trace capture and replay tool, so that addresses used at replay time match the addresses used when the trace was captured. To enable tools to insert these queries, new memory allocation flags must be specified for memory objects that will be bound to buffers accessed via the PhysicalStorageBuffer storage class.

None

None

None

None

This extension provides a new entry point to create render passes in a way that can be easily extended by other extensions through the substructures of render pass creation. The Vulkan 1.0 render pass creation sub-structures do not include sType/pNext members. Additionally, the renderpass begin/next/end commands have been augmented with new extensible structures for passing additional subpass information.

The VkRenderPassMultiviewCreateInfo and VkInputAttachmentAspectReference structures that extended the original VkRenderPassCreateInfo are not accepted into the new creation functions, and instead their parameters are folded into this extension as follows:

The details of these mappings are explained fully in the new structures.

This extension adds support for automatically resolving multisampled depth/stencil attachments in a subpass in a similar manner as for color attachments.

None.

This extension provides the API to enumerate displays and available modes on a given device.

1) Which properties of a mode should be fixed in the mode info vs. settable in some other function when setting the mode? E.g., do we need to double the size of the mode pool to include both stereo and non-stereo modes? YUV and RGB scanout even if they both take RGB input images? BGR vs. RGB input? etc.

PROPOSED RESOLUTION: Many modern displays support at most a handful of resolutions and timings natively. Other “modes” are expected to be supported using scaling hardware on the display engine or GPU. Other properties, such as rotation and mirroring should not require duplicating hardware modes just to express all combinations. Further, these properties may be implemented on a per-display or per-overlay granularity.

To avoid the exponential growth of modes as mutable properties are added, as was the case with EGLConfig/WGL pixel formats/GLXFBConfig, this specification should separate out hardware properties and configurable state into separate objects. Modes and overlay planes will express capabilities of the hardware, while a separate structure will allow applications to configure scaling, rotation, mirroring, color keys, LUT values, alpha masks, etc. for a given swapchain independent of the mode in use. Constraints on these settings will be established by properties of the immutable objects.

Note the resolution of this issue may affect issue 5 as well.

2) What properties of a display itself are useful?

PROPOSED RESOLUTION: This issue is too broad. It was meant to prompt general discussion, but resolving this issue amounts to completing this specification. All interesting properties should be included. The issue will remain as a placeholder since removing it would make it hard to parse existing discussion notes that refer to issues by number.

3) How are multiple overlay planes within a display or mode enumerated?

PROPOSED RESOLUTION: They are referred to by an index. Each display will report the number of overlay planes it contains.

4) Should swapchains be created relative to a mode or a display?

PROPOSED RESOLUTION: When using this extension, swapchains are created relative to a mode and a plane. The mode implies the display object the swapchain will present to. If the specified mode is not the display’s current mode, the new mode will be applied when the first image is presented to the swapchain, and the default operating system mode, if any, will be restored when the swapchain is destroyed.

5) Should users query generic ranges from displays and construct their own modes explicitly using those constraints rather than querying a fixed set of modes (Most monitors only have one real “mode” these days, even though many support relatively arbitrary scaling, either on the monitor side or in the GPU display engine, making “modes” something of a relic/compatibility construct).

PROPOSED RESOLUTION: Expose both. Display info structures will expose a set of predefined modes, as well as any attributes necessary to construct a customized mode.

6) Is it fine if we return the display and display mode handles in the structure used to query their properties?

PROPOSED RESOLUTION: Yes.

7) Is there a possibility that not all displays of a device work with all of the present queues of a device? If yes, how do we determine which displays work with which present queues?

PROPOSED RESOLUTION: No known hardware has such limitations, but determining such limitations is supported automatically using the existing VK_KHR_surface and VK_KHR_swapchain query mechanisms.

8) Should all presentation need to be done relative to an overlay plane, or can a display mode + display be used alone to target an output?

PROPOSED RESOLUTION: Require specifying a plane explicitly.

9) Should displays have an associated window system display, such as an HDC or Display*?

PROPOSED RESOLUTION: No. Displays are independent of any windowing system in use on the system. Further, neither HDC nor Display* refer to a physical display object.

10) Are displays queried from a physical GPU or from a device instance?

PROPOSED RESOLUTION: Developers prefer to query modes directly from the physical GPU so they can use display information as an input to their device selection algorithms prior to device creation. This avoids the need to create dummy device instances to enumerate displays.

This preference must be weighed against the extra initialization that must be done by driver vendors prior to device instance creation to support this usage.

11) Should displays and/or modes be dispatchable objects? If functions are to take displays, overlays, or modes as their first parameter, they must be dispatchable objects as defined in Khronos bug 13529. If they are not added to the list of dispatchable objects, functions operating on them must take some higher-level object as their first parameter. There is no performance case against making them dispatchable objects, but they would be the first extension objects to be dispatchable.

PROPOSED RESOLUTION: Do not make displays or modes dispatchable. They will dispatch based on their associated physical device.

12) Should hardware cursor capabilities be exposed?

PROPOSED RESOLUTION: Defer. This could be a separate extension on top of the base WSI specs.

if they are one physical display device to an end user, but may internally be implemented as two side-by-side displays using the same display engine (and sometimes cabling) resources as two physically separate display devices.

RESOLVED: Tiled displays will appear as a single display object in this API.

14) Should the raw EDID data be included in the display information?

RESOLVED: No. A future extension could be added which reports the EDID if necessary. This may be complicated by the outcome of issue 13.

15) Should min and max scaling factor capabilities of overlays be exposed?

RESOLVED: Yes. This is exposed indirectly by allowing applications to query the min/max position and extent of the source and destination regions from which image contents are fetched by the display engine when using a particular mode and overlay pair.

16) Should devices be able to expose planes that can be moved between displays? If so, how?

RESOLVED: Yes. Applications can determine which displays a given plane supports using vkGetDisplayPlaneSupportedDisplaysKHR.

17) Should there be a way to destroy display modes? If so, does it support destroying “built in” modes?

RESOLVED: Not in this extension. A future extension could add this functionality.

18) What should the lifetime of display and built-in display mode objects be?

