Name Strings

SPV_KHR_compute_shader_derivatives

Contact

To report problems with this extension, please open a new issue at:

https://github.com/KhronosGroup/SPIRV-Headers

Contributors

  • Jean-Noe Morissette, Epic Games

  • Daniel Koch, Nvidia

  • John Kessenich, Google

  • Tobias Hector, AMD

  • Alan Baker, Google

  • Stu Smith, AMD

  • Samuel (Sheng-Wen) Huang, MediaTek

Status

  • Approved by the SPIR-V Working Group: 2024-06-26

  • Approved by the Khronos Board of Promoters: 2024-08-16

Version

Last Modified Date

2024-06-26

Revision

2

Dependencies

This extension is written against the SPIR-V Specification, Version 1.3 Revision 2, Unified.

This extension requires SPIR-V 1.0.

Overview

This extension provides a capability to enable derivatives in any Execution Model that has defined workgroups. There are two new execution modes added which specify which four shader invocations are grouped together.

The new ComputeDerivativeGroupQuadsKHR and ComputeDerivativeGroupLinearKHR capabilities enable the use of OpImageQueryLod, the ImplicitLod instructions, and the Derivative instructions in Execution Models with defined workgroups (GLCompute, Mesh, and Task at the time of writing).

This SPIR-V extension provides support for the GLSL GL_KHR_compute_shader_derivatives extension.

Extension Name

To use this extension within a SPIR-V module, the following OpExtension must be present in the module:

OpExtension "SPV_KHR_compute_shader_derivatives"

Modifications to the SPIR-V Specification, Version 1.3

(Modify Section 2.2.4, Control Flow)

(Modify the definition of Derivative Group, to include more Execution Models)

Derivative Group: Defined only for the Fragment Execution Model and any Execution Model that has defined workgroups. In the Fragment execution model this is the set of invocations collectively processing a single point, line, or triangle, including any helper invocations. In other execution models, this is a single workgroup.

(Modify Section 2.19, Derivatives)

(Replace the first sentence:)

Derivatives appear only in the Fragment Execution Model.

(with the following:)

Derivatives appear in the Fragment and any Execution Model that has defined workgroups.

(Modify Section 3.6, Execution Mode)

(add new rows to the Execution Mode table)

Execution Mode Extra Operands Enabling Capabilities

5289

DerivativeGroupQuadsKHR
Specifies that shader derivatives are evaluated over 2x2 groups of invocations. See the Vulkan or OpenGL API specifications for more detail. Valid with any Execution Model that has defined workgroups.

ComputeDerivativeGroupQuadsKHR

5290

DerivativeGroupLinearKHR
Specifies that shader derivatives are evaluated over groups of four invocations with consecutive LocalInvocationIndex values. See the Vulkan or OpenGL API specifications for more detail. Valid with any Execution Model that has defined workgroups.

ComputeDerivativeGroupLinearKHR
(Modify Section 3.31, Capability, adding a new row to the Capability table)
Capability Implicitly Declares Enabled by Extension

5288

ComputeDerivativeGroupQuadsKHR
Uses the DerivativeGroupQuadsKHR execution mode.

Shader

SPV_KHR_compute_shader_derivatives

5350

ComputeDerivativeGroupLinearKHR
Uses the DerivativeGroupLinearKHR execution mode.

Shader

SPV_KHR_compute_shader_derivatives
(Modify Section 3.32.10, Image Instructions)

(Modify the description of the following instructions to allow them in any Execution Model that has defined workgroups in addition to the Fragment Execution Model)

  • OpImageSampleImplicitLod

  • OpImageSampleDrefImplicitLod

  • OpImageSampleProjImplicitLod

  • OpImageSampleProjDrefImplicitLod

  • OpImageQueryLod

  • OpImageSparseSampleImplicitLod

  • OpImageSparseSampleDrefImplicitLod

    • This instruction is only valid in the Fragment and any Execution Model that has defined workgroups. In addition, it consumes an implicit derivative that can be affected by code motion.

(Modify Section 3.32.16, Derivative Instructions)

(Modify the description of the following instructions to allow them in any Execution Model that has defined workgroup in addition to the Fragment Execution Model)

  • OpDPdx

  • OpDPdy

  • OpFwidth

  • OpDPdxFine

  • OpDPdyFine

  • OpFwidthFine

  • OpDPdxCoarse

  • OpDPdyCoarse

  • OpFwidthCoarse

    • This instruction is only valid in the Fragment and any Execution Model that has defined workgroups.

(Modify the existing descriptions of OpDPd{x,y}{Fine,Course}, prefacing the existing language that talks about partial derivatives relative to the window x or y coordinate with "In the Fragment Execution Model:")

(Add the following to the descriptions of OpDPd{x,y}{Fine,Course}, describing how partial derivatives work in any Execution Model that has defined workgroups)

In any Execution Model that has defined workgroups:
Result is the partial derivative of P evaluated over groups of four invocations. Selection of the four invocations is determined by the DerivativeGroup*KHR execution mode that was specified for the entry point. For these instructions to be used, one of the derivative group modes must be specified.

Validation Rules

An OpExtension must be added to the SPIR-V for validation layers to check legal use of this extension:

OpExtension "SPV_KHR_compute_shader_derivatives"

  • An entry point cannot have both the DerivativeGroupQuadsKHR and DerivativeGroupLinearKHR execution modes specified.

  • The DerivativeGroupQuadsKHR and DerivativeGroupLinearKHR execution modes can be used on entry points with any Execution Model that has defined workgroups.

Issues

None yet!

Revision History

Rev Date Author Changes

1

2023-02-28

Jean-Noe Morissette

Internal revisions

2

2024-06-26

Daniel Koch

Update overview to clarify supported execution models