© Copyright 2014-2019 The Khronos Group Inc. All Rights Reserved.

This specification is protected by copyright laws and contains material proprietary to the Khronos Group, Inc. It or any components may not be reproduced, republished, distributed, transmitted, displayed, broadcast, or otherwise exploited in any manner without the express prior written permission of Khronos Group. You may use this specification for implementing the functionality therein, without altering or removing any trademark, copyright or other notice from the specification, but the receipt or possession of this specification does not convey any rights to reproduce, disclose, or distribute its contents, or to manufacture, use, or sell anything that it may describe, in whole or in part.

Khronos Group grants express permission to any current Promoter, Contributor or Adopter member of Khronos to copy and redistribute UNMODIFIED versions of this specification in any fashion, provided that NO CHARGE is made for the specification and the latest available update of the specification for any version of the API is used whenever possible. Such distributed specification may be reformatted AS LONG AS the contents of the specification are not changed in any way. The specification may be incorporated into a product that is sold as long as such product includes significant independent work developed by the seller. A link to the current version of this specification on the Khronos Group website should be included whenever possible with specification distributions.

Khronos Group makes no, and expressly disclaims any, representations or warranties, express or implied, regarding this specification, including, without limitation, any implied warranties of merchantability or fitness for a particular purpose or noninfringement of any intellectual property. Khronos Group makes no, and expressly disclaims any, warranties, express or implied, regarding the correctness, accuracy, completeness, timeliness, and reliability of the specification. Under no circumstances will the Khronos Group, or any of its Promoters, Contributors or Members or their respective partners, officers, directors, employees, agents, or representatives be liable for any damages, whether direct, indirect, special or consequential damages for lost revenues, lost profits, or otherwise, arising from or in connection with these materials.

Khronos, SYCL, SPIR, WebGL, EGL, COLLADA, StreamInput, OpenVX, OpenKCam, glTF, OpenKODE, OpenVG, OpenWF, OpenSL ES, OpenMAX, OpenMAX AL, OpenMAX IL and OpenMAX DL are trademarks and WebCL is a certification mark of the Khronos Group Inc. OpenCL is a trademark of Apple Inc. and OpenGL and OpenML are registered trademarks and the OpenGL ES and OpenGL SC logos are trademarks of Silicon Graphics International used under license by Khronos. All other product names, trademarks, and/or company names are used solely for identification and belong to their respective owners.

Table of Contents

Contributors and Acknowledgments

Connor Abbott, Intel

Ben Ashbaugh, Intel

Alexey Bader, Intel

Alan Baker, Google

Dan Baker, Oxide Games

Kenneth Benzie, Codeplay

Gordon Brown, Codeplay

Pat Brown, NVIDIA

Diana Po-Yu Chen, MediaTek

Stephen Clarke, Imagination

Patrick Doane, Blizzard Entertainment

Stefanus Du Toit, Google

Tim Foley, Intel

Ben Gaster, Qualcomm

Alexander Galazin, ARM

Christopher Gautier, ARM

Neil Henning, AMD

Kerch Holt, NVIDIA

Lee Howes, Qualcomm

Roy Ju, MediaTek

Ronan Keryell, Xilinx

John Kessenich, Google

Daniel Koch, NVIDIA

Ashwin Kolhe, NVIDIA

Raun Krisch, Intel

Graeme Leese, Broadcom

Yuan Lin, NVIDIA

Yaxun Liu, AMD

Victor Lomuller, Codeplay

Timothy Lottes, Epic Games

John McDonald, Valve

Mariusz Merecki, Intel

David Neto, Google

Boaz Ouriel, Intel

Christophe Riccio, Unity

Andrew Richards, Codeplay

Ian Romanick, Intel

Graham Sellers, AMD

Robert Simpson, Qualcomm

Bartosz Sochacki, Intel

Nikos Stavropoulos, Think Silicon

Brian Sumner, AMD

Andrew Woloszyn, Google

Ruihao Zhang, Qualcomm

Weifeng Zhang, Qualcomm

Note
 Up-to-date HTML and PDF versions of this specification may be found at the Khronos SPIR-V Registry. (https://www.khronos.org/registry/spir-v/)

1. Introduction

Abstract

SPIR-V is a simple binary intermediate language for graphical shaders and compute kernels. A SPIR-V module contains multiple entry points with potentially shared functions in the entry point’s call trees. Each function contains a control-flow graph (CFG) of basic blocks, with optional instructions to express structured control flow. Load/store instructions are used to access declared variables, which includes all input/output (IO). Intermediate results bypassing load/store use static single-assignment (SSA) representation. Data objects are represented logically, with hierarchical type information: There is no flattening of aggregates or assignment to physical register banks, etc. Selectable addressing models establish whether general pointer operations may be used, or if memory access is purely logical.

This document fully defines SPIR-V, a Khronos-standard binary intermediate language for representing graphical-shader stages and compute kernels for multiple client APIs.

This is a unified specification, specifying all versions since and including version 1.0.

1.1. Goals

SPIR-V has the following goals:

  • Provide a simple binary intermediate language for all functionality appearing in Khronos shaders/kernels.

  • Have a concise, transparent, self-contained specification (sections Specification and Binary Form).

  • Map easily to other intermediate languages.

  • Be the form passed by a client API into a driver to set shaders/kernels.

  • Support multiple execution environments, specified by client APIs.

  • Can be targeted by new front ends for novel high-level languages.

  • Allow the first steps of compilation and reflection to be done offline.

  • Be low-level enough to require a reverse-engineering step to reconstruct source code.

  • Improve portability by enabling shared tools to generate or operate on it.

  • Reduce compile time during application run time. (Eliminating most of the compile time during application run time is not a goal of this intermediate language. Target-specific register allocation and scheduling are still expected to take significant time.)

  • Allow some optimizations to be done offline.

1.2. Execution Environment and Client API

SPIR-V is adaptable to multiple execution environments: A SPIR-V module is consumed by an execution environment, as specified by a client API. The full set of rules needed to consume SPIR-V in a particular environment comes from the combination of SPIR-V and that environment’s client API specification. The client API will specify its SPIR-V execution environment as well as extra rules, limitations, capabilities, etc. required by the form of SPIR-V it can validly consume.

1.3. About this document

This document aims to:

  • Include everything needed to fully understand, create, and consume SPIR-V. However:

    • Extended instruction sets can be imported and come with their own specifications.

    • Client API-specific rules are documented in client API specifications.

  • Separate expository and specification language. The specification-proper is in Specification and Binary Form.

1.4. Extendability

SPIR-V can be extended by multiple vendors or parties simultaneously:

  • Using the OpExtension instruction to require new semantics that must be supported. Such new semantics would come from an extension specification.

  • Reserving (registering) ranges of the token values, as described further below.

  • Aided by instruction skipping, also further described below.

Enumeration Token Values. It is easy to extend all the types, storage classes, opcodes, decorations, etc. by adding to the token values.

Registration. Ranges of token values in the Binary Form section can be pre-allocated to numerous vendors/parties. This allows combining multiple independent extensions without conflict. To register ranges, use the https://github.com/KhronosGroup/SPIRV-Headers repository, and submit pull requests against the include/spirv/spir-v.xml file.

Extended Instructions. Sets of extended instructions can be provided and specified in separate specifications. Multiple sets of extended instructions can be imported without conflict, as the extended instructions are selected by {set id, instruction number} pairs.

Instruction Skipping. Tools are encouraged to skip opcodes for features they are not required to process. This is trivially enabled by the word count in an instruction, which makes it easier to add new instructions without breaking existing tools.

1.5. Debuggability

SPIR-V can decorate, with a text string, virtually anything created in the shader: types, variables, functions, etc. This is required for externally visible symbols, and also allowed for naming the result of any instruction. This can be used to aid in understandability when disassembling or debugging lowered versions of SPIR-V.

Location information (file names, lines, and columns) can be interleaved with the instruction stream to track the origin of each instruction.

1.6. Design Principles

Regularity. All instructions start with a word count. This allows walking a SPIR-V module without decoding each opcode. All instructions have an opcode that dictates for all operands what kind of operand they are. For instructions with a variable number of operands, the number of variable operands is known by subtracting the number of non-variable words from the instruction’s word count.

Non Combinatorial. There is no combinatorial type explosion or need for large encode/decode tables for types. Rather, types are parameterized. Image types declare their dimensionality, arrayness, etc. all orthogonally, which greatly simplify code. This is done similarly for other types. It also applies to opcodes. Operations are orthogonal to scalar/vector size, but not to integer vs. floating-point differences.

Modeless. After a given execution model (e.g., pipeline stage) is specified, internal operation is essentially modeless: Generally, it will follow the rule: "same spelling, same semantics", and does not have mode bits that modify semantics. If a change to SPIR-V modifies semantics, it should use a different spelling. This makes consumers of SPIR-V much more robust. There are execution modes declared, but these generally affect the way the module interacts with its execution environment, not its internal semantics. Capabilities are also declared, but this is to declare the subset of functionality that is used, not to change any semantics of what is used.

Declarative. SPIR-V declares externally-visible modes like "writes depth", rather than having rules that require deduction from full shader inspection. It also explicitly declares what addressing modes, execution model, extended instruction sets, etc. will be used. See Language Capabilities for more information.

SSA. All results of intermediate operations are strictly SSA. However, declared variables reside in memory and use load/store for access, and such variables can be stored to multiple times.

IO. Some storage classes are for input/output (IO) and, fundamentally, IO will be done through load/store of variables declared in these storage classes.

1.7. Static Single Assignment (SSA)

SPIR-V includes a phi instruction to allow the merging together of intermediate results from split control flow. This allows split control flow without load/store to memory. SPIR-V is flexible in the degree to which load/store is used; it is possible to use control flow with no phi-instructions, while still staying in SSA form, by using memory load/store.

Some storage classes are for IO and, fundamentally, IO will be done through load/store, and initial load and final store can never be eliminated. Other storage classes are shader local and can have their load/store eliminated. It can be considered an optimization to largely eliminate such loads/stores by moving them into intermediate results in SSA form.

1.8. Built-In Variables

SPIR-V identifies built-in variables from a high-level language with an enumerant decoration. This assigns any unusual semantics to the variable. Built-in variables must otherwise be declared with their correct SPIR-V type and treated the same as any other variable.

1.9. Specialization

Specialization enables offline creation of a portable SPIR-V module based on constant values that won’t be known until a later point in time. For example, to size a fixed array with a constant not known during creation of a module, but known when the module will be lowered to the target architecture.

See Specialization in the next section for more details.

1.10. Example

The SPIR-V form is binary, not human readable, and fully described in Binary Form. This is an example disassembly to give a basic idea of what SPIR-V looks like:

GLSL fragment shader:

#version 450

in vec4 color1;
in vec4 multiplier;
noperspective in vec4 color2;
out vec4 color;

struct S {
    bool b;
    vec4 v[5];
    int i;
};

uniform blockName {
    S s;
    bool cond;
};

void main()
{
    vec4 scale = vec4(1.0, 1.0, 2.0, 1.0);

    if (cond)
        color = color1 + s.v[2];
    else
        color = sqrt(color2) * scale;

    for (int i = 0; i < 4; ++i)
        color *= multiplier;
}

Corresponding SPIR-V:

; Magic:     0x07230203 (SPIR-V)
; Version:   0x00010000 (Version: 1.0.0)
; Generator: 0x00080001 (Khronos Glslang Reference Front End; 1)
; Bound:     63
; Schema:    0

               OpCapability Shader
          %1 = OpExtInstImport "GLSL.std.450"
               OpMemoryModel Logical GLSL450
               OpEntryPoint Fragment %4 "main" %31 %33 %42 %57
               OpExecutionMode %4 OriginLowerLeft

; Debug information
               OpSource GLSL 450
               OpName %4 "main"
               OpName %9 "scale"
               OpName %17 "S"
               OpMemberName %17 0 "b"
               OpMemberName %17 1 "v"
               OpMemberName %17 2 "i"
               OpName %18 "blockName"
               OpMemberName %18 0 "s"
               OpMemberName %18 1 "cond"
               OpName %20 ""
               OpName %31 "color"
               OpName %33 "color1"
               OpName %42 "color2"
               OpName %48 "i"
               OpName %57 "multiplier"

; Annotations (non-debug)
               OpDecorate %15 ArrayStride 16
               OpMemberDecorate %17 0 Offset 0
               OpMemberDecorate %17 1 Offset 16
               OpMemberDecorate %17 2 Offset 96
               OpMemberDecorate %18 0 Offset 0
               OpMemberDecorate %18 1 Offset 112
               OpDecorate %18 Block
               OpDecorate %20 DescriptorSet 0
               OpDecorate %42 NoPerspective

; All types, variables, and constants
          %2 = OpTypeVoid
          %3 = OpTypeFunction %2                      ; void ()
          %6 = OpTypeFloat 32                         ; 32-bit float
          %7 = OpTypeVector %6 4                      ; vec4
          %8 = OpTypePointer Function %7              ; function-local vec4*
         %10 = OpConstant %6 1
         %11 = OpConstant %6 2
         %12 = OpConstantComposite %7 %10 %10 %11 %10 ; vec4(1.0, 1.0, 2.0, 1.0)
         %13 = OpTypeInt 32 0                         ; 32-bit int, sign-less
         %14 = OpConstant %13 5
         %15 = OpTypeArray %7 %14
         %16 = OpTypeInt 32 1
         %17 = OpTypeStruct %13 %15 %16
         %18 = OpTypeStruct %17 %13
         %19 = OpTypePointer Uniform %18
         %20 = OpVariable %19 Uniform
         %21 = OpConstant %16 1
         %22 = OpTypePointer Uniform %13
         %25 = OpTypeBool
         %26 = OpConstant %13 0
         %30 = OpTypePointer Output %7
         %31 = OpVariable %30 Output
         %32 = OpTypePointer Input %7
         %33 = OpVariable %32 Input
         %35 = OpConstant %16 0
         %36 = OpConstant %16 2
         %37 = OpTypePointer Uniform %7
         %42 = OpVariable %32 Input
         %47 = OpTypePointer Function %16
         %55 = OpConstant %16 4
         %57 = OpVariable %32 Input

; All functions
          %4 = OpFunction %2 None %3                  ; main()
          %5 = OpLabel
          %9 = OpVariable %8 Function
         %48 = OpVariable %47 Function
               OpStore %9 %12
         %23 = OpAccessChain %22 %20 %21              ; location of cond
         %24 = OpLoad %13 %23                         ; load 32-bit int from cond
         %27 = OpINotEqual %25 %24 %26                ; convert to bool
               OpSelectionMerge %29 None              ; structured if
               OpBranchConditional %27 %28 %41        ; if cond
         %28 = OpLabel                                ; then
         %34 = OpLoad %7 %33
         %38 = OpAccessChain %37 %20 %35 %21 %36      ; s.v[2]
         %39 = OpLoad %7 %38
         %40 = OpFAdd %7 %34 %39
               OpStore %31 %40
               OpBranch %29
         %41 = OpLabel                                ; else
         %43 = OpLoad %7 %42
         %44 = OpExtInst %7 %1 Sqrt %43               ; extended instruction sqrt
         %45 = OpLoad %7 %9
         %46 = OpFMul %7 %44 %45
               OpStore %31 %46
               OpBranch %29
         %29 = OpLabel                                ; endif
               OpStore %48 %35
               OpBranch %49
         %49 = OpLabel
               OpLoopMerge %51 %52 None               ; structured loop
               OpBranch %53
         %53 = OpLabel
         %54 = OpLoad %16 %48
         %56 = OpSLessThan %25 %54 %55                ; i < 4 ?
               OpBranchConditional %56 %50 %51        ; body or break
         %50 = OpLabel                                ; body
         %58 = OpLoad %7 %57
         %59 = OpLoad %7 %31
         %60 = OpFMul %7 %59 %58
               OpStore %31 %60
               OpBranch %52
         %52 = OpLabel                                ; continue target
         %61 = OpLoad %16 %48
         %62 = OpIAdd %16 %61 %21                     ; ++i
               OpStore %48 %62
               OpBranch %49                           ; loop back
         %51 = OpLabel                                ; loop merge point
               OpReturn
               OpFunctionEnd

2. Specification

2.1. Language Capabilities

A SPIR-V module is consumed by a client API that needs to support the features used by that SPIR-V module. Features are classified through capabilities. Capabilities used by a particular SPIR-V module must be declared early in that module with the OpCapability instruction. Then:

  • A validator can validate that the module uses only its declared capabilities.

  • A client API is allowed to reject modules declaring capabilities it does not support.

All available capabilities and their dependencies form a capability hierarchy, fully listed in the capability section. Only top-level capabilities need to be explicitly declared; their dependencies are implicitly declared.

When an instruction, enumerant, or other feature specifies multiple enabling capabilities, only one such capability needs to be declared to use the feature. This declaration does not itself imply anything about the presence of the other enabling capabilities: The execution environment needs to support only the declared capability.

The SPIR-V specification provides universal capability-specific validation rules, in the validation section. Additionally, each client API must include the following:

  • Which capabilities in the capability section it supports or requires, and hence allows in a SPIR-V module.

  • Any additional validation rules it has beyond those specified by the SPIR-V specification.

  • Required limits, if they are beyond the Universal Limits.

2.2. Terms

2.2.1. Instructions

Word: 32 bits.

<id>: A numerical name; the name used to refer to an object, a type, a function, a label, etc. An <id> always consumes one word. The <id>s defined by a module obey SSA.

Result <id>: Most instructions define a result, named by an <id> explicitly provided in the instruction. The Result <id> is used as an operand in other instructions to refer to the instruction that defined it.

Literal String: A nul-terminated stream of characters consuming an integral number of words. The character set is Unicode in the UTF-8 encoding scheme. The UTF-8 octets (8-bit bytes) are packed four per word, following the little-endian convention (i.e., the first octet is in the lowest-order 8 bits of the word). The final word contains the string’s nul-termination character (0), and all contents past the end of the string in the final word are padded with 0.

Literal Number: A numeric value consuming one or more words. An instruction will determine what type a literal will be interpreted as. When the type’s bit width is larger than one word, the literal’s low-order words appear first. When the type’s bit width is less than 32-bits, the literal’s value appears in the low-order bits of the word, and the high-order bits must be 0 for a floating-point type, or 0 for an integer type with Signedness of 0, or sign extended when Signedness is 1. (Similarly for the remaining bits of widths larger than 32 bits but not a multiple of 32 bits.)

Literal: A Literal String or a Literal Number.

Operand: A one-word argument to an instruction. E.g., it could be an <id>, or a (part of a) literal. Which form it holds is always explicitly known from the opcode.

Immediate: Operand(s) directly holding a literal value rather than an <id>. Immediate values larger than one word will consume multiple operands, one per word. That is, operand counting is always done per word, not per immediate.

WordCount: The complete number of words taken by an instruction, including the word holding the word count and opcode, and any optional operands. An instruction’s word count is the total space taken by the instruction.

Instruction: After a header, a module is simply a linear list of instructions. An instruction contains a word count, an opcode, an optional Result <id>, an optional <id> of the instruction’s type, and a variable list of operands. All instruction opcodes and semantics are listed in Instructions.

Decoration: Auxiliary information such as built-in variable, stream numbers, invariance, interpolation type, relaxed precision, etc., added to <id>s or structure-type members through Decorations. Decorations are enumerated in Decoration in the Binary Form section.

Object: An instantiation of a non-void type, either as the Result <id> of an operation, or created through OpVariable.

Memory Object: An object created through OpVariable. Such an object can die on function exit, if it was a function variable, or exist for the duration of an entry point.

Memory Object Declaration: An OpVariable, or an OpFunctionParameter of pointer type.

Intermediate Object or Intermediate Value or Intermediate Result: An object created by an operation (not memory allocated by OpVariable) and dying on its last consumption.

Constant Instruction: Either a specialization-constant instruction or a fixed constant instruction: Instructions that start "OpConstant" or "OpSpec".

[a, b]: This square-bracket notation means the range from a to b, inclusive of a and b. Parentheses exclude their end point, so, for example, (a, b] means a to b excluding a but including b.

2.2.2. Types

Boolean type: The type returned by OpTypeBool.

Integer type: Any width signed or unsigned type from OpTypeInt. By convention, the lowest-order bit will be referred to as bit-number 0, and the highest-order bit as bit-number Width - 1.

Floating-point type: Any width type from OpTypeFloat.

Numerical type: An integer type or a floating-point type.

Scalar: A single instance of a numerical type or Boolean type. Scalars will also be called components when being discussed either by themselves or in the context of the contents of a vector.

Vector: An ordered homogeneous collection of two or more scalars. Vector sizes are quite restrictive and dependent on the execution model.

Matrix: An ordered homogeneous collection of vectors. When vectors are part of a matrix, they will also be called columns. Matrix sizes are quite restrictive and dependent on the execution model.

Array: An ordered homogeneous collection of any non-void-type objects. When an object is part of an array, it will also be called an element. Array sizes are generally not restricted.

Structure: An ordered heterogeneous collection of any non-void types. When an object is part of a structure, it will also be called a member.

Aggregate: A structure or an array.

Composite: An aggregate, a matrix, or a vector.

Image: A traditional texture or image; SPIR-V has this single name for these. An image type is declared with OpTypeImage. An image does not include any information about how to access, filter, or sample it.

Sampler: Settings that describe how to access, filter, or sample an image. Can come either from literal declarations of settings or be an opaque reference to externally bound settings. A sampler does not include an image.

Sampled Image: An image combined with a sampler, enabling filtered accesses of the image’s contents.

Concrete Type: A numerical scalar, vector, or matrix type, or OpTypePointer when using a Physical addressing model, or any aggregate containing only these types.

Abstract Type: An OpTypeVoid or OpTypeBool, or OpTypePointer when using the Logical addressing model, or any aggregate type containing any of these.

Opaque Type: A type that is, or contains, or points to, or contains pointers to, any of the following types:

Variable pointer: A pointer that results from one of the following instructions:

Additionally, any OpAccessChain, OpInBoundsAccessChain, or OpCopyObject that takes a variable pointer as an operand also produces a variable pointer. An OpFunctionParameter of pointer type is a variable pointer if any OpFunctionCall to the function statically passes a variable pointer as the value of the parameter.

2.2.3. Computation

Remainder: When dividing a by b, a remainder r is defined to be a value that satisfies r + q × b = a where q is a whole number and |r| < |b|.

2.2.4. Module

Module: A single unit of SPIR-V. It can contain multiple entry points, but only one set of capabilities.

Entry Point: A function in a module where execution begins. A single entry point is limited to a single execution model. An entry point is declared using OpEntryPoint.

Execution Model: A graphical-pipeline stage or OpenCL kernel. These are enumerated in Execution Model.

Execution Mode: Modes of operation relating to the interface or execution environment of the module. These are enumerated in Execution Mode. Generally, modes do not change the semantics of instructions within a SPIR-V module.

Vertex Processor: Any stage or execution model that processes vertices: Vertex, tessellation control, tessellation evaluation, and geometry. Explicitly excludes fragment and compute execution models.

2.2.5. Control Flow

Block: A contiguous sequence of instructions starting with an OpLabel, ending with a termination instruction. A block has no additional label or termination instructions.

Branch Instruction: One of the following, used as a termination instruction:

Termination Instruction: One of the following, used to terminate blocks:

Dominate: A block A dominates a block B, where A and B are in the same function, if every path from the function’s entry point to block B includes block A. A strictly dominates B only if A dominates B and A and B are different blocks.

Post Dominate: A block B post dominates a block A, where A and B are in the same function, if every path from A to a function-return instruction goes through block B.

Control-Flow Graph: The graph formed by a function’s blocks and branches. The blocks are the graph’s nodes, and the branches the graph’s edges.

CFG: Control-flow graph.

Back Edge: If a depth-first traversal is done on a function’s CFG, starting from the first block of the function, a back edge is a branch to a previously visited block. A back-edge block is the block containing such a branch.

Merge Instruction: One of the following, used before a branch instruction to declare structured control flow:

Header Block: A block containing a merge instruction.

Loop Header: A header block whose merge instruction is an OpLoopMerge.

Merge Block: A block declared by the Merge Block operand of a merge instruction.

Break Block: A block containing a branch to the Merge Block of a loop header’s merge instruction.

Continue Block: A block containing a branch to an OpLoopMerge instruction’s Continue Target.

Return Block: A block containing an OpReturn or OpReturnValue branch.

Invocation: A single execution of an entry point in a SPIR-V module, operating only on the amount of data explicitly exposed by the semantics of the instructions. (Any implicit operation on additional instances of data would comprise additional invocations.) For example, in compute execution models, a single invocation operates only on a single work item, or, in a vertex execution model, a single invocation operates only on a single vertex.

Subgroup: Invocations are partitioned into subgroups, where invocations within a subgroup can synchronize and share data with each other efficiently. In compute models, the current workgroup is a superset of the subgroup.

Invocation Group: The complete set of invocations collectively processing a particular compute workgroup or graphical operation, where the scope of a "graphical operation" is implementation dependent, but at least as large as a single point, line, triangle, or patch, and at most as large as a single rendering command, as defined by the client API.

Derivative Group: Defined only for the Fragment Execution Model: The set of invocations collectively processing a single point, line, or triangle, including any helper invocations.

Dynamic Instance: Within a single invocation, a single static instruction can be executed multiple times, giving multiple dynamic instances of that instruction. This can happen when the instruction is executed in a loop, or in a function called from multiple call sites, or combinations of multiple of these. Different loop iterations and different dynamic function-call-site chains yield different dynamic instances of such an instruction. Dynamic instances are distinguished by the control-flow path within an invocation, not by which invocation executed it. That is, different invocations of an entry point execute the same dynamic instances of an instruction when they follow the same control-flow path, starting from that entry point.

Dynamically Uniform: An <id> is dynamically uniform for a dynamic instance consuming it when its value is the same for all invocations (in the invocation group) that execute that dynamic instance.

Uniform Control Flow: Uniform control flow (or converged control flow) occurs when all invocations in the invocation group or derivative group execute the same control-flow path (and hence the same sequence of dynamic instances of instructions). Uniform control flow is the initial state at the entry point, and lasts until a conditional branch takes different control paths for different invocations (non-uniform or divergent control flow). Such divergence can reconverge, with all the invocations once again executing the same control-flow path, and this re-establishes the existence of uniform control flow. If control flow is uniform upon entry into a header block, and all invocations leave that dynamic instance of the header block’s control-flow construct via the header block’s declared merge block, then control flow reconverges to be uniform at that merge block.

2.3. Physical Layout of a SPIR-V Module and Instruction

A SPIR-V module is a single linear stream of words. The first words are shown in the following table:

Table 1. First Words of Physical Layout
Word Number Contents

0

Magic Number.

1

Version number. The bytes are, high-order to low-order:

0 | Major Number | Minor Number | 0

Hence, version 1.3 is the value 0x00010300.

2

Generator’s magic number. It is associated with the tool that generated the module. Its value does not affect any semantics, and is allowed to be 0. Using a non-0 value is encouraged, and can be registered with Khronos at https://www.khronos.org/registry/spir-v/api/spir-v.xml.

3

Bound; where all <id>s in this module are guaranteed to satisfy

0 < id < Bound

Bound should be small, smaller is better, with all <id> in a module being densely packed and near 0.

4

0 (Reserved for instruction schema, if needed.)

5

First word of instruction stream, see below.

All remaining words are a linear sequence of instructions.

Each instruction is a stream of words:

Table 2. Instruction Physical Layout
Instruction Word Number Contents

0

Opcode: The 16 high-order bits are the WordCount of the instruction. The 16 low-order bits are the opcode enumerant.

1

Optional instruction type <id> (presence determined by opcode).

.

Optional instruction Result <id> (presence determined by opcode).

.

Operand 1 (if needed)

.

Operand 2 (if needed)

…

…

WordCount - 1

Operand N (N is determined by WordCount minus the 1 to 3 words used for the opcode, instruction type <id>, and instruction Result <id>).

Instructions are variable length due both to having optional instruction type <id> and Result <id> words as well as a variable number of operands. The details for each specific instruction are given in the Binary Form section.

2.4. Logical Layout of a Module

The instructions of a SPIR-V module must be in the following order. For sections earlier than function definitions, it is invalid to use instructions other than those indicated.

  1. All OpCapability instructions.

  2. Optional OpExtension instructions (extensions to SPIR-V).

  3. Optional OpExtInstImport instructions.

  4. The single required OpMemoryModel instruction.

  5. All entry point declarations, using OpEntryPoint.

  6. All execution-mode declarations, using OpExecutionMode or OpExecutionModeId.

  7. These debug instructions, which must be grouped in the following order:

    1. all OpString, OpSourceExtension, OpSource, and OpSourceContinued, without forward references.

    2. all OpName and all OpMemberName

    3. all OpModuleProcessed instructions

  8. All annotation instructions:

    1. all decoration instructions (OpDecorate, OpMemberDecorate, OpGroupDecorate, OpGroupMemberDecorate, and OpDecorationGroup).

  9. All type declarations (OpTypeXXX instructions), all constant instructions, and all global variable declarations (all OpVariable instructions whose Storage Class is not Function). This is the preferred location for OpUndef instructions, though they can also appear in function bodies. All operands in all these instructions must be declared before being used. Otherwise, they can be in any order. This section is the first section to allow use of OpLine debug information.

  10. All function declarations ("declarations" are functions without a body; there is no forward declaration to a function with a body). A function declaration is as follows.

    1. Function declaration, using OpFunction.

    2. Function parameter declarations, using OpFunctionParameter.

    3. Function end, using OpFunctionEnd.

  11. All function definitions (functions with a body). A function definition is as follows.

    1. Function definition, using OpFunction.

    2. Function parameter declarations, using OpFunctionParameter.

    3. Block

    4. Block

    5. …

    6. Function end, using OpFunctionEnd.

Within a function definition:

  • A block always starts with an OpLabel instruction. This may be immediately preceded by an OpLine instruction, but the OpLabel is considered as the beginning of the block.

  • A block always ends with a termination instruction (see validation rules for more detail).

  • All OpVariable instructions in a function must have a Storage Class of Function.

  • All OpVariable instructions in a function must be in the first block in the function. These instructions, together with any immediately preceding OpLine instructions, must be the first instructions in that block. (Note the validation rules prevent OpPhi instructions in the first block of a function.)

  • A function definition (starts with OpFunction) can be immediately preceded by an OpLine instruction.

Forward references (an operand <id> that appears before the Result <id> defining it) are allowed for:

In all cases, there is enough type information to enable a single simple pass through a module to transform it. For example, function calls have all the type information in the call, phi-functions don’t change type, and labels don’t have type. The pointer forward reference allows structures to contain pointers to themselves or to be mutually recursive (through pointers), without needing additional type information.

The Validation Rules section lists additional rules that must be satisfied.

2.5. Instructions

Most instructions create a Result <id>, as provided in the Result <id> field of the instruction. These Result <id>s are then referred to by other instructions through their <id> operands. All instruction operands are specified in the Binary Form section.

Instructions are explicit about whether they require immediates, rather than an <id> referring to some other result. This is strictly known just from the opcode.

  • An immediate 32-bit (or smaller) integer is always one operand directly holding a 32-bit two’s-complement value.

  • An immediate 32-bit float is always one operand, directly holding a 32-bit IEEE 754 floating-point representation.

  • An immediate 64-bit float is always two operands, directly holding a 64-bit IEEE 754 representation. The low-order 32 bits appear in the first operand.

2.5.1. SSA Form

A module is always in static single assignment (SSA) form. That is, there is always exactly one instruction resulting in any particular Result <id>. Storing into variables declared in memory is not subject to this; such stores do not create Result <id>s. Accessing declared variables is done through:

  • OpVariable to allocate an object in memory and create a Result <id> that is the name of a pointer to it.

  • OpAccessChain or OpInBoundsAccessChain to create a pointer to a subpart of a composite object in memory.

  • OpLoad through a pointer, giving the loaded object a Result <id> that can then be used as an operand in other instructions.

  • OpStore through a pointer, to write a value. There is no Result <id> for an OpStore.

OpLoad and OpStore instructions can often be eliminated, using intermediate results instead. When this happens in multiple control-flow paths, these values need to be merged again at the path’s merge point. Use OpPhi to merge such values together.

2.6. Entry Point and Execution Model

The OpEntryPoint instruction identifies an entry point with two key things: an execution model and a function definition. Execution models include Vertex, GLCompute, etc. (one for each graphical stage), as well as Kernel for OpenCL kernels. For the complete list, see Execution Model. An OpEntryPoint also supplies a name that can be used externally to identify the entry point, and a declaration of all the Input and Output variables that form its input/output interface.

The static function call graphs rooted at two entry points are allowed to overlap, so that function definitions and global variable definitions can be shared. The execution model and any execution modes associated with an entry point apply to the entire static function call graph rooted at that entry point. This rule implies that a function appearing in both call graphs of two distinct entry points may behave differently in each case. Similarly, variables whose semantics depend on properties of an entry point, e.g. those using the Input Storage Class, may behave differently when used in call graphs rooted in two different entry points.

2.7. Execution Modes

Information like the following is declared with OpExecutionMode instructions. For example,

  • number of invocations (Invocations)

  • vertex-order CCW (VertexOrderCcw)

  • triangle strip generation (OutputTriangleStrip)

  • number of output vertices (OutputVertices)

  • etc.

For a complete list, see Execution Mode.

2.8. Types and Variables

Types are built up hierarchically, using OpTypeXXX instructions. The Result <id> of an OpTypeXXX instruction becomes a type <id> for future use where type <id>s are needed (therefore, OpTypeXXX instructions do not have a type <id>, like most other instructions do).

The "leaves" to start building with are types like OpTypeFloat, OpTypeInt, OpTypeImage, OpTypeEvent, etc. Other types are built up from the Result <id> of these. The numerical types are parameterized to specify bit width and signed vs. unsigned.

Higher-level types are then constructed using opcodes like OpTypeVector, OpTypeMatrix, OpTypeImage, OpTypeArray, OpTypeRuntimeArray, OpTypeStruct, and OpTypePointer. These are parameterized by number of components, array size, member lists, etc. The image types are parameterized by the return type, dimensionality, arrayness, etc. To do sampling or filtering operations, a type from OpTypeSampledImage is used that contains both an image and a sampler. Such a sampled image can be set directly by the client API or combined in a SPIR-V module from an independent image and an independent sampler.

Types are built bottom up: A parameterizing operand in a type must be defined before being used.

Some additional information about the type of an <id> can be provided using the decoration instructions (OpDecorate, OpMemberDecorate, OpGroupDecorate, OpGroupMemberDecorate, and OpDecorationGroup). These can add, for example, Invariant to an <id> created by another instruction. See the full list of Decorations in the Binary Form section.

Two different type <id>s form, by definition, two different types. It is valid to declare multiple aggregate type <id>s having the same opcode and operands. This is to allow multiple instances of aggregate types with the same structure to be decorated differently. (Different decorations are not required; two different aggregate type <id>s are allowed to have identical declarations and decorations, and will still be two different types.) Pointer types are also allowed to have multiple <id>s for the same opcode and operands, to allow for differing decorations (e.g., Volatile) or different decoration values (e.g., different Array Stride values for the ArrayStride). When new pointers are formed, their types must be decorated as needed, so the consumer knows how to generate an access through the pointer. Non-aggregate non-pointer types are different: It is invalid to declare multiple type <id>s for the same scalar, vector, or matrix type. That is, non-aggregate non-pointer type declarations must all have different opcodes or operands. (Note that non-aggregate non-pointer types cannot be decorated in ways that affect their type.)

Variables are declared to be of an already built type, and placed in a Storage Class. Storage classes include UniformConstant, Input, Workgroup, etc. and are fully specified in Storage Class. Variables declared with the Function Storage Class can have their lifetime’s specified within their function using the OpLifetimeStart and OpLifetimeStop instructions.

Intermediate results are typed by the instruction’s type <id>, which must validate with respect to the operation being done.

Built-in variables have special semantics and are declared using OpDecorate or OpMemberDecorate with the BuiltIn Decoration, followed by a BuiltIn enumerant. See the BuiltIn section for details on what can be decorated as a built-in variable.

2.8.1. Unsigned Versus Signed Integers

The integer type, OpTypeInt, is parameterized not only with a size, but also with signedness. There are two typical ways to think about signedness in SPIR-V, both equally valid:

  1. As if all integers are "signless", meaning they are neither signed nor unsigned: All OpTypeInt instructions select a signedness of 0 to conceptually mean "no sign" (rather than "unsigned"). This is useful when translating from a language that does not distinguish between signed and unsigned types. The type of operation (signed or unsigned) to perform is always selected by the choice of opcode.

  2. As if some integers are signed, and some are unsigned: Some OpTypeInt instructions select signedness of 0 to mean "unsigned" and some select signedness of 1 to mean "signed". This is useful when signedness matters to external interface, or when targeting a higher-level language that cares about types being signed and unsigned. The type of operation (signed or unsigned) to perform is still always selected by the choice of opcode, but a small amount of validation can be done where it is non-sensible to use a signed type.

Note in both cases all signed and unsigned operations always work on unsigned types, and the semantics of operation come from the opcode. SPIR-V does not know which way is being used; it is set up to support both ways of thinking.

2.9. Function Calling

To call a function defined in the current module or a function declared to be imported from another module, use OpFunctionCall with an operand that is the <id> of the OpFunction to call, and the <id>s of the arguments to pass. All arguments are passed by value into the called function. This includes pointers, through which a callee object could be modified.

2.10. Extended Instruction Sets

Many operations and/or built-in function calls from high-level languages are represented through extended instruction sets. Extended instruction sets will include things like

  • trigonometric functions: sin(), cos(), …

  • exponentiation functions: exp(), pow(), …

  • geometry functions: reflect(), smoothstep(), …

  • functions having rich performance/accuracy trade-offs

  • etc.

Non-extended instructions, those that are core SPIR-V instructions, are listed in the Binary Form section. Native operations include:

  • Basic arithmetic: +, -, *, min(), scalar * vector, etc.

  • Texturing, to help with back-end decoding and support special code-motion rules.

  • Derivatives, due to special code-motion rules.

Extended instruction sets are specified in independent specifications. They can be referenced (but not specified) in this specification. The separate extended instruction set specification will specify instruction opcodes, semantics, and instruction names.

To use an extended instruction set, first import it by name string using OpExtInstImport and giving it a Result <id>:

<extinst-id> OpExtInstImport "name-of-extended-instruction-set"

The "name-of-extended-instruction-set" is a literal string. The standard convention for this string is

"<source language name>.<package name>.<version>"

For example "GLSL.std.450" could be the name of the core built-in functions for GLSL versions 450 and earlier.

Note
 There is nothing precluding having two "mirror" sets of instructions with different names but the same opcode values, which could, for example, let modifying just the import statement to change a performance/accuracy trade off.

Then, to call a specific extended instruction, use OpExtInst:

OpExtInst <extinst-id> instruction-number operand0, operand1, ...

Extended instruction-set specifications will provide semantics for each "instruction-number". It is up to the specific specification what the overloading rules are on operand type. The specification must be clear on its semantics, and producers/consumers of it must follow those semantics.

By convention, it is recommended that all external specifications include an enum {…} listing all the "instruction-numbers", and a mapping between these numbers and a string representing the instruction name. However, there are no requirements that instruction name strings are provided or mangled.

Note
 Producing and consuming extended instructions can be done entirely through numbers (no string parsing). An extended instruction set specification provides opcode enumerant values for the instructions, and these will be produced by the front end and consumed by the back end.

2.11. Structured Control Flow

SPIR-V can explicitly declare structured control-flow constructs using merge instructions. These explicitly declare a header block before the control flow diverges and a merge block where control flow subsequently converges. These blocks delimit constructs that must nest, and can only be entered and exited in structured ways, as per the following.

Structured control-flow declarations must satisfy the following rules:

  • the merge block declared by a header block cannot be a merge block declared by any other header block

  • each header block must strictly dominate its merge block, unless the merge block is unreachable in the CFG

  • all CFG back edges must branch to a loop header, with each loop header having exactly one back edge branching to it

  • for a given loop header, its OpLoopMerge Continue Target, and corresponding back-edge block:

    • the loop header must dominate the Continue Target, unless the Continue Target is unreachable in the CFG

    • the Continue Target must dominate the back-edge block

    • the back-edge block must post dominate the Continue Target

A structured control-flow construct is then defined as one of:

  • a selection construct: the set of blocks dominated by a selection header, minus the set of blocks dominated by the header’s merge block

  • a continue construct: the set of blocks dominated by an OpLoopMerge’s Continue Target and post dominated by the corresponding back-edge block

  • a loop construct: the set of blocks dominated by a loop header, minus the set of blocks dominated by the loop’s merge block, minus the loop’s corresponding continue construct

  • a case construct: the set of blocks dominated by an OpSwitch Target or Default, minus the set of blocks dominated by the OpSwitch’s merge block (this construct is only defined for those OpSwitch Target or Default that are not equal to the OpSwitch’s corresponding merge block)

The above structured control-flow constructs must satisfy the following rules:

  • when a construct contains another header block, it also contains that header’s corresponding merge block if that merge block is reachable in the CFG

  • all branches into a construct from reachable blocks outside the construct must be to the header block

  • the only blocks in a construct that can branch outside the construct are

    • a block branching to the construct’s merge block

    • a block branching from one case construct to another, for the same OpSwitch

    • a back-edge block

    • a continue block for the innermost loop it is nested inside of

    • a break block for the innermost loop it is nested inside of

    • a return block

  • additionally for switches:

    • an OpSwitch block dominates all its defined case constructs

    • each case construct has at most one branch to another case construct

    • each case construct is branched to by at most one other case construct

    • if Target T1 branches to Target T2, or if Target T1 branches to the Default and the Default branches to Target T2, then T1 must immediately precede T2 in the list of the OpSwitch Target operands

2.12. Specialization

Specialization is intended for constant objects that will not have known constant values until after initial generation of a SPIR-V module. Such objects are called specialization constants.