RESOLVED: The lifetime of the instance. These objects cannot be destroyed. A future extension may be added to expose a way to destroy these objects and/or support display hotplug.

19) Should persistent mode for smart panels be enabled/disabled at swapchain creation time, or on a per-present basis.

RESOLVED: On a per-present basis.

This extension provides an API to create a swapchain directly on a device’s display without any underlying window system.

None

None

1) Should swapchains sharing images each hold a reference to the images, or should it be up to the application to destroy the swapchains and images in an order that avoids the need for reference counting?

RESOLVED: Take a reference. The lifetime of presentable images is already complex enough.

2) Should the srcRect and dstRect parameters be specified as part of the present command, or at swapchain creation time?

RESOLVED: As part of the presentation command. This allows moving and scaling the image on the screen without the need to respecify the mode or create a new swapchain and presentable images.

3) Should srcRect and dstRect be specified as rects, or separate offset/extent values?

RESOLVED: As rects. Specifying them separately might make it easier for hardware to expose support for one but not the other, but in such cases applications must just take care to obey the reported capabilities and not use non-zero offsets or extents that require scaling, as appropriate.

4) How can applications create multiple swapchains that use the same images?

RESOLVED: By calling vkCreateSharedSwapchainsKHR.

An earlier resolution used vkCreateSwapchainKHR, chaining multiple VkSwapchainCreateInfoKHR structures through pNext. In order to allow each swapchain to also allow other extension structs, a level of indirection was used: VkSwapchainCreateInfoKHR::pNext pointed to a different structure, which had both sType and pNext members for additional extensions, and also had a pointer to the next VkSwapchainCreateInfoKHR structure. The number of swapchains to be created could only be found by walking this linked list of alternating structures, and the pSwapchains out parameter was reinterpreted to be an array of VkSwapchainKHR handles.

Another option considered was a method to specify a “shared” swapchain when creating a new swapchain, such that groups of swapchains using the same images could be built up one at a time. This was deemed unusable because drivers need to know all of the displays an image will be used on when determining which internal formats and layouts to use for that image.

This extension is based off the VK_AMD_draw_indirect_count extension. This extension allows an application to source the number of draw calls for indirect draw calls from a buffer. This enables applications to generate arbitrary amounts of draw commands and execute them without host intervention.

This extension provides a new physical device query which allows retrieving information about the driver implementation, allowing applications to determine which physical device corresponds to which particular vendor’s driver, and which conformance test suite version the driver implementation is compliant with.

None.

None.

None.

None.

None.

An application using external memory may wish to synchronize access to that memory using fences. This extension enables an application to export fence payload to and import fence payload from POSIX file descriptors.

None.

None.

This extension borrows concepts, semantics, and language from VK_KHR_external_semaphore_fd. That extension’s issues apply equally to this extension.

An application using external memory may wish to synchronize access to that memory using fences. This extension enables an application to export fence payload to and import fence payload from Windows handles.

None.

None.

This extension borrows concepts, semantics, and language from VK_KHR_external_semaphore_win32. That extension’s issues apply equally to this extension.

1) Should D3D12 fence handle types be supported, like they are for semaphores?

RESOLVED: No. Doing so would require extending the fence signal and wait operations to provide values to signal / wait for, like VkD3D12FenceSubmitInfoKHR does. A D3D12 fence can be signaled by importing it into a VkSemaphore instead of a VkFence, and applications can check status or wait on the D3D12 fence using non-Vulkan APIs. The convenience of being able to do these operations on VkFence objects doesn’t justify the extra API complexity.

An application may wish to reference device memory in multiple Vulkan logical devices or instances, in multiple processes, and/or in multiple APIs. This extension enables an application to export POSIX file descriptor handles from Vulkan memory objects and to import Vulkan memory objects from POSIX file descriptor handles exported from other Vulkan memory objects or from similar resources in other APIs.

None.

None.

1) Does the application need to close the file descriptor returned by vkGetMemoryFdKHR?

RESOLVED: Yes, unless it is passed back in to a driver instance to import the memory. A successful get call transfers ownership of the file descriptor to the application, and a successful import transfers it back to the driver. Destroying the original memory object will not close the file descriptor or remove its reference to the underlying memory resource associated with it.

2) Do drivers ever need to expose multiple file descriptors per memory object?

RESOLVED: No. This would indicate there are actually multiple memory objects, rather than a single memory object.

3) How should the valid size and memory type for POSIX file descriptor memory handles created outside of Vulkan be specified?

RESOLVED: The valid memory types are queried directly from the external handle. The size will be specified by future extensions that introduce such external memory handle types.

An application may wish to reference device memory in multiple Vulkan logical devices or instances, in multiple processes, and/or in multiple APIs. This extension enables an application to export Windows handles from Vulkan memory objects and to import Vulkan memory objects from Windows handles exported from other Vulkan memory objects or from similar resources in other APIs.

None.

None.

1) Do applications need to call CloseHandle() on the values returned from vkGetMemoryWin32HandleKHR when handleType is VK_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32_BIT_KHR?

RESOLVED: Yes, unless it is passed back in to another driver instance to import the object. A successful get call transfers ownership of the handle to the application. Destroying the memory object will not destroy the handle or the handle’s reference to the underlying memory resource.

2) Should the language regarding KMT/Windows 7 handles be moved to a separate extension so that it can be deprecated over time?

RESOLVED: No. Support for them can be deprecated by drivers if they choose, by no longer returning them in the supported handle types of the instance level queries.

3) How should the valid size and memory type for windows memory handles created outside of Vulkan be specified?

RESOLVED: The valid memory types are queried directly from the external handle. The size is determined by the associated image or buffer memory requirements for external handle types that require dedicated allocations, and by the size specified when creating the object from which the handle was exported for other external handle types.

An application using external memory may wish to synchronize access to that memory using semaphores. This extension enables an application to export semaphore payload to and import semaphore payload from POSIX file descriptors.

None.

None.

1) Does the application need to close the file descriptor returned by vkGetSemaphoreFdKHR?

