Safe Haskell	Safe-Inferred
Language	Haskell2010

Vulkan.Extensions.VK_KHR_timeline_semaphore

Description

Name

VK_KHR_timeline_semaphore - device extension

VK_KHR_timeline_semaphore

Name String: VK_KHR_timeline_semaphore

Extension Type

Device extension

Registered Extension Number

208

Revision

2

Extension and Version Dependencies

Requires support for Vulkan 1.0
Requires VK_KHR_get_physical_device_properties2 to be enabled for any device-level functionality

Deprecation state

Promoted to Vulkan 1.2

Contact

Jason Ekstrand @jekstrand%0A*Here describe the issue or question you have about the VK_KHR_timeline_semaphore extension*

Other Extension Metadata

Last Modified Date: 2019-06-12

IP Status

No known IP claims.

Interactions and External Dependencies

This extension interacts with VK_KHR_external_semaphore_capabilities
This extension interacts with VK_KHR_external_semaphore
This extension interacts with VK_KHR_external_semaphore_win32
Promoted to Vulkan 1.2 Core

Contributors

Jeff Bolz, NVIDIA
Yuriy O’Donnell, Epic Games
Jason Ekstrand, Intel
Jesse Hall, Google
James Jones, NVIDIA
Jeff Juliano, NVIDIA
Daniel Rakos, AMD
Ray Smith, Arm

Description

This extension introduces a new type of semaphore that has an integer payload identifying a point in a timeline. Such timeline semaphores support the following operations:

Host query - A host operation that allows querying the payload of the timeline semaphore.
Host wait - A host operation that allows a blocking wait for a timeline semaphore to reach a specified value.
Host signal - A host operation that allows advancing the timeline semaphore to a specified value.
Device wait - A device operation that allows waiting for a timeline semaphore to reach a specified value.
Device signal - A device operation that allows advancing the timeline semaphore to a specified value.

Promotion to Vulkan 1.2

All functionality in this extension is included in core Vulkan 1.2, with the KHR suffix omitted. The original type, enum and command names are still available as aliases of the core functionality.

New Commands

getSemaphoreCounterValueKHR

New Structures

SemaphoreSignalInfoKHR

SemaphoreWaitInfoKHR
Extending PhysicalDeviceFeatures2, DeviceCreateInfo:
- PhysicalDeviceTimelineSemaphoreFeaturesKHR
Extending PhysicalDeviceProperties2:
- PhysicalDeviceTimelineSemaphorePropertiesKHR
Extending SemaphoreCreateInfo, PhysicalDeviceExternalSemaphoreInfo:
- SemaphoreTypeCreateInfoKHR
Extending SubmitInfo, BindSparseInfo:
- TimelineSemaphoreSubmitInfoKHR

New Enums

SemaphoreTypeKHR

SemaphoreWaitFlagBitsKHR

New Bitmasks

SemaphoreWaitFlagsKHR

New Enum Constants

KHR_TIMELINE_SEMAPHORE_EXTENSION_NAME

Issues

1) Do we need a new object type for this?

RESOLVED: No, we just introduce a new type of semaphore object, as VK_KHR_external_semaphore_win32 already uses semaphores as the destination for importing D3D12 fence objects, which are semantically close/identical to the proposed synchronization primitive.

2) What type of payload the new synchronization primitive has?

RESOLVED: A 64-bit unsigned integer that can only be set to strictly increasing values by signal operations and is not changed by wait operations.

3) Does the new synchronization primitive have the same signal-before-wait requirement as the existing semaphores do?

RESOLVED: No. Timeline semaphores support signaling and waiting entirely asynchronously. It is the responsibility of the client to avoid deadlock.

4) Does the new synchronization primitive allow resetting its payload?

RESOLVED: No, allowing the payload value to “go backwards” is problematic. Applications looking for reset behavior should create a new instance of the synchronization primitive instead.

5) How do we enable host waits on the synchronization primitive?

RESOLVED: Both a non-blocking query of the current payload value of the synchronization primitive, and a blocking wait operation are provided.

6) How do we enable device waits and signals on the synchronization primitive?

RESOLVED: Similar to VK_KHR_external_semaphore_win32, this extension introduces a new structure that can be chained to SubmitInfo to specify the values signaled semaphores should be set to, and the values waited semaphores need to reach.

7) Can the new synchronization primitive be used to synchronize presentation and swapchain image acquisition operations?

RESOLVED: Some implementations may have problems with supporting that directly, thus it is not allowed in this extension.

8) Do we want to support external sharing of the new synchronization primitive type?

RESOLVED: Yes. Timeline semaphore specific external sharing capabilities can be queried using getPhysicalDeviceExternalSemaphoreProperties by chaining the new SemaphoreTypeCreateInfoKHR structure to its pExternalSemaphoreInfo structure. This allows having a different set of external semaphore handle types supported for timeline semaphores vs. binary semaphores.

9) Do we need to add a host signal operation for the new synchronization primitive type?

RESOLVED: Yes. This helps in situations where one host thread submits a workload but another host thread has the information on when the workload is ready to be executed.

10) How should the new synchronization primitive interact with the ordering requirements of the original Semaphore?

RESOLVED: Prior to calling any command which may cause a wait operation on a binary semaphore, the client must ensure that the semaphore signal operation that has been submitted for execution and any semaphore signal operations on which it depends (if any) must have also been submitted for execution.

11) Should we have separate feature bits for different sub-features of timeline semaphores?

RESOLVED: No. The only feature which cannot be supported universally is timeline semaphore import/export. For import/export, the client is already required to query available external handle types via getPhysicalDeviceExternalSemaphoreProperties and provide the semaphore type by adding a SemaphoreTypeCreateInfoKHR structure to the pNext chain of PhysicalDeviceExternalSemaphoreInfo so no new feature bit is required.