A SPIR-V module containing specialization constants can consume one or more externally provided specializations: A set of final constant values for some subset of the module’s specialization constants. Applying these final constant values yields a new module having fewer remaining specialization constants. A module also contains default values for any specialization constants that never get externally specialized.

Note
 No optimizing transforms are required to make a specialized module functionally correct. The specializing transform is straightforward and explicitly defined below.
Note
 Ad hoc specializing should not be done through constants (OpConstant or OpConstantComposite) that get overwritten: A SPIR-V → SPIR-V transform might want to do something irreversible with the value of such a constant, unconstrained from the possibility that its value could be later changed.

Within a module, a Specialization Constant is declared with one of these instructions:

The literal operands to OpSpecConstant are the default numerical specialization constants. Similarly, the "True" and "False" parts of OpSpecConstantTrue and OpSpecConstantFalse provide the default Boolean specialization constants. These default values make an external specialization optional. However, such a default constant is applied only after all external specializations are complete, and none contained a specialization for it.

An external specialization is provided as a logical list of pairs. Each pair is a SpecId Decoration of a scalar specialization instruction along with its specialization constant. The numeric values are exactly what the operands would be to a corresponding OpConstant instruction. Boolean values are true if non-zero and false if zero.

Specializing a module is straightforward. The following specialization-constant instructions can be updated with specialization constants, and replaced in place, leaving everything else in the module exactly the same:

           OpSpecConstantTrue -> OpConstantTrue or OpConstantFalse
          OpSpecConstantFalse -> OpConstantTrue or OpConstantFalse
               OpSpecConstant -> OpConstant
      OpSpecConstantComposite -> OpConstantComposite

The OpSpecConstantOp instruction is specialized by executing the operation and replacing the instruction with the result. The result can be expressed in terms of a constant instruction that is not a specialization-constant instruction. (Note, however, this resulting instruction might not have the same size as the original instruction, so is not a "replaced in place" operation.)

When applying an external specialization, the following (and only the following) must be modified to be non-specialization-constant instructions:

  • specialization-constant instructions with values provided by the specialization

  • specialization-constant instructions that consume nothing but non-specialization constant instructions (including those that the partial specialization transformed from specialization-constant instructions; these are in order, so it is a single pass to do so)

A full specialization can also be done, when requested or required, in which all specialization-constant instructions will be modified to non-specialization-constant instructions, using the default values where required.

2.13. Linkage

The ability to have partially linked modules and libraries is provided as part of the Linkage capability.

By default, functions and global variables are private to a module and cannot be accessed by other modules. However, a module may be written to export or import functions and global (module scope) variables. Imported functions and global variable definitions are resolved at linkage time. A module is considered to be partially linked if it depends on imported values.

Within a module, imported or exported values are decorated using the Linkage Attributes Decoration. This decoration assigns the following linkage attributes to decorated values:

Note
 When resolving imported functions, the Function Control and all Function Parameter Attributes are taken from the function definition, and not from the function declaration.

2.14. Relaxed Precision

The RelaxedPrecision Decoration allows 32-bit integer and 32-bit floating-point operations to execute with a relaxed precision of somewhere between 16 and 32 bits.

For a floating-point operation, operating at relaxed precision means that the minimum requirements for range and precision are as follows:

  • the floating point range may be as small as (-214, 214)

  • the floating point magnitude range may be as small as (2-14, 214)

  • the relative floating point precision may be as small as 2-10

Relative floating-point precision is defined as the worst case (i.e. largest) ratio of the smallest step in relation to the value for all non-zero values:

Precisionrelative = (abs(v1 - v2)min / abs(v1))max for v1 ≠ 0, v2 ≠ 0, v1 ≠ v2

For integer operations, operating at relaxed precision means that the operation will be evaluated by an operation in which, for some N, 16 ≤ N ≤ 32:

  • the operation is executed as though its type were N bits in size, and

  • the result is zero or sign extended to 32 bits as determined by the signedness of the result type of the operation.

The RelaxedPrecision Decoration can be applied to:

  • The <id> of a variable, where the variable’s type is a scalar, vector, or matrix, or an array of scalar, vector, or matrix. In all cases, the components in the type must be a 32-bit numerical type.

  • The Result <id> of an instruction that operates on numerical types, meaning the instruction is to operate at relaxed precision.

  • The Result <id> of an instruction that reads or filters from an image. E.g. OpImageSampleExplicitLod, meaning the instruction is to operate at relaxed precision.

  • The Result <id> of an OpFunction meaning the function’s returned result is at relaxed precision. It cannot be applied to OpTypeFunction or to an OpFunction whose return type is OpTypeVoid.

  • A structure-type member (through OpMemberDecorate).

When applied to a variable or structure member, all loads and stores from the decorated object may be treated as though they were decorated with RelaxedPrecision. Loads may also be decorated with RelaxedPrecision, in which case they are treated as operating at relaxed precision.

All loads and stores involving relaxed precision still read and write 32 bits of data, respectively. Floating-point data read or written in such a manner is written in full 32-bit floating-point format. However, a load or store might reduce the precision (as allowed by RelaxedPrecision) of the destination value.

For debugging portability of floating-point operations, OpQuantizeToF16 may be used to explicitly reduce the precision of a relaxed-precision result to 16-bit precision. (Integer-result precision can be reduced, for example, using left- and right-shift opcodes.)

For image-sampling operations, decorations can appear on both the sampling instruction and the image variable being sampled. If either is decorated, they both should be decorated, and when both are decorated their decorations must match. If only one is decorated, the sampling instruction can behave either as if both were decorated or neither were decorated.

2.15. Debug Information

Debug information is supplied with:

A module will not lose any semantics when all such instructions are removed.

2.15.1. Function-Name Mangling

There is no functional dependency on how functions are named. Signature-typing information is explicitly provided, without any need for name "unmangling". (Valid modules can be created without inclusion of mangled names.)

By convention, for debugging purposes, modules with OpSource Source Language of OpenCL use the Itanium name-mangling standard.

2.16. Validation Rules

2.16.1. Universal Validation Rules

All modules must obey the following, or it is an invalid module:

  • The stream of instructions must be ordered as described in the Logical Layout section.

  • Any use of a feature described by a capability in the capability section requires that capability to be declared, either directly, or as an "implicitly declares" capability on a capability that is declared.

  • Non-structure types (scalars, vectors, arrays, etc.) with the same operand parameterization cannot be type aliases. For non-structures, two type <id>s match if-and-only-if the types match.

  • If the Logical addressing model is selected and the VariablePointers capability is not declared:

  • If the Logical addressing model is selected and the VariablePointers or VariablePointersStorageBuffer capability is declared (in addition to what is allowed above by the Logical addressing model):

    • OpVariable can allocate an object whose type is a pointer type, if the Storage Class of the OpVariable is one of the following:

      • Function

      • Private

    • A pointer can be the Object operand of OpStore or result of OpLoad, if the storage class the pointer is stored to or loaded from is one of the following:

      • Function

      • Private

    • A pointer type can be the:

      • Result Type of OpFunction

      • Result Type of OpFunctionCall

      • Return Type of OpTypeFunction

    • A pointer can be a variable pointer or an operand to one of:

    • A variable pointer must point to one of the following storage classes:

      • StorageBuffer

      • Workgroup (if the VariablePointers capability is declared)

    • If the VariablePointers capability is not declared, a variable pointer must be selected from pointers pointing into the same structure or be OpConstantNull.

    • A pointer operand to OpFunctionCall can point into the storage class:

      • StorageBuffer

    • For pointer operands to OpFunctionCall, the memory object declaration-restriction is removed for the following storage classes:

      • StorageBuffer

      • Workgroup

    • The instructions OpPtrEqual and OpPtrNotEqual can be used only when the Storage Class of the operands' OpTypePointer declaration is

      • StorageBuffer when the VariablePointersStorageBuffer capability is explicitly or implicitly declared, or

      • Workgroup, which can be used only if the VariablePointers capability was declared.

  • A variable pointer with the Logical addressing model cannot

    • be an operand to an OpArrayLength instruction

    • point to an object that is or contains an OpTypeMatrix

    • point to a column, or a component in a column, within an OpTypeMatrix

  • SSA

    • Each <id> must appear exactly once as the Result <id> of an instruction.

    • The definition of an SSA <id> should dominate all uses of it, with the following exceptions:

      • Function calls may call functions not yet defined. However, note that the function’s argument and return types will already be known at the call site.

      • An OpPhi can consume definitions that do not dominate it.

  • Entry Point

  • Functions

    • A function declaration (an OpFunction with no basic blocks), must have a Linkage Attributes Decoration with the Import Linkage Type.

    • A function definition (an OpFunction with basic blocks) cannot be decorated with the Import Linkage Type.

    • A function cannot have both a declaration and a definition (no forward declarations).

  • Global (Module Scope) Variables

    • It is illegal to initialize an imported variable. This means that a module-scope OpVariable with initialization value cannot be marked with the Import Linkage Type.

  • Control-Flow Graph (CFG)

    • Blocks exist only within a function.

    • The first block in a function definition is the entry point of that function and cannot be the target of any branch. (Note this means it will have no OpPhi instructions.)

    • The order of blocks in a function must satisfy the rule that blocks appear before all blocks they dominate.

    • Each block starts with a label.

      • A label is made by OpLabel.

      • This includes the first block of a function (OpFunction is not a label).

      • Labels are used only to form blocks.

    • The last instruction of each block is a termination instruction.

    • Termination instructions can only appear as the last instruction in a block.

    • OpLabel instructions can only appear within a function.

    • All branches within a function must be to labels in that function.

  • All OpFunctionCall Function operands are an <id> of an OpFunction in the same module.

  • Data rules

    • Scalar floating-point types can be parameterized only as 32 bit, plus any additional sizes enabled by capabilities.

    • Scalar integer types can be parameterized only as 32 bit, plus any additional sizes enabled by capabilities.

    • Vector types can only be parameterized with numerical types or the OpTypeBool type.

    • Vector types for can only be parameterized as having 2, 3, or 4 components, plus any additional sizes enabled by capabilities.

    • Matrix types can only be parameterized with floating-point types.

    • Matrix types can only be parameterized as having only 2, 3, or 4 columns.

    • Specialization constants (see Specialization) are limited to integers, Booleans, floating-point numbers, and vectors of these.

    • Forward reference operands in an OpTypeStruct

    • All OpSampledImage instructions must be in the same block in which their Result <id> are consumed. Result <id> from OpSampledImage instructions must not appear as operands to OpPhi instructions or OpSelect instructions, or any instructions other than the image lookup and query instructions specified to take an operand whose type is OpTypeSampledImage.

    • Instructions for extracting a scalar image or scalar sampler out of a composite must only use dynamically-uniform indexes. They must be in the same block in which their Result <id> are consumed. Such Result <id> must not appear as operands to OpPhi instructions or OpSelect instructions, or any instructions other than the image instructions specified to operate on them.

  • Decoration rules

    • The Linkage Attributes Decoration cannot be applied to functions targeted by an OpEntryPoint instruction.

    • A BuiltIn Decoration can only be applied as follows:

      • When applied to a structure-type member, all members of that structure type must also be decorated with BuiltIn. (No allowed mixing of built-in variables and non-built-in variables within a single structure.)

      • When applied to a structure-type member, that structure type cannot be contained as a member of another structure type.

      • There is at most one object per Storage Class that can contain a structure type containing members decorated with BuiltIn, consumed per entry-point.

  • OpLoad and OpStore can only consume objects whose type is a pointer.

  • A Result <id> resulting from an instruction within a function can only be used in that function.

  • A function call must have the same number of arguments as the function definition (or declaration) has parameters, and their respective types must match.

  • An instruction requiring a specific number of operands must have that many operands. The word count must agree.

  • Each opcode specifies its own requirements for number and type of operands, and these must be followed.

  • Atomic access rules

    • The pointers taken by atomic operation instructions must be a pointer into one of the following Storage Classes:

      • Uniform when used with the BufferBlock Decoration

      • StorageBuffer

      • Workgroup

      • CrossWorkgroup

      • Generic

      • AtomicCounter

      • Image

      • Function

  • It is invalid to have a construct that uses the StorageBuffer Storage Class and a construct that uses the Uniform Storage Class with the BufferBlock Decoration in the same SPIR-V module.

  • All XfbStride Decorations must be the same for all objects decorated with the same XfbBuffer XFB Buffer Number.

  • All Stream Decorations must be the same for all objects decorated with the same XfbBuffer XFB Buffer Number.

2.16.2. Validation Rules for Shader Capabilities

  • CFG:

    • Loops must be structured, having an OpLoopMerge instruction in their header.

    • Selections must be structured, having an OpSelectionMerge instruction in their header.

  • Entry point and execution model

  • Composite objects in the StorageBuffer, Uniform, and PushConstant Storage Classes must be explicitly laid out. The following apply to all the aggregate and matrix types describing such an object, recursively through their nested types:

    • Each structure-type member must have an Offset decoration.

    • Each array type must have an ArrayStride decoration, unless it is an array that contains a structure decorated with Block or BufferBlock, in which case it must not have an ArrayStride decoration.

    • Each structure-type member that is a matrix or array-of-matrices must have be decorated with

    • The ArrayStride, MatrixStride, and Offset decorations must be large enough to hold the size of the objects they affect (that is, specifying overlap is invalid). Each ArrayStride and MatrixStride must be greater than zero, and no two members of a given structure can be assigned to the same Offset.

    • Each OpPtrAccessChain must have a Base whose type is decorated with ArrayStride.

    • When an array-element pointer is derived from an array (e.g., using OpAccessChain), and the resulting element-pointer type is decorated with ArrayStride, its Array Stride must match the Array Stride of the array’s type. If the array’s type is not decorated with ArrayStride, the derived array-element pointer also cannot be decorated with ArrayStride.

  • For structure objects in the Input and Output Storage Classes, the following apply:

    • When applied to structure-type members, the decorations Noperspective, Flat, Patch, Centroid, and Sample can only be applied to the top-level members of the structure type. (Nested objects' types cannot be structures whose members are decorated with these decorations.)

  • Data Rules

  • Decorations

    • At most one of Noperspective or Flat decorations can be applied to the same object or member.

    • At most one of Patch, Centroid, or Sample decorations can be applied to the same object or member.

    • At most one of RowMajor and ColMajor decorations can be applied to a structure type.

    • At most one of Block and BufferBlock decorations can be applied to a structure type.

    • Block and BufferBlock decorations cannot decorate a structure type that is nested at any level inside another structure type decorated with Block or BufferBlock.

    • The FPRoundingMode decoration can be applied only to a width-only conversion instruction whose only uses are Object operands of OpStore instructions storing through a pointer to a 16-bit floating-point object in the StorageBuffer, Uniform, or Output Storage Classes.

  • All <id> used for Scope and Memory Semantics must be of an OpConstant.

  • Atomic access rules

    • The pointers taken by atomic operation instructions are further restricted to not point into the Function storage class.

2.16.3. Validation Rules for Kernel Capabilities

  • The Signedness in OpTypeInt must always be 0.

2.17. Universal Limits

These quantities are minimum limits for all implementations and validators. Implementations are allowed to support larger quantities. Client APIs may impose larger minimums. See Language Capabilities.

Validators must either

  • inform when these limits are crossed, or

  • be explicitly parameterized with larger limits.

Table 3. Limits

Limited Entity

Minimum Limit

Decimal

Hexadecimal

Characters in a literal string

65,535

FFFF

Result <id> bound

See Physical Layout for the shader-specific bound.

4,194,303

3FFFFF

Control-flow nesting depth

Measured per function, in program order, counting the maximum number of OpBranch, OpBranchConditional, or OpSwitch that are seen without yet seeing their corresponding Merge Block, as declared by OpSelectionMerge or OpLoopMerge.

1023

3FF

Global variables (Storage Class other than Function)

65,535

FFFF

Local variables (Function Storage Class)

524,287

7FFFF

Decorations per target <id>

Number of entries in the Decoration table.

Execution modes per entry point

255

FF

Indexes for OpAccessChain, OpInBoundsAccessChain, OpPtrAccessChain, OpInBoundsPtrAccessChain, OpCompositeExtract, and OpCompositeInsert

255

FF

Number of function parameters, per function declaration

255

FF

OpFunctionCall actual arguments

255

FF

OpExtInst actual arguments

255

FF

OpSwitch (literal, label) pairs

16,383

3FFF

OpTypeStruct members

16,383

3FFF

Structure nesting depth

255

FF

2.18. Memory Model

A memory model is chosen using a single OpMemoryModel instruction near the beginning of the module. This selects both an addressing model and a memory model.

The Logical addressing model means pointers are abstract, having no physical size or numeric value. In this mode, pointers can only be created from existing objects, and they cannot be stored into an object, unless additional capabilities, e.g., VariablePointers, are declared to add such functionality.

The non-Logical addressing models allow physical pointers to be formed. OpVariable can be used to create objects that hold pointers. These are declared for a specific Storage Class. Pointers for one Storage Class cannot be used to access objects in another Storage Class. However, they can be converted with conversion opcodes. Any particular addressing model must describe the bit width of pointers for each of the storage classes.

2.18.1. Memory Layout

When memory is shared between a SPIR-V module and its client API, its contents are transparent, and must be agreed on. For example, the Offset, MatrixStride, and ArrayStride Decorations can partially define how the memory is laid out. In addition, the following are always true, applied recursively as needed, of the offsets within the memory buffer:

  • a vector consumes contiguous memory with lower-numbered components appearing in smaller offsets than higher-numbered components, and with component 0 starting at the vector’s Offset Decoration, if present

  • in an array, lower-numbered elements appear at smaller offsets than higher-numbered elements, with element 0 starting at the Offset Decoration for the array, if present

  • in a matrix, lower-numbered columns appear at smaller offsets than higher-numbered columns, and lower-numbered components within the matrix’s vectors appearing at smaller offsets than high-numbered components, with component 0 of column 0 starting at the Offset Decoration, if present (the RowMajor and ColMajor Decorations dictate what is contiguous)

2.18.2. Aliasing

Two memory object declarations are said to alias if they can be accessed (in bounds) such that both accesses address the same memory locations. If two memory operations access the same locations, and at least one of them performs a write, then those accesses must be ordered according to the memory consistency model specified by the client API.

Alias management depends on the memory model:

  • The Simple and GLSL memory models can assume that aliasing is generally not present between the memory object declarations. Specifically, the consumer is free to assume aliasing is not present between memory object declarations, unless the memory object declarations explicitly indicate they alias. Aliasing is indicated by applying the Aliased decoration to a memory object declaration’s <id>. Applying Restrict is allowed, but has no effect. Only those memory object declarations decorated with Aliased may alias each other.

  • The OpenCL memory model must, unless otherwise proven, assume that memory object declarations might alias each other. An implementation may assume that memory object declarations decorated with Restrict will not alias any other memory object declaration. Applying Aliased is allowed, but has no effect.

The Aliased decoration can be used to express that certain memory object declarations may alias. Referencing the following table, a memory object declaration P may alias another declared pointer Q if within a single row:

  • P is an instruction with opcode and storage class from the first pair of columns, and

  • Q is an instruction with opcode and storage class from the second pair of columns.

First Storage Class

First Instruction(s)

Second Instructions

Second Storage Classes

CrossWorkgroup

OpFunctionParameter, OpVariable

OpFunctionParameter, OpVariable

CrossWorkgroup, Generic

Function

OpFunctionParameter

OpFunctionParameter, OpVariable

Function, Generic

Function

OpVariable

OpFunctionParameter

Function, Generic

Generic

OpFunctionParameter

OpFunctionParameter, OpVariable

CrossWorkgroup, Function, Generic, Workgroup

Image

OpFunctionParameter, OpVariable

OpFunctionParameter, OpVariable

Image, StorageBuffer, Uniform, UniformConstant

Output

OpFunctionParameter

OpFunctionParameter, OpVariable

Output

Private

OpFunctionParameter

OpFunctionParameter, OpVariable

Private

StorageBuffer

OpFunctionParameter, OpVariable

OpFunctionParameter, OpVariable

Image, StorageBuffer, Uniform, UniformConstant

Uniform

OpFunctionParameter, OpVariable

OpFunctionParameter, OpVariable

Image, StorageBuffer, Uniform, UniformConstant

UniformConstant

OpFunctionParameter, OpVariable

OpFunctionParameter, OpVariable

Image, StorageBuffer, Uniform, UniformConstant

Workgroup

OpFunctionParameter

OpFunctionParameter, OpVariable

Workgroup, Generic

Workgroup

OpVariable

OpFunctionParameter

Workgroup, Generic

In addition to the above table, memory object declarations in the CrossWorkgroup, Function, Input, Output, Private, or Workgroup storage classes must also have matching pointee types for aliasing to be present. In all other cases the decoration is ignored.

Because aliasing, as described above, only applies to memory object declarations, a consumer cannot make any assumptions about whether or not memory regions of non memory object declarations overlap. As such, a consumer must perform dependency analysis on non memory object declarations if it wishes to reorder instructions affecting memory. Behavior is undefined when operations on two memory object declarations access the same memory location, with at least one of them performing a write, and at least one of the memory object declarations does not have the Aliased decoration.

It is invalid to apply both Restrict and Aliased to the same <id>.

2.18.3. Null pointers

A "null pointer" can be formed from an OpConstantNull instruction with a pointer result type. The resulting pointer value is abstract, and will not equal the pointer value formed from any declared object or access chain into a declared object. Behavior is undefined when loading or storing through an OpConstantNull value.

2.19. Derivatives

Derivatives appear only in the Fragment Execution Model. They can be implicit or explicit. Some image instructions consume implicit derivatives, while the derivative instructions compute explicit derivatives. In all cases, derivatives are well defined only if the derivative group has uniform control flow.

2.20. Code Motion

Texturing instructions in the Fragment Execution Model that rely on an implicit derivative cannot be moved into control flow that is not known to be uniform control flow within each derivative group.

2.21. Deprecation

A feature may be marked as deprecated by a version of the specification or extension to the specification. Features marked as deprecated in one version of the specification are still present in that version, but future versions may reduce their support or completely remove them. Deprecating before removing allows applications time to transition away from the deprecated feature. Once the feature is removed, all tokens used exclusively by that feature will be reserved and any use of those tokens will become invalid.

2.22. Unified Specification

This document specifies all versions of SPIR-V.

There are three kinds of entries in the tables of enumerated tokens:

  • Reservation: These say Reserved in the enabling capabilities. They often contain token names only, lacking a semantic description. They are invalid SPIR-V for any version, serving only to reserve the tokens. They may identify enabling capabilities and extensions, in which case any listed extensions might add the tokens. See the listed extensions for additional information.

  • Conditional: These say Missing before or Missing after in the enabling capabilities. They are invalid SPIR-V for the missing versions. They may identify enabling capabilities and extensions, in which case any listed extensions might add the tokens for some of the missing versions. See the listed extensions for additional information. For versions not identified as missing, the tokens are valid SPIR-V, subject to any listed enabling capabilities.

  • Universal: These have no mention of what version they are missing in, or of being reserved. They are valid in all versions of SPIR-V.

3. Binary Form

This section contains the exact form for all instructions, starting with the numerical values for all fields. See Physical Layout for the order words appear in.

3.1. Magic Number

Magic number for a SPIR-V module.

Tip
 Endianness: A module is defined as a stream of words, not a stream of bytes. However, if stored as a stream of bytes (e.g., in a file), the magic number can be used to deduce what endianness to apply to convert the byte stream back to a word stream.
Magic Number

0x07230203

3.2. Source Language

The source language is for debug purposes only, with no semantics that affect the meaning of other parts of the module. Used by OpSource.

Source Language

0

Unknown

1

ESSL

2

GLSL

3

OpenCL_C

4

OpenCL_CPP

5

HLSL

3.3. Execution Model

Used by OpEntryPoint.

Execution Model Enabling Capabilities

0

Vertex
Vertex shading stage.

Shader

1

TessellationControl
Tessellation control (or hull) shading stage.

Tessellation

2

TessellationEvaluation
Tessellation evaluation (or domain) shading stage.

Tessellation

3

Geometry
Geometry shading stage.

Geometry

4

Fragment
Fragment shading stage.

Shader

5

GLCompute
Graphical compute shading stage.

Shader

6

Kernel
Compute kernel.

Kernel

5267

TaskNV

MeshShadingNV

5268

MeshNV

MeshShadingNV

5313

RayGenerationNV

RayTracingNV

5314

IntersectionNV

RayTracingNV

5315

AnyHitNV

RayTracingNV

5316

ClosestHitNV

RayTracingNV

5317

MissNV

RayTracingNV

5318

CallableNV

RayTracingNV

3.4. Addressing Model

Used by OpMemoryModel.

Addressing Model Enabling Capabilities

0

Logical

1

Physical32
Indicates a 32-bit module, where the address width is equal to 32 bits.

Addresses

2

Physical64
Indicates a 64-bit module, where the address width is equal to 64 bits.

Addresses

5348

PhysicalStorageBuffer64EXT

PhysicalStorageBufferAddressesEXT

Also see extension: SPV_EXT_physical_storage_buffer

3.5. Memory Model

Used by OpMemoryModel.

Memory Model Enabling Capabilities

0

Simple
No shared memory consistency issues.

Shader

1

GLSL450
Memory model needed by later versions of GLSL and ESSL. Works across multiple versions.

Shader

2

OpenCL
OpenCL memory model.

Kernel

3

VulkanKHR

VulkanMemoryModelKHR

3.6. Execution Mode

Declare the modes an entry point will execute in. Used by OpExecutionMode and OpExecutionModeId.

Execution Mode Extra Operands Enabling Capabilities

0

Invocations
Number of times to invoke the geometry stage for each input primitive received. The default is to run once for each input primitive. It is invalid to specify a value greater than the target-dependent maximum. Only valid with the Geometry Execution Model.

Literal Number
Number of invocations

Geometry

1

SpacingEqual
Requests the tessellation primitive generator to divide edges into a collection of equal-sized segments. Only valid with one of the tessellation Execution Models.

Tessellation

2

SpacingFractionalEven
Requests the tessellation primitive generator to divide edges into an even number of equal-length segments plus two additional shorter fractional segments. Only valid with one of the tessellation Execution Models.

Tessellation

3

SpacingFractionalOdd
Requests the tessellation primitive generator to divide edges into an odd number of equal-length segments plus two additional shorter fractional segments. Only valid with one of the tessellation Execution Models.

Tessellation

4

VertexOrderCw
Requests the tessellation primitive generator to generate triangles in clockwise order. Only valid with one of the tessellation Execution Models.

Tessellation

5

VertexOrderCcw
Requests the tessellation primitive generator to generate triangles in counter-clockwise order. Only valid with one of the tessellation Execution Models.

Tessellation

6

PixelCenterInteger
Pixels appear centered on whole-number pixel offsets. E.g., the coordinate (0.5, 0.5) appears to move to (0.0, 0.0). Only valid with the Fragment Execution Model. If a Fragment entry point does not have this set, pixels appear centered at offsets of (0.5, 0.5) from whole numbers

Shader

7

OriginUpperLeft
The coordinates decorated by FragCoord appear to originate in the upper left, and increase toward the right and downward. Only valid with the Fragment Execution Model.

Shader

8

OriginLowerLeft
The coordinates decorated by FragCoord appear to originate in the lower left, and increase toward the right and upward. Only valid with the Fragment Execution Model.

Shader

9

EarlyFragmentTests
Fragment tests are to be performed before fragment shader execution. Only valid with the Fragment Execution Model.

Shader

10

PointMode
Requests the tessellation primitive generator to generate a point for each distinct vertex in the subdivided primitive, rather than to generate lines or triangles. Only valid with one of the tessellation Execution Models.

Tessellation

11

Xfb
This stage will run in transform feedback-capturing mode and this module is responsible for describing the transform-feedback setup. See the XfbBuffer, Offset, and XfbStride Decorations.

TransformFeedback

12

DepthReplacing
This mode must be declared if and only if this entry point dynamically writes the FragDepth-decorated variable. Only valid with the Fragment Execution Model.

Shader

14

DepthGreater
Indicates that per-fragment tests may assume that any FragDepth built in-decorated value written by the shader will be greater-than-or-equal to the fragment’s interpolated depth value (given by the z component of the FragCoord built in-decorated variable). Other stages of the pipeline use the written value as normal. Only valid with the Fragment execution model.

Shader

15

DepthLess
Indicates that per-fragment tests may assume that any FragDepth built in-decorated value written by the shader will be less than the fragment’s interpolated depth value (given by the z component of the FragCoord built in-decorated variable). Other stages of the pipeline use the written value as normal. Only valid with the Fragment execution model.

Shader

16

DepthUnchanged
Indicates that per-fragment tests may assume that any FragDepth built in-decorated value written by the shader will be the same as the fragment’s interpolated depth value (given by the z component of the FragCoord built in-decorated variable). Other stages of the pipeline use the written value as normal. Only valid with the Fragment execution model.

Shader

17

LocalSize
Indicates the work-group size in the x, y, and z dimensions. Only valid with the GLCompute or Kernel Execution Models.

Literal Number
x size

Literal Number
y size

Literal Number
z size

18

LocalSizeHint
A hint to the compiler, which indicates the most likely to be used work-group size in the x, y, and z dimensions. Only valid with the Kernel Execution Model.

Literal Number
x size

Literal Number
y size

Literal Number
z size

Kernel

19

InputPoints
Stage input primitive is points. Only valid with the Geometry Execution Model.

Geometry

20

InputLines
Stage input primitive is lines. Only valid with the Geometry Execution Model.

Geometry

21

InputLinesAdjacency
Stage input primitive is lines adjacency. Only valid with the Geometry Execution Model.

Geometry

22

Triangles
For a geometry stage, input primitive is triangles. For a tessellation stage, requests the tessellation primitive generator to generate triangles. Only valid with the Geometry or one of the tessellation Execution Models.

Geometry, Tessellation

23

InputTrianglesAdjacency
Geometry stage input primitive is triangles adjacency. Only valid with the Geometry Execution Model.

Geometry

24

Quads
Requests the tessellation primitive generator to generate quads. Only valid with one of the tessellation Execution Models.

Tessellation

25

Isolines
Requests the tessellation primitive generator to generate isolines. Only valid with one of the tessellation Execution Models.

Tessellation

26

OutputVertices
For a geometry stage, the maximum number of vertices the shader will ever emit in a single invocation. For a tessellation-control stage, the number of vertices in the output patch produced by the tessellation control shader, which also specifies the number of times the tessellation control shader is invoked. Only valid with the Geometry or one of the tessellation Execution Models.

Literal Number
Vertex count

Geometry, Tessellation, MeshShadingNV

27

OutputPoints
Stage output primitive is points. Only valid with the Geometry Execution Model.

Geometry, MeshShadingNV

28

OutputLineStrip
Stage output primitive is line strip. Only valid with the Geometry Execution Model.

Geometry

29

OutputTriangleStrip
Stage output primitive is triangle strip. Only valid with the Geometry Execution Model.

Geometry

30

VecTypeHint
A hint to the compiler, which indicates that most operations used in the entry point are explicitly vectorized using a particular vector type. The 16 high-order bits of Vector Type operand specify the number of components of the vector. The 16 low-order bits of Vector Type operand specify the data type of the vector.

These are the legal data type values:
0 represents an 8-bit integer value.
1 represents a 16-bit integer value.
2 represents a 32-bit integer value.
3 represents a 64-bit integer value.
4 represents a 16-bit float value.
5 represents a 32-bit float value.
6 represents a 64-bit float value.

Only valid with the Kernel Execution Model.

Literal Number
Vector type

Kernel

31

ContractionOff
Indicates that floating-point-expressions contraction is disallowed. Only valid with the Kernel Execution Model.

Kernel

33

Initializer
Indicates that this entry point is a module initializer.

Kernel

Missing before version 1.1.

34

Finalizer
Indicates that this entry point is a module finalizer.

Kernel

Missing before version 1.1.

35

SubgroupSize
Indicates that this entry point requires the specified Subgroup Size.

Literal Number
Subgroup Size

SubgroupDispatch

Missing before version 1.1.

36

SubgroupsPerWorkgroup
Indicates that this entry point requires the specified number of Subgroups Per Workgroup.

Literal Number
Subgroups Per Workgroup

SubgroupDispatch

Missing before version 1.1.

37

SubgroupsPerWorkgroupId
Indicates that this entry point requires the specified number of Subgroups Per Workgroup.

Specified as an Id.

<id>
Subgroups Per Workgroup

SubgroupDispatch

Missing before version 1.2.

38

LocalSizeId
Indicates the work-group size in the x, y, and z dimensions. Only valid with the GLCompute or Kernel Execution Models.

Specified as Ids.

<id>
x size

<id>
y size

<id>
z size

Missing before version 1.2.

39

LocalSizeHintId
A hint to the compiler, which indicates the most likely to be used work-group size in the x, y, and z dimensions. Only valid with the Kernel Execution Model.

Specified as an Id.

<id>
Local Size Hint

Kernel

Missing before version 1.2.

4446

PostDepthCoverage

SampleMaskPostDepthCoverage

Reserved.

Also see extension: SPV_KHR_post_depth_coverage

4459

DenormPreserve
Any denormalized value input into a shader or potentially generated by any instruction in a shader must be preserved. Denormalized values obtained via unpacking an integer into a vector of values with smaller bit width and interpreting those values as floating-point numbers must be preserved.

Only affects instructions operating on a floating-point type whose component width is Target Width.

Literal Number
Target Width

DenormPreserve

Missing before version 1.4.

Also see extension: SPV_KHR_float_controls

4460

DenormFlushToZero
Any denormalized value input into a shader or potentially generated by any instruction in a shader must be flushed to zero. Denormalized values obtained via unpacking an integer into a vector of values with smaller bit width and interpreting those values as floating-point numbers must be flushed to zero.

Only affects instructions operating on a floating-point type whose component width is Target Width.

Literal Number
Target Width

DenormFlushToZero

Missing before version 1.4.

Also see extension: SPV_KHR_float_controls

4461

SignedZeroInfNanPreserve
The implementation must not perform optimizations on floating-point instructions that do not preserve sign of a zero, or assume that operands and results are not NaNs or infinities. Bit patterns for NaNs might not be preserved.

Only affects instructions operating on a floating-point type whose component width is Target Width.

Literal Number
Target Width

SignedZeroInfNanPreserve

Missing before version 1.4.

Also see extension: SPV_KHR_float_controls

4462

RoundingModeRTE
The default rounding mode for floating-point arithmetic and conversions instructions must be round to nearest even. If an instruction is decorated with FPRoundingMode or defines a rounding mode in its description, that rounding mode is applied and RoundingModeRTE is ignored.

Only affects instructions operating on a floating-point type whose component width is Target Width.

Literal Number
Target Width

RoundingModeRTE

Missing before version 1.4.

Also see extension: SPV_KHR_float_controls

4463

RoundingModeRTZ
The default rounding mode for floating-point arithmetic and conversions instructions must be round toward zero. If an instruction is decorated with FPRoundingMode or defines a rounding mode in its description, that rounding mode is applied and RoundingModeRTZ is ignored.

Only affects instructions operating on a floating-point type whose component width is Target Width.

Literal Number
Target Width

RoundingModeRTZ

Missing before version 1.4.

Also see extension: SPV_KHR_float_controls

5027

StencilRefReplacingEXT

StencilExportEXT

Reserved.

Also see extension: SPV_EXT_shader_stencil_export

5269

OutputLinesNV

MeshShadingNV

Reserved.

Also see extension: SPV_NV_mesh_shader

5270

OutputPrimitivesNV

Literal Number
Primitive count

MeshShadingNV

Reserved.

Also see extension: SPV_NV_mesh_shader

5289

DerivativeGroupQuadsNV

ComputeDerivativeGroupQuadsNV

Reserved.

Also see extension: SPV_NV_compute_shader_derivatives

5290

DerivativeGroupLinearNV

ComputeDerivativeGroupLinearNV

Reserved.

Also see extension: SPV_NV_compute_shader_derivatives

5298

OutputTrianglesNV

MeshShadingNV

Reserved.

Also see extension: SPV_NV_mesh_shader

3.7. Storage Class

Class of storage for declared variables (does not include intermediate values). Used by:

Storage Class Enabling Capabilities

0

UniformConstant
Shared externally, visible across all functions in all invocations in all work groups. Graphics uniform memory. OpenCL constant memory. Variables declared with this storage class are read-only. They may have initializers, as allowed by the client API.

1

Input
Input from pipeline. Visible across all functions in the current invocation. Variables declared with this storage class are read-only, and cannot have initializers.

2

Uniform
Shared externally, visible across all functions in all invocations in all work groups. Graphics uniform blocks and buffer blocks.

Shader

3

Output
Output to pipeline. Visible across all functions in the current invocation.

Shader

4

Workgroup
Shared across all invocations within a work group. Visible across all functions. The OpenGL "shared" storage qualifier. OpenCL local memory.

5

CrossWorkgroup
Visible across all functions of all invocations of all work groups. OpenCL global memory.

6

Private
Visible to all functions in the current invocation. Regular global memory.

Shader

7

Function
Visible only within the declaring function of the current invocation. Regular function memory.

8

Generic
For generic pointers, which overload the Function, Workgroup, and CrossWorkgroup Storage Classes.

GenericPointer

9

PushConstant
For holding push-constant memory, visible across all functions in all invocations in all work groups. Intended to contain a small bank of values pushed from the client API. Variables declared with this storage class are read-only, and cannot have initializers.

Shader

10

AtomicCounter
For holding atomic counters. Visible across all functions of the current invocation. Atomic counter-specific memory.

AtomicStorage

11

Image
For holding image memory.

12

StorageBuffer
Shared externally, readable and writable, visible across all functions in all invocations in all work groups. Graphics storage buffers (buffer blocks).

Shader

Missing before version 1.3.

Also see extensions: SPV_KHR_storage_buffer_storage_class, SPV_KHR_variable_pointers

5328

CallableDataNV

RayTracingNV

Also see extension: SPV_NV_ray_tracing

5329

IncomingCallableDataNV

RayTracingNV

Also see extension: SPV_NV_ray_tracing

5338

RayPayloadNV

RayTracingNV

Also see extension: SPV_NV_ray_tracing

5339

HitAttributeNV

RayTracingNV

Also see extension: SPV_NV_ray_tracing

5342

IncomingRayPayloadNV

RayTracingNV

Also see extension: SPV_NV_ray_tracing

5343

ShaderRecordBufferNV

RayTracingNV

Also see extension: SPV_NV_ray_tracing

5349

PhysicalStorageBufferEXT

PhysicalStorageBufferAddressesEXT

Also see extension: SPV_EXT_physical_storage_buffer

3.8. Dim

Dimensionality of an image. The listed Array capabilities are required if the type’s Arrayed operand is 1. The listed Image capabilities are required if the type’s Sampled operand is 2. Used by OpTypeImage.

Dim Enabling Capabilities

0

1D

Sampled1D, Image1D

1

2D

Shader, Kernel, ImageMSArray

2

3D

3

Cube

Shader, ImageCubeArray

4

Rect

SampledRect, ImageRect

5

Buffer

SampledBuffer, ImageBuffer

6

SubpassData

InputAttachment

3.9. Sampler Addressing Mode

Addressing mode for creating constant samplers. Used by OpConstantSampler.

Sampler Addressing Mode Enabling Capabilities

0

None
The image coordinates used to sample elements of the image refer to a location inside the image, otherwise the results are undefined.

Kernel

1

ClampToEdge
Out-of-range image coordinates are clamped to the extent.

Kernel

2

Clamp
Out-of-range image coordinates will return a border color.

Kernel

3

Repeat
Out-of-range image coordinates are wrapped to the valid range. Can only be used with normalized coordinates.

Kernel

4

RepeatMirrored
Flip the image coordinate at every integer junction. Can only be used with normalized coordinates.

Kernel

3.10. Sampler Filter Mode

Filter mode for creating constant samplers. Used by OpConstantSampler.

Sampler Filter Mode Enabling Capabilities

0

Nearest
Use filter nearest mode when performing a read image operation.

Kernel

1

Linear
Use filter linear mode when performing a read image operation.

Kernel

3.11. Image Format

Declarative image format. Used by OpTypeImage.