RESOLVED: Yes, unless it is passed back in to a driver instance to import the semaphore. A successful get call transfers ownership of the file descriptor to the application, and a successful import transfers it back to the driver. Destroying the original semaphore object will not close the file descriptor or remove its reference to the underlying semaphore resource associated with it.

An application using external memory may wish to synchronize access to that memory using semaphores. This extension enables an application to export semaphore payload to and import semaphore payload from Windows handles.

None.

None.

1) Do applications need to call CloseHandle() on the values returned from vkGetSemaphoreWin32HandleKHR when handleType is VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_WIN32_BIT_KHR?

RESOLVED: Yes, unless it is passed back in to another driver instance to import the object. A successful get call transfers ownership of the handle to the application. Destroying the semaphore object will not destroy the handle or the handle’s reference to the underlying semaphore resource.

2) Should the language regarding KMT/Windows 7 handles be moved to a separate extension so that it can be deprecated over time?

RESOLVED: No. Support for them can be deprecated by drivers if they choose, by no longer returning them in the supported handle types of the instance level queries.

3) Should applications be allowed to specify additional object attributes for shared handles?

RESOLVED: Yes. Applications will be allowed to provide similar attributes to those they would to any other handle creation API.

4) How do applications communicate the desired fence values to use with D3D12_FENCE-based Vulkan semaphores?

RESOLVED: There are a couple of options. The values for the signaled and reset states could be communicated up front when creating the object and remain static for the life of the Vulkan semaphore, or they could be specified using auxiliary structures when submitting semaphore signal and wait operations, similar to what is done with the keyed mutex extensions. The latter is more flexible and consistent with the keyed mutex usage, but the former is a much simpler API.

Since Vulkan tends to favor flexibility and consistency over simplicity, a new structure specifying D3D12 fence acquire and release values is added to the vkQueueSubmit function.

This extension provides new entry points to query device display properties and capabilities in a way that can be easily extended by other extensions, without introducing any further entry points. This extension can be considered the VK_KHR_display equivalent of the VK_KHR_get_physical_device_properties2 extension.

None.

None.

1) What should this extension be named?

RESOLVED: VK_KHR_get_display_properties2. Other alternatives:

2) Should extensible input structs be added for these new functions:

RESOLVED:

3) Should additional display query functions be extended?

RESOLVED:

This extension provides new entry points to query device surface capabilities in a way that can be easily extended by other extensions, without introducing any further entry points. This extension can be considered the VK_KHR_surface equivalent of the VK_KHR_get_physical_device_properties2 extension.

None.

None.

1) What should this extension be named?

RESOLVED: VK_KHR_get_surface_capabilities2. Other alternatives:

2) Should additional WSI query functions be extended?

RESOLVED:

On some implementations, setting the VK_IMAGE_CREATE_MUTABLE_FORMAT_BIT on image creation can cause access to that image to perform worse than an equivalent image created without VK_IMAGE_CREATE_MUTABLE_FORMAT_BIT because the implementation does not know what view formats will be paired with the image.

This extension allows an application to provide the list of all formats that can be used with an image when it is created. The implementation may then be able to create a more efficient image that supports the subset of formats required by the application without having to support all formats in the format compatibility class of the image format.

None.

None.

None.

This extension allows framebuffers to be created without the need for creating images first, allowing more flexibility in how they are used, and avoiding the need for many of the confusing compatibility rules.

Framebuffers are now created with a small amount of additional metadata about the image views that will be used in VkFramebufferAttachmentsCreateInfoKHR, and the actual image views are provided at render pass begin time via VkRenderPassAttachmentBeginInfoKHR.

This device extension extends vkQueuePresentKHR, from the VK_KHR_swapchain extension, allowing an application to specify a list of rectangular, modified regions of each image to present. This should be used in situations where an application is only changing a small portion of the presentable images within a swapchain, since it enables the presentation engine to avoid wasting time presenting parts of the surface that have not changed.

This extension is leveraged from the EGL_KHR_swap_buffers_with_damage extension.

None.

None.

None.

None.

1) How should we handle steroescopic-3D swapchains? We need to add a layer for each rectangle. One approach is to create another struct containing the VkRect2D plus layer, and have VkPresentRegionsKHR point to an array of that struct. Another approach is to have two parallel arrays, pRectangles and pLayers, where pRectangles[i] and pLayers[i] must be used together. Which approach should we use, and if the array of a new structure, what should that be called?

RESOLVED: Create a new structure, which is a VkRect2D plus a layer, and will be called VkRectLayerKHR.

2) Where is the origin of the VkRectLayerKHR?

RESOLVED: The upper left corner of the presentable image(s) of the swapchain, per the definition of framebuffer coordinates.

3) Does the rectangular region, VkRectLayerKHR, specify pixels of the swapchain’s image(s), or of the surface?

RESOLVED: Of the image(s). Some presentation engines may scale the pixels of a swapchain’s image(s) to the size of the surface. The size of the swapchain’s image(s) will be consistent, where the size of the surface may vary over time.

4) What if all of the rectangles for a given swapchain contain a width and/or height of zero?

RESOLVED: The application is indicating that no pixels changed since the last present. The presentation engine may use such a hint and not update any pixels for the swapchain. However, all other semantics of vkQueuePresentKHR must still be honored, including waiting for semaphores to signal.

The VK_KHR_performance_query extension adds a mechanism to allow querying of performance counters for use in applications and by profiling tools.

Each queue family may expose counters that can be enabled on a queue of that family. We extend VkQueryType to add a new query type for performance queries, and chain a structure on VkQueryPoolCreateInfo to specify the performance queries to enable.

None.

1) Should this extension include a mechanism to begin a query in command buffer A and end the query in command buffer B?

RESOLVED No - queries are tied to command buffer creation and thus have to be encapsulated within a single command buffer.

2) Should this extension include a mechanism to begin and end queries globally on the queue, not using the existing command buffer commands?

RESOLVED No - for the same reasoning as the resolution of 1).

3) Should this extension expose counters that require multiple passes?

RESOLVED Yes - users should re-submit a command buffer with the same commands in it multiple times, specifying the pass to count as the query parameter in VkPerformanceQuerySubmitInfoKHR.

4) How to handle counters across parallel workloads?

RESOLVED In the spirit of Vulkan, a counter description flag VK_PERFORMANCE_COUNTER_DESCRIPTION_CONCURRENTLY_IMPACTED_KHR denotes that the accuracy of a counter result is affected by parallel workloads.

5) How to handle secondary command buffers?

RESOLVED Secondary command buffers inherit any counter pass index specified in the parent primary command buffer. Note: this is no longer an issue after change from issue 10 resolution

6) What commands does the profiling lock have to be held for?

RESOLVED For any command buffer that is being queried with a performance query pool, the profiling lock must be held while that command buffer is in the recording, executable, or pending state.

7) Should we support vkCmdCopyQueryPoolResults?

RESOLVED Yes.

8) Should we allow performance queries to interact with multiview?

RESOLVED Yes, but the performance queries must be performed once for each pass per view.

9) Should a queryCount > 1 be usable for performance queries?

RESOLVED Yes. Some vendors will have costly performance counter query pool creation, and would rather if a certain set of counters were to be used multiple times that a queryCount > 1 can be used to amortize the instantiation cost.

10) Should we introduce an indirect mechanism to set the counter pass index?

RESOLVED Specify the counter pass index at submit time instead to avoid requiring re-recording of command buffers when multiple counter passes needed.

The following example shows how to find what performance counters a queue family supports, setup a query pool to record these performance counters, how to add the query pool to the command buffer to record information, and how to get the results from the query pool.

When a pipeline is created, its state and shaders are compiled into zero or more device-specific executables, which are used when executing commands against that pipeline. This extension adds a mechanism to query properties and statistics about the different executables produced by the pipeline compilation process. This is intended to be used by debugging and performance tools to allow them to provide more detailed information to the user. Certain compile-time shader statistics provided through this extension may be useful to developers for debugging or performance analysis.

1) What should we call the pieces of the pipeline which are produced by the compilation process and about which you can query properties and statistics?

RESOLVED: Call them "executables". The name "binary" was used in early drafts of the extension but it was determined that "pipeline binary" could have a fairly broad meaning (such as a binary serialized form of an entire pipeline) and was too big of a namespace for the very specific needs of this extension.

This extension allows descriptors to be written into the command buffer, while the implementation is responsible for managing their memory. Push descriptors may enable easier porting from older APIs and in some cases can be more efficient than writing descriptors into descriptor sets.

None.

None.

None.

None.

VK_KHR_sampler_mirror_clamp_to_edge extends the set of sampler address modes to include an additional mode (VK_SAMPLER_ADDRESS_MODE_MIRROR_CLAMP_TO_EDGE) that effectively uses a texture map twice as large as the original image in which the additional half of the new image is a mirror image of the original image.

This new mode relaxes the need to generate images whose opposite edges match by using the original image to generate a matching “mirror image”. This mode allows the texture to be mirrored only once in the negative s, t, and r directions.

Creating a sampler with the new address mode in each dimension

1) Why are both KHR and core versions of the VK_SAMPLER_ADDRESS_MODE_MIRROR_CLAMP_TO_EDGE token present?

RESOLVED: This functionality was intended to be required in Vulkan 1.0. We realized shortly before public release that not all implementations could support it, and moved the functionality into an optional extension, but did not apply the KHR extension suffix. Adding a KHR-suffixed alias of the non-suffixed enum has been done to comply with our own naming rules.

In a related change, before spec revision 1.1.121 this extension was hardwiring into the spec Makefile so it was always included with the Specification, even in the core-only versions. This has now been reverted, and it is treated as any other extension.

This extension allows image memory barriers for depth/stencil images to have just one of the VK_IMAGE_ASPECT_DEPTH_BIT or VK_IMAGE_ASPECT_STENCIL_BIT aspect bits set, rather than require both. This allows their layouts to be set independently. To support depth/stencil images with different layouts for the depth and stencil aspects, the depth/stencil attachment interface has been updated to support a separate layout for stencil.

None.

None.

None.

None.

None.

This extension advertises the SPIR-V Int64Atomics capability for Vulkan, which allows a shader to contain 64-bit atomic operations on signed and unsigned integers. The supported operations include OpAtomicMin, OpAtomicMax, OpAtomicAnd, OpAtomicOr, OpAtomicXor, OpAtomicAdd, OpAtomicExchange, and OpAtomicCompareExchange.

This extension advertises the SPIR-V ShaderClockKHR capability for Vulkan, which allows a shader to query a real-time or monotonically incrementing counter at the subgroup level or across the device level. The two valid SPIR-V scopes for OpReadClockKHR are Subgroup and Device.

When using GLSL source-based shading languages, the clockRealtime*EXT() timing functions map to the OpReadClockKHR instruction with a scope of Device, and the clock*ARB() timing functions map to the OpReadClockKHR instruction with a scope of Subgroup.

None.

The VK_KHR_shader_float16_int8 extension allows use of 16-bit floating-point types and 8-bit integer types in shaders for arithmetic operations.

It introduces two new optional features shaderFloat16 and shaderInt8 which directly map to the Float16 and the Int8 SPIR-V capabilities. The VK_KHR_shader_float16_int8 extension also specifies precision requirements for half-precision floating-point SPIR-V operations. This extension does not enable use of 8-bit integer types or 16-bit floating-point types in any shader input and output interfaces and therefore does not supersede the VK_KHR_8bit_storage or VK_KHR_16bit_storage extensions.

The VK_KHR_shader_float_controls extension enables efficient use of floating-point computations through the ability to query and override the implementation’s default behavior for rounding modes, denormals, signed zero, and infinity.

1) Which instructions must flush denorms?

RESOLVED: Only floating-point conversion, floating-point arithmetic, floating-point relational (except OpIsNaN, OpIsInf), and floating-point GLSL.std.450 extended instructions must flush denormals.

2) What is the denorm behavior for intermediate results?

RESOLVED: When a SPIR-V instruction is implemented as a sequence of other instructions: - in the DenormFlushToZero execution mode the intermediate instructions may flush denormals, the final result of the sequence must not be denormal. - in the DenormPreserve execution mode denormals must be preserved throughout the whole sequence.