Image Format Enabling Capabilities

0

Unknown

1

Rgba32f

Shader

2

Rgba16f

Shader

3

R32f

Shader

4

Rgba8

Shader

5

Rgba8Snorm

Shader

6

Rg32f

StorageImageExtendedFormats

7

Rg16f

StorageImageExtendedFormats

8

R11fG11fB10f

StorageImageExtendedFormats

9

R16f

StorageImageExtendedFormats

10

Rgba16

StorageImageExtendedFormats

11

Rgb10A2

StorageImageExtendedFormats

12

Rg16

StorageImageExtendedFormats

13

Rg8

StorageImageExtendedFormats

14

R16

StorageImageExtendedFormats

15

R8

StorageImageExtendedFormats

16

Rgba16Snorm

StorageImageExtendedFormats

17

Rg16Snorm

StorageImageExtendedFormats

18

Rg8Snorm

StorageImageExtendedFormats

19

R16Snorm

StorageImageExtendedFormats

20

R8Snorm

StorageImageExtendedFormats

21

Rgba32i

Shader

22

Rgba16i

Shader

23

Rgba8i

Shader

24

R32i

Shader

25

Rg32i

StorageImageExtendedFormats

26

Rg16i

StorageImageExtendedFormats

27

Rg8i

StorageImageExtendedFormats

28

R16i

StorageImageExtendedFormats

29

R8i

StorageImageExtendedFormats

30

Rgba32ui

Shader

31

Rgba16ui

Shader

32

Rgba8ui

Shader

33

R32ui

Shader

34

Rgb10a2ui

StorageImageExtendedFormats

35

Rg32ui

StorageImageExtendedFormats

36

Rg16ui

StorageImageExtendedFormats

37

Rg8ui

StorageImageExtendedFormats

38

R16ui

StorageImageExtendedFormats

39

R8ui

StorageImageExtendedFormats

3.12. Image Channel Order

Image channel order returned by OpImageQueryOrder.

Image Channel Order Enabling Capabilities

0

R

Kernel

1

A

Kernel

2

RG

Kernel

3

RA

Kernel

4

RGB

Kernel

5

RGBA

Kernel

6

BGRA

Kernel

7

ARGB

Kernel

8

Intensity

Kernel

9

Luminance

Kernel

10

Rx

Kernel

11

RGx

Kernel

12

RGBx

Kernel

13

Depth

Kernel

14

DepthStencil

Kernel

15

sRGB

Kernel

16

sRGBx

Kernel

17

sRGBA

Kernel

18

sBGRA

Kernel

19

ABGR

Kernel

3.13. Image Channel Data Type

Image channel data type returned by OpImageQueryFormat.

Image Channel Data Type Enabling Capabilities

0

SnormInt8

Kernel

1

SnormInt16

Kernel

2

UnormInt8

Kernel

3

UnormInt16

Kernel

4

UnormShort565

Kernel

5

UnormShort555

Kernel

6

UnormInt101010

Kernel

7

SignedInt8

Kernel

8

SignedInt16

Kernel

9

SignedInt32

Kernel

10

UnsignedInt8

Kernel

11

UnsignedInt16

Kernel

12

UnsignedInt32

Kernel

13

HalfFloat

Kernel

14

Float

Kernel

15

UnormInt24

Kernel

16

UnormInt101010_2

Kernel

3.14. Image Operands

Additional operands to sampling, or getting texels from, an image. Bits that are set can indicate whether an additional operand follows, as described by the table. If there are multiple following operands indicated, they are ordered: Those indicated by smaller-numbered bits appear first. At least one bit must be set (None is invalid).

This value is a literal mask; it can be formed by combining the bits from multiple rows in the table below.

Used by:

Image Operands Enabling Capabilities

0x0

None

0x1

Bias
A following operand is the bias added to the implicit level of detail. Only valid with implicit-lod instructions. It must be a floating-point type scalar. This can only be used with an OpTypeImage that has a Dim operand of 1D, 2D, 3D, or Cube, and the MS operand must be 0.

Shader

0x2

Lod
A following operand is the explicit level-of-detail to use. Only valid with explicit-lod instructions. For sampling operations, it must be a floating-point type scalar. For fetch operations, it must be an integer type scalar. This can only be used with an OpTypeImage that has a Dim operand of 1D, 2D, 3D, or Cube, and the MS operand must be 0.

0x4

Grad
Two following operands are dx followed by dy. These are explicit derivatives in the x and y direction to use in computing level of detail. Each is a scalar or vector containing (du/dx[, dv/dx] [, dw/dx]) and (du/dy[, dv/dy] [, dw/dy]). The number of components of each must equal the number of components in Coordinate, minus the array layer component, if present. Only valid with explicit-lod instructions. They must be a scalar or vector of floating-point type. This can only be used with an OpTypeImage that has an MS operand of 0. It is invalid to set both the Lod and Grad bits.

0x8

ConstOffset
A following operand is added to (u, v, w) before texel lookup. It must be an <id> of an integer-based constant instruction of scalar or vector type. It is invalid for these to be outside a target-dependent allowed range. The number of components must equal the number of components in Coordinate, minus the array layer component, if present. Not valid with the Cube dimension.

0x10

Offset
A following operand is added to (u, v, w) before texel lookup. It must be a scalar or vector of integer type. It is invalid for these to be outside a target-dependent allowed range. The number of components must equal the number of components in Coordinate, minus the array layer component, if present. Not valid with the Cube dimension.

ImageGatherExtended

0x20

ConstOffsets
A following operand is Offsets. Offsets must be an <id> of a constant instruction making an array of size four of vectors of two integer components. Each gathered texel is identified by adding one of these array elements to the (u, v) sampled location. It is invalid for these to be outside a target-dependent allowed range. Only valid with OpImageGather or OpImageDrefGather. Not valid with the Cube dimension.

ImageGatherExtended

0x40

Sample
A following operand is the sample number of the sample to use. Only valid with OpImageFetch, OpImageRead, OpImageWrite, OpImageSparseFetch, and OpImageSparseRead. It is invalid to have a Sample operand if the underlying OpTypeImage has MS of 0. It must be an integer type scalar.

0x80

MinLod
A following operand is the minimum level-of-detail to use when accessing the image. Only valid with Implicit instructions and Grad instructions. It must be a floating-point type scalar. This can only be used with an OpTypeImage that has a Dim operand of 1D, 2D, 3D, or Cube, and the MS operand must be 0.

MinLod

0x100

MakeTexelAvailableKHR

VulkanMemoryModelKHR

0x200

MakeTexelVisibleKHR

VulkanMemoryModelKHR

0x400

NonPrivateTexelKHR

VulkanMemoryModelKHR

0x800

VolatileTexelKHR

VulkanMemoryModelKHR

0x1000

SignExtend
The texel value is converted to the target value via sign extension. Only valid when the texel type is a scalar or vector of integer type.

Missing before version 1.4.

0x2000

ZeroExtend
The texel value is converted to the target value via zero extension. Only valid when the texel type is a scalar or vector of integer type.

Missing before version 1.4.

3.15. FP Fast Math Mode

Enables fast math operations which are otherwise unsafe.

This value is a literal mask; it can be formed by combining the bits from multiple rows in the table below.

FP Fast Math Mode Enabling Capabilities

0x0

None

0x1

NotNaN
Assume parameters and result are not NaN.

Kernel

0x2

NotInf
Assume parameters and result are not +/- Inf.

Kernel

0x4

NSZ
Treat the sign of a zero parameter or result as insignificant.

Kernel

0x8

AllowRecip
Allow the usage of reciprocal rather than perform a division.

Kernel

0x10

Fast
Allow algebraic transformations according to real-number associative and distributive algebra. This flag implies all the others.

Kernel

3.16. FP Rounding Mode

Associate a rounding mode to a floating-point conversion instruction.

FP Rounding Mode

0

RTE
Round to nearest even.

1

RTZ
Round towards zero.

2

RTP
Round towards positive infinity.

3

RTN
Round towards negative infinity.

3.17. Linkage Type

Associate a linkage type to functions or global variables. See linkage.

Linkage Type Enabling Capabilities

0

Export
Accessible by other modules as well.

Linkage

1

Import
A declaration of a global variable or a function that exists in another module.

Linkage

3.18. Access Qualifier

Defines the access permissions.

Used by OpTypeImage and OpTypePipe.

Access Qualifier Enabling Capabilities

0

ReadOnly
A read-only object.

Kernel

1

WriteOnly
A write-only object.

Kernel

2

ReadWrite
A readable and writable object.

Kernel

3.19. Function Parameter Attribute

Adds additional information to the return type and to each parameter of a function.

Function Parameter Attribute Enabling Capabilities

0

Zext
Value should be zero extended if needed.

Kernel

1

Sext
Value should be sign extended if needed.

Kernel

2

ByVal
This indicates that the pointer parameter should really be passed by value to the function. Only valid for pointer parameters (not for ret value).

Kernel

3

Sret
Indicates that the pointer parameter specifies the address of a structure that is the return value of the function in the source program. Only applicable to the first parameter which must be a pointer parameters.

Kernel

4

NoAlias
Indicates that the memory pointed to by a pointer parameter is not accessed via pointer values which are not derived from this pointer parameter. Only valid for pointer parameters. Not valid on return values.

Kernel

5

NoCapture
The callee does not make a copy of the pointer parameter into a location that is accessible after returning from the callee. Only valid for pointer parameters. Not valid on return values.

Kernel

6

NoWrite
Can only read the memory pointed to by a pointer parameter. Only valid for pointer parameters. Not valid on return values.

Kernel

7

NoReadWrite
Cannot dereference the memory pointed to by a pointer parameter. Only valid for pointer parameters. Not valid on return values.

Kernel

3.20. Decoration

Used by:

Decoration Extra Operands Enabling Capabilities

0

RelaxedPrecision
Allow reduced precision operations. To be used as described in Relaxed Precision.

Shader

1

SpecId
Apply to a scalar specialization constant. Forms the external linkage for setting a specialized value. See specialization.

Literal Number
Specialization Constant ID

Shader, Kernel

2

Block
Apply to a structure type to establish it is a non-SSBO-like shader-interface block.

Shader

3

BufferBlock
Deprecated (use Block-decorated StorageBuffer Storage Class objects).
Apply to a structure type to establish it is an SSBO-like shader-interface block.

Shader

Missing after version 1.3.

4

RowMajor
Applies only to a member of a structure type. Only valid on a matrix or array whose most basic element is a matrix. Indicates that components within a row are contiguous in memory.

Matrix

5

ColMajor
Applies only to a member of a structure type. Only valid on a matrix or array whose most basic element is a matrix. Indicates that components within a column are contiguous in memory.

Matrix

6

ArrayStride
Apply to an array type to specify the stride, in bytes, of the array’s elements. Can also apply to a pointer type to an array element, to specify the stride of the array that the element resides in. Must not be applied to any other type.

Literal Number
Array Stride

Shader

7

MatrixStride
Applies only to a member of a structure type. Only valid on a matrix or array whose most basic element is a matrix. Specifies the stride of rows in a RowMajor-decorated matrix, or columns in a ColMajor-decorated matrix.

Literal Number
Matrix Stride

Matrix

8

GLSLShared
Apply to a structure type to get GLSL shared memory layout.

Shader

9

GLSLPacked
Apply to a structure type to get GLSL packed memory layout.

Shader

10

CPacked
Apply to a structure type, to marks it as "packed", indicating that the alignment of the structure is one and that there is no padding between structure members.

Kernel

11

BuiltIn
Indicates which built-in variable an object represents. See BuiltIn for more information.

BuiltIn

13

NoPerspective
Must only be used on a memory object declaration or a member of a structure type. Indicates that linear, non-perspective correct, interpolation must be used. Only valid for the Input and Output Storage Classes.

Shader

14

Flat
Must only be used on a memory object declaration or a member of a structure type. Indicates no interpolation will be done. The non-interpolated value will come from a vertex, as specified by the client API. Only valid for the Input and Output Storage Classes.

Shader

15

Patch
Must only be used on a memory object declaration or a member of a structure type. Indicates a tessellation patch. Only valid for the Input and Output Storage Classes. Invalid to use on objects or types referenced by non-tessellation Execution Models.

Tessellation

16

Centroid
Must only be used on a memory object declaration or a member of a structure type. When used with multi-sampling rasterization, allows a single interpolation location for an entire pixel. The interpolation location must lie in both the pixel and in the primitive being rasterized. Only valid for the Input and Output Storage Classes.

Shader

17

Sample
Must only be used on a memory object declaration or a member of a structure type. When used with multi-sampling rasterization, requires per-sample interpolation. The interpolation locations must be the locations of the samples lying in both the pixel and in the primitive being rasterized. Only valid for the Input and Output Storage Classes.

SampleRateShading

18

Invariant
Apply to a variable, to indicate expressions computing its value be done invariant with respect to other modules computing the same expressions.

Shader

19

Restrict
Apply to a memory object declaration, to indicate the compiler may compile as if there is no aliasing. See the Aliasing section for more detail.

20

Aliased
Apply to a memory object declaration, to indicate the compiler is to generate accesses to the variable that work correctly in the presence of aliasing. See the Aliasing section for more detail.

21

Volatile
Must be applied only to memory object declarations or members of a structure type. Any such memory object declaration, or any memory object declaration that contains such a structure type, must be one of:
- A storage image (see OpTypeImage).
- A block in the StorageBuffer storage class, or in the Uniform storage class with the BufferBlock decoration.
This indicates the memory holding the variable is volatile memory. Accesses to volatile memory cannot be eliminated, duplicated, or combined with other accesses.

22

Constant
Indicates that a global variable is constant and will never be modified. Only allowed on global variables.

Kernel

23

Coherent
Must be applied only to memory object declarations or members of a structure type. Any such memory object declaration, or any memory object declaration that contains such a structure type, must be one of:
- A storage image (see OpTypeImage).
- A block in the StorageBuffer storage class, or in the Uniform storage class with the BufferBlock decoration.
This indicates the memory backing the object is coherent.

24

NonWritable
Must be applied only to memory object declarations or members of a structure type. Any such memory object declaration, or any memory object declaration that contains such a structure type, must be one of:
- A storage image (see OpTypeImage).
- A block in the StorageBuffer storage class, or in the Uniform storage class with the BufferBlock decoration.
- Missing before version 1.4: An object in the Private or Function storage classes.
This decoration indicates the memory holding the variable is not writable, and that this module does not write to it. It does not prevent the use of initializers on a declaration.

25

NonReadable
Must be applied only to memory object declarations or members of a structure type. Any such memory object declaration, or any memory object declaration that contains such a structure type, must be one of:
- A storage image (see OpTypeImage).
- A block in the StorageBuffer storage class, or in the Uniform storage class with the BufferBlock decoration.
This indicates the memory holding the variable is not readable, and that this module does not read from it.

26

Uniform
Apply to an object. Asserts that, for each dynamic instance of the instruction that computes the result, all active invocations in the invocation’s Subgroup scope will compute the same result value.

Shader

27

UniformId
Apply to an object. Asserts that, for each dynamic instance of the instruction that computes the result, all active invocations in the Execution scope compute the same result value. Execution must not be Invocation.

Scope <id>
Execution

Shader

Missing before version 1.4.

28

SaturatedConversion
Indicates that a conversion to an integer type which is outside the representable range of Result Type will be clamped to the nearest representable value of Result Type. NaN will be converted to 0.

This decoration can only be applied to conversion instructions to integer types, not including the OpSatConvertUToS and OpSatConvertSToU instructions.

Kernel

29

Stream
Must only be used on a memory object declaration or a member of a structure type. Indicates the stream number to put an output on. Only valid for the Output Storage Class and the Geometry Execution Model.

Literal Number
Stream Number

GeometryStreams

30

Location
Apply to a variable or a structure-type member. Forms the main linkage for Storage Class Input and Output variables:
- between the client API and vertex-stage inputs,
- between consecutive programmable stages, or
- between fragment-stage outputs and the client API.
Also can tag variables or structure-type members in the UniformConstant Storage Class for linkage with the client API.
Only valid for the Input, Output, and UniformConstant Storage Classes.

Literal Number
Location

Shader

31

Component
Must only be used on a memory object declaration or a member of a structure type. Indicates which component within a Location will be taken by the decorated entity. Only valid for the Input and Output Storage Classes.

Literal Number
Component

Shader

32

Index
Apply to a variable to identify a blend equation input index, used as specified by the client API. Only valid for the Output Storage Class and the Fragment Execution Model.

Literal Number
Index

Shader

33

Binding
Apply to a variable. Part of the main linkage between the client API and SPIR-V modules for memory buffers, images, etc. See the client API specification for more detail.

Literal Number
Binding Point

Shader

34

DescriptorSet
Apply to a variable. Part of the main linkage between the client API and SPIR-V modules for memory buffers, images, etc. See the client API specification for more detail.

Literal Number
Descriptor Set

Shader

35

Offset
Apply to a structure-type member. This gives the byte offset of the member relative to the beginning of the structure. Can be used, for example, by both uniform and transform-feedback buffers. It must not cause any overlap of the structure’s members, or overflow of a transform-feedback buffer’s XfbStride.

Literal Number
Byte Offset

Shader

36

XfbBuffer
Must only be used on a memory object declaration or a member of a structure type. Indicates which transform-feedback buffer an output is written to. Only valid for the Output Storage Classes of vertex processing Execution Models.

Literal Number
XFB Buffer Number

TransformFeedback

37

XfbStride
Apply to anything XfbBuffer is applied to. Specifies the stride, in bytes, of transform-feedback buffer vertices. If the transform-feedback buffer is capturing any double-precision components, the stride must be a multiple of 8, otherwise it must be a multiple of 4.

Literal Number
XFB Stride

TransformFeedback

38

FuncParamAttr
Indicates a function return value or parameter attribute.

Function Parameter Attribute
Function Parameter Attribute

Kernel

39

FPRoundingMode
Indicates a floating-point rounding mode.

FP Rounding Mode
Floating-Point Rounding Mode

40

FPFastMathMode
Indicates a floating-point fast math flag.

FP Fast Math Mode
Fast-Math Mode

Kernel

41

LinkageAttributes
Associate linkage attributes to values. Only valid on OpFunction or global (module scope) OpVariable. See linkage.

Literal String
Name

Linkage Type
Linkage Type

Linkage

42

NoContraction
Apply to an arithmetic instruction to indicate the operation cannot be combined with another instruction to form a single operation. For example, if applied to an OpFMul, that multiply can’t be combined with an addition to yield a fused multiply-add operation. Furthermore, such operations are not allowed to reassociate; e.g., add(a + add(b+c)) cannot be transformed to add(add(a+b) + c).

Shader

43

InputAttachmentIndex
Apply to a variable to provide an input-target index (as specified by the client API). Only valid in the Fragment Execution Model and for variables of type OpTypeImage with a Dim operand of SubpassData.

Literal Number
Attachment Index

InputAttachment

44

Alignment
Apply to a pointer. This declares a known minimum alignment the pointer has.

Literal Number
Alignment

Kernel

45

MaxByteOffset
Apply to a pointer. This declares a known maximum byte offset this pointer will be incremented by from the point of the decoration. This is a guaranteed upper bound when applied to OpFunctionParameter.

Literal Number
Max Byte Offset

Addresses

Missing before version 1.1.

46

AlignmentId
Apply to a pointer. This declares a known minimum alignment the pointer has.

Specified as an Id.

<id>
Alignment

Kernel

Missing before version 1.2.

47

MaxByteOffsetId
Apply to a pointer. This declares a known maximum byte offset this pointer will be incremented by from the point of the decoration. This is a guaranteed upper bound when applied to OpFunctionParameter.

Specified as an Id.

<id>
Max Byte Offset

Addresses

Missing before version 1.2.

4469

NoSignedWrap
Apply to an instruction to indicate that it does not cause signed integer wrapping to occur, in the form of overflow or underflow.

It can decorate only the following instructions:
- OpIAdd
- OpISub
- OpIMul
- OpShiftLeftLogical
- OpSNegate
- OpExtInst for instruction numbers specified in the extended instruction-set specifications as accepting this decoration.

If an instruction decorated with NoSignedWrap does overflow or underflow, the behavior is undefined.

Missing before version 1.4.

Also see extension: SPV_KHR_no_integer_wrap_decoration

4470

NoUnsignedWrap
Apply to an instruction to indicate that it does not cause unsigned integer wrapping to occur, in the form of overflow or underflow.

It can decorate only the following instructions:
- OpIAdd
- OpISub
- OpIMul
- OpShiftLeftLogical
- OpExtInst for instruction numbers specified in the extended instruction-set specifications as accepting this decoration.

If an instruction decorated with NoUnsignedWrap does overflow or underflow, the behavior is undefined.

Missing before version 1.4.

Also see extension: SPV_KHR_no_integer_wrap_decoration

4999

ExplicitInterpAMD

Reserved.

Also see extension: SPV_AMD_shader_explicit_vertex_parameter

5248

OverrideCoverageNV

SampleMaskOverrideCoverageNV

Reserved.

Also see extension: SPV_NV_sample_mask_override_coverage

5250

PassthroughNV

GeometryShaderPassthroughNV

Reserved.

Also see extension: SPV_NV_geometry_shader_passthrough

5252

ViewportRelativeNV

ShaderViewportMaskNV

Reserved.

5256

SecondaryViewportRelativeNV

Literal Number
Offset

ShaderStereoViewNV

Reserved.

Also see extension: SPV_NV_stereo_view_rendering

5271

PerPrimitiveNV

MeshShadingNV

Reserved.

Also see extension: SPV_NV_mesh_shader

5272

PerViewNV

MeshShadingNV

Reserved.

Also see extension: SPV_NV_mesh_shader

5273

PerTaskNV

MeshShadingNV

Reserved.

Also see extension: SPV_NV_mesh_shader

5285

PerVertexNV

FragmentBarycentricNV

Reserved.

Also see extension: SPV_NV_fragment_shader_barycentric

5300

NonUniformEXT

ShaderNonUniformEXT

5634

CounterBuffer
The <id> of a counter buffer associated with the decorated buffer. It can decorate only a variable in the Uniform storage class. Counter Buffer must be a variable in the Uniform storage class.

<id>
Counter Buffer

Missing before version 1.4.

5634

HlslCounterBufferGOOGLE

<id>
Counter Buffer

Reserved.

Also see extension: SPV_GOOGLE_hlsl_functionality1

5635

UserSemantic
A string describing a user-defined semantic intent of what it decorates. Semantic is case insensitive. It can decorate only a variable or a member of a structure type. If decorating a variable, it must be in the Input or Output storage classes.

Literal String
Semantic

Missing before version 1.4.

5635

HlslSemanticGOOGLE

Literal String
Semantic

Reserved.

Also see extension: SPV_GOOGLE_hlsl_functionality1

5355

RestrictPointerEXT

PhysicalStorageBufferAddressesEXT

Reserved.

Also see extension: SPV_EXT_physical_storage_buffer

5356

AliasedPointerEXT

PhysicalStorageBufferAddressesEXT

Reserved.

Also see extension: SPV_EXT_physical_storage_buffer

3.21. BuiltIn

Used when Decoration is BuiltIn. Apply to:

  • the result <id> of the OpVariable declaration of the built-in variable, or

  • a structure-type member, if the built-in is a member of a structure, or

  • a constant instruction, if the built-in is a constant.

As stated per entry below, these have additional semantics and constraints specified by the client API.

BuiltIn Enabling Capabilities

0

Position
Output vertex position from a vertex processing Execution Model. See the client API specification for more detail.

Shader

1

PointSize
Output point size from a vertex processing Execution Model. See the client API specification for more detail.

Shader

3

ClipDistance
Array of clip distances. See the client API specification for more detail.

ClipDistance

4

CullDistance
Array of clip distances. See the client API specification for more detail.

CullDistance

5

VertexId
Input vertex ID to a Vertex Execution Model. See the client API specification for more detail.

Shader

6

InstanceId
Input instance ID to a Vertex Execution Model. See the client API specification for more detail.

Shader

7

PrimitiveId
Primitive ID in a Geometry Execution Model. See the client API specification for more detail.

Geometry, Tessellation, RayTracingNV

8

InvocationId
Invocation ID, input to Geometry and TessellationControl Execution Model. See the client API specification for more detail.

Geometry, Tessellation

9

Layer
Layer output by a Geometry Execution Model, input to a Fragment Execution Model, for multi-layer framebuffer. See the client API specification for more detail.

Geometry

10

ViewportIndex
Viewport Index output by a Geometry stage, input to a Fragment Execution Model. See the client API specification for more detail.

MultiViewport

11

TessLevelOuter
Output patch outer levels in a TessellationControl Execution Model. See the client API specification for more detail.

Tessellation

12

TessLevelInner
Output patch inner levels in a TessellationControl Execution Model. See the client API specification for more detail.

Tessellation

13

TessCoord
Input vertex position in TessellationEvaluation Execution Model. See the client API specification for more detail.

Tessellation

14

PatchVertices
Input patch vertex count in a tessellation Execution Model. See the client API specification for more detail.

Tessellation

15

FragCoord
Coordinates (x, y, z, 1/w) of the current fragment, input to the Fragment Execution Model. See the client API specification for more detail.

Shader

16

PointCoord
Coordinates within a point, input to the Fragment Execution Model. See the client API specification for more detail.

Shader

17

FrontFacing
Face direction, input to the Fragment Execution Model. See the client API specification for more detail.

Shader

18

SampleId
Input sample number to the Fragment Execution Model. See the client API specification for more detail.

SampleRateShading

19

SamplePosition
Input sample position to the Fragment Execution Model. See the client API specification for more detail.

SampleRateShading

20

SampleMask
Input or output sample mask to the Fragment Execution Model. See the client API specification for more detail.

Shader

22

FragDepth
Output fragment depth from the Fragment Execution Model. See the client API specification for more detail.

Shader

23

HelperInvocation
Input whether a helper invocation, to the Fragment Execution Model. See the client API specification for more detail.

Shader

24

NumWorkgroups
Number of workgroups in GLCompute or Kernel Execution Models. See the client API specification for more detail.

25

WorkgroupSize
Work-group size in GLCompute or Kernel Execution Models. See the client API specification for more detail.

26

WorkgroupId
Work-group ID in GLCompute or Kernel Execution Models. See the client API specification for more detail.

27

LocalInvocationId
Local invocation ID in GLCompute or Kernel Execution Models. See the client API specification for more detail.

28

GlobalInvocationId
Global invocation ID in GLCompute or Kernel Execution Models. See the client API specification for more detail.

29

LocalInvocationIndex
Local invocation index in GLCompute Execution Models. See the client API specification for more detail.

Work-group Linear ID in Kernel Execution Models. See the client API specification for more detail.

30

WorkDim
Work dimensions in Kernel Execution Models. See the client API specification for more detail.

Kernel

31

GlobalSize
Global size in Kernel Execution Models. See the client API specification for more detail.

Kernel

32

EnqueuedWorkgroupSize
Enqueued work-group size in Kernel Execution Models. See the client API specification for more detail.

Kernel

33

GlobalOffset
Global offset in Kernel Execution Models. See the client API specification for more detail.

Kernel

34

GlobalLinearId
Global linear ID in Kernel Execution Models. See the client API specification for more detail.

Kernel

36

SubgroupSize
Subgroup size. See the client API specification for more detail.

Kernel, GroupNonUniform, SubgroupBallotKHR

37

SubgroupMaxSize
Subgroup maximum size in Kernel Execution Models. See the client API specification for more detail.

Kernel

38

NumSubgroups
Number of subgroups in GLCompute or Kernel Execution Models. See the client API specification for more detail.

Kernel, GroupNonUniform

39

NumEnqueuedSubgroups
Number of enqueued subgroups in Kernel Execution Models. See the client API specification for more detail.

Kernel

40

SubgroupId
Subgroup ID in GLCompute or Kernel Execution Models. See the client API specification for more detail.

Kernel, GroupNonUniform

41

SubgroupLocalInvocationId
Subgroup local invocation ID. See the client API specification for more detail.

Kernel, GroupNonUniform, SubgroupBallotKHR

42

VertexIndex
Vertex index. See the client API specification for more detail.

Shader

43

InstanceIndex
Instance index. See the client API specification for more detail.

Shader

4416

SubgroupEqMask
Subgroup invocations bitmask where bit index == SubgroupLocalInvocationId.
See the client API specification for more detail.

SubgroupBallotKHR, GroupNonUniformBallot

Missing before version 1.3.

4417

SubgroupGeMask
Subgroup invocations bitmask where bit index >= SubgroupLocalInvocationId.
See the client API specification for more detail.

SubgroupBallotKHR, GroupNonUniformBallot

Missing before version 1.3.

4418

SubgroupGtMask
Subgroup invocations bitmask where bit index > SubgroupLocalInvocationId.
See the client API specification for more detail.

SubgroupBallotKHR, GroupNonUniformBallot

Missing before version 1.3.

4419

SubgroupLeMask
Subgroup invocations bitmask where bit index <= SubgroupLocalInvocationId.
See the client API specification for more detail.

SubgroupBallotKHR, GroupNonUniformBallot

Missing before version 1.3.

4420

SubgroupLtMask
Subgroup invocations bitmask where bit index < SubgroupLocalInvocationId.
See the client API specification for more detail.

SubgroupBallotKHR, GroupNonUniformBallot

Missing before version 1.3.

4416

SubgroupEqMaskKHR

SubgroupBallotKHR, GroupNonUniformBallot

Missing before version 1.3.

Also see extension: SPV_KHR_shader_ballot

4417

SubgroupGeMaskKHR

SubgroupBallotKHR, GroupNonUniformBallot

Missing before version 1.3.

Also see extension: SPV_KHR_shader_ballot

4418

SubgroupGtMaskKHR

SubgroupBallotKHR, GroupNonUniformBallot

Missing before version 1.3.

Also see extension: SPV_KHR_shader_ballot

4419

SubgroupLeMaskKHR

SubgroupBallotKHR, GroupNonUniformBallot

Missing before version 1.3.

Also see extension: SPV_KHR_shader_ballot

4420

SubgroupLtMaskKHR

SubgroupBallotKHR, GroupNonUniformBallot

Missing before version 1.3.

Also see extension: SPV_KHR_shader_ballot

4424

BaseVertex
Base vertex component of vertex ID.
See the client API specification for more detail.

DrawParameters

Missing before version 1.3.

Also see extension: SPV_KHR_shader_draw_parameters

4425

BaseInstance
Base instance component of instance ID.
See the client API specification for more detail.

DrawParameters

Missing before version 1.3.

Also see extension: SPV_KHR_shader_draw_parameters

4426

DrawIndex
Contains the index of the draw currently being processed.
See the client API specification for more detail.

DrawParameters, MeshShadingNV

Missing before version 1.3.

Also see extensions: SPV_KHR_shader_draw_parameters, SPV_NV_mesh_shader

4438

DeviceIndex
Input device index of the logical device.
See the client API specification for more detail.

DeviceGroup

Missing before version 1.3.

Also see extension: SPV_KHR_device_group

4440

ViewIndex
Input view index of the view currently being rendered to.
See the client API specification for more detail.

MultiView

Missing before version 1.3.

Also see extension: SPV_KHR_multiview

4992

BaryCoordNoPerspAMD

Reserved.

Also see extension: SPV_AMD_shader_explicit_vertex_parameter

4993

BaryCoordNoPerspCentroidAMD

Reserved.

Also see extension: SPV_AMD_shader_explicit_vertex_parameter

4994

BaryCoordNoPerspSampleAMD

Reserved.

Also see extension: SPV_AMD_shader_explicit_vertex_parameter

4995

BaryCoordSmoothAMD

Reserved.

Also see extension: SPV_AMD_shader_explicit_vertex_parameter

4996

BaryCoordSmoothCentroidAMD

Reserved.

Also see extension: SPV_AMD_shader_explicit_vertex_parameter

4997

BaryCoordSmoothSampleAMD

Reserved.

Also see extension: SPV_AMD_shader_explicit_vertex_parameter

4998

BaryCoordPullModelAMD

Reserved.

Also see extension: SPV_AMD_shader_explicit_vertex_parameter

5014

FragStencilRefEXT

StencilExportEXT

Reserved.

Also see extension: SPV_EXT_shader_stencil_export

5253

ViewportMaskNV

ShaderViewportMaskNV, MeshShadingNV

Reserved.

Also see extensions: SPV_NV_viewport_array2, SPV_NV_mesh_shader

5257

SecondaryPositionNV

ShaderStereoViewNV

Reserved.

Also see extension: SPV_NV_stereo_view_rendering

5258

SecondaryViewportMaskNV

ShaderStereoViewNV

Reserved.

Also see extension: SPV_NV_stereo_view_rendering

5261

PositionPerViewNV

PerViewAttributesNV, MeshShadingNV

Reserved.

Also see extensions: SPV_NVX_multiview_per_view_attributes, SPV_NV_mesh_shader

5262

ViewportMaskPerViewNV

PerViewAttributesNV, MeshShadingNV

Reserved.

Also see extensions: SPV_NVX_multiview_per_view_attributes, SPV_NV_mesh_shader

5264

FullyCoveredEXT

FragmentFullyCoveredEXT

Reserved.

Also see extension: SPV_EXT_fragment_fully_covered

5274

TaskCountNV

MeshShadingNV

Reserved.

Also see extension: SPV_NV_mesh_shader

5275

PrimitiveCountNV

MeshShadingNV

Reserved.

Also see extension: SPV_NV_mesh_shader

5276

PrimitiveIndicesNV

MeshShadingNV

Reserved.

Also see extension: SPV_NV_mesh_shader

5277

ClipDistancePerViewNV

MeshShadingNV

Reserved.

Also see extension: SPV_NV_mesh_shader

5278

CullDistancePerViewNV

MeshShadingNV

Reserved.

Also see extension: SPV_NV_mesh_shader

5279

LayerPerViewNV

MeshShadingNV

Reserved.

Also see extension: SPV_NV_mesh_shader

5280

MeshViewCountNV

MeshShadingNV

Reserved.

Also see extension: SPV_NV_mesh_shader

5281

MeshViewIndicesNV

MeshShadingNV

Reserved.

Also see extension: SPV_NV_mesh_shader

5286

BaryCoordNV

FragmentBarycentricNV

Reserved.

Also see extension: SPV_NV_fragment_shader_barycentric

5287

BaryCoordNoPerspNV

FragmentBarycentricNV

Reserved.

Also see extension: SPV_NV_fragment_shader_barycentric

5292

FragSizeEXT

FragmentDensityEXT, ShadingRateNV

Reserved.

Also see extensions: SPV_EXT_fragment_invocation_density, SPV_NV_shading_rate

5292

FragmentSizeNV

ShadingRateNV, FragmentDensityEXT

Reserved.

Also see extensions: SPV_NV_shading_rate, SPV_EXT_fragment_invocation_density

5293

FragInvocationCountEXT

FragmentDensityEXT, ShadingRateNV

Reserved.

Also see extensions: SPV_EXT_fragment_invocation_density, SPV_NV_shading_rate

5293

InvocationsPerPixelNV

ShadingRateNV, FragmentDensityEXT

Reserved.

Also see extensions: SPV_NV_shading_rate, SPV_EXT_fragment_invocation_density

5319

LaunchIdNV

RayTracingNV

Also see extension: SPV_NV_ray_tracing

5320

LaunchSizeNV

RayTracingNV

Also see extension: SPV_NV_ray_tracing

5321

WorldRayOriginNV

RayTracingNV

Also see extension: SPV_NV_ray_tracing

5322

WorldRayDirectionNV

RayTracingNV

Also see extension: SPV_NV_ray_tracing

5323

ObjectRayOriginNV

RayTracingNV

Also see extension: SPV_NV_ray_tracing

5324

ObjectRayDirectionNV

RayTracingNV

Also see extension: SPV_NV_ray_tracing

5325

RayTminNV

RayTracingNV

Also see extension: SPV_NV_ray_tracing

5326

RayTmaxNV

RayTracingNV

Also see extension: SPV_NV_ray_tracing

5327

InstanceCustomIndexNV

RayTracingNV

Also see extension: SPV_NV_ray_tracing

5330

ObjectToWorldNV

RayTracingNV

Also see extension: SPV_NV_ray_tracing

5331

WorldToObjectNV

RayTracingNV

Also see extension: SPV_NV_ray_tracing

5332

HitTNV

RayTracingNV

Also see extension: SPV_NV_ray_tracing

5333

HitKindNV

RayTracingNV

Also see extension: SPV_NV_ray_tracing

5351

IncomingRayFlagsNV

RayTracingNV

Also see extension: SPV_NV_ray_tracing

3.22. Selection Control

This value is a literal mask; it can be formed by combining the bits from multiple rows in the table below.

Used by OpSelectionMerge.

Selection Control

0x0

None

0x1

Flatten
Strong request, to the extent possible, to remove the control flow for this selection.

0x2

DontFlatten
Strong request, to the extent possible, to keep this selection as control flow.

3.23. Loop Control

Loop controls. Bits that are set can indicate whether an additional operand follows, as described by the table. If there are multiple following operands indicated, they are ordered: Those indicated by smaller-numbered bits appear first.

This value is a literal mask; it can be formed by combining the bits from multiple rows in the table below.

Used by OpLoopMerge.

Loop Control Enabling Capabilities

0x0

None

0x1

Unroll
Strong request, to the extent possible, to unroll or unwind this loop.
This must not be used with the DontUnroll bit.

0x2

DontUnroll
Strong request, to the extent possible, to keep this loop as a loop, without unrolling.

0x4

DependencyInfinite
Guarantees that there are no dependencies between loop iterations.

Missing before version 1.1.

0x8

DependencyLength
Guarantees that there are no dependencies between a number of loop iterations, specified as a subsequent literal-number operand to the instruction. The literal number is treated as unsigned.

Missing before version 1.1.

0x10

MinIterations
Unchecked assertion that the loop will execute at least a given number of iterations, specified as a subsequent literal-number operand to the instruction. The literal number is treated as unsigned.

Missing before version 1.4.

0x20

MaxIterations
Unchecked assertion that the loop will execute at most a given number of iterations, specified as a subsequent literal-number operand to the instruction. The literal number is treated as unsigned.

Missing before version 1.4.

0x40

IterationMultiple
Unchecked assertion that the loop will execute a multiple of a given number of iterations, specified as a subsequent literal-number operand to the instruction. The literal number is treated as unsigned. It must also be greater than 0.

Missing before version 1.4.

0x80

PeelCount
Request that the loop be peeled by a given number of loop iterations, specified as a subsequent literal-number operand to the instruction. The literal number is treated as unsigned.
This must not be used with the DontUnroll bit.

Missing before version 1.4.

0x100

PartialCount
Request that the loop be partially unrolled by a given number of loop iterations, specified as a subsequent literal-number operand to the instruction. The literal number is treated as unsigned.
This must not be used with the DontUnroll bit.

Missing before version 1.4.

3.24. Function Control

This value is a literal mask; it can be formed by combining the bits from multiple rows in the table below.

Used by OpFunction.

Function Control

0x0

None

0x1

Inline
Strong request, to the extent possible, to inline the function.

0x2

DontInline
Strong request, to the extent possible, to not inline the function.

0x4

Pure
Compiler can assume this function has no side effect, but might read global memory or read through dereferenced function parameters. Always computes the same result for the same argument values.

0x8

Const
Compiler can assume this function has no side effects, and will not access global memory or dereference function parameters. Always computes the same result for the same argument values.

3.25. Memory Semantics <id>

Must be an <id> of a 32-bit integer scalar.

Memory semantics define memory-order constraints, and on what storage classes those constraints apply to. The memory order constrains the allowed orders in which memory operations in this invocation can made visible to another invocation. The storage classes specify to which subsets of memory these constraints are to be applied. Storage classes not selected are not being constrained.

Despite being a mask and allowing multiple bits to be combined, it is invalid for more than one of these four bits to be set: Acquire, Release, AcquireRelease, or SequentiallyConsistent. Requesting both Acquire and Release semantics is done by setting the AcquireRelease bit, not by setting two bits.

This value is a mask; it can be formed by combining the bits from multiple rows in the table below.

Used by:

Memory Semantics Enabling Capabilities

0x0

None (Relaxed)

0x2

Acquire
All memory operations provided in program order after this memory operation will execute after this memory operation.

0x4

Release
All memory operations provided in program order before this memory operation will execute before this memory operation.

0x8

AcquireRelease
Has the properties of both Acquire and Release semantics. It is used for read-modify-write operations.

0x10

SequentiallyConsistent
All observers will see this memory access in the same order with respect to other sequentially-consistent memory accesses from this invocation.

0x40

UniformMemory
Apply the memory-ordering constraints to StorageBuffer or Uniform Storage Class memory.

Shader

0x80

SubgroupMemory
Apply the memory-ordering constraints to subgroup memory.

0x100

WorkgroupMemory
Apply the memory-ordering constraints to Workgroup Storage Class memory.

0x200

CrossWorkgroupMemory
Apply the memory-ordering constraints to CrossWorkgroup Storage Class memory.

0x400

AtomicCounterMemory
Apply the memory-ordering constraints to AtomicCounter Storage Class memory.

AtomicStorage

0x800

ImageMemory
Apply the memory-ordering constraints to image contents (types declared by OpTypeImage), or to accesses done through pointers to the Image Storage Class.

0x1000

OutputMemoryKHR

VulkanMemoryModelKHR

0x2000

MakeAvailableKHR

VulkanMemoryModelKHR

0x4000

MakeVisibleKHR

VulkanMemoryModelKHR

3.26. Memory Operands

Additional operands to the listed memory instructions. Bits that are set can indicate whether an additional operand follows, as described by the table. If there are multiple following operands indicated, they are ordered: Those indicated by smaller-numbered bits appear first. An instruction needing two masks must first provide the first mask followed by the first mask’s additional operands, and then provide the second mask followed by the second mask’s additional operands.

This value is a literal mask; it can be formed by combining the bits from multiple rows in the table below.