3) Do denorm and rounding mode controls apply to OpSpecConstantOp?

RESOLVED: Yes, except when the opcode is OpQuantizeToF16.

4) The SPIR-V specification says that OpConvertFToU and OpConvertFToS unconditionally round towards zero. Do the rounding mode controls specified through the execution modes apply to them?

RESOLVED: No, these instructions unconditionally round towards zero.

5) Do any of the "Pack" GLSL.std.450 instructions count as conversion instructions and have the rounding mode apply?

RESOLVED: No, only instructions listed in the section "3.32.11. Conversion Instructions" of the SPIR-V specification count as conversion instructions.

6) When using inf/nan-ignore mode, what is expected of OpIsNan and OpIsInf?

RESOLVED: These instructions must always accurately detect inf/nan if it is passed to them.

The original versions of VK_KHR_shader_float_controls shipped with booleans named “separateDenormSettings” and “separateRoundingModeSettings”, which at first glance could have indicated “they can all independently set, or not”. However the spec language as written indicated that the 32-bit value could always be set independently, and only the 16- and 64-bit controls needed to be the same if these values were VK_FALSE.

As a result of this slight disparity, and lack of test coverage for this facet of the extension, we ended up with two different behaviors in the wild, where some implementations worked as written, and others worked based on the naming. As these are hard limits in hardware with reasons for exposure as written, it was not possible to standardise on a single way to make this work within the existing API.

No known users of this part of the extension exist in the wild, and as such the Vulkan WG took the unusual step of retroactively changing the once boolean value into a tri-state enum, breaking source compatibility. This was however done in such a way as to retain ABI compatibility, in case any code using this did exist; with the numerical values 0 and 1 retaining their original specified meaning, and a new value signifying the additional “all need to be set together” state. If any applications exist today, compiled binaries will continue to work as written in most cases, but will need changes before the code can be recompiled.

This extension enables the Non Uniform Group Operations in SPIR-V to support 8-bit integer, 16-bit integer, 64-bit integer, 16-bit floating-point, and vectors of these types.

None.

None.

None.

None.

None.

None.

This extension extends VK_KHR_swapchain to enable creation of a shared presentable image. This allows the application to use the image while the presention engine is accessing it, in order to reduce the latency between rendering and presentation.

None.

None.

1) Should we allow a Vulkan WSI swapchain to toggle between normal usage and shared presentation usage?

RESOLVED: No. WSI swapchains are typically recreated with new properties instead of having their properties changed. This can also save resources, assuming that fewer images are needed for shared presentation, and assuming that most VR applications do not need to switch between normal and shared usage.

2) Should we have a query for determining how the presentation engine refresh is triggered?

RESOLVED: Yes. This is done via which presentation modes a surface supports.

3) Should the object representing a shared presentable image be an extension of a VkSwapchainKHR or a separate object?

RESOLVED: Extension of a swapchain due to overlap in creation properties and to allow common functionality between shared and normal presentable images and swapchains.

4) What should we call the extension and the new structures it creates?

RESOLVED: Shared presentable image / shared present.

5) Should the minImageCount and presentMode values of the VkSwapchainCreateInfoKHR be ignored, or required to be compatible values?

RESOLVED: minImageCount must be set to 1, and presentMode should be set to either VK_PRESENT_MODE_SHARED_DEMAND_REFRESH_KHR or VK_PRESENT_MODE_SHARED_CONTINUOUS_REFRESH_KHR.

6) What should the layout of the shared presentable image be?

RESOLVED: After acquiring the shared presentable image, the application must transition it to the VK_IMAGE_LAYOUT_SHARED_PRESENT_KHR layout prior to it being used. After this initial transition, any image usage that was requested during swapchain creation can be performed on the image without layout transitions being performed.

7) Do we need a new API for the trigger to refresh new content?

RESOLVED: vkQueuePresentKHR to act as API to trigger a refresh, as will allow combination with other compatible extensions to vkQueuePresentKHR.

8) How should an application detect a VK_ERROR_OUT_OF_DATE_KHR error on a swapchain using the VK_PRESENT_MODE_SHARED_CONTINUOUS_REFRESH_KHR present mode?

RESOLVED: Introduce vkGetSwapchainStatusKHR to allow applications to query the status of a swapchain using a shared presentation mode.

9) What should subsequent calls to vkQueuePresentKHR for VK_PRESENT_MODE_SHARED_CONTINUOUS_REFRESH_KHR swapchains be defined to do?

RESOLVED: State that implementations may use it as a hint for updated content.

10) Can the ownership of a shared presentable image be transferred to a different queue?

RESOLVED: No. It is not possible to transfer ownership of a shared presentable image obtained from a swapchain created using VK_SHARING_MODE_EXCLUSIVE after it has been presented.

11) How should vkQueueSubmit behave if a command buffer uses an image from a VK_ERROR_OUT_OF_DATE_KHR swapchain?

RESOLVED: vkQueueSubmit is expected to return the VK_ERROR_DEVICE_LOST error.

12) Can Vulkan provide any guarantee on the order of rendering, to enable beam chasing?

RESOLVED: This could be achieved via use of render passes to ensure strip rendering.

This extension allows the use of SPIR-V 1.4 shader modules. SPIR-V 1.4’s new features primarily make it an easier target for compilers from high-level languages, rather than exposing new hardware functionality.

SPIR-V 1.4 incorporates features that are also available separately as extensions. SPIR-V 1.4 shader modules do not need to enable those extensions with the OpExtension opcode, since they are integral parts of SPIR-V 1.4.

SPIR-V 1.4 introduces new floating point execution mode capabilities, also available via SPV_KHR_float_controls. Implementations are not required to support all of these new capabilities; support can be queried using VkPhysicalDeviceFloatControlsPropertiesKHR from the VK_KHR_shader_float_controls extension.

None.

None.

None.

None.

None.

1. Should we have an extension specific to this SPIR-V version, or add a version-generic query for SPIR-V version? SPIR-V 1.4 doesn’t need any other API changes.

RESOLVED: Just expose SPIR-V 1.4.