Used by:

Memory Operands Enabling Capabilities

0x0

None

0x1

Volatile
This access cannot be eliminated, duplicated, or combined with other accesses.

0x2

Aligned
This access has a known alignment, provided as a literal in the next operand.

0x4

Nontemporal
Hints that the accessed address is not likely to be accessed again in the near future.

0x8

MakePointerAvailableKHR

VulkanMemoryModelKHR

0x10

MakePointerVisibleKHR

VulkanMemoryModelKHR

0x20

NonPrivatePointerKHR

VulkanMemoryModelKHR

3.27. Scope <id>

Must be an <id> of a 32-bit integer scalar. Its value must be one of the values in the table below.

The execution scope or memory scope of an operation. When used as a memory scope, it specifies the distance of synchronization from the current invocation. When used as an execution scope, it specifies the set of executing invocations taking part in the operation. Used by:

Scope Enabling Capabilities

0

CrossDevice
Scope crosses multiple devices.

1

Device
Scope is the current device.

2

Workgroup
Scope is the current workgroup.

3

Subgroup
Scope is the current subgroup.

4

Invocation
Scope is the current Invocation.

5

QueueFamilyKHR

VulkanMemoryModelKHR

3.28. Group Operation

Defines the class of workgroup or subgroup operation. Used by:

Group Operation Enabling Capabilities

0

Reduce
A reduction operation for all values of a specific value X specified by invocations within a workgroup.

Kernel, GroupNonUniformArithmetic, GroupNonUniformBallot

1

InclusiveScan
A binary operation with an identity I and n (where n is the size of the workgroup) elements[a0, a1, … an-1] resulting in [a0, (a0 op a1), …(a0 op a1 op … op an-1)]

Kernel, GroupNonUniformArithmetic, GroupNonUniformBallot

2

ExclusiveScan
A binary operation with an identity I and n (where n is the size of the workgroup) elements[a0, a1, … an-1] resulting in [I, a0, (a0 op a1), … (a0 op a1 op … op an-2)].

Kernel, GroupNonUniformArithmetic, GroupNonUniformBallot

3

ClusteredReduce

GroupNonUniformClustered

Missing before version 1.3.

6

PartitionedReduceNV

GroupNonUniformPartitionedNV

Reserved.

Also see extension: SPV_NV_shader_subgroup_partitioned

7

PartitionedInclusiveScanNV

GroupNonUniformPartitionedNV

Reserved.

Also see extension: SPV_NV_shader_subgroup_partitioned

8

PartitionedExclusiveScanNV

GroupNonUniformPartitionedNV

Reserved.

Also see extension: SPV_NV_shader_subgroup_partitioned

3.29. Kernel Enqueue Flags

Specify when the child kernel begins execution.

Note: Implementations are not required to honor this flag. Implementations may not schedule kernel launch earlier than the point specified by this flag, however. Used by OpEnqueueKernel.

Kernel Enqueue Flags Enabling Capabilities

0

NoWait
Indicates that the enqueued kernels do not need to wait for the parent kernel to finish execution before they begin execution.

Kernel

1

WaitKernel
Indicates that all work-items of the parent kernel must finish executing and all immediate side effects committed before the enqueued child kernel may begin execution.

Note: Immediate meaning not side effects resulting from child kernels. The side effects would include stores to global memory and pipe reads and writes.

Kernel

2

WaitWorkGroup
Indicates that the enqueued kernels wait only for the workgroup that enqueued the kernels to finish before they begin execution.

Note: This acts as a memory synchronization point between work-items in a work-group and child kernels enqueued by work-items in the work-group.

Kernel

3.30. Kernel Profiling Info

Specify the profiling information to be queried. Used by OpCaptureEventProfilingInfo.

This value is a mask; it can be formed by combining the bits from multiple rows in the table below.

Kernel Profiling Info Enabling Capabilities

0x0

None

0x1

CmdExecTime
Indicates that the profiling info queried is the execution time.

Kernel

3.31. Capability

Capabilities a module can declare it uses.

All used capabilities must be declared, either explicitly with OpCapability or implicitly through the Implicitly Declares column. The Implicitly Declares column lists additional capabilities that are all implicitly declared when the Capability entry is explicitly or implicitly declared. It is not necessary, but allowed, to explicitly declare an implicitly declared capability.

See the capabilities section for more detail. Used by OpCapability.

Capability Implicitly Declares

0

Matrix
Uses OpTypeMatrix.

1

Shader
Uses Vertex, Fragment, or GLCompute Execution Models.

Matrix

2

Geometry
Uses the Geometry Execution Model.

Shader

3

Tessellation
Uses the TessellationControl or TessellationEvaluation Execution Models.

Shader

4

Addresses
Uses physical addressing, non-logical addressing modes.

5

Linkage
Uses partially linked modules and libraries.

6

Kernel
Uses the Kernel Execution Model.

7

Vector16
Uses OpTypeVector to declare 8 component or 16 component vectors.

Kernel

8

Float16Buffer
Allows a 16-bit OpTypeFloat instruction for the sole purpose of creating an OpTypePointer to a 16-bit float. Pointers to a 16-bit float cannot be dereferenced directly, they must only be dereferenced via an extended instruction. All other uses of 16-bit OpTypeFloat are disallowed.

Kernel

9

Float16
Uses OpTypeFloat to declare the 16-bit floating-point type.

10

Float64
Uses OpTypeFloat to declare the 64-bit floating-point type.

11

Int64
Uses OpTypeInt to declare 64-bit integer types.

12

Int64Atomics
Uses atomic instructions on 64-bit integer types.

Int64

13

ImageBasic
Uses OpTypeImage or OpTypeSampler in a Kernel.

Kernel

14

ImageReadWrite
Uses OpTypeImage with the ReadWrite access qualifier.

ImageBasic

15

ImageMipmap
Uses non-zero Lod Image Operands.

ImageBasic

17

Pipes
Uses OpTypePipe, OpTypeReserveId or pipe instructions.

Kernel

18

Groups
Uses group instructions.

19

DeviceEnqueue
Uses OpTypeQueue, OpTypeDeviceEvent, and device side enqueue instructions.

Kernel

20

LiteralSampler
Samplers are made from literals within the module. See OpConstantSampler.

Kernel

21

AtomicStorage
Uses the AtomicCounter Storage Class, allowing use of only the OpAtomicLoad, OpAtomicIIncrement, and OpAtomicIDecrement instructions.

Shader

22

Int16
Uses OpTypeInt to declare 16-bit integer types.

23

TessellationPointSize
Tessellation stage exports point size.

Tessellation

24

GeometryPointSize
Geometry stage exports point size

Geometry

25

ImageGatherExtended
Uses texture gather with non-constant or independent offsets

Shader

27

StorageImageMultisample
Uses multi-sample images for non-sampled images.

Shader

28

UniformBufferArrayDynamicIndexing
Block-decorated arrays in uniform storage classes use dynamically uniform indexing.

Shader

29

SampledImageArrayDynamicIndexing
Arrays of sampled images use dynamically uniform indexing.

Shader

30

StorageBufferArrayDynamicIndexing
Arrays in the StorageBuffer Storage Class, or BufferBlock-decorated arrays, use dynamically uniform indexing.

Shader

31

StorageImageArrayDynamicIndexing
Arrays of non-sampled images are accessed with dynamically uniform indexing.

Shader

32

ClipDistance
Uses the ClipDistance BuiltIn.

Shader

33

CullDistance
Uses the CullDistance BuiltIn.

Shader

34

ImageCubeArray
Uses the Cube Dim with the Arrayed operand in OpTypeImage, without a sampler.

SampledCubeArray

35

SampleRateShading
Uses per-sample rate shading.

Shader

36

ImageRect
Uses the Rect Dim without a sampler.

SampledRect

37

SampledRect
Uses the Rect Dim with a sampler.

Shader

38

GenericPointer
Uses the Generic Storage Class.

Addresses

39

Int8
Uses OpTypeInt to declare 8-bit integer types.

40

InputAttachment
Uses the SubpassData Dim.

Shader

41

SparseResidency
Uses OpImageSparse… instructions.

Shader

42

MinLod
Uses the MinLod Image Operand.

Shader

43

Sampled1D
Uses the 1D Dim with a sampler.

44

Image1D
Uses the 1D Dim without a sampler.

Sampled1D

45

SampledCubeArray
Uses the Cube Dim with the Arrayed operand in OpTypeImage, with a sampler.

Shader

46

SampledBuffer
Uses the Buffer Dim with a sampler.

47

ImageBuffer
Uses the Buffer Dim without a sampler.

SampledBuffer

48

ImageMSArray
An MS operand in OpTypeImage indicates multisampled, used without a sampler.

Shader

49

StorageImageExtendedFormats
One of a large set of more advanced image formats are used, namely one of those in the Image Format table listed as requiring this capability.

Shader

50

ImageQuery
The sizes, number of samples, or lod, etc. are queried.

Shader

51

DerivativeControl
Uses fine or coarse-grained derivatives, e.g., OpDPdxFine.

Shader

52

InterpolationFunction
Uses one of the InterpolateAtCentroid, InterpolateAtSample, or InterpolateAtOffset GLSL.std.450 extended instructions.

Shader

53

TransformFeedback
Uses the Xfb Execution Mode.

Shader

54

GeometryStreams
Uses multiple numbered streams for geometry-stage output.

Geometry

55

StorageImageReadWithoutFormat
OpImageRead can use the Unknown Image Format.

Shader

56

StorageImageWriteWithoutFormat
OpImageWrite can use the Unknown Image Format.

Shader

57

MultiViewport
Multiple viewports are used.

Geometry

58

SubgroupDispatch
Uses subgroup dispatch instructions.

DeviceEnqueue

Missing before version 1.1.

59

NamedBarrier
Uses OpTypeNamedBarrier.

Kernel

Missing before version 1.1.

60

PipeStorage
Uses OpTypePipeStorage.

Pipes

Missing before version 1.1.

61

GroupNonUniform

Missing before version 1.3.

62

GroupNonUniformVote

GroupNonUniform

Missing before version 1.3.

63

GroupNonUniformArithmetic

GroupNonUniform

Missing before version 1.3.

64

GroupNonUniformBallot

GroupNonUniform

Missing before version 1.3.

65

GroupNonUniformShuffle

GroupNonUniform

Missing before version 1.3.

66

GroupNonUniformShuffleRelative

GroupNonUniform

Missing before version 1.3.

67

GroupNonUniformClustered

GroupNonUniform

Missing before version 1.3.

68

GroupNonUniformQuad

GroupNonUniform

Missing before version 1.3.

4423

SubgroupBallotKHR

Reserved.

Also see extension: SPV_KHR_shader_ballot

4427

DrawParameters

Shader

Missing before version 1.3.

Also see extension: SPV_KHR_shader_draw_parameters

4431

SubgroupVoteKHR

Reserved.

Also see extension: SPV_KHR_subgroup_vote

4433

StorageBuffer16BitAccess
Allows 16-bit OpTypeFloat and OpTypeInt for the sole purpose of creating an OpTypePointer to a 16-bit floating-point or 16-bit integer member of an object. The object must be in the StorageBuffer Storage Class, or be in the Uniform storage class and have the BufferBlock decoration.

An object of a 16-bit type produced by dereferencing such a pointer may be the result of a width-only conversion instruction (OpFConvert, OpSConvert, or OpUConvert) from a 32-bit type or of an OpLoad, and may be used as an operand to a width-only conversion instruction to a 32-bit type or as the object operand of an OpStore.

Other uses of 16-bit types are not enabled by this capability.

Missing before version 1.3.

Also see extension: SPV_KHR_16bit_storage

4433

StorageUniformBufferBlock16

Missing before version 1.3.

Also see extension: SPV_KHR_16bit_storage

4434

UniformAndStorageBuffer16BitAccess
Allows 16-bit OpTypeFloat and OpTypeInt for the sole purpose of creating an OpTypePointer to a 16-bit floating-point or 16-bit integer member of an object. The object must be in the StorageBuffer or Uniform Storage Classes.

An object of a 16-bit type produced by dereferencing such a pointer may be the result of a width-only conversion instruction from a 32-bit type or of an OpLoad, and may be used as an operand to a width-only conversion instruction to a 32-bit type or as the object operand of an OpStore.

Other uses of 16-bit types are not enabled by this capability.

StorageBuffer16BitAccess, StorageUniformBufferBlock16

Missing before version 1.3.

Also see extension: SPV_KHR_16bit_storage

4434

StorageUniform16

StorageBuffer16BitAccess, StorageUniformBufferBlock16

Missing before version 1.3.

Also see extension: SPV_KHR_16bit_storage

4435

StoragePushConstant16
Allows 16-bit OpTypeFloat and OpTypeInt for the sole purpose of creating an OpTypePointer to a 16-bit floating-point or 16-bit integer object in the PushConstant Storage Class.

An object of a 16-bit type produced by dereferencing such a pointer may only be the result of a width-only conversion instruction from a 32-bit type or of an OpLoad.

Other uses of 16-bit types are not enabled by this capability.

Missing before version 1.3.

Also see extension: SPV_KHR_16bit_storage

4436

StorageInputOutput16
Allows 16-bit OpTypeFloat and OpTypeInt for the sole purpose of creating an OpTypePointer to a 16-bit floating-point or 16-bit integer object in the Input or Output Storage Classes.

An object of a 16-bit type produced by dereferencing such a pointer may only be the result of a width-only conversion instruction from a 32-bit type or of an OpLoad, and may be used as an operand to a width-only conversion instruction to a 32-bit type or as the object operand of an OpStore.

Other uses of 16-bit types are not enabled by this capability.

Missing before version 1.3.

Also see extension: SPV_KHR_16bit_storage

4437

DeviceGroup

Missing before version 1.3.

Also see extension: SPV_KHR_device_group

4439

MultiView

Shader

Missing before version 1.3.

Also see extension: SPV_KHR_multiview

4441

VariablePointersStorageBuffer
Allow variable pointers, each confined to a single Block-decorated struct in the StorageBuffer storage class.

Shader

Missing before version 1.3.

Also see extension: SPV_KHR_variable_pointers

4442

VariablePointers
Allow variable pointers.

VariablePointersStorageBuffer

Missing before version 1.3.

Also see extension: SPV_KHR_variable_pointers

4445

AtomicStorageOps

Reserved.

Also see extension: SPV_KHR_shader_atomic_counter_ops

4447

SampleMaskPostDepthCoverage

Reserved.

Also see extension: SPV_KHR_post_depth_coverage

4448

StorageBuffer8BitAccess

Reserved.

Also see extension: SPV_KHR_8bit_storage

4449

UniformAndStorageBuffer8BitAccess

StorageBuffer8BitAccess

Reserved.

Also see extension: SPV_KHR_8bit_storage

4450

StoragePushConstant8

Reserved.

Also see extension: SPV_KHR_8bit_storage

4464

DenormPreserve
Uses the DenormPreserve execution mode.

Missing before version 1.4.

Also see extension: SPV_KHR_float_controls

4465

DenormFlushToZero
Uses the DenormFlushToZero execution mode.

Missing before version 1.4.

Also see extension: SPV_KHR_float_controls

4466

SignedZeroInfNanPreserve
Uses the SignedZeroInfNanPreserve execution mode.

Missing before version 1.4.

Also see extension: SPV_KHR_float_controls

4467

RoundingModeRTE
Uses the RoundingModeRTE execution mode.

Missing before version 1.4.

Also see extension: SPV_KHR_float_controls

4468

RoundingModeRTZ
Uses the RoundingModeRTZ execution mode.

Missing before version 1.4.

Also see extension: SPV_KHR_float_controls

5008

Float16ImageAMD

Shader

Reserved.

Also see extension: SPV_AMD_gpu_shader_half_float_fetch

5009

ImageGatherBiasLodAMD

Shader

Reserved.

Also see extension: SPV_AMD_texture_gather_bias_lod

5010

FragmentMaskAMD

Shader

Reserved.

Also see extension: SPV_AMD_shader_fragment_mask

5013

StencilExportEXT

Shader

Reserved.

Also see extension: SPV_EXT_shader_stencil_export

5015

ImageReadWriteLodAMD

Shader

Reserved.

Also see extension: SPV_AMD_shader_image_load_store_lod

5249

SampleMaskOverrideCoverageNV

SampleRateShading

Reserved.

Also see extension: SPV_NV_sample_mask_override_coverage

5251

GeometryShaderPassthroughNV

Geometry

Reserved.

Also see extension: SPV_NV_geometry_shader_passthrough

5254

ShaderViewportIndexLayerEXT

MultiViewport

Reserved.

Also see extension: SPV_EXT_shader_viewport_index_layer

5254

ShaderViewportIndexLayerNV

MultiViewport

Reserved.

Also see extension: SPV_NV_viewport_array2

5255

ShaderViewportMaskNV

ShaderViewportIndexLayerNV

Reserved.

Also see extension: SPV_NV_viewport_array2

5259

ShaderStereoViewNV

ShaderViewportMaskNV

Reserved.

Also see extension: SPV_NV_stereo_view_rendering

5260

PerViewAttributesNV

MultiView

Reserved.

Also see extension: SPV_NVX_multiview_per_view_attributes

5265

FragmentFullyCoveredEXT

Shader

Reserved.

Also see extension: SPV_EXT_fragment_fully_covered

5266

MeshShadingNV

Shader

Reserved.

Also see extension: SPV_NV_mesh_shader

5301

ShaderNonUniformEXT

Shader

Reserved.

Also see extension: SPV_EXT_descriptor_indexing

5302

RuntimeDescriptorArrayEXT

Shader

Reserved.

Also see extension: SPV_EXT_descriptor_indexing

5303

InputAttachmentArrayDynamicIndexingEXT

InputAttachment

Reserved.

Also see extension: SPV_EXT_descriptor_indexing

5304

UniformTexelBufferArrayDynamicIndexingEXT

SampledBuffer

Reserved.

Also see extension: SPV_EXT_descriptor_indexing

5305

StorageTexelBufferArrayDynamicIndexingEXT

ImageBuffer

Reserved.

Also see extension: SPV_EXT_descriptor_indexing

5306

UniformBufferArrayNonUniformIndexingEXT

ShaderNonUniformEXT

Reserved.

Also see extension: SPV_EXT_descriptor_indexing

5307

SampledImageArrayNonUniformIndexingEXT

ShaderNonUniformEXT

Reserved.

Also see extension: SPV_EXT_descriptor_indexing

5308

StorageBufferArrayNonUniformIndexingEXT

ShaderNonUniformEXT

Reserved.

Also see extension: SPV_EXT_descriptor_indexing

5309

StorageImageArrayNonUniformIndexingEXT

ShaderNonUniformEXT

Reserved.

Also see extension: SPV_EXT_descriptor_indexing

5310

InputAttachmentArrayNonUniformIndexingEXT

InputAttachment, ShaderNonUniformEXT

Reserved.

Also see extension: SPV_EXT_descriptor_indexing

5311

UniformTexelBufferArrayNonUniformIndexingEXT

SampledBuffer, ShaderNonUniformEXT

Reserved.

Also see extension: SPV_EXT_descriptor_indexing

5312

StorageTexelBufferArrayNonUniformIndexingEXT

ImageBuffer, ShaderNonUniformEXT

Reserved.

Also see extension: SPV_EXT_descriptor_indexing

5340

RayTracingNV

Shader

Reserved.

Also see extension: SPV_NV_ray_tracing

5568

SubgroupShuffleINTEL

Reserved.

Also see extension: SPV_INTEL_subgroups

5569

SubgroupBufferBlockIOINTEL

Reserved.

Also see extension: SPV_INTEL_subgroups

5570

SubgroupImageBlockIOINTEL

Reserved.

Also see extension: SPV_INTEL_subgroups

5579

SubgroupImageMediaBlockIOINTEL

Reserved.

Also see extension: SPV_INTEL_media_block_io

5696

SubgroupAvcMotionEstimationINTEL

Reserved.

Also see extension: SPV_INTEL_device_side_avc_motion_estimation

5697

SubgroupAvcMotionEstimationIntraINTEL

Reserved.

Also see extension: SPV_INTEL_device_side_avc_motion_estimation

5698

SubgroupAvcMotionEstimationChromaINTEL

Reserved.

Also see extension: SPV_INTEL_device_side_avc_motion_estimation

5297

GroupNonUniformPartitionedNV

Reserved.

Also see extension: SPV_NV_shader_subgroup_partitioned

5345

VulkanMemoryModelKHR

Reserved.

Also see extension: SPV_KHR_vulkan_memory_model

5346

VulkanMemoryModelDeviceScopeKHR

Reserved.

Also see extension: SPV_KHR_vulkan_memory_model

5282

ImageFootprintNV

Reserved.

Also see extension: SPV_NV_shader_image_footprint

5284

FragmentBarycentricNV

Reserved.

Also see extension: SPV_NV_fragment_shader_barycentric

5288

ComputeDerivativeGroupQuadsNV

Reserved.

Also see extension: SPV_NV_compute_shader_derivatives

5350

ComputeDerivativeGroupLinearNV

Reserved.

Also see extension: SPV_NV_compute_shader_derivatives

5291

FragmentDensityEXT

Shader

Reserved.

Also see extensions: SPV_EXT_fragment_invocation_density, SPV_NV_shading_rate

5291

ShadingRateNV

Shader

Reserved.

Also see extensions: SPV_NV_shading_rate, SPV_EXT_fragment_invocation_density

5347

PhysicalStorageBufferAddressesEXT

Shader

Reserved.

Also see extension: SPV_EXT_physical_storage_buffer

5357

CooperativeMatrixNV

Shader

Reserved.

Also see extension: SPV_NV_cooperative_matrix

3.32. Instructions

Form for each instruction:

Opcode Name (name-alias, name-alias, …)

Instruction description.

Word Count is the high-order 16 bits of word 0 of the instruction, holding its total WordCount. If the instruction takes a variable number of operands, Word Count will also say "+ variable", after stating the minimum size of the instruction.

Opcode is the low-order 16 bits of word 0 of the instruction, holding its opcode enumerant.

Results, when present, are any Result <id> or Result Type created by the instruction. Each one is always 32 bits.

Operands, when present, are any literals, other instruction’s Result <id>, etc., consumed by the instruction. Each one is always 32 bits.

Capability Enabling Capabilities
(when needed)

Word Count

Opcode

Results

Operands

3.32.1. Miscellaneous Instructions

OpNop

This has no semantic impact and can safely be removed from a module.

1

0

OpUndef

Make an intermediate object whose value is undefined.

Result Type is the type of object to make.

Each consumption of Result <id> yields an arbitrary, possibly different bit pattern or abstract value resulting in possibly different concrete, abstract, or opaque values.

3

1

<id>
Result Type

Result <id>

OpSizeOf

Computes the run-time size of the type pointed to by Pointer

Result Type must be a 32-bit integer type scalar.

Pointer must point to a concrete type.

Capability:
Addresses

Missing before version 1.1.

4

321

<id>
Result Type

Result <id>

<id>
Pointer

OpFragmentMaskFetchAMD

TBD

Capability:
FragmentMaskAMD

Reserved.

5

5011

<id>
Result Type

Result <id>

<id>
Image

<id>
Coordinate

OpFragmentFetchAMD

TBD

Capability:
FragmentMaskAMD

Reserved.

6

5012

<id>
Result Type

Result <id>

<id>
Image

<id>
Coordinate

<id>
Fragment Index

OpWritePackedPrimitiveIndices4x8NV

TBD

Capability:
MeshShadingNV

Reserved.

3

5299

<id>
Index Offset

<id>
Packed Indices

OpReportIntersectionNV

TBD

Capability:
RayTracingNV

5

5334

<id>
Result Type

Result <id>

<id>
Hit

<id>
HitKind

OpIgnoreIntersectionNV

TBD

Capability:
RayTracingNV

1

5335

OpTerminateRayNV

TBD

Capability:
RayTracingNV

1

5336

OpTraceNV

TBD

Capability:
RayTracingNV

12

5337

<id>
Accel

<id>
Ray Flags

<id>
Cull Mask

<id>
SBT Offset

<id>
SBT Stride

<id>
Miss Index

<id>
Ray Origin

<id>
Ray Tmin

<id>
Ray Direction

<id>
Ray Tmax

<id>
PayloadId

OpExecuteCallableNV

TBD

Capability:
RayTracingNV

3

5344

<id>
SBT Index

<id>
Callable DataId

OpCooperativeMatrixLoadNV

TBD

Capability:
CooperativeMatrixNV

Reserved.

6 + variable

5359

<id>
Result Type

Result <id>

<id>
Pointer

<id>
Stride

<id>
Column Major

Optional
Memory Operands

OpCooperativeMatrixStoreNV

TBD

Capability:
CooperativeMatrixNV

Reserved.

5 + variable

5360

<id>
Pointer

<id>
Object

<id>
Stride

<id>
Column Major

Optional
Memory Operands

OpCooperativeMatrixMulAddNV

TBD

Capability:
CooperativeMatrixNV

Reserved.

6

5361

<id>
Result Type

Result <id>

<id>
A

<id>
B

<id>
C

OpCooperativeMatrixLengthNV

TBD

Capability:
CooperativeMatrixNV

Reserved.

4

5362

<id>
Result Type

Result <id>

<id>
Type

3.32.2. Debug Instructions

OpSourceContinued

Continue specifying the Source text from the previous instruction. This has no semantic impact and can safely be removed from a module.

Continued Source is a continuation of the source text in the previous Source.

The previous instruction must be an OpSource or an OpSourceContinued instruction. As is true for all literal strings, the previous instruction’s string was nul terminated. That terminating 0 word from the previous instruction is not part of the source text; the first character of Continued Source logically immediately follows the last character of Source before its nul.

2 + variable

2

Literal String
Continued Source

OpSource

Document what source language and text this module was translated from. This has no semantic impact and can safely be removed from a module.

Version is the version of the source language. This literal operand is limited to a single word.

File is an OpString instruction and is the source-level file name.

Source is the text of the source-level file.

Each client API specifies what form the Version operand takes, per source language.

3 + variable

3

Source Language

Literal Number
Version

Optional
<id>
File

Optional
Literal String
Source

OpSourceExtension

Document an extension to the source language. This has no semantic impact and can safely be removed from a module.

Extension is a string describing a source-language extension. Its form is dependent on the how the source language describes extensions.

2 + variable

4

Literal String
Extension

OpName

Assign a name string to another instruction’s Result <id>. This has no semantic impact and can safely be removed from a module.

Target is the Result <id> to assign a name to. It can be the Result <id> of any other instruction; a variable, function, type, intermediate result, etc.

Name is the string to assign.

3 + variable

5

<id>
Target

Literal String
Name

OpMemberName

Assign a name string to a member of a structure type. This has no semantic impact and can safely be removed from a module.

Type is the <id> from an OpTypeStruct instruction.

Member is the number of the member to assign in the structure. The first member is member 0, the next is member 1, … This literal operand is limited to a single word.

Name is the string to assign to the member.

4 + variable

6

<id>
Type

Literal Number
Member

Literal String
Name

OpString

Assign a Result <id> to a string for use by other debug instructions (see OpLine and OpSource). This has no semantic impact and can safely be removed from a module. (Removal also requires removal of all instructions referencing Result <id>.)

String is the literal string being assigned a Result <id>.

3 + variable

7

Result <id>

Literal String
String

OpLine

Add source-level location information. This has no semantic impact and can safely be removed from a module.

This location information applies to the instructions physically following this instruction, up to the first occurrence of any of the following: the next end of block, the next OpLine instruction, or the next OpNoLine instruction.

File must be an OpString instruction and is the source-level file name.

Line is the source-level line number. This literal operand is limited to a single word.

Column is the source-level column number. This literal operand is limited to a single word.

OpLine can generally immediately precede other instructions, with the following exceptions:

- it may not be used until after the annotation instructions,
(see the Logical Layout section)

- cannot be the last instruction in a block, which is defined to end with a termination instruction

- if a branch merge instruction is used, the last OpLine in the block must be before its merge instruction

4

8

<id>
File

Literal Number
Line

Literal Number
Column

OpNoLine

Discontinue any source-level location information that might be active from a previous OpLine instruction. This has no semantic impact and can safely be removed from a module.

This instruction can only appear after the annotation instructions (see the Logical Layout section). It cannot be the last instruction in a block, or the second-to-last instruction if the block has a merge instruction. There is not a requirement that there is a preceding OpLine instruction.

1

317

OpModuleProcessed

Document a process that was applied to a module. This has no semantic impact and can safely be removed from a module.

Process is a string describing a process and/or tool (processor) that did the processing. Its form is dependent on the processor.

Missing before version 1.1.

2 + variable

330

Literal String
Process

3.32.3. Annotation Instructions

OpDecorate

Add a Decoration to another <id>.

Target is the <id> to decorate. It can potentially be any <id> that is a forward reference. A set of decorations can be grouped together by having multiple decoration instructions targeting the same OpDecorationGroup instruction.

This instruction is only valid when the Decoration operand is a decoration that takes no Extra Operands, or takes Extra Operands that are not <id> operands.

3 + variable

71

<id>
Target

Decoration

Literal, Literal, …
See Decoration.

OpMemberDecorate

Add a Decoration to a member of a structure type.

Structure type is the <id> of a type from OpTypeStruct.

Member is the number of the member to decorate in the type. The first member is member 0, the next is member 1, …

Note: See OpDecorate for creating groups of decorations for consumption by OpGroupMemberDecorate

4 + variable

72

<id>
Structure Type

Literal Number
Member

Decoration

Literal, Literal, …
See Decoration.

OpDecorationGroup

Deprecated (directly use non-group decoration instructions instead).

A collector for Decorations from OpDecorate and OpDecorateId instructions. All such decoration instructions targeting this OpDecorationGroup instruction must precede it. Subsequent OpGroupDecorate and OpGroupMemberDecorate instructions that consume this instruction’s Result <id> will apply these decorations to their targets.

2

73

Result <id>

OpGroupDecorate

Deprecated (directly use non-group decoration instructions instead).

Add a group of Decorations to another <id>.

Decoration Group is the <id> of an OpDecorationGroup instruction.

Targets is a list of <id>s to decorate with the groups of decorations. The Targets list must not include the <id> of any OpDecorationGroup instruction.

2 + variable

74

<id>
Decoration Group

<id>, <id>, …
Targets

OpGroupMemberDecorate

Deprecated (directly use non-group decoration instructions instead).

Add a group of Decorations to members of structure types.

Decoration Group is the <id> of an OpDecorationGroup instruction.

Targets is a list of (<id>, Member) pairs to decorate with the groups of decorations. Each <id> in the pair must be a target structure type, and the associated Member is the number of the member to decorate in the type. The first member is member 0, the next is member 1, …

2 + variable

75

<id>
Decoration Group

<id>, literal,
<id>, literal,
…
Targets

OpDecorateId

Add a Decoration to another <id>, using <id>s as Extra Operands.

Target is the <id> to decorate. It can potentially be any <id> that is a forward reference. A set of decorations can be grouped together by having multiple decoration instructions targeting the same OpDecorationGroup instruction.

This instruction is only valid when the Decoration operand is a decoration that takes Extra Operands that are <id> operands. All such <id> Extra Operands must be constant instructions or OpVariable instructions.

Missing before version 1.2.

3 + variable

332

<id>
Target

Decoration

<id>, <id>, …
See Decoration.

OpDecorateString (OpDecorateStringGOOGLE)

Add a string Decoration to another <id>.

Target is the <id> to decorate. It can potentially be any <id> that is a forward reference, except it must not be the <id> of an OpDecorationGroup.

Decoration is a decoration that takes at least one Literal String operand, and has only Literal String operands.

Missing before version 1.4.

4 + variable

5632

<id>
Target

Decoration

Literal String
See Decoration.

Optional Literal Strings
See Decoration.

OpMemberDecorateString (OpMemberDecorateStringGOOGLE)

Add a string Decoration to a member of a structure type.

Structure Type is the <id> of an OpTypeStruct.

Member is the number of the member to decorate in the type. The first member is member 0, the next is member 1, …

Decoration is a decoration that takes at least one Literal String operand, and has only Literal String operands.

Missing before version 1.4.

5 + variable

5633

<id>
Struct Type

Literal Number
Member

Decoration

Literal String
See Decoration.

Optional Literal Strings
See Decoration.

3.32.4. Extension Instructions

OpExtension

Declare use of an extension to SPIR-V. This allows validation of additional instructions, tokens, semantics, etc.

Name is the extension’s name string.

2 + variable

10

Literal String
Name

OpExtInstImport

Import an extended set of instructions. It can be later referenced by the Result <id>.

Name is the extended instruction-set’s name string. There must be an external specification defining the semantics for this extended instruction set.

See Extended Instruction Sets for more information.

3 + variable

11

Result <id>

Literal String
Name

OpExtInst

Execute an instruction in an imported set of extended instructions.

Result Type is as defined, per Instruction, in the external specification for Set.

Set is the result of an OpExtInstImport instruction.

Instruction is the enumerant of the instruction to execute within Set. This literal operand is limited to a single word. The semantics of the instruction must be defined in the external specification for Set.

Operand 1, … are the operands to the extended instruction.

5 + variable

12

<id>
Result Type

Result <id>

<id>
Set

Literal Number
Instruction

<id>, <id>, …
Operand 1,
Operand 2,
…

3.32.5. Mode-Setting Instructions

OpMemoryModel

Set addressing model and memory model for the entire module.

Addressing Model selects the module’s Addressing Model.

Memory Model selects the module’s memory model, see Memory Model.

3

14

Addressing Model

Memory Model

OpEntryPoint

Declare an entry point, its execution model, and its interface.

Execution Model is the execution model for the entry point and its static call tree. See Execution Model.

Entry Point must be the Result <id> of an OpFunction instruction.

Name is a name string for the entry point. A module cannot have two OpEntryPoint instructions with the same Execution Model and the same Name string.

Interface is a list of <id> of global OpVariable instructions. These declare the set of global variables from a module that form the interface of this entry point. The set of Interface <id> must be equal to or a superset of the global OpVariable Result <id> referenced by the entry point’s static call tree, within the interface’s storage classes. Before version 1.4, the interface’s storage classes are limited to the Input and Output storage classes. Starting with version 1.4, the interface’s storage classes are all storage classes used in declaring all global variables referenced by the entry point’s call tree.

Interface <id> are forward references. Before version 1.4, duplication of these <id> is tolerated. Starting with version 1.4, an <id> must not appear more than once.

4 + variable

15

Execution Model

<id>
Entry Point

Literal String
Name

<id>, <id>, …
Interface

OpExecutionMode

Declare an execution mode for an entry point.

Entry Point must be the Entry Point <id> operand of an OpEntryPoint instruction.

Mode is the execution mode. See Execution Mode.

This instruction is only valid when the Mode operand is an execution mode that takes no Extra Operands, or takes Extra Operands that are not <id> operands.

3 + variable

16

<id>
Entry Point

Execution Mode
Mode

Literal, Literal, …
See Execution Mode

OpCapability

Declare a capability used by this module.

Capability is the capability declared by this instruction. There are no restrictions on the order in which capabilities are declared.

See the capabilities section for more detail.

2

17

Capability
Capability

OpExecutionModeId

Declare an execution mode for an entry point, using <id>s as Extra Operands.

Entry Point must be the Entry Point <id> operand of an OpEntryPoint instruction.

Mode is the execution mode. See Execution Mode.

This instruction is only valid when the Mode operand is an execution mode that takes Extra Operands that are <id> operands. All such <id> Extra Operands must be constant instructions.

Missing before version 1.2.

3 + variable

331

<id>
Entry Point

Execution Mode
Mode

<id>, <id>, …
See Execution Mode

3.32.6. Type-Declaration Instructions

OpTypeVoid

Declare the void type.

2

19

Result <id>

OpTypeBool

Declare the Boolean type. Values of this type can only be either true or false. There is no physical size or bit pattern defined for these values. If they are stored (in conjunction with OpVariable), they can only be used with logical addressing operations, not physical, and only with non-externally visible shader Storage Classes: Workgroup, CrossWorkgroup, Private, and Function.

2

20

Result <id>

OpTypeInt

Declare a new integer type.

Width specifies how many bits wide the type is. This literal operand is limited to a single word. The bit pattern of a signed integer value is two’s complement.

Signedness specifies whether there are signed semantics to preserve or validate.
0 indicates unsigned, or no signedness semantics
1 indicates signed semantics.
In all cases, the type of operation of an instruction comes from the instruction’s opcode, not the signedness of the operands.

4

21

Result <id>

Literal Number
Width

Literal Number
Signedness

OpTypeFloat

Declare a new floating-point type.

Width specifies how many bits wide the type is. The bit pattern of a floating-point value is as described by the IEEE 754 standard.

3

22

Result <id>

Literal Number
Width

OpTypeVector

Declare a new vector type.

Component Type is the type of each component in the resulting type. It must be a scalar type.

Component Count is the number of components in the resulting type. It must be at least 2.

Components are numbered consecutively, starting with 0.

4

23

Result <id>

<id>
Component Type

Literal Number
Component Count

OpTypeMatrix

Declare a new matrix type.

Column Type is the type of each column in the matrix. It must be vector type.

Column Count is the number of columns in the new matrix type. It must be at least 2.

Matrix columns are numbered consecutively, starting with 0. This is true independently of any Decorations describing the memory layout of a matrix (e.g., RowMajor or MatrixStride).

Capability:
Matrix

4

24

Result <id>

<id>
Column Type

Literal Number
Column Count

OpTypeImage

Declare a new image type. Consumed, for example, by OpTypeSampledImage. This type is opaque: values of this type have no defined physical size or bit pattern.

Sampled Type is the type of the components that result from sampling or reading from this image type. Must be a scalar numerical type or OpTypeVoid.

Dim is the image dimensionality (Dim).

Depth is whether or not this image is a depth image. (Note that whether or not depth comparisons are actually done is a property of the sampling opcode, not of this type declaration.)
0 indicates not a depth image
1 indicates a depth image
2 means no indication as to whether this is a depth or non-depth image

Arrayed must be one of the following indicated values:
0 indicates non-arrayed content
1 indicates arrayed content

MS must be one of the following indicated values:
0 indicates single-sampled content
1 indicates multisampled content

Sampled indicates whether or not this image will be accessed in combination with a sampler, and must be one of the following values:
0 indicates this is only known at run time, not at compile time
1 indicates will be used with sampler
2 indicates will be used without a sampler (a storage image)

Image Format is the Image Format, which can be Unknown, as specified by the client API.

If Dim is SubpassData, Sampled must be 2, Image Format must be Unknown, and the Execution Model must be Fragment.

Access Qualifier is an image Access Qualifier.

9 + variable

25

Result <id>

<id>
Sampled Type

Dim

Literal Number
Depth

Literal Number
Arrayed

Literal Number
MS

Literal Number
Sampled

Image Format

Optional
Access Qualifier

OpTypeSampler

Declare the sampler type. Consumed by OpSampledImage. This type is opaque: values of this type have no defined physical size or bit pattern.

2

26

Result <id>

OpTypeSampledImage

Declare a sampled image type, the Result Type of OpSampledImage, or an externally combined sampler and image. This type is opaque: values of this type have no defined physical size or bit pattern.

Image Type must be an OpTypeImage. It is the type of the image in the combined sampler and image type.

3

27

Result <id>

<id>
Image Type

OpTypeArray

Declare a new array type: a dynamically-indexable ordered aggregate of elements all having the same type.

Element Type is the type of each element in the array.

Length is the number of elements in the array. It must be at least 1. Length must come from a constant instruction of an integer-type scalar whose value is at least 1.

Array elements are number consecutively, starting with 0.

4

28

Result <id>

<id>
Element Type

<id>
Length

OpTypeRuntimeArray

Declare a new run-time array type. Its length is not known at compile time.

Element Type is the type of each element in the array. It must be a concrete type.

See OpArrayLength for getting the Length of an array of this type.

Capability:
Shader

3

29

Result <id>

<id>
Element Type

OpTypeStruct

Declare a new structure type: an aggregate of zero or more potentially heterogeneous members.

Member N type is the type of member N of the structure. The first member is member 0, the next is member 1, …

If an operand is not yet defined, it must be defined by an OpTypePointer, where the type pointed to is an OpTypeStruct.

2 + variable

30

Result <id>

<id>, <id>, …
Member 0 type,
member 1 type,
…

OpTypeOpaque

Declare a structure type with no body specified.

Capability:
Kernel

3 + variable

31

Result <id>

Literal String
The name of the opaque type.

OpTypePointer

Declare a new pointer type.

Storage Class is the Storage Class of the memory holding the object pointed to. If there was a forward reference to this type from an OpTypeForwardPointer, the Storage Class of that instruction must equal the Storage Class of this instruction.

Type is the type of the object pointed to.

4

32

Result <id>

Storage Class

<id>
Type

OpTypeFunction

Declare a new function type.

OpFunction will use this to declare the return type and parameter types of a function. OpFunction is the only valid use of OpTypeFunction.

Return Type is the type of the return value of functions of this type. It must be a concrete or abstract type, or a pointer to such a type. If the function has no return value, Return Type must be OpTypeVoid.

Parameter N Type is the type <id> of the type of parameter N. It must not be OpTypeVoid

3 + variable

33

Result <id>

<id>
Return Type

<id>, <id>, …
Parameter 0 Type,
Parameter 1 Type,
…

OpTypeEvent

Declare an OpenCL event type.