Most new SPIR-V versions introduce optionally-required capabilities or have implementation-defined limits, and would need more API and specification changes specific to that version to make them available in Vulkan. For example, to support the subgroup features added in SPIR-V 1.3 required introducing VkPhysicalDeviceSubgroupProperties to allow querying the supported subgroup operation categories, maximum supported subgroup size, etc. While we could expose the parts of a new SPIR-V version that don’t need accompanying changes generically, we’ll still end up writing extensions specific to each version for the remaining parts. Thus the generic mechanism won’t reduce future spec-writing effort. In addition, making it clear which parts of a future version are supported by the generic mechanism and which can’t be used without specific support would be difficult to get right ahead of time.

2. Can different stages of the same pipeline use shaders with different SPIR-V versions?

RESOLVED: Yes.

Mixing SPIR-V versions 1.0-1.3 in the same pipeline has not been disallowed, so it would be inconsistent to disallow mixing 1.4 with previous versions.. SPIR-V 1.4 does not introduce anything that should cause new difficulties here.

3. Must Vulkan extensions corresponding to SPIR-V extensions that were promoted to core in 1.4 be enabled in order to use that functionality in a SPIR-V 1.4 module?

RESOLVED: No, with caveats.

The SPIR-V 1.4 module does not need to declare the SPIR-V extensions, since the functionality is now part of core, so there is no need to enable the Vulkan extension that allows SPIR-V modules to declare the SPIR-V extension. However, when the functionality that is now core in SPIR-V 1.4 is optionally supported, the query for support is provided by a Vulkan extension, and that query can only be used if the extension is enabled.

This applies to any SPIR-V version; specifically for SPIR-V 1.4 this only applies to the functionality from SPV_KHR_float_controls, which was made available in Vulkan by VK_KHR_shader_float_controls. Even though the extension was promoted in SPIR-V 1.4, the capabilities are still optional in implementations that support VK_KHR_spirv_1_4.

A SPIR-V 1.4 module doesn’t need to enable SPV_KHR_float_controls in order to use the capabilities, so if the application has a priori knowledge that the implementation supports the capabilities, it doesn’t need to enable VK_KHR_shader_float_controls. However, if it doesn’t have this knowledge and has to query for support at runtime, it must enable VK_KHR_shader_float_controls in order to use VkPhysicalDeviceFloatControlsPropertiesKHR.

The VK_KHR_surface extension is an instance extension. It introduces VkSurfaceKHR objects, which abstract native platform surface or window objects for use with Vulkan. It also provides a way to determine whether a queue family in a physical device supports presenting to particular surface.

Separate extensions for each platform provide the mechanisms for creating VkSurfaceKHR objects, but once created they may be used in this and other platform-independent extensions, in particular the VK_KHR_swapchain extension.

1) Should this extension include a method to query whether a physical device supports presenting to a specific window or native surface on a given platform?

RESOLVED: Yes. Without this, applications would need to create a device instance to determine whether a particular window can be presented to. Knowing that a device supports presentation to a platform in general is not sufficient, as a single machine might support multiple seats, or instances of the platform that each use different underlying physical devices. Additionally, on some platforms, such as the X Window System, different drivers and devices might be used for different windows depending on which section of the desktop they exist on.

2) Should the vkGetPhysicalDeviceSurfaceCapabilitiesKHR, vkGetPhysicalDeviceSurfaceFormatsKHR, and vkGetPhysicalDeviceSurfacePresentModesKHR functions from VK_KHR_swapchain be modified to operate on physical devices and moved to this extension to implement the resolution of issue 1?

RESOLVED: No, separate query functions are needed, as the purposes served are similar but incompatible. The vkGetPhysicalDeviceSurface*KHR functions return information that could potentially depend on an initialized device. For example, the formats supported for presentation to the surface might vary depending on which device extensions are enabled. The query introduced to resolve issue 1 should be used only to query generic driver or platform properties. The physical device parameter is intended to serve only as an identifier rather than a stateful object.

3) Should Vulkan include support Xlib or XCB as the API for accessing the X Window System platform?

RESOLVED: Both. XCB is a more modern and efficient API, but Xlib usage is deeply ingrained in many applications and likely will remain in use for the foreseeable future. Not all drivers necessarily need to support both, but including both as options in the core specification will probably encourage support, which should in turn ease adoption of the Vulkan API in older codebases. Additionally, the performance improvements possible with XCB likely will not have a measurable impact on the performance of Vulkan presentation and other minimal window system interactions defined here.

4) Should the GBM platform be included in the list of platform enums?

RESOLVED: Deferred, and will be addressed with a platform-specific extension to be written in the future.

This extension extends VkSurfaceCapabilities2KHR, providing applications a way to query whether swapchains can be created with the VK_SWAPCHAIN_CREATE_PROTECTED_BIT_KHR flag set.

Vulkan 1.1 added (optional) support for protect memory and protected resources including buffers (VK_BUFFER_CREATE_PROTECTED_BIT), images (VK_IMAGE_CREATE_PROTECTED_BIT), and swapchains (VK_SWAPCHAIN_CREATE_PROTECTED_BIT_KHR). However, on implementations which support multiple windowing systems, not all window systems may be able to provide a protected display path.

This extension provides a way to query if a protected swapchain created for a surface (and thus a specific windowing system) can be displayed on screen. It extends the existing VkSurfaceCapabilities2KHR structure with a new VkSurfaceProtectedCapabilitiesKHR structure from which the application can obtain information about support for protected swapchain creation through vkGetPhysicalDeviceSurfaceCapabilities2KHR.

None.

None.

None.

None.

The VK_KHR_swapchain extension is the device-level companion to the VK_KHR_surface extension. It introduces VkSwapchainKHR objects, which provide the ability to present rendering results to a surface.

1) Does this extension allow the application to specify the memory backing of the presentable images?

RESOLVED: No. Unlike standard images, the implementation will allocate the memory backing of the presentable image.

2) What operations are allowed on presentable images?

RESOLVED: This is determined by the image usage flags specified when creating the presentable image’s swapchain.

3) Does this extension support MSAA presentable images?