Capability:
Kernel

2

34

Result <id>

OpTypeDeviceEvent

Declare an OpenCL device-side event type.

Capability:
DeviceEnqueue

2

35

Result <id>

OpTypeReserveId

Declare an OpenCL reservation id type.

Capability:
Pipes

2

36

Result <id>

OpTypeQueue

Declare an OpenCL queue type.

Capability:
DeviceEnqueue

2

37

Result <id>

OpTypePipe

Declare an OpenCL pipe type.

Qualifier is the pipe access qualifier.

Capability:
Pipes

3

38

Result <id>

Access Qualifier
Qualifier

OpTypeForwardPointer

Declare the Storage Class for a forward reference to a pointer.

Pointer Type is a forward reference to the result of an OpTypePointer. The type of object the pointer points to is declared by the OpTypePointer instruction, not this instruction. Subsequent OpTypeStruct instructions can use Pointer Type as an operand.

Storage Class is the Storage Class of the memory holding the object pointed to.

Capability:
Addresses, PhysicalStorageBufferAddressesEXT

3

39

<id>
Pointer Type

Storage Class

OpTypePipeStorage

Declare the OpenCL pipe-storage type.

Capability:
PipeStorage

Missing before version 1.1.

2

322

Result <id>

OpTypeNamedBarrier

Declare the named-barrier type.

Capability:
NamedBarrier

Missing before version 1.1.

2

327

Result <id>

OpTypeAccelerationStructureNV

TBD

Capability:
RayTracingNV

2

5341

Result <id>

OpTypeCooperativeMatrixNV

TBD

Capability:
CooperativeMatrixNV

Reserved.

6

5358

Result <id>

<id>
Component Type

Scope <id>
Execution

<id>
Rows

<id>
Columns

3.32.7. Constant-Creation Instructions

OpConstantTrue

Declare a true Boolean-type scalar constant.

Result Type must be the scalar Boolean type.

3

41

<id>
Result Type

Result <id>

OpConstantFalse

Declare a false Boolean-type scalar constant.

Result Type must be the scalar Boolean type.

3

42

<id>
Result Type

Result <id>

OpConstant

Declare a new integer-type or floating-point-type scalar constant.

Result Type must be a scalar integer type or floating-point type.

Value is the bit pattern for the constant. Types 32 bits wide or smaller take one word. Larger types take multiple words, with low-order words appearing first.

3 + variable

43

<id>
Result Type

Result <id>

Literal, Literal, …
Value

OpConstantComposite

Declare a new composite constant.

Result Type must be a composite type, whose top-level members/elements/components/columns have the same type as the types of the Constituents. The ordering must be the same between the top-level types in Result Type and the Constituents.

Constituents will become members of a structure, or elements of an array, or components of a vector, or columns of a matrix. There must be exactly one Constituent for each top-level member/element/component/column of the result. The Constituents must appear in the order needed by the definition of the Result Type. The Constituents must all be <id>s of other constant declarations or an OpUndef.

3 + variable

44

<id>
Result Type

Result <id>

<id>, <id>, …
Constituents

OpConstantSampler

Declare a new sampler constant.

Result Type must be OpTypeSampler.

Sampler Addressing Mode is the addressing mode; a literal from Sampler Addressing Mode.

Param is one of:
0: Non Normalized
1: Normalized

Sampler Filter Mode is the filter mode; a literal from Sampler Filter Mode.

Capability:
LiteralSampler

6

45

<id>
Result Type

Result <id>

Sampler Addressing Mode

Literal Number
Param

Sampler Filter Mode

OpConstantNull

Declare a new null constant value.

The null value is type dependent, defined as follows:
- Scalar Boolean: false
- Scalar integer: 0
- Scalar floating point: +0.0 (all bits 0)
- All other scalars: Abstract
- Composites: Members are set recursively to the null constant according to the null value of their constituent types.

Result Type must be one of the following types:
- Scalar or vector Boolean type
- Scalar or vector integer type
- Scalar or vector floating-point type
- Pointer type
- Event type
- Device side event type
- Reservation id type
- Queue type
- Composite type

3

46

<id>
Result Type

Result <id>

OpSpecConstantTrue

Declare a Boolean-type scalar specialization constant with a default value of true.

This instruction can be specialized to become either an OpConstantTrue or OpConstantFalse instruction.

Result Type must be the scalar Boolean type.

See Specialization.

3

48

<id>
Result Type

Result <id>

OpSpecConstantFalse

Declare a Boolean-type scalar specialization constant with a default value of false.

This instruction can be specialized to become either an OpConstantTrue or OpConstantFalse instruction.

Result Type must be the scalar Boolean type.

See Specialization.

3

49

<id>
Result Type

Result <id>

OpSpecConstant

Declare a new integer-type or floating-point-type scalar specialization constant.

Result Type must be a scalar integer type or floating-point type.

Value is the bit pattern for the default value of the constant. Types 32 bits wide or smaller take one word. Larger types take multiple words, with low-order words appearing first.

This instruction can be specialized to become an OpConstant instruction.

See Specialization.

3 + variable

50

<id>
Result Type

Result <id>

Literal, Literal, …
Value

OpSpecConstantComposite

Declare a new composite specialization constant.

Result Type must be a composite type, whose top-level members/elements/components/columns have the same type as the types of the Constituents. The ordering must be the same between the top-level types in Result Type and the Constituents.

Constituents will become members of a structure, or elements of an array, or components of a vector, or columns of a matrix. There must be exactly one Constituent for each top-level member/element/component/column of the result. The Constituents must appear in the order needed by the definition of the type of the result. The Constituents must be the <id> of other specialization constant or constant declarations.

This instruction will be specialized to an OpConstantComposite instruction.

See Specialization.

3 + variable

51

<id>
Result Type

Result <id>

<id>, <id>, …
Constituents

OpSpecConstantOp

Declare a new specialization constant that results from doing an operation.

Result Type must be the type required by the Result Type of Opcode.

Opcode must be one of the following opcodes. This literal operand is limited to a single word.
OpSConvert, OpUConvert (missing before version 1.4), OpFConvert
OpSNegate, OpNot
OpIAdd, OpISub
OpIMul, OpUDiv, OpSDiv, OpUMod, OpSRem, OpSMod
OpShiftRightLogical, OpShiftRightArithmetic, OpShiftLeftLogical
OpBitwiseOr, OpBitwiseXor, OpBitwiseAnd
OpVectorShuffle, OpCompositeExtract, OpCompositeInsert
OpLogicalOr, OpLogicalAnd, OpLogicalNot,
OpLogicalEqual, OpLogicalNotEqual
OpSelect
OpIEqual, OpINotEqual
OpULessThan, OpSLessThan
OpUGreaterThan, OpSGreaterThan
OpULessThanEqual, OpSLessThanEqual
OpUGreaterThanEqual, OpSGreaterThanEqual

If the Shader capability was declared, the following opcode is also valid:
OpQuantizeToF16

If the Kernel capability was declared, the following opcodes are also valid:
OpConvertFToS, OpConvertSToF
OpConvertFToU, OpConvertUToF
OpUConvert
OpConvertPtrToU, OpConvertUToPtr
OpGenericCastToPtr, OpPtrCastToGeneric
OpBitcast
OpFNegate
OpFAdd, OpFSub
OpFMul, OpFDiv
OpFRem, OpFMod
OpAccessChain, OpInBoundsAccessChain
OpPtrAccessChain, OpInBoundsPtrAccessChain

Operands are the operands required by opcode, and satisfy the semantics of opcode. In addition, all Operands must be either:
- the <id>s of other constant instructions, or
- OpUndef, when allowed by opcode, or
- for the AccessChain named opcodes, their Base is allowed to be a global (module scope) OpVariable instruction.

See Specialization.

4 + variable

52

<id>
Result Type

Result <id>

Literal Number
Opcode

<id>, <id>, …
Operands

3.32.8. Memory Instructions

OpVariable

Allocate an object in memory, resulting in a pointer to it, which can be used with OpLoad and OpStore.

Result Type must be an OpTypePointer. Its Type operand is the type of object in memory.

Storage Class is the Storage Class of the memory holding the object. It cannot be Generic. It must be the same as the Storage Class operand of the Result Type.

Initializer is optional. If Initializer is present, it will be the initial value of the variable’s memory content. Initializer must be an <id> from a constant instruction or a global (module scope) OpVariable instruction. Initializer must have the same type as the type pointed to by Result Type.

4 + variable

59

<id>
Result Type

Result <id>

Storage Class

Optional
<id>
Initializer

OpImageTexelPointer

Form a pointer to a texel of an image. Use of such a pointer is limited to atomic operations.

Result Type must be an OpTypePointer whose Storage Class operand is Image. Its Type operand must be a scalar numerical type or OpTypeVoid.

Image must have a type of OpTypePointer with Type OpTypeImage. The Sampled Type of the type of Image must be the same as the Type pointed to by Result Type. The Dim operand of Type cannot be SubpassData.

Coordinate and Sample specify which texel and sample within the image to form a pointer to.

Coordinate must be a scalar or vector of integer type. It must have the number of components specified below, given the following Arrayed and Dim operands of the type of the OpTypeImage.

If Arrayed is 0:
1D: scalar
2D: 2 components
3D: 3 components
Cube: 3 components
Rect: 2 components
Buffer: scalar

If Arrayed is 1:
1D: 2 components
2D: 3 components
Cube: 3 components; the face and layer combine into the 3rd component, layer_face, such that face is layer_face % 6 and layer is floor(layer_face / 6)

Sample must be an integer type scalar. It specifies which sample to select at the given coordinate. It must be a valid <id> for the value 0 if the OpTypeImage has MS of 0.

6

60

<id>
Result Type

Result <id>

<id>
Image

<id>
Coordinate

<id>
Sample

OpLoad

Load through a pointer.

Result Type is the type of the loaded object. It must be a type with fixed size; i.e., it cannot be, nor include, any OpTypeRuntimeArray types.

Pointer is the pointer to load through. Its type must be an OpTypePointer whose Type operand is the same as Result Type.

If present, any Memory Operands must begin with a memory operand literal. If not present, it is the same as specifying the memory operand None.

4 + variable

61

<id>
Result Type

Result <id>

<id>
Pointer

Optional
Memory Operands

OpStore

Store through a pointer.

Pointer is the pointer to store through. Its type must be an OpTypePointer whose Type operand is the same as the type of Object.

Object is the object to store.

If present, any Memory Operands must begin with a memory operand literal. If not present, it is the same as specifying the memory operand None.

3 + variable

62

<id>
Pointer

<id>
Object

Optional
Memory Operands

OpCopyMemory

Copy from the memory pointed to by Source to the memory pointed to by Target. Both operands must be non-void pointers and having the same <id> Type operand in their OpTypePointer type declaration. Matching Storage Class is not required. The amount of memory copied is the size of the type pointed to. The copied type must have a fixed size; i.e., it cannot be, nor include, any OpTypeRuntimeArray types.

If present, any Memory Operands must begin with a memory operand literal. If not present, it is the same as specifying the memory operand None. Before version 1.4, at most one memory operands mask can be provided. Starting with version 1.4 two masks can be provided, as described in Memory Operands. If no masks or only one mask is present, it applies to both Source and Target. If two masks are present, the first applies to Target and the second applies to Source.

3 + variable

63

<id>
Target

<id>
Source

Optional
Memory Operands

Optional
Memory Operands

OpCopyMemorySized

Copy from the memory pointed to by Source to the memory pointed to by Target.

Size is the number of bytes to copy. It must have a scalar integer type. If it is a constant instruction, the constant value cannot be 0. It is invalid for both the constant’s type to have Signedness of 1 and to have the sign bit set. Otherwise, as a run-time value, Size is treated as unsigned, and if its value is 0, no memory access will be made.

If present, any Memory Operands must begin with a memory operand literal. If not present, it is the same as specifying the memory operand None. Before version 1.4, at most one memory operands mask can be provided. Starting with version 1.4 two masks can be provided, as described in Memory Operands. If no masks or only one mask is present, it applies to both Source and Target. If two masks are present, the first applies to Target and the second applies to Source.

Capability:
Addresses

4 + variable

64

<id>
Target

<id>
Source

<id>
Size

Optional
Memory Operands

Optional
Memory Operands

OpAccessChain

Create a pointer into a composite object that can be used with OpLoad and OpStore.

Result Type must be an OpTypePointer. Its Type operand must be the type reached by walking the Base’s type hierarchy down to the last provided index in Indexes, and its Storage Class operand must be the same as the Storage Class of Base.

Base must be a pointer, pointing to the base of a composite object.

Indexes walk the type hierarchy to the desired depth, potentially down to scalar granularity. The first index in Indexes will select the top-level member/element/component/element of the base composite. All composite constituents use zero-based numbering, as described by their OpType… instruction. The second index will apply similarly to that result, and so on. Once any non-composite type is reached, there must be no remaining (unused) indexes.

Each index in Indexes
- must be a scalar integer type,
- is treated as a signed count, and
- must be an OpConstant when indexing into a structure.

4 + variable

65

<id>
Result Type

Result <id>

<id>
Base

<id>, <id>, …
Indexes

OpInBoundsAccessChain

Has the same semantics as OpAccessChain, with the addition that the resulting pointer is known to point within the base object.

4 + variable

66

<id>
Result Type

Result <id>

<id>
Base

<id>, <id>, …
Indexes

OpPtrAccessChain

Has the same semantics as OpAccessChain, with the addition of the Element operand.

Element is used to do an initial dereference of Base: Base is treated as the address of an element in an array, and a new element address is computed from Base and Element to become the OpAccessChain Base to dereference as per OpAccessChain. This computed Base has the same type as the originating Base.

To compute the new element address, Element is treated as a signed count of elements E, relative to the original Base element B, and the address of element B + E is computed using enough precision to avoid overflow and underflow. For objects in the Uniform, StorageBuffer, or PushConstant storage classes, the element’s address or location is calculated using a stride, which will be the Base-type’s Array Stride when the Base type is decorated with ArrayStride. For all other objects, the implementation will calculate the element’s address or location.

With one exception, undefined behavior results when B + E is not an element in the same array (same innermost array, if array types are nested) as B. The exception being that the result is still well defined when B + E = L, where L is the length of the array: the address computation for element L is done with the same stride as any other B + E computation that stays within the array.

Note: If Base is typed to be a pointer to an array and the desired operation is to select an element of that array, OpAccessChain should be directly used, as its first Index will select the array element.

Capability:
Addresses, VariablePointers, VariablePointersStorageBuffer, PhysicalStorageBufferAddressesEXT

5 + variable

67

<id>
Result Type

Result <id>

<id>
Base

<id>
Element

<id>, <id>, …
Indexes

OpArrayLength

Length of a run-time array.

Result Type must be an OpTypeInt with 32-bit Width and 0 Signedness.

Structure must be a pointer to an OpTypeStruct whose last member is a run-time array.

Array member is the index of the last member of the structure that Structure points to. That member’s type must be from OpTypeRuntimeArray.

Capability:
Shader

5

68

<id>
Result Type

Result <id>

<id>
Structure

Literal Number
Array member

OpGenericPtrMemSemantics

Result is a valid Memory Semantics which includes mask bits set for the Storage Class for the specific (non-Generic) Storage Class of Pointer.

Pointer must point to Generic Storage Class.

Result Type must be an OpTypeInt with 32-bit Width and 0 Signedness.

Capability:
Kernel

4

69

<id>
Result Type

Result <id>

<id>
Pointer

OpInBoundsPtrAccessChain

Has the same semantics as OpPtrAccessChain, with the addition that the resulting pointer is known to point within the base object.

Capability:
Addresses

5 + variable

70

<id>
Result Type

Result <id>

<id>
Base

<id>
Element

<id>, <id>, …
Indexes

OpPtrEqual

Result is true if Operand 1 and Operand 2 have the same value. Result is false if Operand 1 and Operand 2 have different values.

Result Type must be a Boolean type scalar.

The types of Operand 1 and Operand 2 must be OpTypePointer of the same type.

Missing before version 1.4.

5

401

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpPtrNotEqual

Result is true if Operand 1 and Operand 2 have different values. Result is false if Operand 1 and Operand 2 have the same value.

Result Type must be a Boolean type scalar.

The types of Operand 1 and Operand 2 must be OpTypePointer of the same type.

Missing before version 1.4.

5

402

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpPtrDiff

Element-number subtraction: The number of elements to add to Operand 2 to get to Operand 1.

Result Type must be an integer type scalar. It will be computed as a signed value, as negative differences are allowed, independently of the signed bit in the type. The result will equal the low-order N bits of the correct result R, where R is computed with enough precision to avoid overflow and underflow and Result Type has a bitwidth of N bits.

The units of Result Type are a count of elements. I.e., the same value you would use as the Element operand to OpPtrAccessChain.

The types of Operand 1 and Operand 2 must be OpTypePointer of exactly the same type, and point to a type that can be aggregated into an array. For an array of length L, Operand 1 and Operand 2 can point to any element in the range [0, L], where element L is outside the array but has a representative address computed with the same stride as elements in the array. Additionally, Operand 1 must be a valid Base operand of OpPtrAccessChain. Behavior is undefined if Operand 1 and Operand 2 are not pointers to element numbers in [0, L] in the same array.

Capability:
Addresses, VariablePointers, VariablePointersStorageBuffer

Missing before version 1.4.

5

403

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

3.32.9. Function Instructions

OpFunction

Add a function. This instruction must be immediately followed by one OpFunctionParameter instruction per each formal parameter of this function. This function’s body or declaration will terminate with the next OpFunctionEnd instruction.

Result Type must be the same as the Return Type declared in Function Type.

Function Type is the result of an OpTypeFunction, which declares the types of the return value and parameters of the function.

5

54

<id>
Result Type

Result <id>

Function Control

<id>
Function Type

OpFunctionParameter

Declare a formal parameter of the current function.

Result Type is the type of the parameter.

This instruction must immediately follow an OpFunction or OpFunctionParameter instruction. The order of contiguous OpFunctionParameter instructions is the same order arguments will be listed in an OpFunctionCall instruction to this function. It is also the same order in which Parameter Type operands are listed in the OpTypeFunction of the Function Type operand for this function’s OpFunction instruction.

3

55

<id>
Result Type

Result <id>

OpFunctionEnd

Last instruction of a function.

1

56

OpFunctionCall

Call a function.

Result Type is the type of the return value of the function. It must be the same as the Return Type operand of the Function Type operand of the Function operand.

Function is an OpFunction instruction. This could be a forward reference.

Argument N is the object to copy to parameter N of Function.

Note: A forward call is possible because there is no missing type information: Result Type must match the Return Type of the function, and the calling argument types must match the formal parameter types.

4 + variable

57

<id>
Result Type

Result <id>

<id>
Function

<id>, <id>, …
Argument 0,
Argument 1,
…

3.32.10. Image Instructions

OpSampledImage

Create a sampled image, containing both a sampler and an image.

Result Type must be the OpTypeSampledImage type whose Image Type operand is the type of Image.

Image is an object whose type is an OpTypeImage, whose Sampled operand is 0 or 1, and whose Dim operand is not SubpassData.

Sampler must be an object whose type is OpTypeSampler.

5

86

<id>
Result Type

Result <id>

<id>
Image

<id>
Sampler

OpImageSampleImplicitLod

Sample an image with an implicit level of detail.

Result Type must be a vector of four components of floating-point type or integer type. Its components must be the same as Sampled Type of the underlying OpTypeImage (unless that underlying Sampled Type is OpTypeVoid).

Sampled Image must be an object whose type is OpTypeSampledImage.

Coordinate must be a scalar or vector of floating-point type. It contains (u[, v] … [, array layer]) as needed by the definition of Sampled Image. It may be a vector larger than needed, but all unused components will appear after all used components.

Image Operands encodes what operands follow, as per Image Operands.

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

Capability:
Shader

5 + variable

87

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

Optional
Image Operands

Optional
<id>, <id>, …

OpImageSampleExplicitLod

Sample an image using an explicit level of detail.

Result Type must be a vector of four components of floating-point type or integer type. Its components must be the same as Sampled Type of the underlying OpTypeImage (unless that underlying Sampled Type is OpTypeVoid).

Sampled Image must be an object whose type is OpTypeSampledImage.

Coordinate must be a scalar or vector of floating-point type or integer type. It contains (u[, v] … [, array layer]) as needed by the definition of Sampled Image. Unless the Kernel capability is being used, it must be floating point. It may be a vector larger than needed, but all unused components will appear after all used components.

Image Operands encodes what operands follow, as per Image Operands. At least one operand setting the level of detail must be present.

7 + variable

88

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

Image Operands

<id>

Optional
<id>, <id>, …

OpImageSampleDrefImplicitLod

Sample an image doing depth-comparison with an implicit level of detail.

Result Type must be a scalar of integer type or floating-point type. It must be the same as Sampled Type of the underlying OpTypeImage.

Sampled Image must be an object whose type is OpTypeSampledImage.

Coordinate must be a scalar or vector of floating-point type. It contains (u[, v] … [, array layer]) as needed by the definition of Sampled Image. It may be a vector larger than needed, but all unused components will appear after all used components.

Dref is the depth-comparison reference value.

Image Operands encodes what operands follow, as per Image Operands.

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

Capability:
Shader

6 + variable

89

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

<id>
Dref

Optional
Image Operands

Optional
<id>, <id>, …

OpImageSampleDrefExplicitLod

Sample an image doing depth-comparison using an explicit level of detail.

Result Type must be a scalar of integer type or floating-point type. It must be the same as Sampled Type of the underlying OpTypeImage.

Sampled Image must be an object whose type is OpTypeSampledImage.

Coordinate must be a scalar or vector of floating-point type. It contains (u[, v] … [, array layer]) as needed by the definition of Sampled Image. It may be a vector larger than needed, but all unused components will appear after all used components.

Dref is the depth-comparison reference value.

Image Operands encodes what operands follow, as per Image Operands. At least one operand setting the level of detail must be present.

Capability:
Shader

8 + variable

90

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

<id>
Dref

Image Operands

<id>

Optional
<id>, <id>, …

OpImageSampleProjImplicitLod

Sample an image with with a project coordinate and an implicit level of detail.

Result Type must be a vector of four components of floating-point type or integer type. Its components must be the same as Sampled Type of the underlying OpTypeImage (unless that underlying Sampled Type is OpTypeVoid).

Sampled Image must be an object whose type is OpTypeSampledImage. The Dim operand of the underlying OpTypeImage must be 1D, 2D, 3D, or Rect, and the Arrayed and MS operands must be 0.

Coordinate is a floating-point vector containing (u [, v] [, w], q), as needed by the definition of Sampled Image, with the q component consumed for the projective division. That is, the actual sample coordinate will be (u/q [, v/q] [, w/q]), as needed by the definition of Sampled Image. It may be a vector larger than needed, but all unused components will appear after all used components.

Image Operands encodes what operands follow, as per Image Operands.

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

Capability:
Shader

5 + variable

91

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

Optional
Image Operands

Optional
<id>, <id>, …

OpImageSampleProjExplicitLod

Sample an image with a project coordinate using an explicit level of detail.

Result Type must be a vector of four components of floating-point type or integer type. Its components must be the same as Sampled Type of the underlying OpTypeImage (unless that underlying Sampled Type is OpTypeVoid).

Sampled Image must be an object whose type is OpTypeSampledImage. The Dim operand of the underlying OpTypeImage must be 1D, 2D, 3D, or Rect, and the Arrayed and MS operands must be 0.

Coordinate is a floating-point vector containing (u [, v] [, w], q), as needed by the definition of Sampled Image, with the q component consumed for the projective division. That is, the actual sample coordinate will be (u/q [, v/q] [, w/q]), as needed by the definition of Sampled Image. It may be a vector larger than needed, but all unused components will appear after all used components.

Image Operands encodes what operands follow, as per Image Operands. At least one operand setting the level of detail must be present.

Capability:
Shader

7 + variable

92

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

Image Operands

<id>

Optional
<id>, <id>, …

OpImageSampleProjDrefImplicitLod

Sample an image with a project coordinate, doing depth-comparison, with an implicit level of detail.

Result Type must be a scalar of integer type or floating-point type. It must be the same as Sampled Type of the underlying OpTypeImage.

Sampled Image must be an object whose type is OpTypeSampledImage. The Dim operand of the underlying OpTypeImage must be 1D, 2D, 3D, or Rect, and the Arrayed and MS operands must be 0.

Coordinate is a floating-point vector containing (u [, v] [, w], q), as needed by the definition of Sampled Image, with the q component consumed for the projective division. That is, the actual sample coordinate will be (u/q [, v/q] [, w/q]), as needed by the definition of Sampled Image. It may be a vector larger than needed, but all unused components will appear after all used components.

Dref /q is the depth-comparison reference value.

Image Operands encodes what operands follow, as per Image Operands.

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

Capability:
Shader

6 + variable

93

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

<id>
Dref

Optional
Image Operands

Optional
<id>, <id>, …

OpImageSampleProjDrefExplicitLod

Sample an image with a project coordinate, doing depth-comparison, using an explicit level of detail.

Result Type must be a scalar of integer type or floating-point type. It must be the same as Sampled Type of the underlying OpTypeImage.

Sampled Image must be an object whose type is OpTypeSampledImage. The Dim operand of the underlying OpTypeImage must be 1D, 2D, 3D, or Rect, and the Arrayed and MS operands must be 0.

Coordinate is a floating-point vector containing (u [, v] [, w], q), as needed by the definition of Sampled Image, with the q component consumed for the projective division. That is, the actual sample coordinate will be (u/q [, v/q] [, w/q]), as needed by the definition of Sampled Image. It may be a vector larger than needed, but all unused components will appear after all used components.

Dref /q is the depth-comparison reference value.

Image Operands encodes what operands follow, as per Image Operands. At least one operand setting the level of detail must be present.

Capability:
Shader

8 + variable

94

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

<id>
Dref

Image Operands

<id>

Optional
<id>, <id>, …

OpImageFetch

Fetch a single texel from an image whose Sampled operand is 1.

Result Type must be a vector of four components of floating-point type or integer type. Its components must be the same as Sampled Type of the underlying OpTypeImage (unless that underlying Sampled Type is OpTypeVoid).

Image must be an object whose type is OpTypeImage. Its Dim operand cannot be Cube, and its Sampled operand must be 1.

Coordinate is an integer scalar or vector containing (u[, v] … [, array layer]) as needed by the definition of Sampled Image.

Image Operands encodes what operands follow, as per Image Operands.

5 + variable

95

<id>
Result Type

Result <id>

<id>
Image

<id>
Coordinate

Optional
Image Operands

Optional
<id>, <id>, …

OpImageGather

Gathers the requested component from four texels.

Result Type must be a vector of four components of floating-point type or integer type. Its components must be the same as Sampled Type of the underlying OpTypeImage (unless that underlying Sampled Type is OpTypeVoid). It has one component per gathered texel.

Sampled Image must be an object whose type is OpTypeSampledImage. Its OpTypeImage must have a Dim of 2D, Cube, or Rect.

Coordinate must be a scalar or vector of floating-point type. It contains (u[, v] … [, array layer]) as needed by the definition of Sampled Image.

Component is the component number that will be gathered from all four texels. It must be 0, 1, 2 or 3.

Image Operands encodes what operands follow, as per Image Operands.

Capability:
Shader

6 + variable

96

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

<id>
Component

Optional
Image Operands

Optional
<id>, <id>, …

OpImageDrefGather

Gathers the requested depth-comparison from four texels.

Result Type must be a vector of four components of floating-point type or integer type. Its components must be the same as Sampled Type of the underlying OpTypeImage (unless that underlying Sampled Type is OpTypeVoid). It has one component per gathered texel.

Sampled Image must be an object whose type is OpTypeSampledImage. Its OpTypeImage must have a Dim of 2D, Cube, or Rect.

Coordinate must be a scalar or vector of floating-point type. It contains (u[, v] … [, array layer]) as needed by the definition of Sampled Image.

Dref is the depth-comparison reference value.

Image Operands encodes what operands follow, as per Image Operands.

Capability:
Shader

6 + variable

97

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

<id>
Dref

Optional
Image Operands

Optional
<id>, <id>, …

OpImageRead

Read a texel from an image without a sampler.

Result Type must be a scalar or vector of floating-point type or integer type. Its component type must be the same as Sampled Type of the OpTypeImage (unless that Sampled Type is OpTypeVoid).

Image must be an object whose type is OpTypeImage with a Sampled operand of 0 or 2. If the Sampled operand is 2, then some dimensions require a capability; e.g., Image1D, ImageRect, or ImageBuffer. If the Arrayed operand is 1, then additional capabilities may be required; e.g., ImageCubeArray, or ImageMSArray.

Coordinate is an integer scalar or vector containing non-normalized texel coordinates (u[, v] … [, array layer]) as needed by the definition of Image. If the coordinates are outside the image, the memory location that is accessed is undefined.

When the Image Dim operand is SubpassData, Coordinate is relative to the current fragment location. That is, the integer value (rounded down) of the current fragment’s window-relative (x, y) coordinate is added to (u, v).

When the Image Dim operand is not SubpassData, the Image Format must not be Unknown, unless the StorageImageReadWithoutFormat Capability was declared.

Image Operands encodes what operands follow, as per Image Operands.

5 + variable

98

<id>
Result Type

Result <id>

<id>
Image

<id>
Coordinate

Optional
Image Operands

Optional
<id>, <id>, …

OpImageWrite

Write a texel to an image without a sampler.

Image must be an object whose type is OpTypeImage with a Sampled operand of 0 or 2. If the Sampled operand is 2, then some dimensions require a capability; e.g., Image1D, ImageRect, or ImageBuffer. If the Arrayed operand is 1, then additional capabilities may be required; e.g., ImageCubeArray, or ImageMSArray. Its Dim operand cannot be SubpassData.

Coordinate is an integer scalar or vector containing non-normalized texel coordinates (u[, v] … [, array layer]) as needed by the definition of Image. If the coordinates are outside the image, the memory location that is accessed is undefined.

Texel is the data to write. Its component type must be the same as Sampled Type of the OpTypeImage (unless that Sampled Type is OpTypeVoid).

The Image Format must not be Unknown, unless the StorageImageWriteWithoutFormat Capability was declared.

Image Operands encodes what operands follow, as per Image Operands.

4 + variable

99

<id>
Image

<id>
Coordinate

<id>
Texel

Optional
Image Operands

Optional
<id>, <id>, …

OpImage

Extract the image from a sampled image.

Result Type must be OpTypeImage.

Sampled Image must have type OpTypeSampledImage whose Image Type is the same as Result Type.

4

100

<id>
Result Type

Result <id>

<id>
Sampled Image

OpImageQueryFormat

Query the image format of an image created with an Unknown Image Format.

Result Type must be a scalar integer type. The resulting value is an enumerant from Image Channel Data Type.

Image must be an object whose type is OpTypeImage.

Capability:
Kernel

4

101

<id>
Result Type

Result <id>

<id>
Image

OpImageQueryOrder

Query the channel order of an image created with an Unknown Image Format.

Result Type must be a scalar integer type. The resulting value is an enumerant from Image Channel Order.

Image must be an object whose type is OpTypeImage.

Capability:
Kernel

4

102

<id>
Result Type

Result <id>

<id>
Image

OpImageQuerySizeLod

Query the dimensions of Image for mipmap level for Level of Detail.

Result Type must be an integer type scalar or vector. The number of components must be
1 for the 1D dimensionality,
2 for the 2D and Cube dimensionalities,
3 for the 3D dimensionality,
plus 1 more if the image type is arrayed. This vector is filled in with (width [, height] [, depth] [, elements]) where elements is the number of layers in an image array, or the number of cubes in a cube-map array.

Image must be an object whose type is OpTypeImage. Its Dim operand must be one of 1D, 2D, 3D, or Cube, and its MS must be 0. See OpImageQuerySize for querying image types without level of detail. This operation is allowed on an image decorated as NonReadable. See the client API specification for additional image type restrictions.

Level of Detail is used to compute which mipmap level to query, as specified by the client API.

Capability:
Kernel, ImageQuery

5

103

<id>
Result Type

Result <id>

<id>
Image

<id>
Level of Detail

OpImageQuerySize

Query the dimensions of Image, with no level of detail.

Result Type must be an integer type scalar or vector. The number of components must be:
1 for the 1D and Buffer dimensionalities,
2 for the 2D, Cube, and Rect dimensionalities,
3 for the 3D dimensionality,
plus 1 more if the image type is arrayed. This vector is filled in with (width [, height] [, elements]) where elements is the number of layers in an image array or the number of cubes in a cube-map array.

Image must be an object whose type is OpTypeImage. Its Dim operand must be one of those listed under Result Type, above. Additionally, if its Dim is 1D, 2D, 3D, or Cube, it must also have either an MS of 1 or a Sampled of 0 or 2. There is no implicit level-of-detail consumed by this instruction. See OpImageQuerySizeLod for querying images having level of detail. This operation is allowed on an image decorated as NonReadable. See the client API specification for additional image type restrictions.

Capability:
Kernel, ImageQuery

4

104

<id>
Result Type

Result <id>

<id>
Image

OpImageQueryLod

Query the mipmap level and the level of detail for a hypothetical sampling of Image at Coordinate using an implicit level of detail.

Result Type must be a two-component floating-point type vector.
The first component of the result will contain the mipmap array layer.
The second component of the result will contain the implicit level of detail relative to the base level.

Sampled Image must be an object whose type is OpTypeSampledImage. Its Dim operand must be one of 1D, 2D, 3D, or Cube.

Coordinate must be a scalar or vector of floating-point type or integer type. It contains (u[, v] … ) as needed by the definition of Sampled Image, not including any array layer index. Unless the Kernel capability is being used, it must be floating point.

If called on an incomplete image, the results are undefined.

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

Capability:
ImageQuery

5

105

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

OpImageQueryLevels

Query the number of mipmap levels accessible through Image.

Result Type must be a scalar integer type. The result is the number of mipmap levels,as specified by the client API.

Image must be an object whose type is OpTypeImage. Its Dim operand must be one of 1D, 2D, 3D, or Cube. See the client API specification for additional image type restrictions.

Capability:
Kernel, ImageQuery

4

106

<id>
Result Type

Result <id>

<id>
Image

OpImageQuerySamples

Query the number of samples available per texel fetch in a multisample image.

Result Type must be a scalar integer type. The result is the number of samples.

Image must be an object whose type is OpTypeImage. Its Dim operand must be one of 2D and MS of 1.

Capability:
Kernel, ImageQuery

4

107

<id>
Result Type

Result <id>

<id>
Image

OpImageSparseSampleImplicitLod

Sample a sparse image with an implicit level of detail.

Result Type must be an OpTypeStruct with two members. The first member’s type must be an integer type scalar. It will hold a Residency Code that can be passed to OpImageSparseTexelsResident. The second member must be a vector of four components of floating-point type or integer type. Its components must be the same as Sampled Type of the underlying OpTypeImage (unless that underlying Sampled Type is OpTypeVoid).

Sampled Image must be an object whose type is OpTypeSampledImage.

Coordinate must be a scalar or vector of floating-point type. It contains (u[, v] … [, array layer]) as needed by the definition of Sampled Image. It may be a vector larger than needed, but all unused components will appear after all used components.

Image Operands encodes what operands follow, as per Image Operands.

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

Capability:
SparseResidency

5 + variable

305

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

Optional
Image Operands

Optional
<id>, <id>, …

OpImageSparseSampleExplicitLod

Sample a sparse image using an explicit level of detail.

Result Type must be an OpTypeStruct with two members. The first member’s type must be an integer type scalar. It will hold a Residency Code that can be passed to OpImageSparseTexelsResident. The second member must be a vector of four components of floating-point type or integer type. Its components must be the same as Sampled Type of the underlying OpTypeImage (unless that underlying Sampled Type is OpTypeVoid).

Sampled Image must be an object whose type is OpTypeSampledImage.

Coordinate must be a scalar or vector of floating-point type or integer type. It contains (u[, v] … [, array layer]) as needed by the definition of Sampled Image. Unless the Kernel capability is being used, it must be floating point. It may be a vector larger than needed, but all unused components will appear after all used components.

Image Operands encodes what operands follow, as per Image Operands. At least one operand setting the level of detail must be present.

Capability:
SparseResidency

7 + variable

306

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

Image Operands

<id>

Optional
<id>, <id>, …

OpImageSparseSampleDrefImplicitLod

Sample a sparse image doing depth-comparison with an implicit level of detail.

Result Type must be an OpTypeStruct with two members. The first member’s type must be an integer type scalar. It will hold a Residency Code that can be passed to OpImageSparseTexelsResident. The second member must be a scalar of integer type or floating-point type. It must be the same as Sampled Type of the underlying OpTypeImage.

Sampled Image must be an object whose type is OpTypeSampledImage.

Coordinate must be a scalar or vector of floating-point type. It contains (u[, v] … [, array layer]) as needed by the definition of Sampled Image. It may be a vector larger than needed, but all unused components will appear after all used components.

Dref is the depth-comparison reference value.

Image Operands encodes what operands follow, as per Image Operands.

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

Capability:
SparseResidency

6 + variable

307

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

<id>
Dref

Optional
Image Operands

Optional
<id>, <id>, …

OpImageSparseSampleDrefExplicitLod

Sample a sparse image doing depth-comparison using an explicit level of detail.

Result Type must be an OpTypeStruct with two members. The first member’s type must be an integer type scalar. It will hold a Residency Code that can be passed to OpImageSparseTexelsResident. The second member must be a scalar of integer type or floating-point type. It must be the same as Sampled Type of the underlying OpTypeImage.

Sampled Image must be an object whose type is OpTypeSampledImage.

Coordinate must be a scalar or vector of floating-point type. It contains (u[, v] … [, array layer]) as needed by the definition of Sampled Image. It may be a vector larger than needed, but all unused components will appear after all used components.

Dref is the depth-comparison reference value.

Image Operands encodes what operands follow, as per Image Operands. At least one operand setting the level of detail must be present.

Capability:
SparseResidency

8 + variable

308

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

<id>
Dref

Image Operands

<id>

Optional
<id>, <id>, …

OpImageSparseSampleProjImplicitLod

Sample a sparse image with a projective coordinate and an implicit level of detail.

Capability:
SparseResidency

Reserved.

5 + variable

309

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

Optional
Image Operands

Optional
<id>, <id>, …

OpImageSparseSampleProjExplicitLod

Sample a sparse image with a projective coordinate using an explicit level of detail.

Capability:
SparseResidency

Reserved.

7 + variable

310

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

Image Operands

<id>

Optional
<id>, <id>, …

OpImageSparseSampleProjDrefImplicitLod

Sample a sparse image with a projective coordinate, doing depth-comparison, with an implicit level of detail.

Capability:
SparseResidency

Reserved.

6 + variable

311

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

<id>
Dref

Optional
Image Operands

Optional
<id>, <id>, …

OpImageSparseSampleProjDrefExplicitLod

Sample a sparse image with a projective coordinate, doing depth-comparison, using an explicit level of detail.

Capability:
SparseResidency

Reserved.

8 + variable

312

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

<id>
Dref

Image Operands

<id>

Optional
<id>, <id>, …

OpImageSparseFetch

Fetch a single texel from a sampled sparse image.

Result Type must be an OpTypeStruct with two members. The first member’s type must be an integer type scalar. It will hold a Residency Code that can be passed to OpImageSparseTexelsResident. The second member must be a vector of four components of floating-point type or integer type. Its components must be the same as Sampled Type of the underlying OpTypeImage (unless that underlying Sampled Type is OpTypeVoid).

Image must be an object whose type is OpTypeImage. Its Dim operand cannot be Cube.

Coordinate is an integer scalar or vector containing (u[, v] … [, array layer]) as needed by the definition of Sampled Image.

Image Operands encodes what operands follow, as per Image Operands.

Capability:
SparseResidency

5 + variable

313

<id>
Result Type

Result <id>

<id>
Image

<id>
Coordinate

Optional
Image Operands

Optional
<id>, <id>, …

OpImageSparseGather

Gathers the requested component from four texels of a sparse image.

Result Type must be an OpTypeStruct with two members. The first member’s type must be an integer type scalar. It will hold a Residency Code that can be passed to OpImageSparseTexelsResident. The second member must be a vector of four components of floating-point type or integer type. Its components must be the same as Sampled Type of the underlying OpTypeImage (unless that underlying Sampled Type is OpTypeVoid). It has one component per gathered texel.

Sampled Image must be an object whose type is OpTypeSampledImage. Its OpTypeImage must have a Dim of 2D, Cube, or Rect.

Coordinate must be a scalar or vector of floating-point type. It contains (u[, v] … [, array layer]) as needed by the definition of Sampled Image.

Component is the component number that will be gathered from all four texels. It must be 0, 1, 2 or 3.

Image Operands encodes what operands follow, as per Image Operands.

Capability:
SparseResidency

6 + variable

314

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

<id>
Component

Optional
Image Operands

Optional
<id>, <id>, …

OpImageSparseDrefGather

Gathers the requested depth-comparison from four texels of a sparse image.

Result Type must be an OpTypeStruct with two members. The first member’s type must be an integer type scalar. It will hold a Residency Code that can be passed to OpImageSparseTexelsResident. The second member must be a vector of four components of floating-point type or integer type. Its components must be the same as Sampled Type of the underlying OpTypeImage (unless that underlying Sampled Type is OpTypeVoid). It has one component per gathered texel.

Sampled Image must be an object whose type is OpTypeSampledImage. Its OpTypeImage must have a Dim of 2D, Cube, or Rect.

Coordinate must be a scalar or vector of floating-point type. It contains (u[, v] … [, array layer]) as needed by the definition of Sampled Image.

Dref is the depth-comparison reference value.

Image Operands encodes what operands follow, as per Image Operands.

Capability:
SparseResidency

6 + variable

315

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

<id>
Dref

Optional
Image Operands

Optional
<id>, <id>, …

OpImageSparseTexelsResident

Translates a Resident Code into a Boolean. Result is false if any of the texels were in uncommitted texture memory, and true otherwise.

Result Type must be a Boolean type scalar.

Resident Code is a value from an OpImageSparse… instruction that returns a resident code.

Capability:
SparseResidency

4

316

<id>
Result Type

Result <id>

<id>
Resident Code

OpImageSparseRead

Read a texel from a sparse image without a sampler.

Result Type must be an OpTypeStruct with two members. The first member’s type must be an integer type scalar. It will hold a Residency Code that can be passed to OpImageSparseTexelsResident. The second member must be a scalar or vector of floating-point type or integer type. Its component type must be the same as Sampled Type of the OpTypeImage (unless that Sampled Type is OpTypeVoid).

Image must be an object whose type is OpTypeImage with a Sampled operand of 2.

Coordinate is an integer scalar or vector containing non-normalized texel coordinates (u[, v] … [, array layer]) as needed by the definition of Image. If the coordinates are outside the image, the memory location that is accessed is undefined.

The Image Dim operand must not be SubpassData. The Image Format must not be Unknown unless the StorageImageReadWithoutFormat Capability was declared.

Image Operands encodes what operands follow, as per Image Operands.

Capability:
SparseResidency

5 + variable

320

<id>
Result Type

Result <id>

<id>
Image

<id>
Coordinate

Optional
Image Operands

Optional
<id>, <id>, …

OpImageSampleFootprintNV

TBD

Capability:
ImageFootprintNV

Reserved.

7 + variable

5283

<id>
Result Type

Result <id>

<id>
Sampled Image

<id>
Coordinate

<id>
Granularity

<id>
Coarse

Optional
Image Operands

Optional
<id>, <id>, …

3.32.11. Conversion Instructions

OpConvertFToU

Convert value numerically from floating point to unsigned integer, with round toward 0.0.

Result Type must be a scalar or vector of integer type, whose Signedness operand is 0.

Float Value must be a scalar or vector of floating-point type. It must have the same number of components as Result Type.

Results are computed per component.

4

109

<id>
Result Type

Result <id>

<id>
Float Value

OpConvertFToS

Convert value numerically from floating point to signed integer, with round toward 0.0.

Result Type must be a scalar or vector of integer type.

Float Value must be a scalar or vector of floating-point type. It must have the same number of components as Result Type.

Results are computed per component.

4

110

<id>
Result Type

Result <id>

<id>
Float Value

OpConvertSToF

Convert value numerically from signed integer to floating point.

Result Type must be a scalar or vector of floating-point type.

Signed Value must be a scalar or vector of integer type. It must have the same number of components as Result Type.

Results are computed per component.

4

111

<id>
Result Type

Result <id>

<id>
Signed Value

OpConvertUToF

Convert value numerically from unsigned integer to floating point.

Result Type must be a scalar or vector of floating-point type.

Unsigned Value must be a scalar or vector of integer type. It must have the same number of components as Result Type.

Results are computed per component.

4

112

<id>
Result Type

Result <id>

<id>
Unsigned Value

OpUConvert

Convert unsigned width. This is either a truncate or a zero extend.

Result Type must be a scalar or vector of integer type, whose Signedness operand is 0.

Unsigned Value must be a scalar or vector of integer type. It must have the same number of components as Result Type. The component width cannot equal the component width in Result Type.

Results are computed per component.

4

113

<id>
Result Type

Result <id>

<id>
Unsigned Value

OpSConvert

Convert signed width. This is either a truncate or a sign extend.

Result Type must be a scalar or vector of integer type.

Signed Value must be a scalar or vector of integer type. It must have the same number of components as Result Type. The component width cannot equal the component width in Result Type.

Results are computed per component.

4

114

<id>
Result Type

Result <id>

<id>
Signed Value

OpFConvert

Convert value numerically from one floating-point width to another width.

Result Type must be a scalar or vector of floating-point type.

Float Value must be a scalar or vector of floating-point type. It must have the same number of components as Result Type. The component width cannot equal the component width in Result Type.

Results are computed per component.

4

115

<id>
Result Type

Result <id>

<id>
Float Value

OpQuantizeToF16

Quantize a floating-point value to what is expressible by a 16-bit floating-point value.

Result Type must be a scalar or vector of floating-point type. The component width must be 32 bits.

Value is the value to quantize. The type of Value must be the same as Result Type.

If Value is an infinity, the result is the same infinity. If Value is a NaN, the result is a NaN, but not necessarily the same NaN. If Value is positive with a magnitude too large to represent as a 16-bit floating-point value, the result is positive infinity. If Value is negative with a magnitude too large to represent as a 16-bit floating-point value, the result is negative infinity. If the magnitude of Value is too small to represent as a normalized 16-bit floating-point value, the result may be either +0 or -0.

The RelaxedPrecision Decoration has no effect on this instruction.

Results are computed per component.

Capability:
Shader

4

116

<id>
Result Type

Result <id>

<id>
Value

OpConvertPtrToU

Bit pattern-preserving conversion of a pointer to an unsigned scalar integer of possibly different bit width.

Result Type must be a scalar of integer type, whose Signedness operand is 0.

Pointer must be an OpTypePointer. If the bit width of Pointer is smaller than that of Result Type, the conversion will zero extend Pointer. If the bit width of Pointer is larger than that of Result Type, the conversion will truncate Pointer. For same bit width Pointer and Result Type, this is the same as OpBitcast.

Capability:
Addresses, PhysicalStorageBufferAddressesEXT

4

117

<id>
Result Type

Result <id>

<id>
Pointer

OpSatConvertSToU

Convert a signed integer to unsigned integer. Converted values outside the representable range of Result Type are clamped to the nearest representable value of Result Type.

Result Type must be a scalar or vector of integer type.

Signed Value must be a scalar or vector of integer type. It must have the same number of components as Result Type.

Results are computed per component.

Capability:
Kernel

4

118

<id>
Result Type

Result <id>

<id>
Signed Value

OpSatConvertUToS

Convert an unsigned integer to signed integer. Converted values outside the representable range of Result Type are clamped to the nearest representable value of Result Type.

Result Type must be a scalar or vector of integer type.

Unsigned Value must be a scalar or vector of integer type. It must have the same number of components as Result Type.

Results are computed per component.

Capability:
Kernel

4

119

<id>
Result Type

Result <id>

<id>
Unsigned Value

OpConvertUToPtr

Bit pattern-preserving conversion of an unsigned scalar integer to a pointer.

Result Type must be an OpTypePointer.

Integer Value must be a scalar of integer type, whose Signedness operand is 0. If the bit width of Integer Value is smaller than that of Result Type, the conversion will zero extend Integer Value. If the bit width of Integer Value is larger than that of Result Type, the conversion will truncate Integer Value. For same-width Integer Value and Result Type, this is the same as OpBitcast.

Capability:
Addresses, PhysicalStorageBufferAddressesEXT

4

120

<id>
Result Type

Result <id>

<id>
Integer Value

OpPtrCastToGeneric

Convert a pointer’s Storage Class to Generic.

Result Type must be an OpTypePointer. Its Storage Class must be Generic.

Pointer must point to the Workgroup, CrossWorkgroup, or Function Storage Class.

Result Type and Pointer must point to the same type.

Capability:
Kernel

4

121

<id>
Result Type

Result <id>

<id>
Pointer

OpGenericCastToPtr

Convert a pointer’s Storage Class to a non-Generic class.

Result Type must be an OpTypePointer. Its Storage Class must be Workgroup, CrossWorkgroup, or Function.

Pointer must point to the Generic Storage Class.

Result Type and Pointer must point to the same type.

Capability:
Kernel

4

122

<id>
Result Type

Result <id>

<id>
Pointer

OpGenericCastToPtrExplicit

Attempts to explicitly convert Pointer to Storage storage-class pointer value.

Result Type must be an OpTypePointer. Its Storage Class must be Storage.

Pointer must have a type of OpTypePointer whose Type is the same as the Type of Result Type.Pointer must point to the Generic Storage Class. If the cast fails, the instruction result is an OpConstantNull pointer in the Storage Storage Class.

Storage must be one of the following literal values from Storage Class: Workgroup, CrossWorkgroup, or Function.

Capability:
Kernel

5

123

<id>
Result Type

Result <id>

<id>
Pointer

Storage Class
Storage

OpBitcast

Bit pattern-preserving type conversion.

Result Type must be an OpTypePointer, or a scalar or vector of numerical-type.

Operand must have a type of OpTypePointer, or a scalar or vector of numerical-type. It must be a different type than Result Type.

If Result Type is a pointer, Operand must be a pointer or integer scalar. If Operand is a pointer, Result Type must be a pointer or integer scalar.

If Result Type has the same number of components as Operand, they must also have the same component width, and results are computed per component.

If Result Type has a different number of components than Operand, the total number of bits in Result Type must equal the total number of bits in Operand. Let L be the type, either Result Type or Operand’s type, that has the larger number of components. Let S be the other type, with the smaller number of components. The number of components in L must be an integer multiple of the number of components in S. The first component (that is, the only or lowest-numbered component) of S maps to the first components of L, and so on, up to the last component of S mapping to the last components of L. Within this mapping, any single component of S (mapping to multiple components of L) maps its lower-ordered bits to the lower-numbered components of L.

4

124

<id>
Result Type

Result <id>

<id>
Operand

3.32.12. Composite Instructions

OpVectorExtractDynamic

Extract a single, dynamically selected, component of a vector.

Result Type must be a scalar type.

Vector must have a type OpTypeVector whose Component Type is Result Type.

Index must be a scalar integer 0-based index of which component of Vector to extract.

The value read is undefined if Index’s value is less than zero or greater than or equal to the number of components in Vector.

5

77

<id>
Result Type

Result <id>

<id>
Vector

<id>
Index

OpVectorInsertDynamic

Make a copy of a vector, with a single, variably selected, component modified.

Result Type must be an OpTypeVector.

Vector must have the same type as Result Type and is the vector that the non-written components will be copied from.

Component is the value that will be supplied for the component selected by Index. It must have the same type as the type of components in Result Type.

Index must be a scalar integer 0-based index of which component to modify.

What is written is undefined if Index’s value is less than zero or greater than or equal to the number of components in Vector.

6

78

<id>
Result Type

Result <id>

<id>
Vector

<id>
Component

<id>
Index

OpVectorShuffle

Select arbitrary components from two vectors to make a new vector.

Result Type must be an OpTypeVector. The number of components in Result Type must be the same as the number of Component operands.

Vector 1 and Vector 2 must both have vector types, with the same Component Type as Result Type. They do not have to have the same number of components as Result Type or with each other. They are logically concatenated, forming a single vector with Vector 1’s components appearing before Vector 2’s. The components of this logical vector are logically numbered with a single consecutive set of numbers from 0 to N - 1, where N is the total number of components.

Components are these logical numbers (see above), selecting which of the logically numbered components form the result. They can select the components in any order and can repeat components. The first component of the result is selected by the first Component operand, the second component of the result is selected by the second Component operand, etc. A Component literal may also be FFFFFFFF, which means the corresponding result component has no source and is undefined. All Component literals must either be FFFFFFFF or in [0, N - 1] (inclusive).

Note: A vector “swizzle” can be done by using the vector for both Vector operands, or using an OpUndef for one of the Vector operands.

5 + variable

79

<id>
Result Type

Result <id>

<id>
Vector 1

<id>
Vector 2

Literal, Literal, …
Components

OpCompositeConstruct

Construct a new composite object from a set of constituent objects that will fully form it.

Result Type must be a composite type, whose top-level members/elements/components/columns have the same type as the types of the operands, with one exception. The exception is that for constructing a vector, the operands may also be vectors with the same component type as the Result Type component type. When constructing a vector, the total number of components in all the operands must equal the number of components in Result Type.

Constituents will become members of a structure, or elements of an array, or components of a vector, or columns of a matrix. There must be exactly one Constituent for each top-level member/element/component/column of the result, with one exception. The exception is that for constructing a vector, a contiguous subset of the scalars consumed can be represented by a vector operand instead. The Constituents must appear in the order needed by the definition of the type of the result. When constructing a vector, there must be at least two Constituent operands.

3 + variable

80

<id>
Result Type

Result <id>

<id>, <id>, …
Constituents

OpCompositeExtract

Extract a part of a composite object.

Result Type must be the type of object selected by the last provided index. The instruction result is the extracted object.

Composite is the composite to extract from.

Indexes walk the type hierarchy, potentially down to component granularity, to select the part to extract. All indexes must be in bounds. All composite constituents use zero-based numbering, as described by their OpType… instruction.

4 + variable

81

<id>
Result Type

Result <id>

<id>
Composite

Literal, Literal, …
Indexes

OpCompositeInsert

Make a copy of a composite object, while modifying one part of it.

Result Type must be the same type as Composite.

Object is the object to use as the modified part.

Composite is the composite to copy all but the modified part from.

Indexes walk the type hierarchy of Composite to the desired depth, potentially down to component granularity, to select the part to modify. All indexes must be in bounds. All composite constituents use zero-based numbering, as described by their OpType… instruction. The type of the part selected to modify must match the type of Object.

5 + variable

82

<id>
Result Type

Result <id>

<id>
Object

<id>
Composite

Literal, Literal, …
Indexes

OpCopyObject

Make a copy of Operand. There are no pointer dereferences involved.

Result Type must equal Operand type. There are no other restrictions on the types.

4

83

<id>
Result Type

Result <id>

<id>
Operand

OpTranspose

Transpose a matrix.

Result Type must be an OpTypeMatrix.

Matrix must be an object of type OpTypeMatrix. The number of columns and the column size of Matrix must be the reverse of those in Result Type. The types of the scalar components in Matrix and Result Type must be the same.

Matrix must have of type of OpTypeMatrix.

Capability:
Matrix

4

84

<id>
Result Type

Result <id>

<id>
Matrix

OpCopyLogical

Make a logical copy of Operand. There are no pointer dereferences involved.

Result Type must not equal the type of Operand (see OpCopyObject), but Result Type must logically match the Operand type.

Logically match is recursively defined by these three rules:
1. They must be either both be OpTypeArray or both be OpTypeStruct
2. If they are OpTypeArray:
- they must have the same Length operand, and
- their Element Type operands must be either the same or must logically match.
3. If they are OpTypeStruct:
- they must have the same number of Member type, and
- Member N type for the same N in the two types must be either the same or must logically match.

Missing before version 1.4.

4

400

<id>
Result Type

Result <id>

<id>
Operand

3.32.13. Arithmetic Instructions

OpSNegate

Signed-integer subtract of Operand from zero.

Result Type must be a scalar or vector of integer type.

Operand’s type must be a scalar or vector of integer type. It must have the same number of components as Result Type. The component width must equal the component width in Result Type.

Results are computed per component.

4

126

<id>
Result Type

Result <id>

<id>
Operand

OpFNegate

Floating-point subtract of Operand from zero.

Result Type must be a scalar or vector of floating-point type.

The type of Operand must be the same as Result Type.

Results are computed per component.

4

127

<id>
Result Type

Result <id>

<id>
Operand

OpIAdd

Integer addition of Operand 1 and Operand 2.

Result Type must be a scalar or vector of integer type.

The type of Operand 1 and Operand 2 must be a scalar or vector of integer type. They must have the same number of components as Result Type. They must have the same component width as Result Type.

The resulting value will equal the low-order N bits of the correct result R, where N is the component width and R is computed with enough precision to avoid overflow and underflow.

Results are computed per component.

5

128

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpFAdd

Floating-point addition of Operand 1 and Operand 2.

Result Type must be a scalar or vector of floating-point type.

The types of Operand 1 and Operand 2 both must be the same as Result Type.

Results are computed per component.

5

129

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpISub

Integer subtraction of Operand 2 from Operand 1.

Result Type must be a scalar or vector of integer type.

The type of Operand 1 and Operand 2 must be a scalar or vector of integer type. They must have the same number of components as Result Type. They must have the same component width as Result Type.

The resulting value will equal the low-order N bits of the correct result R, where N is the component width and R is computed with enough precision to avoid overflow and underflow.

Results are computed per component.

5

130

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpFSub

Floating-point subtraction of Operand 2 from Operand 1.

Result Type must be a scalar or vector of floating-point type.

The types of Operand 1 and Operand 2 both must be the same as Result Type.

Results are computed per component.

5

131

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpIMul

Integer multiplication of Operand 1 and Operand 2.

Result Type must be a scalar or vector of integer type.

The type of Operand 1 and Operand 2 must be a scalar or vector of integer type. They must have the same number of components as Result Type. They must have the same component width as Result Type.

The resulting value will equal the low-order N bits of the correct result R, where N is the component width and R is computed with enough precision to avoid overflow and underflow.

Results are computed per component.

5

132

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpFMul

Floating-point multiplication of Operand 1 and Operand 2.

Result Type must be a scalar or vector of floating-point type.

The types of Operand 1 and Operand 2 both must be the same as Result Type.

Results are computed per component.

5

133

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpUDiv

Unsigned-integer division of Operand 1 divided by Operand 2.

Result Type must be a scalar or vector of integer type, whose Signedness operand is 0.

The types of Operand 1 and Operand 2 both must be the same as Result Type.

Results are computed per component. The resulting value is undefined if Operand 2 is 0.

5

134

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpSDiv

Signed-integer division of Operand 1 divided by Operand 2.

Result Type must be a scalar or vector of integer type.

The type of Operand 1 and Operand 2 must be a scalar or vector of integer type. They must have the same number of components as Result Type. They must have the same component width as Result Type.

Results are computed per component. The resulting value is undefined if Operand 2 is 0.

5

135

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpFDiv

Floating-point division of Operand 1 divided by Operand 2.

Result Type must be a scalar or vector of floating-point type.

The types of Operand 1 and Operand 2 both must be the same as Result Type.

Results are computed per component. The resulting value is undefined if Operand 2 is 0.

5

136

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpUMod

Unsigned modulo operation of Operand 1 modulo Operand 2.

Result Type must be a scalar or vector of integer type, whose Signedness operand is 0.

The types of Operand 1 and Operand 2 both must be the same as Result Type.

Results are computed per component. The resulting value is undefined if Operand 2 is 0.

5

137

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpSRem

Signed remainder operation for the remainder whose sign matches the sign of Operand 1.

Result Type must be a scalar or vector of integer type.

The type of Operand 1 and Operand 2 must be a scalar or vector of integer type. They must have the same number of components as Result Type. They must have the same component width as Result Type.

Results are computed per component. The resulting value is undefined if Operand 2 is 0. Otherwise, the result is the remainder r of Operand 1 divided by Operand 2 where if r ≠ 0, the sign of r is the same as the sign of Operand 1.

5

138

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpSMod

Signed remainder operation for the remainder whose sign matches the sign of Operand 2.

Result Type must be a scalar or vector of integer type.

The type of Operand 1 and Operand 2 must be a scalar or vector of integer type. They must have the same number of components as Result Type. They must have the same component width as Result Type.

Results are computed per component. The resulting value is undefined if Operand 2 is 0. Otherwise, the result is the remainder r of Operand 1 divided by Operand 2 where if r ≠ 0, the sign of r is the same as the sign of Operand 2.

5

139

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpFRem

The floating-point remainder whose sign matches the sign of Operand 1.

Result Type must be a scalar or vector of floating-point type.

The types of Operand 1 and Operand 2 both must be the same as Result Type.

Results are computed per component. The resulting value is undefined if Operand 2 is 0. Otherwise, the result is the remainder r of Operand 1 divided by Operand 2 where if r ≠ 0, the sign of r is the same as the sign of Operand 1.

5

140

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpFMod

The floating-point remainder whose sign matches the sign of Operand 2.

Result Type must be a scalar or vector of floating-point type.

The types of Operand 1 and Operand 2 both must be the same as Result Type.

Results are computed per component. The resulting value is undefined if Operand 2 is 0. Otherwise, the result is the remainder r of Operand 1 divided by Operand 2 where if r ≠ 0, the sign of r is the same as the sign of Operand 2.

5

141

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpVectorTimesScalar

Scale a floating-point vector.

Result Type must be a vector of floating-point type.

The type of Vector must be the same as Result Type. Each component of Vector is multiplied by Scalar.

Scalar must have the same type as the Component Type in Result Type.

5

142

<id>
Result Type

Result <id>

<id>
Vector

<id>
Scalar

OpMatrixTimesScalar

Scale a floating-point matrix.

Result Type must be an OpTypeMatrix whose Column Type is a vector of floating-point type.

The type of Matrix must be the same as Result Type. Each component in each column in Matrix is multiplied by Scalar.

Scalar must have the same type as the Component Type in Result Type.

Capability:
Matrix

5

143

<id>
Result Type

Result <id>

<id>
Matrix

<id>
Scalar

OpVectorTimesMatrix

Linear-algebraic Vector X Matrix.

Result Type must be a vector of floating-point type.

Vector must be a vector with the same Component Type as the Component Type in Result Type. Its number of components must equal the number of components in each column in Matrix.

Matrix must be a matrix with the same Component Type as the Component Type in Result Type. Its number of columns must equal the number of components in Result Type.

Capability:
Matrix

5

144

<id>
Result Type

Result <id>

<id>
Vector

<id>
Matrix

OpMatrixTimesVector

Linear-algebraic Matrix X Vector.

Result Type must be a vector of floating-point type.

Matrix must be an OpTypeMatrix whose Column Type is Result Type.

Vector must be a vector with the same Component Type as the Component Type in Result Type. Its number of components must equal the number of columns in Matrix.

Capability:
Matrix

5

145

<id>
Result Type

Result <id>

<id>
Matrix

<id>
Vector

OpMatrixTimesMatrix

Linear-algebraic multiply of LeftMatrix X RightMatrix.

Result Type must be an OpTypeMatrix whose Column Type is a vector of floating-point type.

LeftMatrix must be a matrix whose Column Type is the same as the Column Type in Result Type.

RightMatrix must be a matrix with the same Component Type as the Component Type in Result Type. Its number of columns must equal the number of columns in Result Type. Its columns must have the same number of components as the number of columns in LeftMatrix.

Capability:
Matrix

5

146

<id>
Result Type

Result <id>

<id>
LeftMatrix

<id>
RightMatrix

OpOuterProduct

Linear-algebraic outer product of Vector 1 and Vector 2.

Result Type must be an OpTypeMatrix whose Column Type is a vector of floating-point type.

Vector 1 must have the same type as the Column Type in Result Type.

Vector 2 must be a vector with the same Component Type as the Component Type in Result Type. Its number of components must equal the number of columns in Result Type.

Capability:
Matrix

5

147

<id>
Result Type

Result <id>

<id>
Vector 1

<id>
Vector 2

OpDot

Dot product of Vector 1 and Vector 2.

Result Type must be a floating-point type scalar.

Vector 1 and Vector 2 must be vectors of the same type, and their component type must be Result Type.

5

148

<id>
Result Type

Result <id>

<id>
Vector 1

<id>
Vector 2

OpIAddCarry

Result is the unsigned integer addition of Operand 1 and Operand 2, including its carry.

Result Type must be from OpTypeStruct. The struct must have two members, and the two members must be the same type. The member type must be a scalar or vector of integer type, whose Signedness operand is 0.

Operand 1 and Operand 2 must have the same type as the members of Result Type. These are consumed as unsigned integers.

Results are computed per component.

Member 0 of the result gets the low-order bits (full component width) of the addition.

Member 1 of the result gets the high-order (carry) bit of the result of the addition. That is, it gets the value 1 if the addition overflowed the component width, and 0 otherwise.

5

149

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpISubBorrow

Result is the unsigned integer subtraction of Operand 2 from Operand 1, and what it needed to borrow.

Result Type must be from OpTypeStruct. The struct must have two members, and the two members must be the same type. The member type must be a scalar or vector of integer type, whose Signedness operand is 0.

Operand 1 and Operand 2 must have the same type as the members of Result Type. These are consumed as unsigned integers.

Results are computed per component.

Member 0 of the result gets the low-order bits (full component width) of the subtraction. That is, if Operand 1 is larger than Operand 2, member 0 gets the full value of the subtraction; if Operand 2 is larger than Operand 1, member 0 gets 2w + Operand 1 - Operand 2, where w is the component width.

Member 1 of the result gets 0 if Operand 1 ≥ Operand 2, and gets 1 otherwise.

5

150

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpUMulExtended

Result is the full value of the unsigned integer multiplication of Operand 1 and Operand 2.

Result Type must be from OpTypeStruct. The struct must have two members, and the two members must be the same type. The member type must be a scalar or vector of integer type, whose Signedness operand is 0.

Operand 1 and Operand 2 must have the same type as the members of Result Type. These are consumed as unsigned integers.

Results are computed per component.

Member 0 of the result gets the low-order bits of the multiplication.

Member 1 of the result gets the high-order bits of the multiplication.

5

151

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpSMulExtended

Result is the full value of the signed integer multiplication of Operand 1 and Operand 2.

Result Type must be from OpTypeStruct. The struct must have two members, and the two members must be the same type. The member type must be a scalar or vector of integer type.

Operand 1 and Operand 2 must have the same type as the members of Result Type. These are consumed as signed integers.

Results are computed per component.

Member 0 of the result gets the low-order bits of the multiplication.

Member 1 of the result gets the high-order bits of the multiplication.

5

152

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

3.32.14. Bit Instructions

OpShiftRightLogical

Shift the bits in Base right by the number of bits specified in Shift. The most-significant bits will be zero filled.

Result Type must be a scalar or vector of integer type.

The type of each Base and Shift must be a scalar or vector of integer type. Base and Shift must have the same number of components. The number of components and bit width of the type of Base must be the same as in Result Type.

Shift is consumed as an unsigned integer. The result is undefined if Shift is greater than or equal to the bit width of the components of Base.

Results are computed per component.

5

194

<id>
Result Type

Result <id>

<id>
Base

<id>
Shift

OpShiftRightArithmetic

Shift the bits in Base right by the number of bits specified in Shift. The most-significant bits will be filled with the sign bit from Base.

Result Type must be a scalar or vector of integer type.

The type of each Base and Shift must be a scalar or vector of integer type. Base and Shift must have the same number of components. The number of components and bit width of the type of Base must be the same as in Result Type.

Shift is treated as unsigned. The result is undefined if Shift is greater than or equal to the bit width of the components of Base.

Results are computed per component.

5

195

<id>
Result Type

Result <id>

<id>
Base

<id>
Shift

OpShiftLeftLogical

Shift the bits in Base left by the number of bits specified in Shift. The least-significant bits will be zero filled.

Result Type must be a scalar or vector of integer type.

The type of each Base and Shift must be a scalar or vector of integer type. Base and Shift must have the same number of components. The number of components and bit width of the type of Base must be the same as in Result Type.

Shift is treated as unsigned. The result is undefined if Shift is greater than or equal to the bit width of the components of Base.

The number of components and bit width of Result Type must match those Base type. All types must be integer types.

Results are computed per component.

5

196

<id>
Result Type

Result <id>

<id>
Base

<id>
Shift

OpBitwiseOr

Result is 1 if either Operand 1 or Operand 2 is 1. Result is 0 if both Operand 1 and Operand 2 are 0.

Results are computed per component, and within each component, per bit.

Result Type must be a scalar or vector of integer type. The type of Operand 1 and Operand 2 must be a scalar or vector of integer type. They must have the same number of components as Result Type. They must have the same component width as Result Type.

5

197

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpBitwiseXor

Result is 1 if exactly one of Operand 1 or Operand 2 is 1. Result is 0 if Operand 1 and Operand 2 have the same value.

Results are computed per component, and within each component, per bit.

Result Type must be a scalar or vector of integer type. The type of Operand 1 and Operand 2 must be a scalar or vector of integer type. They must have the same number of components as Result Type. They must have the same component width as Result Type.

5

198

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpBitwiseAnd

Result is 1 if both Operand 1 and Operand 2 are 1. Result is 0 if either Operand 1 or Operand 2 are 0.

Results are computed per component, and within each component, per bit.

Result Type must be a scalar or vector of integer type. The type of Operand 1 and Operand 2 must be a scalar or vector of integer type. They must have the same number of components as Result Type. They must have the same component width as Result Type.

5

199

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpNot

Complement the bits of Operand.

Results are computed per component, and within each component, per bit.

Result Type must be a scalar or vector of integer type.

Operand’s type must be a scalar or vector of integer type. It must have the same number of components as Result Type. The component width must equal the component width in Result Type.

4

200

<id>
Result Type

Result <id>

<id>
Operand

OpBitFieldInsert

Make a copy of an object, with a modified bit field that comes from another object.

Results are computed per component.

Result Type must be a scalar or vector of integer type.

The type of Base and Insert must be the same as Result Type.

Any result bits numbered outside [Offset, Offset + Count - 1] (inclusive) will come from the corresponding bits in Base.

Any result bits numbered in [Offset, Offset + Count - 1] come, in order, from the bits numbered [0, Count - 1] of Insert.

Count must be an integer type scalar. Count is the number of bits taken from Insert. It will be consumed as an unsigned value. Count can be 0, in which case the result will be Base.

Offset must be an integer type scalar. Offset is the lowest-order bit of the bit field. It will be consumed as an unsigned value.

The resulting value is undefined if Count or Offset or their sum is greater than the number of bits in the result.

Capability:
Shader

7

201

<id>
Result Type

Result <id>

<id>
Base

<id>
Insert

<id>
Offset

<id>
Count

OpBitFieldSExtract

Extract a bit field from an object, with sign extension.

Results are computed per component.

Result Type must be a scalar or vector of integer type.

The type of Base must be the same as Result Type.

If Count is greater than 0: The bits of Base numbered in [Offset, Offset + Count - 1] (inclusive) become the bits numbered [0, Count - 1] of the result. The remaining bits of the result will all be the same as bit Offset + Count - 1 of Base.

Count must be an integer type scalar. Count is the number of bits extracted from Base. It will be consumed as an unsigned value. Count can be 0, in which case the result will be 0.

Offset must be an integer type scalar. Offset is the lowest-order bit of the bit field to extract from Base. It will be consumed as an unsigned value.

The resulting value is undefined if Count or Offset or their sum is greater than the number of bits in the result.

Capability:
Shader

6

202

<id>
Result Type

Result <id>

<id>
Base

<id>
Offset

<id>
Count

OpBitFieldUExtract

Extract a bit field from an object, without sign extension.

The semantics are the same as with OpBitFieldSExtract with the exception that there is no sign extension. The remaining bits of the result will all be 0.

Capability:
Shader

6

203

<id>
Result Type

Result <id>

<id>
Base

<id>
Offset

<id>
Count

OpBitReverse

Reverse the bits in an object.

Results are computed per component.

Result Type must be a scalar or vector of integer type.

The type of Base must be the same as Result Type.

The bit-number n of the result will be taken from bit-number Width - 1 - n of Base, where Width is the OpTypeInt operand of the Result Type.

Capability:
Shader

4

204

<id>
Result Type

Result <id>

<id>
Base

OpBitCount

Count the number of set bits in an object.

Results are computed per component.

Result Type must be a scalar or vector of integer type. The components must be wide enough to hold the unsigned Width of Base as an unsigned value. That is, no sign bit is needed or counted when checking for a wide enough result width.

Base must be a scalar or vector of integer type. It must have the same number of components as Result Type.

The result is the unsigned value that is the number of bits in Base that are 1.

4

205

<id>
Result Type

Result <id>

<id>
Base

3.32.15. Relational and Logical Instructions

OpAny

Result is true if any component of Vector is true, otherwise result is false.

Result Type must be a Boolean type scalar.

Vector must be a vector of Boolean type.

4

154

<id>
Result Type

Result <id>

<id>
Vector

OpAll

Result is true if all components of Vector are true, otherwise result is false.

Result Type must be a Boolean type scalar.

Vector must be a vector of Boolean type.

4

155

<id>
Result Type

Result <id>

<id>
Vector

OpIsNan

Result is true if x is an IEEE NaN, otherwise result is false.

Result Type must be a scalar or vector of Boolean type.

x must be a scalar or vector of floating-point type. It must have the same number of components as Result Type.

Results are computed per component.

4

156

<id>
Result Type

Result <id>

<id>
x

OpIsInf

Result is true if x is an IEEE Inf, otherwise result is false

Result Type must be a scalar or vector of Boolean type.

x must be a scalar or vector of floating-point type. It must have the same number of components as Result Type.

Results are computed per component.

4

157

<id>
Result Type

Result <id>

<id>
x

OpIsFinite

Result is true if x is an IEEE finite number, otherwise result is false.

Result Type must be a scalar or vector of Boolean type.

x must be a scalar or vector of floating-point type. It must have the same number of components as Result Type.

Results are computed per component.

Capability:
Kernel

4

158

<id>
Result Type

Result <id>

<id>
x

OpIsNormal

Result is true if x is an IEEE normal number, otherwise result is false.

Result Type must be a scalar or vector of Boolean type.

x must be a scalar or vector of floating-point type. It must have the same number of components as Result Type.

Results are computed per component.

Capability:
Kernel

4

159

<id>
Result Type

Result <id>

<id>
x

OpSignBitSet

Result is true if x has its sign bit set, otherwise result is false.

Result Type must be a scalar or vector of Boolean type.

x must be a scalar or vector of floating-point type. It must have the same number of components as Result Type.

Results are computed per component.

Capability:
Kernel

4

160

<id>
Result Type

Result <id>

<id>
x

OpLessOrGreater

Result is true if x < y or x > y, where IEEE comparisons are used, otherwise result is false.

Result Type must be a scalar or vector of Boolean type.

x must be a scalar or vector of floating-point type. It must have the same number of components as Result Type.

y must have the same type as x.

Results are computed per component.

Capability:
Kernel

5

161

<id>
Result Type

Result <id>

<id>
x

<id>
y

OpOrdered

Result is true if both x == x and y == y are true, where IEEE comparison is used, otherwise result is false.

Result Type must be a scalar or vector of Boolean type.

x must be a scalar or vector of floating-point type. It must have the same number of components as Result Type.

y must have the same type as x.

Results are computed per component.

Capability:
Kernel

5

162

<id>
Result Type

Result <id>

<id>
x

<id>
y

OpUnordered

Result is true if either x or y is an IEEE NaN, otherwise result is false.

Result Type must be a scalar or vector of Boolean type.

x must be a scalar or vector of floating-point type. It must have the same number of components as Result Type.

y must have the same type as x.

Results are computed per component.

Capability:
Kernel

5

163

<id>
Result Type

Result <id>

<id>
x

<id>
y

OpLogicalEqual

Result is true if Operand 1 and Operand 2 have the same value. Result is false if Operand 1 and Operand 2 have different values.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 must be the same as Result Type.

The type of Operand 2 must be the same as Result Type.

Results are computed per component.

5

164

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpLogicalNotEqual

Result is true if Operand 1 and Operand 2 have different values. Result is false if Operand 1 and Operand 2 have the same value.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 must be the same as Result Type.

The type of Operand 2 must be the same as Result Type.

Results are computed per component.

5

165

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpLogicalOr

Result is true if either Operand 1 or Operand 2 is true. Result is false if both Operand 1 and Operand 2 are false.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 must be the same as Result Type.

The type of Operand 2 must be the same as Result Type.

Results are computed per component.

5

166

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpLogicalAnd

Result is true if both Operand 1 and Operand 2 are true. Result is false if either Operand 1 or Operand 2 are false.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 must be the same as Result Type.

The type of Operand 2 must be the same as Result Type.

Results are computed per component.

5

167

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpLogicalNot

Result is true if Operand is false. Result is false if Operand is true.

Result Type must be a scalar or vector of Boolean type.

The type of Operand must be the same as Result Type.

Results are computed per component.

4

168

<id>
Result Type

Result <id>

<id>
Operand

OpSelect

Select between two objects. Before version 1.4, results are only computed per component.

Before version 1.4, Result Type must be a pointer, scalar, or vector. Starting with version 1.4, Result Type can additionally be a composite type other than a vector.

The types of Object 1 and Object 2 must be the same as Result Type.

Condition must be a scalar or vector of Boolean type.

If Condition is a scalar and true, the result is Object 1. If Condition is a scalar and false, the result is Object 2.

If Condition is a vector, Result Type must be a vector with the same number of components as Condition and the result is a mix of Object 1 and Object 2: When a component of Condition is true, the corresponding component in the result is taken from Object 1, otherwise it is taken from Object 2.

6

169

<id>
Result Type

Result <id>

<id>
Condition

<id>
Object 1

<id>
Object 2

OpIEqual

Integer comparison for equality.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of integer type. They must have the same component width, and they must have the same number of components as Result Type.

Results are computed per component.

5

170

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpINotEqual

Integer comparison for inequality.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of integer type. They must have the same component width, and they must have the same number of components as Result Type.

Results are computed per component.

5

171

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpUGreaterThan

Unsigned-integer comparison if Operand 1 is greater than Operand 2.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of integer type. They must have the same component width, and they must have the same number of components as Result Type.

Results are computed per component.

5

172

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpSGreaterThan

Signed-integer comparison if Operand 1 is greater than Operand 2.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of integer type. They must have the same component width, and they must have the same number of components as Result Type.

Results are computed per component.

5

173

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpUGreaterThanEqual

Unsigned-integer comparison if Operand 1 is greater than or equal to Operand 2.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of integer type. They must have the same component width, and they must have the same number of components as Result Type.

Results are computed per component.

5

174

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpSGreaterThanEqual

Signed-integer comparison if Operand 1 is greater than or equal to Operand 2.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of integer type. They must have the same component width, and they must have the same number of components as Result Type.

Results are computed per component.

5

175

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpULessThan

Unsigned-integer comparison if Operand 1 is less than Operand 2.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of integer type. They must have the same component width, and they must have the same number of components as Result Type.

Results are computed per component.

5

176

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpSLessThan

Signed-integer comparison if Operand 1 is less than Operand 2.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of integer type. They must have the same component width, and they must have the same number of components as Result Type.

Results are computed per component.

5

177

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpULessThanEqual

Unsigned-integer comparison if Operand 1 is less than or equal to Operand 2.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of integer type. They must have the same component width, and they must have the same number of components as Result Type.

Results are computed per component.

5

178

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpSLessThanEqual

Signed-integer comparison if Operand 1 is less than or equal to Operand 2.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of integer type. They must have the same component width, and they must have the same number of components as Result Type.

Results are computed per component.

5

179

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpFOrdEqual

Floating-point comparison for being ordered and equal.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of floating-point type. They must have the same type, and they must have the same number of components as Result Type.

Results are computed per component.

5

180

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpFUnordEqual

Floating-point comparison for being unordered or equal.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of floating-point type. They must have the same type, and they must have the same number of components as Result Type.

Results are computed per component.

5

181

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpFOrdNotEqual

Floating-point comparison for being ordered and not equal.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of floating-point type. They must have the same type, and they must have the same number of components as Result Type.

Results are computed per component.

5

182

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpFUnordNotEqual

Floating-point comparison for being unordered or not equal.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of floating-point type. They must have the same type, and they must have the same number of components as Result Type.

Results are computed per component.

5

183

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpFOrdLessThan

Floating-point comparison if operands are ordered and Operand 1 is less than Operand 2.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of floating-point type. They must have the same type, and they must have the same number of components as Result Type.

Results are computed per component.

5

184

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpFUnordLessThan

Floating-point comparison if operands are unordered or Operand 1 is less than Operand 2.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of floating-point type. They must have the same type, and they must have the same number of components as Result Type.

Results are computed per component.

5

185

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpFOrdGreaterThan

Floating-point comparison if operands are ordered and Operand 1 is greater than Operand 2.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of floating-point type. They must have the same type, and they must have the same number of components as Result Type.

Results are computed per component.

5

186

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpFUnordGreaterThan

Floating-point comparison if operands are unordered or Operand 1 is greater than Operand 2.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of floating-point type. They must have the same type, and they must have the same number of components as Result Type.

Results are computed per component.

5

187

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpFOrdLessThanEqual

Floating-point comparison if operands are ordered and Operand 1 is less than or equal to Operand 2.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of floating-point type. They must have the same type, and they must have the same number of components as Result Type.

Results are computed per component.

5

188

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpFUnordLessThanEqual

Floating-point comparison if operands are unordered or Operand 1 is less than or equal to Operand 2.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of floating-point type. They must have the same type, and they must have the same number of components as Result Type.

Results are computed per component.

5

189

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpFOrdGreaterThanEqual

Floating-point comparison if operands are ordered and Operand 1 is greater than or equal to Operand 2.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of floating-point type. They must have the same type, and they must have the same number of components as Result Type.

Results are computed per component.

5

190

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

OpFUnordGreaterThanEqual

Floating-point comparison if operands are unordered or Operand 1 is greater than or equal to Operand 2.

Result Type must be a scalar or vector of Boolean type.

The type of Operand 1 and Operand 2 must be a scalar or vector of floating-point type. They must have the same type, and they must have the same number of components as Result Type.

Results are computed per component.