RESOLVED: No. Presentable images are always single-sampled. Multi-sampled rendering must use regular images. To present the rendering results the application must manually resolve the multi- sampled image to a single-sampled presentable image prior to presentation.

4) Does this extension support stereo/multi-view presentable images?

RESOLVED: Yes. The number of views associated with a presentable image is determined by the imageArrayLayers specified when creating a swapchain. All presentable images in a given swapchain use the same array size.

5) Are the layers of stereo presentable images half-sized?

RESOLVED: No. The image extents always match those requested by the application.

6) Do the “present” and “acquire next image” commands operate on a queue? If not, do they need to include explicit semaphore objects to interlock them with queue operations?

RESOLVED: The present command operates on a queue. The image ownership operation it represents happens in order with other operations on the queue, so no explicit semaphore object is required to synchronize its actions.

Applications may want to acquire the next image in separate threads from those in which they manage their queue, or in multiple threads. To make such usage easier, the acquire next image command takes a semaphore to signal as a method of explicit synchronization. The application must later queue a wait for this semaphore before queuing execution of any commands using the image.

7) Does vkAcquireNextImageKHR block if no images are available?

RESOLVED: The command takes a timeout parameter. Special values for the timeout are 0, which makes the call a non-blocking operation, and UINT64_MAX, which blocks indefinitely. Values in between will block for up to the specified time. The call will return when an image becomes available or an error occurs. It may, but is not required to, return before the specified timeout expires if the swapchain becomes out of date.

8) Can multiple presents be queued using one vkQueuePresentKHR call?

RESOLVED: Yes. VkPresentInfoKHR contains a list of swapchains and corresponding image indices that will be presented. When supported, all presentations queued with a single vkQueuePresentKHR call will be applied atomically as one operation. The same swapchain must not appear in the list more than once. Later extensions may provide applications stronger guarantees of atomicity for such present operations, and/or allow them to query whether atomic presentation of a particular group of swapchains is possible.

9) How do the presentation and acquire next image functions notify the application the targeted surface has changed?

RESOLVED: Two new result codes are introduced for this purpose:

These can be returned by both vkAcquireNextImageKHR and vkQueuePresentKHR.

10) Does the vkAcquireNextImageKHR command return a semaphore to the application via an output parameter, or accept a semaphore to signal from the application as an object handle parameter?

RESOLVED: Accept a semaphore to signal as an object handle. This avoids the need to specify whether the application must destroy the semaphore or whether it is owned by the swapchain, and if the latter, what its lifetime is and whether it can be re-used for other operations once it is received from vkAcquireNextImageKHR.

11) What types of swapchain queuing behavior should be exposed? Options include swap interval specification, mailbox/most recent vs. FIFO queue management, targeting specific vertical blank intervals or absolute times for a given present operation, and probably others. For some of these, whether they are specified at swapchain creation time or as per-present parameters needs to be decided as well.

RESOLVED: The base swapchain extension will expose 3 possible behaviors (of which, FIFO will always be supported):

Not all surfaces will support all of these modes, so the modes supported will be returned using a surface info query. All surfaces must support the FIFO queue mode. Applications must choose one of these modes up front when creating a swapchain. Switching modes can be accomplished by recreating the swapchain.

12) Can VK_PRESENT_MODE_MAILBOX_KHR provide non-blocking guarantees for vkAcquireNextImageKHR? If so, what is the proper criteria?

RESOLVED: Yes. The difficulty is not immediately obvious here. Naively, if at least 3 images are requested, mailbox mode should always have an image available for the application if the application does not own any images when the call to vkAcquireNextImageKHR was made. However, some presentation engines may have more than one “current” image, and would still need to block in some cases. The right requirement appears to be that if the application allocates the surface’s minimum number of images + 1 then it is guaranteed non-blocking behavior when it does not currently own any images.

13) Is there a way to create and initialize a new swapchain for a surface that has generated a VK_SUBOPTIMAL_KHR return code while still using the old swapchain?

RESOLVED: Not as part of this specification. This could be useful to allow the application to create an “optimal” replacement swapchain and rebuild all its command buffers using it in a background thread at a low priority while continuing to use the “suboptimal” swapchain in the main thread. It could probably use the same “atomic replace” semantics proposed for recreating direct-to-device swapchains without incurring a mode switch. However, after discussion, it was determined some platforms probably could not support concurrent swapchains for the same surface though, so this will be left out of the base KHR extensions. A future extension could add this for platforms where it is supported.

14) Should there be a special value for VkSurfaceCapabilitiesKHR::maxImageCount to indicate there are no practical limits on the number of images in a swapchain?

RESOLVED: Yes. There where often be cases where there is no practical limit to the number of images in a swapchain other than the amount of available resources (I.e., memory) in the system. Trying to derive a hard limit from things like memory size is prone to failure. It is better in such cases to leave it to applications to figure such soft limits out via trial/failure iterations.

15) Should there be a special value for VkSurfaceCapabilitiesKHR::currentExtent to indicate the size of the platform surface is undefined?

RESOLVED: Yes. On some platforms (Wayland, for example), the surface size is defined by the images presented to it rather than the other way around.

16) Should there be a special value for VkSurfaceCapabilitiesKHR::maxImageExtent to indicate there is no practical limit on the surface size?

RESOLVED: No. It seems unlikely such a system would exist. 0 could be used to indicate the platform places no limits on the extents beyond those imposed by Vulkan for normal images, but this query could just as easily return those same limits, so a special “unlimited” value does not seem useful for this field.

17) How should surface rotation and mirroring be exposed to applications? How do they specify rotation and mirroring transforms applied prior to presentation?

RESOLVED: Applications can query both the supported and current transforms of a surface. Both are specified relative to the device’s “natural” display rotation and direction. The supported transforms indicates which orientations the presentation engine accepts images in. For example, a presentation engine that does not support transforming surfaces as part of presentation, and which is presenting to a surface that is displayed with a 90-degree rotation, would return only one supported transform bit: VK_SURFACE_TRANSFORM_ROTATE_90_BIT_KHR. Applications must transform their rendering by the transform they specify when creating the swapchain in preTransform field.