5

191

<id>
Result Type

Result <id>

<id>
Operand 1

<id>
Operand 2

3.32.16. Derivative Instructions

OpDPdx

Same result as either OpDPdxFine or OpDPdxCoarse on P. Selection of which one is based on external factors.

Result Type must be a scalar or vector of floating-point type. The component width must be 32 bits.

The type of P must be the same as Result Type. P is the value to take the derivative of.

This instruction is only valid in the Fragment Execution Model.

Capability:
Shader

4

207

<id>
Result Type

Result <id>

<id>
P

OpDPdy

Same result as either OpDPdyFine or OpDPdyCoarse on P. Selection of which one is based on external factors.

Result Type must be a scalar or vector of floating-point type. The component width must be 32 bits.

The type of P must be the same as Result Type. P is the value to take the derivative of.

This instruction is only valid in the Fragment Execution Model.

Capability:
Shader

4

208

<id>
Result Type

Result <id>

<id>
P

OpFwidth

Result is the same as computing the sum of the absolute values of OpDPdx and OpDPdy on P.

Result Type must be a scalar or vector of floating-point type. The component width must be 32 bits.

The type of P must be the same as Result Type. P is the value to take the derivative of.

This instruction is only valid in the Fragment Execution Model.

Capability:
Shader

4

209

<id>
Result Type

Result <id>

<id>
P

OpDPdxFine

Result is the partial derivative of P with respect to the window x coordinate.Will use local differencing based on the value of P for the current fragment and its immediate neighbor(s).

Result Type must be a scalar or vector of floating-point type. The component width must be 32 bits.

The type of P must be the same as Result Type. P is the value to take the derivative of.

This instruction is only valid in the Fragment Execution Model.

Capability:
DerivativeControl

4

210

<id>
Result Type

Result <id>

<id>
P

OpDPdyFine

Result is the partial derivative of P with respect to the window y coordinate.Will use local differencing based on the value of P for the current fragment and its immediate neighbor(s).

Result Type must be a scalar or vector of floating-point type. The component width must be 32 bits.

The type of P must be the same as Result Type. P is the value to take the derivative of.

This instruction is only valid in the Fragment Execution Model.

Capability:
DerivativeControl

4

211

<id>
Result Type

Result <id>

<id>
P

OpFwidthFine

Result is the same as computing the sum of the absolute values of OpDPdxFine and OpDPdyFine on P.

Result Type must be a scalar or vector of floating-point type. The component width must be 32 bits.

The type of P must be the same as Result Type. P is the value to take the derivative of.

This instruction is only valid in the Fragment Execution Model.

Capability:
DerivativeControl

4

212

<id>
Result Type

Result <id>

<id>
P

OpDPdxCoarse

Result is the partial derivative of P with respect to the window x coordinate. Will use local differencing based on the value of P for the current fragment’s neighbors, and will possibly, but not necessarily, include the value of P for the current fragment. That is, over a given area, the implementation can compute x derivatives in fewer unique locations than would be allowed for OpDPdxFine.

Result Type must be a scalar or vector of floating-point type. The component width must be 32 bits.

The type of P must be the same as Result Type. P is the value to take the derivative of.

This instruction is only valid in the Fragment Execution Model.

Capability:
DerivativeControl

4

213

<id>
Result Type

Result <id>

<id>
P

OpDPdyCoarse

Result is the partial derivative of P with respect to the window y coordinate. Will use local differencing based on the value of P for the current fragment’s neighbors, and will possibly, but not necessarily, include the value of P for the current fragment. That is, over a given area, the implementation can compute y derivatives in fewer unique locations than would be allowed for OpDPdyFine.

Result Type must be a scalar or vector of floating-point type. The component width must be 32 bits.

The type of P must be the same as Result Type. P is the value to take the derivative of.

This instruction is only valid in the Fragment Execution Model.

Capability:
DerivativeControl

4

214

<id>
Result Type

Result <id>

<id>
P

OpFwidthCoarse

Result is the same as computing the sum of the absolute values of OpDPdxCoarse and OpDPdyCoarse on P.

Result Type must be a scalar or vector of floating-point type. The component width must be 32 bits.

The type of P must be the same as Result Type. P is the value to take the derivative of.

This instruction is only valid in the Fragment Execution Model.

Capability:
DerivativeControl

4

215

<id>
Result Type

Result <id>

<id>
P

3.32.17. Control-Flow Instructions

OpPhi

The SSA phi function.

The result is selected based on control flow: If control reached the current block from Parent i, Result Id gets the value that Variable i had at the end of Parent i.

Result Type can be any type.

Operands are a sequence of pairs: (Variable 1, Parent 1 block), (Variable 2, Parent 2 block), … Each Parent i block is the label of an immediate predecessor in the CFG of the current block. There must be exactly one Parent i for each parent block of the current block in the CFG. If Parent i is reachable in the CFG and Variable i is defined in a block, that defining block must dominate Parent i. All Variables must have a type matching Result Type.

Within a block, this instruction must appear before all non-OpPhi instructions (except for OpLine, which can be mixed with OpPhi).

3 + variable

245

<id>
Result Type

Result <id>

<id>, <id>, …
Variable, Parent, …

OpLoopMerge

Declare a structured loop.

This instruction must immediately precede either an OpBranch or OpBranchConditional instruction. That is, it must be the second-to-last instruction in its block.

Merge Block is the label of the merge block for this structured loop.

Continue Target is the label of a block targeted for processing a loop "continue".

Loop Control Parameters appear in Loop Control-table order for any Loop Control setting that requires such a parameter.

See Structured Control Flow for more detail.

4 + variable

246

<id>
Merge Block

<id>
Continue Target

Loop Control

Literal, Literal, …
Loop Control Parameters

OpSelectionMerge

Declare a structured selection.

This instruction must immediately precede either an OpBranchConditional or OpSwitch instruction. That is, it must be the second-to-last instruction in its block.

Merge Block is the label of the merge block for this structured selection.

See Structured Control Flow for more detail.

3

247

<id>
Merge Block

Selection Control

OpLabel

The block label instruction: Any reference to a block is through the Result <id> of its label.

Must be the first instruction of any block, and appears only as the first instruction of a block.

2

248

Result <id>

OpBranch

Unconditional branch to Target Label.

Target Label must be the Result <id> of an OpLabel instruction in the current function.

This instruction must be the last instruction in a block.

2

249

<id>
Target Label

OpBranchConditional

If Condition is true, branch to True Label, otherwise branch to False Label.

Condition must be a Boolean type scalar.

True Label must be an OpLabel in the current function.

False Label must be an OpLabel in the current function.

Branch weights are unsigned 32-bit integer literals. There must be either no Branch Weights or exactly two branch weights. If present, the first is the weight for branching to True Label, and the second is the weight for branching to False Label. The implied probability that a branch is taken is its weight divided by the sum of the two Branch weights. At least one weight must be non-zero. A weight of zero does not imply a branch is dead or permit its removal; branch weights are only hints. The two weights must not overflow a 32-bit unsigned integer when added together.

This instruction must be the last instruction in a block.

4 + variable

250

<id>
Condition

<id>
True Label

<id>
False Label

Literal, Literal, …
Branch weights

OpSwitch

Multi-way branch to one of the operand label <id>.

Selector must have a type of OpTypeInt. Selector will be compared for equality to the Target literals.

Default must be the <id> of a label. If Selector does not equal any of the Target literals, control flow will branch to the Default label <id>.

Target must be alternating scalar integer literals and the <id> of a label. If Selector equals a literal, control flow will branch to the following label <id>. It is invalid for any two literal to be equal to each other. If Selector does not equal any literal, control flow will branch to the Default label <id>. Each literal is interpreted with the type of Selector: The bit width of Selector’s type will be the width of each literal’s type. If this width is not a multiple of 32-bits, the literals must be sign extended when the OpTypeInt Signedness is set to 1. (See Literal Number.)

This instruction must be the last instruction in a block.

3 + variable

251

<id>
Selector

<id>
Default

literal, label <id>,
literal, label <id>,
…
Target

OpKill

Fragment-shader discard.

Ceases all further processing in any invocation that executes it: Only instructions these invocations executed before OpKill will have observable side effects. If this instruction is executed in non-uniform control flow, all subsequent control flow is non-uniform (for invocations that continue to execute).

This instruction must be the last instruction in a block.

This instruction is only valid in the Fragment Execution Model.

Capability:
Shader

1

252

OpReturn

Return with no value from a function with void return type.

This instruction must be the last instruction in a block.

1

253

OpReturnValue

Return a value from a function.

Value is the value returned, by copy, and must match the Return Type operand of the OpTypeFunction type of the OpFunction body this return instruction is in.

This instruction must be the last instruction in a block.

2

254

<id>
Value

OpUnreachable

Declares that this block is not reachable in the CFG.

This instruction must be the last instruction in a block.

1

255

OpLifetimeStart

Declare that an object was not defined before this instruction.

Pointer is a pointer to the object whose lifetime is starting. Its type must be an OpTypePointer with Storage Class Function.

Size must be 0 if Pointer is a pointer to a non-void type or the Addresses capability is not being used. If Size is non-zero, it is the number of bytes of memory whose lifetime is starting. Its type must be an integer type scalar. It is treated as unsigned; if its type has Signedness of 1, its sign bit cannot be set.

Capability:
Kernel

3

256

<id>
Pointer

Literal Number
Size

OpLifetimeStop

Declare that an object is dead after this instruction.

Pointer is a pointer to the object whose lifetime is ending. Its type must be an OpTypePointer with Storage Class Function.

Size must be 0 if Pointer is a pointer to a non-void type or the Addresses capability is not being used. If Size is non-zero, it is the number of bytes of memory whose lifetime is ending. Its type must be an integer type scalar. It is treated as unsigned; if its type has Signedness of 1, its sign bit cannot be set.

Capability:
Kernel

3

257

<id>
Pointer

Literal Number
Size

3.32.18. Atomic Instructions

OpAtomicLoad

Atomically load through Pointer using the given Semantics. All subparts of the value that is loaded will be read atomically with respect to all other atomic accesses to it within Scope.

Result Type must be a scalar of integer type or floating-point type.

Pointer is the pointer to the memory to read. The type of the value pointed to by Pointer must be the same as Result Type.

Memory must be a valid memory Scope.

6

227

<id>
Result Type

Result <id>

<id>
Pointer

Scope <id>
Memory

Memory Semantics <id>
Semantics

OpAtomicStore

Atomically store through Pointer using the given Semantics. All subparts of Value will be written atomically with respect to all other atomic accesses to it within Scope.

Pointer is the pointer to the memory to write. The type it points to must be a scalar of integer type or floating-point type.

Value is the value to write. The type of Value and the type pointed to by Pointer must be the same type.

Memory must be a valid memory Scope.

5

228

<id>
Pointer

Scope <id>
Memory

Memory Semantics <id>
Semantics

<id>
Value

OpAtomicExchange

Perform the following steps atomically with respect to any other atomic accesses within Scope to the same location:
1) load through Pointer to get an Original Value,
2) get a New Value from copying Value, and
3) store the New Value back through Pointer.

The instruction’s result is the Original Value.

Result Type must be a scalar of integer type or floating-point type.

The type of Value must be the same as Result Type. The type of the value pointed to by Pointer must be the same as Result Type.

Memory must be a valid memory Scope.

7

229

<id>
Result Type

Result <id>

<id>
Pointer

Scope <id>
Memory

Memory Semantics <id>
Semantics

<id>
Value

OpAtomicCompareExchange

Perform the following steps atomically with respect to any other atomic accesses within Scope to the same location:
1) load through Pointer to get an Original Value,
2) get a New Value from Value only if Original Value equals Comparator, and
3) store the New Value back through Pointer’only if 'Original Value equaled Comparator.

The instruction’s result is the Original Value.

Result Type must be an integer type scalar.

Use Equal for the memory semantics of this instruction when Value and Original Value compare equal.

Use Unequal for the memory semantics of this instruction when Value and Original Value compare unequal. Unequal cannot be set to Release or Acquire and Release. In addition, Unequal cannot be set to a stronger memory-order then Equal.

The type of Value must be the same as Result Type. The type of the value pointed to by Pointer must be the same as Result Type. This type must also match the type of Comparator.

Memory must be a valid memory Scope.

9

230

<id>
Result Type

Result <id>

<id>
Pointer

Scope <id>
Memory

Memory Semantics <id>
Equal

Memory Semantics <id>
Unequal

<id>
Value

<id>
Comparator

OpAtomicCompareExchangeWeak

Deprecated (use OpAtomicCompareExchange).

Has the same semantics as OpAtomicCompareExchange.

Memory must be a valid memory Scope.

Capability:
Kernel

Missing after version 1.3.

9

231

<id>
Result Type

Result <id>

<id>
Pointer

Scope <id>
Memory

Memory Semantics <id>
Equal

Memory Semantics <id>
Unequal

<id>
Value

<id>
Comparator

OpAtomicIIncrement

Perform the following steps atomically with respect to any other atomic accesses within Scope to the same location:
1) load through Pointer to get an Original Value,
2) get a New Value through integer addition of 1 to Original Value, and
3) store the New Value back through Pointer.

The instruction’s result is the Original Value.

Result Type must be an integer type scalar. The type of the value pointed to by Pointer must be the same as Result Type.

Memory must be a valid memory Scope.

6

232

<id>
Result Type

Result <id>

<id>
Pointer

Scope <id>
Memory

Memory Semantics <id>
Semantics

OpAtomicIDecrement

Perform the following steps atomically with respect to any other atomic accesses within Scope to the same location:
1) load through Pointer to get an Original Value,
2) get a New Value through integer subtraction of 1 from Original Value, and
3) store the New Value back through Pointer.

The instruction’s result is the Original Value.

Result Type must be an integer type scalar. The type of the value pointed to by Pointer must be the same as Result Type.

Memory must be a valid memory Scope.

6

233

<id>
Result Type

Result <id>

<id>
Pointer

Scope <id>
Memory

Memory Semantics <id>
Semantics

OpAtomicIAdd

Perform the following steps atomically with respect to any other atomic accesses within Scope to the same location:
1) load through Pointer to get an Original Value,
2) get a New Value by integer addition of Original Value and Value, and
3) store the New Value back through Pointer.

The instruction’s result is the Original Value.

Result Type must be an integer type scalar.

The type of Value must be the same as Result Type. The type of the value pointed to by Pointer must be the same as Result Type.

Memory must be a valid memory Scope.

7

234

<id>
Result Type

Result <id>

<id>
Pointer

Scope <id>
Memory

Memory Semantics <id>
Semantics

<id>
Value

OpAtomicISub

Perform the following steps atomically with respect to any other atomic accesses within Scope to the same location:
1) load through Pointer to get an Original Value,
2) get a New Value by integer subtraction of Value from Original Value, and
3) store the New Value back through Pointer.

The instruction’s result is the Original Value.

Result Type must be an integer type scalar.

The type of Value must be the same as Result Type. The type of the value pointed to by Pointer must be the same as Result Type.

Memory must be a valid memory Scope.

7

235

<id>
Result Type

Result <id>

<id>
Pointer

Scope <id>
Memory

Memory Semantics <id>
Semantics

<id>
Value

OpAtomicSMin

Perform the following steps atomically with respect to any other atomic accesses within Scope to the same location:
1) load through Pointer to get an Original Value,
2) get a New Value by finding the smallest signed integer of Original Value and Value, and
3) store the New Value back through Pointer.

The instruction’s result is the Original Value.

Result Type must be an integer type scalar.

The type of Value must be the same as Result Type. The type of the value pointed to by Pointer must be the same as Result Type.

Memory must be a valid memory Scope.

7

236

<id>
Result Type

Result <id>

<id>
Pointer

Scope <id>
Memory

Memory Semantics <id>
Semantics

<id>
Value

OpAtomicUMin

Perform the following steps atomically with respect to any other atomic accesses within Scope to the same location:
1) load through Pointer to get an Original Value,
2) get a New Value by finding the smallest unsigned integer of Original Value and Value, and
3) store the New Value back through Pointer.

The instruction’s result is the Original Value.

Result Type must be an integer type scalar.

The type of Value must be the same as Result Type. The type of the value pointed to by Pointer must be the same as Result Type.

Memory must be a valid memory Scope.

7

237

<id>
Result Type

Result <id>

<id>
Pointer

Scope <id>
Memory

Memory Semantics <id>
Semantics

<id>
Value

OpAtomicSMax

Perform the following steps atomically with respect to any other atomic accesses within Scope to the same location:
1) load through Pointer to get an Original Value,
2) get a New Value by finding the largest signed integer of Original Value and Value, and
3) store the New Value back through Pointer.

The instruction’s result is the Original Value.

Result Type must be an integer type scalar.

The type of Value must be the same as Result Type. The type of the value pointed to by Pointer must be the same as Result Type.

Memory must be a valid memory Scope.

7

238

<id>
Result Type

Result <id>

<id>
Pointer

Scope <id>
Memory

Memory Semantics <id>
Semantics

<id>
Value

OpAtomicUMax

Perform the following steps atomically with respect to any other atomic accesses within Scope to the same location:
1) load through Pointer to get an Original Value,
2) get a New Value by finding the largest unsigned integer of Original Value and Value, and
3) store the New Value back through Pointer.

The instruction’s result is the Original Value.

Result Type must be an integer type scalar.

The type of Value must be the same as Result Type. The type of the value pointed to by Pointer must be the same as Result Type.

Memory must be a valid memory Scope.

7

239

<id>
Result Type

Result <id>

<id>
Pointer

Scope <id>
Memory

Memory Semantics <id>
Semantics

<id>
Value

OpAtomicAnd

Perform the following steps atomically with respect to any other atomic accesses within Scope to the same location:
1) load through Pointer to get an Original Value,
2) get a New Value by the bitwise AND of Original Value and Value, and
3) store the New Value back through Pointer.

The instruction’s result is the Original Value.

Result Type must be an integer type scalar.

The type of Value must be the same as Result Type. The type of the value pointed to by Pointer must be the same as Result Type.

Memory must be a valid memory Scope.

7

240

<id>
Result Type

Result <id>

<id>
Pointer

Scope <id>
Memory

Memory Semantics <id>
Semantics

<id>
Value

OpAtomicOr

Perform the following steps atomically with respect to any other atomic accesses within Scope to the same location:
1) load through Pointer to get an Original Value,
2) get a New Value by the bitwise OR of Original Value and Value, and
3) store the New Value back through Pointer.

The instruction’s result is the Original Value.

Result Type must be an integer type scalar.

The type of Value must be the same as Result Type. The type of the value pointed to by Pointer must be the same as Result Type.

Memory must be a valid memory Scope.

7

241

<id>
Result Type

Result <id>

<id>
Pointer

Scope <id>
Memory

Memory Semantics <id>
Semantics

<id>
Value

OpAtomicXor

Perform the following steps atomically with respect to any other atomic accesses within Scope to the same location:
1) load through Pointer to get an Original Value,
2) get a New Value by the bitwise exclusive OR of Original Value and Value, and
3) store the New Value back through Pointer.

The instruction’s result is the Original Value.

Result Type must be an integer type scalar.

The type of Value must be the same as Result Type. The type of the value pointed to by Pointer must be the same as Result Type.

Memory must be a valid memory Scope.

7

242

<id>
Result Type

Result <id>

<id>
Pointer

Scope <id>
Memory

Memory Semantics <id>
Semantics

<id>
Value

OpAtomicFlagTestAndSet

Atomically sets the flag value pointed to by Pointer to the set state.

Pointer must be a pointer to a 32-bit integer type representing an atomic flag.

The instruction’s result is true if the flag was in the set state or false if the flag was in the clear state immediately before the operation.

Result Type must be a Boolean type.

Results are undefined if an atomic flag is modified by an instruction other than OpAtomicFlagTestAndSet or OpAtomicFlagClear

Memory must be a valid memory Scope.

Capability:
Kernel

6

318

<id>
Result Type

Result <id>

<id>
Pointer

Scope <id>
Memory

Memory Semantics <id>
Semantics

OpAtomicFlagClear

Atomically sets the flag value pointed to by Pointer to the clear state.

Pointer must be a pointer to a 32-bit integer type representing an atomic flag.

Memory Semantics cannot be Acquire or AcquireRelease

Results are undefined if an atomic flag is modified by an instruction other than OpAtomicFlagTestAndSet or OpAtomicFlagClear

Memory must be a valid memory Scope.

Capability:
Kernel

4

319

<id>
Pointer

Scope <id>
Memory

Memory Semantics <id>
Semantics

3.32.19. Primitive Instructions

OpEmitVertex

Emits the current values of all output variables to the current output primitive. After execution, the values of all output variables are undefined.

This instruction can only be used when only one stream is present.

Capability:
Geometry

1

218

OpEndPrimitive

Finish the current primitive and start a new one. No vertex is emitted.

This instruction can only be used when only one stream is present.

Capability:
Geometry

1

219

OpEmitStreamVertex

Emits the current values of all output variables to the current output primitive. After execution, the values of all output variables are undefined.

Stream must be an <id> of a constant instruction with a scalar integer type. That constant is the output-primitive stream number.

This instruction can only be used when multiple streams are present.

Capability:
GeometryStreams

2

220

<id>
Stream

OpEndStreamPrimitive

Finish the current primitive and start a new one. No vertex is emitted.

Stream must be an <id> of a constant instruction with a scalar integer type. That constant is the output-primitive stream number.

This instruction can only be used when multiple streams are present.

Capability:
GeometryStreams

2

221

<id>
Stream

3.32.20. Barrier Instructions

OpControlBarrier

Wait for other invocations of this module to reach the current point of execution.

All invocations of this module within Execution scope must reach this point of execution before any invocation will proceed beyond it.

When Execution is Workgroup or larger, behavior is undefined if this instruction is used in control flow that is non-uniform within Execution. When Execution is Subgroup or Invocation, the behavior of this instruction in non-uniform control flow is defined by the client API.

If Semantics is not None, this instruction also serves as an OpMemoryBarrier instruction, and must also perform and adhere to the description and semantics of an OpMemoryBarrier instruction with the same Memory and Semantics operands. This allows atomically specifying both a control barrier and a memory barrier (that is, without needing two instructions). If Semantics is None, Memory is ignored.

Before version 1.3, it is only valid to use this instruction with TessellationControl, GLCompute, or Kernel execution models. There is no such restriction starting with version 1.3.

When used with the TessellationControl execution model, it also implicitly synchronizes the Output Storage Class: Writes to Output variables performed by any invocation executed prior to a OpControlBarrier will be visible to any other invocation after return from that OpControlBarrier.

4

224

Scope <id>
Execution

Scope <id>
Memory

Memory Semantics <id>
Semantics

OpMemoryBarrier

Control the order that memory accesses are observed.

Ensures that memory accesses issued before this instruction will be observed before memory accesses issued after this instruction. This control is ensured only for memory accesses issued by this invocation and observed by another invocation executing within Memory scope.

Semantics declares what kind of memory is being controlled and what kind of control to apply.

To execute both a memory barrier and a control barrier, see OpControlBarrier.

3

225

Scope <id>
Memory

Memory Semantics <id>
Semantics

OpNamedBarrierInitialize

Declare a new named-barrier object.

Result Type must be the type OpTypeNamedBarrier.

Subgroup Count must be a 32-bit integer type scalar representing the number of subgroups that must reach the current point of execution.

Capability:
NamedBarrier

Missing before version 1.1.

4

328

<id>
Result Type

Result <id>

<id>
Subgroup Count

OpMemoryNamedBarrier

Wait for other invocations of this module to reach the current point of execution.

Named Barrier must be the type OpTypeNamedBarrier.

If Semantics is not None, this instruction also serves as an OpMemoryBarrier instruction, and must also perform and adhere to the description and semantics of an OpMemoryBarrier instruction with the same Memory and Semantics operands. This allows atomically specifying both a control barrier and a memory barrier (that is, without needing two instructions). If Semantics None, Memory is ignored.

Capability:
NamedBarrier

Missing before version 1.1.

4

329

<id>
Named Barrier

Scope <id>
Memory

Memory Semantics <id>
Semantics

3.32.21. Group Instructions

OpGroupAsyncCopy

Perform an asynchronous group copy of Num Elements elements from Source to Destination. The asynchronous copy is performed by all work-items in a group.

This instruction returns an event object that can be used by OpGroupWaitEvents to wait for the async copy to finish.

All invocations of this module within Execution must reach this point of execution.

Behavior is undefined if this instruction is used in control flow that is non-uniform within Execution.

Result Type must be an OpTypeEvent object.

Destination must be a pointer to a scalar or vector of floating-point type or integer type.

Destination pointer Storage Class must be Workgroup or CrossWorkgroup.

The type of Source must be the same as Destination.

When Destination pointer Storage Class is Workgroup, the Source pointer Storage Class must be CrossWorkgroup. In this case Stride defines the stride in elements when reading from Source pointer.

When Destination pointer Storage Class is CrossWorkgroup, the Source pointer Storage Class must be Workgroup. In this case Stride defines the stride in elements when writing each element to Destination pointer.

Stride and NumElements must be a 32-bit integer type scalar when the addressing model is Physical32 and 64 bit integer type scalar when the Addressing Model is Physical64.

Event must have a type of OpTypeEvent.

Event can be used to associate the copy with a previous copy allowing an event to be shared by multiple copies. Otherwise Event should be an OpConstantNull.

If Event argument is not OpConstantNull, the event object supplied in event argument will be returned.

Capability:
Kernel

9

259

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Destination

<id>
Source

<id>
Num Elements

<id>
Stride

<id>
Event

OpGroupWaitEvents

Wait for events generated by OpGroupAsyncCopy operations to complete. Events List points to Num Events event objects, which will be released after the wait is performed.

All invocations of this module within Execution must reach this point of execution.

Behavior is undefined if this instruction is used in control flow that is non-uniform within Execution.

Execution must be Workgroup or Subgroup Scope.

Num Events must be a 32-bit integer type scalar.

Events List must be a pointer to OpTypeEvent.

Capability:
Kernel

4

260

Scope <id>
Execution

<id>
Num Events

<id>
Events List

OpGroupAll

Evaluates a predicate for all invocations in the group,resulting in true if predicate evaluates to true for all invocations in the group, otherwise the result is false.

All invocations of this module within Execution must reach this point of execution.

Behavior is undefined if this instruction is used in control flow that is non-uniform within Execution.

Result Type must be a Boolean type.

Execution must be Workgroup or Subgroup Scope.

Predicate must be a Boolean type.

Capability:
Groups

5

261

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Predicate

OpGroupAny

Evaluates a predicate for all invocations in the group,resulting in true if predicate evaluates to true for any invocation in the group, otherwise the result is false.

All invocations of this module within Execution must reach this point of execution.

Behavior is undefined if this instruction is used in control flow that is non-uniform within Execution.

Result Type must be a Boolean type.

Execution must be Workgroup or Subgroup Scope.

Predicate must be a Boolean type.

Capability:
Groups

5

262

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Predicate

OpGroupBroadcast

Return the Value of the invocation identified by the local id LocalId to all invocations in the group.

All invocations of this module within Execution must reach this point of execution.

Behavior is undefined if this instruction is used in control flow that is non-uniform within Execution.

Result Type must be a 32-bit or 64-bit integer type or a 16, 32 or 64 float type scalar.

Execution must be Workgroup or Subgroup Scope.

The type of Value must be the same as Result Type.

LocalId must be an integer datatype. It can be a scalar, or a vector with 2 components or a vector with 3 components. LocalId must be the same for all invocations in the group.

Capability:
Groups

6

263

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Value

<id>
LocalId

OpGroupIAdd

An integer add group operation specified for all values of X specified by invocations in the group.

All invocations of this module within Execution must reach this point of execution.

Behavior is undefined if this instruction is used in control flow that is non-uniform within Execution.

Result Type must be a 32-bit or 64-bit integer type scalar.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is 0.

The type of X must be the same as Result Type.

Capability:
Groups

6

264

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
X

OpGroupFAdd

A floating-point add group operation specified for all values of X specified by invocations in the group.

All invocations of this module within Execution must reach this point of execution.

Behavior is undefined if this instruction is used in control flow that is non-uniform within Execution.

Result Type must be a 16-bit, 32-bit, or 64-bit floating-point type scalar.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is 0.

The type of X must be the same as Result Type.

Capability:
Groups

6

265

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
X

OpGroupFMin

A floating-point minimum group operation specified for all values of X specified by invocations in the group.

All invocations of this module within Execution must reach this point of execution.

Behavior is undefined if this instruction is used in control flow that is non-uniform within Execution.

Result Type must be a 16-bit, 32-bit, or 64-bit floating-point type scalar.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is +INF.

The type of X must be the same as Result Type.

Capability:
Groups

6

266

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
X

OpGroupUMin

An unsigned integer minimum group operation specified for all values of X specified by invocations in the group.

All invocations of this module within Execution must reach this point of execution.

Behavior is undefined if this instruction is used in control flow that is non-uniform within Execution.

Result Type must be a 32-bit or 64-bit integer type scalar.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is UINT_MAX when X is 32 bits wide and ULONG_MAX when X is 64 bits wide.

The type of X must be the same as Result Type.

Capability:
Groups

6

267

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
X

OpGroupSMin

A signed integer minimum group operation specified for all values of X specified by invocations in the group.

All invocations of this module within Execution must reach this point of execution.

Behavior is undefined if this instruction is used in control flow that is non-uniform within Execution.

Result Type must be a 32-bit or 64-bit integer type scalar.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is INT_MAX when X is 32 bits wide and LONG_MAX when X is 64 bits wide.

The type of X must be the same as Result Type.

Capability:
Groups

6

268

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
X

OpGroupFMax

A floating-point maximum group operation specified for all values of X specified by invocations in the group.

All invocations of this module within Execution must reach this point of execution.

Behavior is undefined if this instruction is used in control flow that is non-uniform within Execution.

Result Type must be a 16-bit, 32-bit, or 64-bit floating-point type scalar.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is -INF.

The type of X must be the same as Result Type.

Capability:
Groups

6

269

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
X

OpGroupUMax

An unsigned integer maximum group operation specified for all values of X specified by invocations in the group.

All invocations of this module within Execution must reach this point of execution.

Behavior is undefined if this instruction is used in control flow that is non-uniform within Execution.

Result Type must be a 32-bit or 64-bit integer type scalar.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is 0.

The type of X must be the same as Result Type.

Capability:
Groups

6

270

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
X

OpGroupSMax

A signed integer maximum group operation specified for all values of X specified by invocations in the group.

All invocations of this module within Execution must reach this point of execution.

Behavior is undefined if this instruction is used in control flow that is non-uniform within Execution.

X and Result Type must be a 32-bit or 64-bit OpTypeInt data type.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is INT_MIN when X is 32 bits wide and LONG_MIN when X is 64 bits wide.

The type of X must be the same as Result Type.

Capability:
Groups

6

271

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
X

OpSubgroupBallotKHR

See extension SPV_KHR_shader_ballot

Capability:
SubgroupBallotKHR

Reserved.

4

4421

<id>
Result Type

Result <id>

<id>
Predicate

OpSubgroupFirstInvocationKHR

See extension SPV_KHR_shader_ballot

Capability:
SubgroupBallotKHR

Reserved.

4

4422

<id>
Result Type

Result <id>

<id>
Value

OpSubgroupAllKHR

TBD

Capability:
SubgroupVoteKHR

Reserved.

4

4428

<id>
Result Type

Result <id>

<id>
Predicate

OpSubgroupAnyKHR

TBD

Capability:
SubgroupVoteKHR

Reserved.

4

4429

<id>
Result Type

Result <id>

<id>
Predicate

OpSubgroupAllEqualKHR

TBD

Capability:
SubgroupVoteKHR

Reserved.

4

4430

<id>
Result Type

Result <id>

<id>
Predicate

OpSubgroupReadInvocationKHR

See extension SPV_KHR_shader_ballot

Capability:
SubgroupBallotKHR

Reserved.

5

4432

<id>
Result Type

Result <id>

<id>
Value

<id>
Index

OpGroupIAddNonUniformAMD

TBD

Capability:
Groups

Reserved.

6

5000

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
X

OpGroupFAddNonUniformAMD

TBD

Capability:
Groups

Reserved.

6

5001

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
X

OpGroupFMinNonUniformAMD

TBD

Capability:
Groups

Reserved.

6

5002

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
X

OpGroupUMinNonUniformAMD

TBD

Capability:
Groups

Reserved.

6

5003

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
X

OpGroupSMinNonUniformAMD

TBD

Capability:
Groups

Reserved.

6

5004

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
X

OpGroupFMaxNonUniformAMD

TBD

Capability:
Groups

Reserved.

6

5005

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
X

OpGroupUMaxNonUniformAMD

TBD

Capability:
Groups

Reserved.

6

5006

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
X

OpGroupSMaxNonUniformAMD

TBD

Capability:
Groups

Reserved.

6

5007

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
X

OpSubgroupShuffleINTEL

TBD

Capability:
SubgroupShuffleINTEL

Reserved.

5

5571

<id>
Result Type

Result <id>

<id>
Data

<id>
InvocationId

OpSubgroupShuffleDownINTEL

TBD

Capability:
SubgroupShuffleINTEL

Reserved.

6

5572

<id>
Result Type

Result <id>

<id>
Current

<id>
Next

<id>
Delta

OpSubgroupShuffleUpINTEL

TBD

Capability:
SubgroupShuffleINTEL

Reserved.

6

5573

<id>
Result Type

Result <id>

<id>
Previous

<id>
Current

<id>
Delta

OpSubgroupShuffleXorINTEL

TBD

Capability:
SubgroupShuffleINTEL

Reserved.

5

5574

<id>
Result Type

Result <id>

<id>
Data

<id>
Value

OpSubgroupBlockReadINTEL

TBD

Capability:
SubgroupBufferBlockIOINTEL

Reserved.

4

5575

<id>
Result Type

Result <id>

<id>
Ptr

OpSubgroupBlockWriteINTEL

TBD

Capability:
SubgroupBufferBlockIOINTEL

Reserved.

3

5576

<id>
Ptr

<id>
Data

OpSubgroupImageBlockReadINTEL

TBD

Capability:
SubgroupImageBlockIOINTEL

Reserved.

5

5577

<id>
Result Type

Result <id>

<id>
Image

<id>
Coordinate

OpSubgroupImageBlockWriteINTEL

TBD

Capability:
SubgroupImageBlockIOINTEL

Reserved.

4

5578

<id>
Image

<id>
Coordinate

<id>
Data

OpSubgroupImageMediaBlockReadINTEL

TBD

Capability:
SubgroupImageMediaBlockIOINTEL

Reserved.

7

5580

<id>
Result Type

Result <id>

<id>
Image

<id>
Coordinate

<id>
Width

<id>
Height

OpSubgroupImageMediaBlockWriteINTEL

TBD

Capability:
SubgroupImageMediaBlockIOINTEL

Reserved.

6

5581

<id>
Image

<id>
Coordinate

<id>
Width

<id>
Height

<id>
Data

3.32.22. Device-Side Enqueue Instructions

OpEnqueueMarker

Enqueue a marker command to the queue object specified by Queue. The marker command waits for a list of events to complete, or if the list is empty it waits for all previously enqueued commands in Queue to complete before the marker completes.

Result Type must be a 32-bit integer type scalar. A successful enqueue results in the value 0. A failed enqueue results in a non-0 value.

Queue must be of the type OpTypeQueue.

Num Events specifies the number of event objects in the wait list pointed to by Wait Events and must be a 32-bit integer type scalar, which is treated as an unsigned integer.

Wait Events specifies the list of wait event objects and must be a pointer to OpTypeDeviceEvent.

Ret Event is a pointer to a device event which gets implicitly retained by this instruction. It must have a type of OpTypePointer to OpTypeDeviceEvent. If Ret Event is set to null this instruction becomes a no-op.

Capability:
DeviceEnqueue

7

291

<id>
Result Type

Result <id>

<id>
Queue

<id>
Num Events

<id>
Wait Events

<id>
Ret Event

OpEnqueueKernel

Enqueue the function specified by Invoke and the NDRange specified by ND Range for execution to the queue object specified by Queue.

Result Type must be a 32-bit integer type scalar. A successful enqueue results in the value 0. A failed enqueue results in a non-0 value.

Queue must be of the type OpTypeQueue.

Flags must be an integer type scalar. The content of Flags is interpreted as Kernel Enqueue Flags mask.

The type of ND Range must be an OpTypeStruct whose members are as described by the Result Type of OpBuildNDRange.

Num Events specifies the number of event objects in the wait list pointed to by Wait Events and must be 32-bit integer type scalar, which is treated as an unsigned integer.

Wait Events specifies the list of wait event objects and must be a pointer to OpTypeDeviceEvent.

Ret Event must be a pointer to OpTypeDeviceEvent which gets implicitly retained by this instruction.

Invoke must be an OpFunction whose OpTypeFunction operand has:
- Result Type must be OpTypeVoid.
- The first parameter must have a type of OpTypePointer to an 8-bit OpTypeInt.
- An optional list of parameters, each of which must have a type of OpTypePointer to the Workgroup Storage Class.

Param is the first parameter of the function specified by Invoke and must be a pointer to an 8-bit integer type scalar.

Param Size is the size in bytes of the memory pointed to by Param and must be a 32-bit integer type scalar, which is treated as an unsigned integer.

Param Align is the alignment of Param and must be a 32-bit integer type scalar, which is treated as an unsigned integer.

Each Local Size operand corresponds (in order) to one OpTypePointer to Workgroup Storage Class parameter to the Invoke function, and specifies the number of bytes of Workgroup storage used to back the pointer during the execution of the Invoke function.

Capability:
DeviceEnqueue

13 + variable

292

<id>
Result Type

Result <id>

<id>
Queue

<id>
Flags

<id>
ND Range

<id>
Num Events

<id>
Wait Events

<id>
Ret Event

<id>
Invoke

<id>
Param

<id>
Param Size

<id>
Param Align

<id>, <id>, …
Local Size

OpGetKernelNDrangeSubGroupCount

Returns the number of subgroups in each workgroup of the dispatch (except for the last in cases where the global size does not divide cleanly into work-groups) given the combination of the passed NDRange descriptor specified by ND Range and the function specified by Invoke.

Result Type must be a 32-bit integer type scalar.

The type of ND Range must be an OpTypeStruct whose members are as described by the Result Type of OpBuildNDRange.

Invoke must be an OpFunction whose OpTypeFunction operand has:
- Result Type must be OpTypeVoid.
- The first parameter must have a type of OpTypePointer to an 8-bit OpTypeInt.
- An optional list of parameters, each of which must have a type of OpTypePointer to the Workgroup Storage Class.

Param is the first parameter of the function specified by Invoke and must be a pointer to an 8-bit integer type scalar.

Param Size is the size in bytes of the memory pointed to by Param and must be a 32-bit integer type scalar, which is treated as an unsigned integer.

Param Align is the alignment of Param and must be a 32-bit integer type scalar, which is treated as an unsigned integer.

Capability:
DeviceEnqueue

8

293

<id>
Result Type

Result <id>

<id>
ND Range

<id>
Invoke

<id>
Param

<id>
Param Size

<id>
Param Align

OpGetKernelNDrangeMaxSubGroupSize

Returns the maximum sub-group size for the function specified by Invoke and the NDRange specified by ND Range.

Result Type must be a 32-bit integer type scalar.

The type of ND Range must be an OpTypeStruct whose members are as described by the Result Type of OpBuildNDRange.

Invoke must be an OpFunction whose OpTypeFunction operand has:
- Result Type must be OpTypeVoid.
- The first parameter must have a type of OpTypePointer to an 8-bit OpTypeInt.
- An optional list of parameters, each of which must have a type of OpTypePointer to the Workgroup Storage Class.

Param is the first parameter of the function specified by Invoke and must be a pointer to an 8-bit integer type scalar.

Param Size is the size in bytes of the memory pointed to by Param and must be a 32-bit integer type scalar, which is treated as an unsigned integer.

Param Align is the alignment of Param and must be a 32-bit integer type scalar, which is treated as an unsigned integer.

Capability:
DeviceEnqueue

8

294

<id>
Result Type

Result <id>

<id>
ND Range

<id>
Invoke

<id>
Param

<id>
Param Size

<id>
Param Align

OpGetKernelWorkGroupSize

Returns the maximum work-group size that can be used to execute the function specified by Invoke on the device.

Result Type must be a 32-bit integer type scalar.

Invoke must be an OpFunction whose OpTypeFunction operand has:
- Result Type must be OpTypeVoid.
- The first parameter must have a type of OpTypePointer to an 8-bit OpTypeInt.
- An optional list of parameters, each of which must have a type of OpTypePointer to the Workgroup Storage Class.

Param is the first parameter of the function specified by Invoke and must be a pointer to an 8-bit integer type scalar.

Param Size is the size in bytes of the memory pointed to by Param and must be a 32-bit integer type scalar, which is treated as an unsigned integer.

Param Align is the alignment of Param and must be a 32-bit integer type scalar, which is treated as an unsigned integer.

Capability:
DeviceEnqueue

7

295

<id>
Result Type

Result <id>

<id>
Invoke

<id>
Param

<id>
Param Size

<id>
Param Align

OpGetKernelPreferredWorkGroupSizeMultiple

Returns the preferred multiple of work-group size for the function specified by Invoke. This is a performance hint. Specifying a work-group size that is not a multiple of the value returned by this query as the value of the local work size will not fail to enqueue Invoke for execution unless the work-group size specified is larger than the device maximum.