18) Can surfaces ever not support VK_MIRROR_NONE? Can they support vertical and horizontal mirroring simultaneously? Relatedly, should VK_MIRROR_NONE[_BIT] be zero, or bit one, and should applications be allowed to specify multiple pre and current mirror transform bits, or exactly one?

RESOLVED: Since some platforms may not support presenting with a transform other than the native window’s current transform, and prerotation/mirroring are specified relative to the device’s natural rotation and direction, rather than relative to the surface’s current rotation and direction, it is necessary to express lack of support for no mirroring. To allow this, the MIRROR_NONE enum must occupy a bit in the flags. Since MIRROR_NONE must be a bit in the bitmask rather than a bitmask with no values set, allowing more than one bit to be set in the bitmask would make it possible to describe undefined transforms such as VK_MIRROR_NONE_BIT | VK_MIRROR_HORIZONTAL_BIT, or a transform that includes both “no mirroring” and “horizontal mirroring” simultaneously. Therefore, it is desirable to allow specifying all supported mirroring transforms using only one bit. The question then becomes, should there be a VK_MIRROR_HORIZONTAL_AND_VERTICAL_BIT to represent a simultaneous horizontal and vertical mirror transform? However, such a transform is equivalent to a 180 degree rotation, so presentation engines and applications that wish to support or use such a transform can express it through rotation instead. Therefore, 3 exclusive bits are sufficient to express all needed mirroring transforms.

19) Should support for sRGB be required?

RESOLVED: In the advent of UHD and HDR display devices, proper color space information is vital to the display pipeline represented by the swapchain. The app can discover the supported format/color-space pairs and select a pair most suited to its rendering needs. Currently only the sRGB color space is supported, future extensions may provide support for more color spaces. See issues 23 and 24.

20) Is there a mechanism to modify or replace an existing swapchain with one targeting the same surface?

RESOLVED: Yes. This is described above in the text.

21) Should there be a way to set prerotation and mirroring using native APIs when presenting using a Vulkan swapchain?

RESOLVED: Yes. The transforms that can be expressed in this extension are a subset of those possible on native platforms. If a platform exposes a method to specify the transform of presented images for a given surface using native methods and exposes more transforms or other properties for surfaces than Vulkan supports, it might be impossible, difficult, or inconvenient to set some of those properties using Vulkan KHR extensions and some using the native interfaces. To avoid overwriting properties set using native commands when presenting using a Vulkan swapchain, the application can set the pretransform to “inherit”, in which case the current native properties will be used, or if none are available, a platform-specific default will be used. Platforms that do not specify a reasonable default or do not provide native mechanisms to specify such transforms should not include the inherit bits in the supportedTransforms bitmask they return in VkSurfaceCapabilitiesKHR.

22) Should the content of presentable images be clipped by objects obscuring their target surface?

RESOLVED: Applications can choose which behavior they prefer. Allowing the content to be clipped could enable more optimal presentation methods on some platforms, but some applications might rely on the content of presentable images to perform techniques such as partial updates or motion blurs.

23) What is the purpose of specifying a VkColorSpaceKHR along with VkFormat when creating a swapchain?

RESOLVED: While Vulkan itself is color space agnostic (e.g. even the meaning of R, G, B and A can be freely defined by the rendering application), the swapchain eventually will have to present the images on a display device with specific color reproduction characteristics. If any color space transformations are necessary before an image can be displayed, the color space of the presented image must be known to the swapchain. A swapchain will only support a restricted set of color format and -space pairs. This set can be discovered via vkGetPhysicalDeviceSurfaceFormatsKHR. As it can be expected that most display devices support the sRGB color space, at least one format/color-space pair has to be exposed, where the color space is VK_COLOR_SPACE_SRGB_NONLINEAR_KHR.

24) How are sRGB formats and the sRGB color space related?

RESOLVED: While Vulkan exposes a number of SRGB texture formats, using such formats does not guarantee working in a specific color space. It merely means that the hardware can directly support applying the non-linear transfer functions defined by the sRGB standard color space when reading from or writing to images of that these formats. Still, it is unlikely that a swapchain will expose a *_SRGB format along with any color space other than VK_COLOR_SPACE_SRGB_NONLINEAR_KHR.

On the other hand, non-*_SRGB formats will be very likely exposed in pair with a SRGB color space. This means, the hardware will not apply any transfer function when reading from or writing to such images, yet they will still be presented on a device with sRGB display characteristics. In this case the application is responsible for applying the transfer function, for instance by using shader math.

25) How are the lifetime of surfaces and swapchains targeting them related?

RESOLVED: A surface must outlive any swapchains targeting it. A VkSurfaceKHR owns the binding of the native window to the Vulkan driver.

26) How can the client control the way the alpha channel of swapchain images is treated by the presentation engine during compositing?

RESOLVED: We should add new enum values to allow the client to negotiate with the presentation engine on how to treat image alpha values during the compositing process. Since not all platforms can practically control this through the Vulkan driver, a value of VK_COMPOSITE_ALPHA_INHERIT_BIT_KHR is provided like for surface transforms.

27) Is vkCreateSwapchainKHR the right function to return VK_ERROR_NATIVE_WINDOW_IN_USE_KHR, or should the various platform-specific VkSurfaceKHR factory functions catch this error earlier?

RESOLVED: For most platforms, the VkSurfaceKHR structure is a simple container holding the data that identifies a native window or other object representing a surface on a particular platform. For the surface factory functions to return this error, they would likely need to register a reference on the native objects with the native display server somehow, and ensure no other such references exist. Surfaces were not intended to be that heavyweight.

Swapchains are intended to be the objects that directly manipulate native windows and communicate with the native presentation mechanisms. Swapchains will already need to communicate with the native display server to negotiate allocation and/or presentation of presentable images for a native surface. Therefore, it makes more sense for swapchain creation to be the point at which native object exclusivity is enforced. Platforms may choose to enforce further restrictions on the number of VkSurfaceKHR objects that may be created for the same native window if such a requirement makes sense on a particular platform, but a global requirement is only sensible at the swapchain level.