Result Type must be a 32-bit integer type scalar.

Invoke must be an OpFunction whose OpTypeFunction operand has:
- Result Type must be OpTypeVoid.
- The first parameter must have a type of OpTypePointer to an 8-bit OpTypeInt.
- An optional list of parameters, each of which must have a type of OpTypePointer to the Workgroup Storage Class.

Param is the first parameter of the function specified by Invoke and must be a pointer to an 8-bit integer type scalar.

Param Size is the size in bytes of the memory pointed to by Param and must be a 32-bit integer type scalar, which is treated as an unsigned integer.

Param Align is the alignment of Param and must be a 32-bit integer type scalar, which is treated as an unsigned integer.

Capability:
DeviceEnqueue

7

296

<id>
Result Type

Result <id>

<id>
Invoke

<id>
Param

<id>
Param Size

<id>
Param Align

OpRetainEvent

Increments the reference count of the event object specified by Event.

Event must be an event that was produced by OpEnqueueKernel, OpEnqueueMarker or OpCreateUserEvent.

Capability:
DeviceEnqueue

2

297

<id>
Event

OpReleaseEvent

Decrements the reference count of the event object specified by Event. The event object is deleted once the event reference count is zero, the specific command identified by this event has completed (or terminated) and there are no commands in any device command queue that require a wait for this event to complete.

Event must be an event that was produced by OpEnqueueKernel, OpEnqueueMarker or OpCreateUserEvent.

Capability:
DeviceEnqueue

2

298

<id>
Event

OpCreateUserEvent

Create a user event. The execution status of the created event is set to a value of 2 (CL_SUBMITTED).

Result Type must be OpTypeDeviceEvent.

Capability:
DeviceEnqueue

3

299

<id>
Result Type

Result <id>

OpIsValidEvent

Returns true if the event specified by Event is a valid event, otherwise result is false.

Result Type must be a Boolean type.

Event must have a type of OpTypeDeviceEvent

Capability:
DeviceEnqueue

4

300

<id>
Result Type

Result <id>

<id>
Event

OpSetUserEventStatus

Sets the execution status of a user event specified by Event.Status can be either 0 (CL_COMPLETE) to indicate that this kernel and all its child kernels finished execution successfully, or a negative integer value indicating an error.

Event must have a type of OpTypeDeviceEvent that was produced by OpCreateUserEvent.

Status must have a type of 32-bit OpTypeInt treated as a signed integer.

Capability:
DeviceEnqueue

3

301

<id>
Event

<id>
Status

OpCaptureEventProfilingInfo

Captures the profiling information specified by Profiling Info for the command associated with the event specified by Event in the memory pointed to by Value.The profiling information will be available in the memory pointed to by Value once the command identified by Event has completed.

Event must have a type of OpTypeDeviceEvent that was produced by OpEnqueueKernel or OpEnqueueMarker.

Profiling Info must be an integer type scalar. The content of Profiling Info is interpreted as Kernel Profiling Info mask.

Value must be a pointer to a scalar 8-bit integer type in the CrossWorkgroup Storage Class.

When Profiling Info is CmdExecTime, Value must point to 128-bit memory range.
The first 64 bits contain the elapsed time CL_PROFILING_COMMAND_END - CL_PROFILING_COMMAND_START for the command identified by Event in nanoseconds.
The second 64 bits contain the elapsed time CL_PROFILING_COMMAND_COMPLETE - CL_PROFILING_COMMAND_START for the command identified by Event in nanoseconds.

Note: The behavior of this instruction is undefined when called multiple times for the same event.

Capability:
DeviceEnqueue

4

302

<id>
Event

<id>
Profiling Info

<id>
Value

OpGetDefaultQueue

Returns the default device queue. If a default device queue has not been created, a null queue object is returned.

Result Type must be an OpTypeQueue.

Capability:
DeviceEnqueue

3

303

<id>
Result Type

Result <id>

OpBuildNDRange

Given the global work size specified by GlobalWorkSize, local work size specified by LocalWorkSize and global work offset specified by GlobalWorkOffset, builds a 1D, 2D or 3D ND-range descriptor structure and returns it.

Result Type must be an OpTypeStruct with the following ordered list of members, starting from the first to last:

1) 32-bit integer type scalar, that specifies the number of dimensions used to specify the global work-items and work-items in the work-group.

2) OpTypeArray with 3 elements, where each element is 32-bit integer type scalar when the addressing model is Physical32 and 64-bit integer type scalar when the addressing model is Physical64. This member is an array of per-dimension unsigned values that describe the offset used to calculate the global ID of a work-item.

3) OpTypeArray with 3 elements, where each element is 32-bit integer type scalar when the addressing model is Physical32 and 64-bit integer type scalar when the addressing model is Physical64. This member is an array of per-dimension unsigned values that describe the number of global work-items in the dimensions that will execute the kernel function.

4) OpTypeArray with 3 elements, where each element is 32-bit integer type scalar when the addressing model is Physical32 and 64-bit integer type scalar when the addressing model is Physical64. This member is an array of per-dimension unsigned values that describe the number of work-items that make up a work-group.

GlobalWorkSize must be a scalar or an array with 2 or 3 components. Where the type of each element in the array is 32-bit integer type scalar when the addressing model is Physical32 or 64-bit integer type scalar when the addressing model is Physical64.

The type of LocalWorkSize must be the same as GlobalWorkSize.

The type of GlobalWorkOffset must be the same as GlobalWorkSize.

Capability:
DeviceEnqueue

6

304

<id>
Result Type

Result <id>

<id>
GlobalWorkSize

<id>
LocalWorkSize

<id>
GlobalWorkOffset

OpGetKernelLocalSizeForSubgroupCount

Returns the 1D local size to enqueue Invoke with Subgroup Count subgroups per workgroup.

Result Type must be a 32-bit integer type scalar.

Subgroup Count must be a 32-bit integer type scalar.

Invoke must be an OpFunction whose OpTypeFunction operand has:
- Result Type must be OpTypeVoid.
- The first parameter must have a type of OpTypePointer to an 8-bit OpTypeInt.
- An optional list of parameters, each of which must have a type of OpTypePointer to the Workgroup Storage Class.

Param is the first parameter of the function specified by Invoke and must be a pointer to an 8-bit integer type scalar.

Param Size is the size in bytes of the memory pointed to by Param and must be a 32-bit integer type scalar, which is treated as an unsigned integer.

Param Align is the alignment of Param and must be a 32-bit integer type scalar, which is treated as an unsigned integer.

Capability:
SubgroupDispatch

Missing before version 1.1.

8

325

<id>
Result Type

Result <id>

<id>
Subgroup Count

<id>
Invoke

<id>
Param

<id>
Param Size

<id>
Param Align

OpGetKernelMaxNumSubgroups

Returns the maximum number of subgroups that can be used to execute Invoke on the devce.

Result Type must be a 32-bit integer type scalar.

Invoke must be an OpFunction whose OpTypeFunction operand has:
- Result Type must be OpTypeVoid.
- The first parameter must have a type of OpTypePointer to an 8-bit OpTypeInt.
- An optional list of parameters, each of which must have a type of OpTypePointer to the Workgroup Storage Class.

Param is the first parameter of the function specified by Invoke and must be a pointer to an 8-bit integer type scalar.

Param Size is the size in bytes of the memory pointed to by Param and must be a 32-bit integer type scalar, which is treated as an unsigned integer.

Param Align is the alignment of Param and must be a 32-bit integer type scalar, which is treated as an unsigned integer.

Capability:
SubgroupDispatch

Missing before version 1.1.

7

326

<id>
Result Type

Result <id>

<id>
Invoke

<id>
Param

<id>
Param Size

<id>
Param Align

3.32.23. Pipe Instructions

OpReadPipe

Read a packet from the pipe object specified by Pipe into Pointer. Result is 0 if the operation is successful and a negative value if the pipe is empty.

Result Type must be a 32-bit integer type scalar.

Pipe must have a type of OpTypePipe with ReadOnly access qualifier.

Pointer must have a type of OpTypePointer with the same data type as Pipe and a Generic Storage Class.

Packet Size must be a 32-bit integer type scalar that represents the size in bytes of each packet in the pipe.

Packet Alignment must be a 32-bit integer type scalar that presents the alignment in bytes of each packet in the pipe.

Packet Size and Packet Alignment must satisfy the following:
- 1 <= Packet Alignment <= Packet Size.
- Packet Alignment must evenly divide Packet Size

For concrete types, Packet Alignment should equal Packet Size. For aggregate types, Packet Alignment should be the size of the largest primitive type in the hierarchy of types.

Capability:
Pipes

7

274

<id>
Result Type

Result <id>

<id>
Pipe

<id>
Pointer

<id>
Packet Size

<id>
Packet Alignment

OpWritePipe

Write a packet from Pointer to the pipe object specified by Pipe. Result is 0 if the operation is successful and a negative value if the pipe is full.

Result Type must be a 32-bit integer type scalar.

Pipe must have a type of OpTypePipe with WriteOnly access qualifier.

Pointer must have a type of OpTypePointer with the same data type as Pipe and a Generic Storage Class.

Packet Size must be a 32-bit integer type scalar that represents the size in bytes of each packet in the pipe.

Packet Alignment must be a 32-bit integer type scalar that presents the alignment in bytes of each packet in the pipe.

Packet Size and Packet Alignment must satisfy the following:
- 1 <= Packet Alignment <= Packet Size.
- Packet Alignment must evenly divide Packet Size

For concrete types, Packet Alignment should equal Packet Size. For aggregate types, Packet Alignment should be the size of the largest primitive type in the hierarchy of types.

Capability:
Pipes

7

275

<id>
Result Type

Result <id>

<id>
Pipe

<id>
Pointer

<id>
Packet Size

<id>
Packet Alignment

OpReservedReadPipe

Read a packet from the reserved area specified by Reserve Id and Index of the pipe object specified by Pipe into Pointer. The reserved pipe entries are referred to by indices that go from 0 … Num Packets - 1. Result is 0 if the operation is successful and a negative value otherwise.

Result Type must be a 32-bit integer type scalar.

Pipe must have a type of OpTypePipe with ReadOnly access qualifier.

Reserve Id must have a type of OpTypeReserveId.

Index must be a 32-bit integer type scalar, which is treated as an unsigned value.

Pointer must have a type of OpTypePointer with the same data type as Pipe and a Generic Storage Class.

Packet Size must be a 32-bit integer type scalar that represents the size in bytes of each packet in the pipe.

Packet Alignment must be a 32-bit integer type scalar that presents the alignment in bytes of each packet in the pipe.

Packet Size and Packet Alignment must satisfy the following:
- 1 <= Packet Alignment <= Packet Size.
- Packet Alignment must evenly divide Packet Size

For concrete types, Packet Alignment should equal Packet Size. For aggregate types, Packet Alignment should be the size of the largest primitive type in the hierarchy of types.

Capability:
Pipes

9

276

<id>
Result Type

Result <id>

<id>
Pipe

<id>
Reserve Id

<id>
Index

<id>
Pointer

<id>
Packet Size

<id>
Packet Alignment

OpReservedWritePipe

Write a packet from Pointer into the reserved area specified by Reserve Id and Index of the pipe object specified by Pipe. The reserved pipe entries are referred to by indices that go from 0 … Num Packets - 1. Result is 0 if the operation is successful and a negative value otherwise.

Result Type must be a 32-bit integer type scalar.

Pipe must have a type of OpTypePipe with WriteOnly access qualifier.

Reserve Id must have a type of OpTypeReserveId.

Index must be a 32-bit integer type scalar, which is treated as an unsigned value.

Pointer must have a type of OpTypePointer with the same data type as Pipe and a Generic Storage Class.

Packet Size must be a 32-bit integer type scalar that represents the size in bytes of each packet in the pipe.

Packet Alignment must be a 32-bit integer type scalar that presents the alignment in bytes of each packet in the pipe.

Packet Size and Packet Alignment must satisfy the following:
- 1 <= Packet Alignment <= Packet Size.
- Packet Alignment must evenly divide Packet Size

For concrete types, Packet Alignment should equal Packet Size. For aggregate types, Packet Alignment should be the size of the largest primitive type in the hierarchy of types.

Capability:
Pipes

9

277

<id>
Result Type

Result <id>

<id>
Pipe

<id>
Reserve Id

<id>
Index

<id>
Pointer

<id>
Packet Size

<id>
Packet Alignment

OpReserveReadPipePackets

Reserve Num Packets entries for reading from the pipe object specified by Pipe. Result is a valid reservation ID if the reservation is successful.

Result Type must be an OpTypeReserveId.

Pipe must have a type of OpTypePipe with ReadOnly access qualifier.

Num Packets must be a 32-bit integer type scalar, which is treated as an unsigned value.

Packet Size must be a 32-bit integer type scalar that represents the size in bytes of each packet in the pipe.

Packet Alignment must be a 32-bit integer type scalar that presents the alignment in bytes of each packet in the pipe.

Packet Size and Packet Alignment must satisfy the following:
- 1 <= Packet Alignment <= Packet Size.
- Packet Alignment must evenly divide Packet Size

For concrete types, Packet Alignment should equal Packet Size. For aggregate types, Packet Alignment should be the size of the largest primitive type in the hierarchy of types.

Capability:
Pipes

7

278

<id>
Result Type

Result <id>

<id>
Pipe

<id>
Num Packets

<id>
Packet Size

<id>
Packet Alignment

OpReserveWritePipePackets

Reserve num_packets entries for writing to the pipe object specified by Pipe. Result is a valid reservation ID if the reservation is successful.

Pipe must have a type of OpTypePipe with WriteOnly access qualifier.

Num Packets must be a 32-bit OpTypeInt which is treated as an unsigned value.

Result Type must be an OpTypeReserveId.

Packet Size must be a 32-bit integer type scalar that represents the size in bytes of each packet in the pipe.

Packet Alignment must be a 32-bit integer type scalar that presents the alignment in bytes of each packet in the pipe.

Packet Size and Packet Alignment must satisfy the following:
- 1 <= Packet Alignment <= Packet Size.
- Packet Alignment must evenly divide Packet Size

For concrete types, Packet Alignment should equal Packet Size. For aggregate types, Packet Alignment should be the size of the largest primitive type in the hierarchy of types.

Capability:
Pipes

7

279

<id>
Result Type

Result <id>

<id>
Pipe

<id>
Num Packets

<id>
Packet Size

<id>
Packet Alignment

OpCommitReadPipe

Indicates that all reads to Num Packets associated with the reservation specified by Reserve Id and the pipe object specified by Pipe are completed.

Pipe must have a type of OpTypePipe with ReadOnly access qualifier.

Reserve Id must have a type of OpTypeReserveId.

Packet Size must be a 32-bit integer type scalar that represents the size in bytes of each packet in the pipe.

Packet Alignment must be a 32-bit integer type scalar that presents the alignment in bytes of each packet in the pipe.

Packet Size and Packet Alignment must satisfy the following:
- 1 <= Packet Alignment <= Packet Size.
- Packet Alignment must evenly divide Packet Size

For concrete types, Packet Alignment should equal Packet Size. For aggregate types, Packet Alignment should be the size of the largest primitive type in the hierarchy of types.

Capability:
Pipes

5

280

<id>
Pipe

<id>
Reserve Id

<id>
Packet Size

<id>
Packet Alignment

OpCommitWritePipe

Indicates that all writes to Num Packets associated with the reservation specified by Reserve Id and the pipe object specified by Pipe are completed.

Pipe must have a type of OpTypePipe with WriteOnly access qualifier.

Reserve Id must have a type of OpTypeReserveId.

Packet Size must be a 32-bit integer type scalar that represents the size in bytes of each packet in the pipe.

Packet Alignment must be a 32-bit integer type scalar that presents the alignment in bytes of each packet in the pipe.

Packet Size and Packet Alignment must satisfy the following:
- 1 <= Packet Alignment <= Packet Size.
- Packet Alignment must evenly divide Packet Size

For concrete types, Packet Alignment should equal Packet Size. For aggregate types, Packet Alignment should be the size of the largest primitive type in the hierarchy of types.

Capability:
Pipes

5

281

<id>
Pipe

<id>
Reserve Id

<id>
Packet Size

<id>
Packet Alignment

OpIsValidReserveId

Return true if Reserve Id is a valid reservation id and false otherwise.

Result Type must be a Boolean type.

Reserve Id must have a type of OpTypeReserveId.

Capability:
Pipes

4

282

<id>
Result Type

Result <id>

<id>
Reserve Id

OpGetNumPipePackets

Result is the number of available entries in the pipe object specified by Pipe. The number of available entries in a pipe is a dynamic value. The value returned should be considered immediately stale.

Result Type must be a 32-bit integer type scalar, which should be treated as an unsigned value.

Pipe must have a type of OpTypePipe with ReadOnly or WriteOnly access qualifier.

Packet Size must be a 32-bit integer type scalar that represents the size in bytes of each packet in the pipe.

Packet Alignment must be a 32-bit integer type scalar that presents the alignment in bytes of each packet in the pipe.

Packet Size and Packet Alignment must satisfy the following:
- 1 <= Packet Alignment <= Packet Size.
- Packet Alignment must evenly divide Packet Size

For concrete types, Packet Alignment should equal Packet Size. For aggregate types, Packet Alignment should be the size of the largest primitive type in the hierarchy of types.

Capability:
Pipes

6

283

<id>
Result Type

Result <id>

<id>
Pipe

<id>
Packet Size

<id>
Packet Alignment

OpGetMaxPipePackets

Result is the maximum number of packets specified when the pipe object specified by Pipe was created.

Result Type must be a 32-bit integer type scalar, which should be treated as an unsigned value.

Pipe must have a type of OpTypePipe with ReadOnly or WriteOnly access qualifier.

Packet Size must be a 32-bit integer type scalar that represents the size in bytes of each packet in the pipe.

Packet Alignment must be a 32-bit integer type scalar that presents the alignment in bytes of each packet in the pipe.

Packet Size and Packet Alignment must satisfy the following:
- 1 <= Packet Alignment <= Packet Size.
- Packet Alignment must evenly divide Packet Size

For concrete types, Packet Alignment should equal Packet Size. For aggregate types, Packet Alignment should be the size of the largest primitive type in the hierarchy of types.

Capability:
Pipes

6

284

<id>
Result Type

Result <id>

<id>
Pipe

<id>
Packet Size

<id>
Packet Alignment

OpGroupReserveReadPipePackets

Reserve Num Packets entries for reading from the pipe object specified by Pipe at group level. Result is a valid reservation id if the reservation is successful.

The reserved pipe entries are referred to by indices that go from 0 … Num Packets - 1.

All invocations of this module within Execution must reach this point of execution.

Behavior is undefined if this instruction is used in control flow that is non-uniform within Execution.

Result Type must be an OpTypeReserveId.

Execution must be Workgroup or Subgroup Scope.

Pipe must have a type of OpTypePipe with ReadOnly access qualifier.

Num Packets must be a 32-bit integer type scalar, which is treated as an unsigned value.

Packet Size must be a 32-bit integer type scalar that represents the size in bytes of each packet in the pipe.

Packet Alignment must be a 32-bit integer type scalar that presents the alignment in bytes of each packet in the pipe.

Packet Size and Packet Alignment must satisfy the following:
- 1 <= Packet Alignment <= Packet Size.
- Packet Alignment must evenly divide Packet Size

For concrete types, Packet Alignment should equal Packet Size. For aggregate types, Packet Alignment should be the size of the largest primitive type in the hierarchy of types.

Capability:
Pipes

8

285

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Pipe

<id>
Num Packets

<id>
Packet Size

<id>
Packet Alignment

OpGroupReserveWritePipePackets

Reserve Num Packets entries for writing to the pipe object specified by Pipe at group level. Result is a valid reservation ID if the reservation is successful.

The reserved pipe entries are referred to by indices that go from 0 … Num Packets - 1.

All invocations of this module within Execution must reach this point of execution.

Behavior is undefined if this instruction is used in control flow that is non-uniform within Execution.

Result Type must be an OpTypeReserveId.

Execution must be Workgroup or Subgroup Scope.

Pipe must have a type of OpTypePipe with WriteOnly access qualifier.

Num Packets must be a 32-bit integer type scalar, which is treated as an unsigned value.

Packet Size must be a 32-bit integer type scalar that represents the size in bytes of each packet in the pipe.

Packet Alignment must be a 32-bit integer type scalar that presents the alignment in bytes of each packet in the pipe.

Packet Size and Packet Alignment must satisfy the following:
- 1 <= Packet Alignment <= Packet Size.
- Packet Alignment must evenly divide Packet Size

For concrete types, Packet Alignment should equal Packet Size. For aggregate types, Packet Alignment should be the size of the largest primitive type in the hierarchy of types.

Capability:
Pipes

8

286

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Pipe

<id>
Num Packets

<id>
Packet Size

<id>
Packet Alignment

OpGroupCommitReadPipe

A group level indication that all reads to Num Packets associated with the reservation specified by Reserve Id to the pipe object specified by Pipe are completed.

All invocations of this module within Execution must reach this point of execution.

Behavior is undefined if this instruction is used in control flow that is non-uniform within Execution.

Execution must be Workgroup or Subgroup Scope.

Pipe must have a type of OpTypePipe with ReadOnly access qualifier.

Reserve Id must have a type of OpTypeReserveId.

Packet Size must be a 32-bit integer type scalar that represents the size in bytes of each packet in the pipe.

Packet Alignment must be a 32-bit integer type scalar that presents the alignment in bytes of each packet in the pipe.

Packet Size and Packet Alignment must satisfy the following:
- 1 <= Packet Alignment <= Packet Size.
- Packet Alignment must evenly divide Packet Size

For concrete types, Packet Alignment should equal Packet Size. For aggregate types, Packet Alignment should be the size of the largest primitive type in the hierarchy of types.

Capability:
Pipes

6

287

Scope <id>
Execution

<id>
Pipe

<id>
Reserve Id

<id>
Packet Size

<id>
Packet Alignment

OpGroupCommitWritePipe

A group level indication that all writes to Num Packets associated with the reservation specified by Reserve Id to the pipe object specified by Pipe are completed.

All invocations of this module within Execution must reach this point of execution.

Behavior is undefined if this instruction is used in control flow that is non-uniform within Execution.

Execution must be Workgroup or Subgroup Scope.

Pipe must have a type of OpTypePipe with WriteOnly access qualifier.

Reserve Id must have a type of OpTypeReserveId.

Packet Size must be a 32-bit integer type scalar that represents the size in bytes of each packet in the pipe.

Packet Alignment must be a 32-bit integer type scalar that presents the alignment in bytes of each packet in the pipe.

Packet Size and Packet Alignment must satisfy the following:
- 1 <= Packet Alignment <= Packet Size.
- Packet Alignment must evenly divide Packet Size

For concrete types, Packet Alignment should equal Packet Size. For aggregate types, Packet Alignment should be the size of the largest primitive type in the hierarchy of types.

Capability:
Pipes

6

288

Scope <id>
Execution

<id>
Pipe

<id>
Reserve Id

<id>
Packet Size

<id>
Packet Alignment

OpConstantPipeStorage

Creates a pipe-storage object.

Result Type must be OpTypePipeStorage.

Packet Size must be a 32-bit integer type scalar that represents the size in bytes of each packet in the pipe.

Packet Alignment must be a 32-bit integer type scalar that presents the alignment in bytes of each packet in the pipe.

Packet Size and Packet Alignment must satisfy the following:
- 1 <= Packet Alignment <= Packet Size.
- Packet Alignment must evenly divide Packet Size

For concrete types, Packet Alignment should equal Packet Size. For aggregate types, Packet Alignment should be the size of the largest primitive type in the hierarchy of types.

Capacity is the minimum number of Packet Size blocks the resulting OpTypePipeStorage can hold.

Capability:
PipeStorage

Missing before version 1.1.

6

323

<id>
Result Type

Result <id>

Literal Number
Packet Size

Literal Number
Packet Alignment

Literal Number
Capacity

OpCreatePipeFromPipeStorage

Creates a pipe object from a pipe-storage object.

Result Type must be OpTypePipe.

Pipe Storage must be a pipe-storage object created from OpConstantPipeStorage.

Qualifier is the pipe access qualifier.

Capability:
PipeStorage

Missing before version 1.1.

4

324

<id>
Result Type

Result <id>

<id>
Pipe Storage

3.32.24. Non-Uniform Instructions

OpGroupNonUniformElect

Result is true only in the active invocation with the lowest id in the group, otherwise result is false.

Result Type must be a Boolean type.

Execution must be Workgroup or Subgroup Scope.

Capability:
GroupNonUniform

Missing before version 1.3.

4

333

<id>
Result Type

Result <id>

Scope <id>
Execution

OpGroupNonUniformAll

Evaluates a predicate for all active invocations in the group, resulting in true if predicate evaluates to true for all active invocations in the group, otherwise the result is false.

Result Type must be a Boolean type.

Execution must be Workgroup or Subgroup Scope.

Predicate must be a Boolean type.

Capability:
GroupNonUniformVote

Missing before version 1.3.

5

334

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Predicate

OpGroupNonUniformAny

Evaluates a predicate for all active invocations in the group, resulting in true if predicate evaluates to true for any active invocation in the group, otherwise the result is false.

Result Type must be a Boolean type.

Execution must be Workgroup or Subgroup Scope.

Predicate must be a Boolean type.

Capability:
GroupNonUniformVote

Missing before version 1.3.

5

335

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Predicate

OpGroupNonUniformAllEqual

Evaluates a value for all active invocations in the group. The result is true if Value is equal for all active invocations in the group. Otherwise, the result is false.

Result Type must be a Boolean type.

Execution must be Workgroup or Subgroup Scope.

Value must be a scalar or vector of floating-point type, integer type, or Boolean type. The compare operation is based on this type, and when it is a floating-point type, an ordered-and-equal compare is used.

Capability:
GroupNonUniformVote

Missing before version 1.3.

5

336

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Value

OpGroupNonUniformBroadcast

Return the Value of the invocation identified by the id Id to all active invocations in the group.

Result Type must be a scalar or vector of floating-point type, integer type, or Boolean type.

Execution must be Workgroup or Subgroup Scope.

The type of Value must be the same as Result Type.

Id must be a scalar of integer type, whose Signedness operand is 0.

Id must come from a constant instruction.

The resulting value is undefined if Id is an inactive invocation, or is greater than or equal to the size of the group.

Capability:
GroupNonUniformBallot

Missing before version 1.3.

6

337

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Value

<id>
Id

OpGroupNonUniformBroadcastFirst

Return the Value of the invocation from the active invocation with the lowest id in the group to all active invocations in the group.

Result Type must be a scalar or vector of floating-point type, integer type, or Boolean type.

Execution must be Workgroup or Subgroup Scope.

The type of Value must be the same as Result Type.

Capability:
GroupNonUniformBallot

Missing before version 1.3.

5

338

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Value

OpGroupNonUniformBallot

Returns a bitfield value combining the Predicate value from all invocations in the group that execute the same dynamic instance of this instruction. The bit is set to one if the corresponding invocation is active and the Predicate for that invocation evaluated to true; otherwise, it is set to zero.

Result Type must be a vector of four components of integer type scalar, whose Signedness operand is 0.

Result is a set of bitfields where the first invocation is represented in the lowest bit of the first vector component and the last (up to the size of the group) is the higher bit number of the last bitmask needed to represent all bits of the group invocations.

Execution must be Workgroup or Subgroup Scope.

Predicate must be a Boolean type.

Capability:
GroupNonUniformBallot

Missing before version 1.3.

5

339

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Predicate

OpGroupNonUniformInverseBallot

Evaluates a value for all active invocations in the group, resulting in true if the bit in Value for the corresponding invocation is set to one, otherwise the result is false.

Result Type must be a Boolean type.

Execution must be Workgroup or Subgroup Scope.

Value must be a vector of four components of integer type scalar, whose Signedness operand is 0.

Value must be the same for all invocations that execute the same dynamic instance of this instruction.

Value is a set of bitfields where the first invocation is represented in the lowest bit of the first vector component and the last (up to the size of the group) is the higher bit number of the last bitmask needed to represent all bits of the group invocations.

Capability:
GroupNonUniformBallot

Missing before version 1.3.

5

340

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Value

OpGroupNonUniformBallotBitExtract

Evaluates a value for all active invocations in the group, resulting in true if the bit in Value that corresponds to Index is set to one, otherwise the result is false.

Result Type must be a Boolean type.

Execution must be Workgroup or Subgroup Scope.

Value must be a vector of four components of integer type scalar, whose Signedness operand is 0.

Value is a set of bitfields where the first invocation is represented in the lowest bit of the first vector component and the last (up to the size of the group) is the higher bit number of the last bitmask needed to represent all bits of the group invocations.

Index must be a scalar of integer type, whose Signedness operand is 0.

The resulting value is undefined if Index is greater than or equal to the size of the group.

Capability:
GroupNonUniformBallot

Missing before version 1.3.

6

341

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Value

<id>
Index

OpGroupNonUniformBallotBitCount

A group operation that returns the number of bits that are set to 1 in Value, only considering the bits in Value required to represent all bits of the group’s invocations.

Result Type must be a scalar of integer type, whose Signedness operand is 0.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is 0.

Value must be a vector of four components of integer type scalar, whose Signedness operand is 0.

Value is a set of bitfields where the first invocation is represented in the lowest bit of the first vector component and the last (up to the size of the group) is the higher bit number of the last bitmask needed to represent all bits of the group invocations.

Capability:
GroupNonUniformBallot

Missing before version 1.3.

6

342

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
Value

OpGroupNonUniformBallotFindLSB

Find the least significant bit set to 1 in Value, considering only the bits in Value required to represent all bits of the group’s invocations. If none of the considered bits is set to 1, the result is undefined.

Result Type must be a scalar of integer type, whose Signedness operand is 0.

Execution must be Workgroup or Subgroup Scope.

Value must be a vector of four components of integer type scalar, whose Signedness operand is 0.

Value is a set of bitfields where the first invocation is represented in the lowest bit of the first vector component and the last (up to the size of the group) is the higher bit number of the last bitmask needed to represent all bits of the group invocations.

Capability:
GroupNonUniformBallot

Missing before version 1.3.

5

343

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Value

OpGroupNonUniformBallotFindMSB

Find the most significant bit set to 1 in Value, considering only the bits in Value required to represent all bits of the group’s invocations. If none of the considered bits is set to 1, the result is undefined.

Result Type must be a scalar of integer type, whose Signedness operand is 0.

Execution must be Workgroup or Subgroup Scope.

Value must be a vector of four components of integer type scalar, whose Signedness operand is 0.

Value is a set of bitfields where the first invocation is represented in the lowest bit of the first vector component and the last (up to the size of the group) is the higher bit number of the last bitmask needed to represent all bits of the group invocations.

Capability:
GroupNonUniformBallot

Missing before version 1.3.

5

344

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Value

OpGroupNonUniformShuffle

Return the Value of the invocation identified by the id Id.

Result Type must be a scalar or vector of floating-point type, integer type, or Boolean type.

Execution must be Workgroup or Subgroup Scope.

The type of Value must be the same as Result Type.

Id must be a scalar of integer type, whose Signedness operand is 0.

The resulting value is undefined if Id is an inactive invocation, or is greater than or equal to the size of the group.

Capability:
GroupNonUniformShuffle

Missing before version 1.3.

6

345

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Value

<id>
Id

OpGroupNonUniformShuffleXor

Return the Value of the invocation identified by the current invocation’s id within the group xor’ed with Mask.

Result Type must be a scalar or vector of floating-point type, integer type, or Boolean type.

Execution must be Workgroup or Subgroup Scope.

The type of Value must be the same as Result Type.

Mask must be a scalar of integer type, whose Signedness operand is 0.

The resulting value is undefined if current invocation’s id within the group xor’ed with Mask is an inactive invocation, or is greater than or equal to the size of the group.

Capability:
GroupNonUniformShuffle

Missing before version 1.3.

6

346

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Value

<id>
Mask

OpGroupNonUniformShuffleUp

Return the Value of the invocation identified by the current invocation’s id within the group - Delta.

Result Type must be a scalar or vector of floating-point type, integer type, or Boolean type.

Execution must be Workgroup or Subgroup Scope.

The type of Value must be the same as Result Type.

Delta must be a scalar of integer type, whose Signedness operand is 0.

Delta is treated as unsigned and the resulting value is undefined if Delta is greater than the current invocation’s id within the group or if the selected lane is inactive.

Capability:
GroupNonUniformShuffleRelative

Missing before version 1.3.

6

347

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Value

<id>
Delta

OpGroupNonUniformShuffleDown

Return the Value of the invocation identified by the current invocation’s id within the group + Delta.

Result Type must be a scalar or vector of floating-point type, integer type, or Boolean type.

Execution must be Workgroup or Subgroup Scope.

The type of Value must be the same as Result Type.

Delta must be a scalar of integer type, whose Signedness operand is 0.

Delta is treated as unsigned and the resulting value is undefined if Delta is greater than or equal to the size of the group, or if the current invocation’s id within the group + Delta is either an inactive invocation or greater than or equal to the size of the group.

Capability:
GroupNonUniformShuffleRelative

Missing before version 1.3.

6

348

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Value

<id>
Delta

OpGroupNonUniformIAdd

An integer add group operation of all Value operands contributed active by invocations in the group.

Result Type must be a scalar or vector of integer type.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is 0. If Operation is ClusteredReduce, ClusterSize must be specified.

The type of Value must be the same as Result Type.

ClusterSize is the size of cluster to use. ClusterSize must be a scalar of integer type, whose Signedness operand is 0. ClusterSize must come from a constant instruction. ClusterSize must be at least 1, and must be a power of 2. If ClusterSize is greater than the declared SubGroupSize, executing this instruction results in undefined behavior.

Capability:
GroupNonUniformArithmetic, GroupNonUniformClustered, GroupNonUniformPartitionedNV

Missing before version 1.3.

6 + variable

349

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
Value

Optional
<id>
ClusterSize

OpGroupNonUniformFAdd

A floating point add group operation of all Value operands contributed by active invocations in the group.

Result Type must be a scalar or vector of floating-point type.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is 0. If Operation is ClusteredReduce, ClusterSize must be specified.

The type of Value must be the same as Result Type. The method used to perform the group operation on the contributed Value(s) from active invocations is implementation defined.

ClusterSize is the size of cluster to use. ClusterSize must be a scalar of integer type, whose Signedness operand is 0. ClusterSize must come from a constant instruction. ClusterSize must be at least 1, and must be a power of 2. If ClusterSize is greater than the declared SubGroupSize, executing this instruction results in undefined behavior.

Capability:
GroupNonUniformArithmetic, GroupNonUniformClustered, GroupNonUniformPartitionedNV

Missing before version 1.3.

6 + variable

350

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
Value

Optional
<id>
ClusterSize

OpGroupNonUniformIMul

An integer multiply group operation of all Value operands contributed by active invocations in the group.

Result Type must be a scalar or vector of integer type.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is 1. If Operation is ClusteredReduce, ClusterSize must be specified.

The type of Value must be the same as Result Type.

ClusterSize is the size of cluster to use. ClusterSize must be a scalar of integer type, whose Signedness operand is 0. ClusterSize must come from a constant instruction. ClusterSize must be at least 1, and must be a power of 2. If ClusterSize is greater than the declared SubGroupSize, executing this instruction results in undefined behavior.

Capability:
GroupNonUniformArithmetic, GroupNonUniformClustered, GroupNonUniformPartitionedNV

Missing before version 1.3.

6 + variable

351

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
Value

Optional
<id>
ClusterSize

OpGroupNonUniformFMul

A floating point multiply group operation of all Value operands contributed by active invocations in the group.

Result Type must be a scalar or vector of floating-point type.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is 1. If Operation is ClusteredReduce, ClusterSize must be specified.

The type of Value must be the same as Result Type. The method used to perform the group operation on the contributed Value(s) from active invocations is implementation defined.

ClusterSize is the size of cluster to use. ClusterSize must be a scalar of integer type, whose Signedness operand is 0. ClusterSize must come from a constant instruction. ClusterSize must be at least 1, and must be a power of 2. If ClusterSize is greater than the declared SubGroupSize, executing this instruction results in undefined behavior.

Capability:
GroupNonUniformArithmetic, GroupNonUniformClustered, GroupNonUniformPartitionedNV

Missing before version 1.3.

6 + variable

352

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
Value

Optional
<id>
ClusterSize

OpGroupNonUniformSMin

A signed integer minimum group operation of all Value operands contributed by active invocations in the group.

Result Type must be a scalar or vector of integer type.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is INT_MAX. If Operation is ClusteredReduce, ClusterSize must be specified.

The type of Value must be the same as Result Type.

ClusterSize is the size of cluster to use. ClusterSize must be a scalar of integer type, whose Signedness operand is 0. ClusterSize must come from a constant instruction. ClusterSize must be at least 1, and must be a power of 2. If ClusterSize is greater than the declared SubGroupSize, executing this instruction results in undefined behavior.

Capability:
GroupNonUniformArithmetic, GroupNonUniformClustered, GroupNonUniformPartitionedNV

Missing before version 1.3.

6 + variable

353

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
Value

Optional
<id>
ClusterSize

OpGroupNonUniformUMin

An unsigned integer minimum group operation of all Value operands contributed by active invocations in the group.

Result Type must be a scalar or vector of integer type, whose Signedness operand is 0.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is UINT_MAX. If Operation is ClusteredReduce, ClusterSize must be specified.

The type of Value must be the same as Result Type.

ClusterSize is the size of cluster to use. ClusterSize must be a scalar of integer type, whose Signedness operand is 0. ClusterSize must come from a constant instruction. ClusterSize must be at least 1, and must be a power of 2. If ClusterSize is greater than the declared SubGroupSize, executing this instruction results in undefined behavior.

Capability:
GroupNonUniformArithmetic, GroupNonUniformClustered, GroupNonUniformPartitionedNV

Missing before version 1.3.

6 + variable

354

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
Value

Optional
<id>
ClusterSize

OpGroupNonUniformFMin

A floating point minimum group operation of all Value operands contributed by active invocations in the group.

Result Type must be a scalar or vector of floating-point type.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is +INF. If Operation is ClusteredReduce, ClusterSize must be specified.

The type of Value must be the same as Result Type. The method used to perform the group operation on the contributed Value(s) from active invocations is implementation defined. From the set of Value(s) provided by active invocations within a subgroup, if for any two Values one of them is a NaN, the other is chosen. If all Value(s) that are used by the current invocation are NaN, then the result is an undefined value.

ClusterSize is the size of cluster to use. ClusterSize must be a scalar of integer type, whose Signedness operand is 0. ClusterSize must come from a constant instruction. ClusterSize must be at least 1, and must be a power of 2. If ClusterSize is greater than the declared SubGroupSize, executing this instruction results in undefined behavior.

Capability:
GroupNonUniformArithmetic, GroupNonUniformClustered, GroupNonUniformPartitionedNV

Missing before version 1.3.

6 + variable

355

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
Value

Optional
<id>
ClusterSize

OpGroupNonUniformSMax

A signed integer maximum group operation of all Value operands contributed by active invocations in the group.

Result Type must be a scalar or vector of integer type.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is INT_MIN. If Operation is ClusteredReduce, ClusterSize must be specified.

The type of Value must be the same as Result Type.

ClusterSize is the size of cluster to use. ClusterSize must be a scalar of integer type, whose Signedness operand is 0. ClusterSize must come from a constant instruction. ClusterSize must be at least 1, and must be a power of 2. If ClusterSize is greater than the declared SubGroupSize, executing this instruction results in undefined behavior.

Capability:
GroupNonUniformArithmetic, GroupNonUniformClustered, GroupNonUniformPartitionedNV

Missing before version 1.3.

6 + variable

356

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
Value

Optional
<id>
ClusterSize

OpGroupNonUniformUMax

An unsigned integer maximum group operation of all Value operands contributed by active invocations in the group.

Result Type must be a scalar or vector of integer type, whose Signedness operand is 0.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is 0. If Operation is ClusteredReduce, ClusterSize must be specified.

The type of Value must be the same as Result Type.

ClusterSize is the size of cluster to use. ClusterSize must be a scalar of integer type, whose Signedness operand is 0. ClusterSize must come from a constant instruction. ClusterSize must be at least 1, and must be a power of 2. If ClusterSize is greater than the declared SubGroupSize, executing this instruction results in undefined behavior.

Capability:
GroupNonUniformArithmetic, GroupNonUniformClustered, GroupNonUniformPartitionedNV

Missing before version 1.3.

6 + variable

357

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
Value

Optional
<id>
ClusterSize

OpGroupNonUniformFMax

A floating point maximum group operation of all Value operands contributed by active invocations in by group.

Result Type must be a scalar or vector of floating-point type.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is -INF. If Operation is ClusteredReduce, ClusterSize must be specified.

The type of Value must be the same as Result Type. The method used to perform the group operation on the contributed Value(s) from active invocations is implementation defined. From the set of Value(s) provided by active invocations within a subgroup, if for any two Values one of them is a NaN, the other is chosen. If all Value(s) that are used by the current invocation are NaN, then the result is an undefined value.

ClusterSize is the size of cluster to use. ClusterSize must be a scalar of integer type, whose Signedness operand is 0. ClusterSize must come from a constant instruction. ClusterSize must be at least 1, and must be a power of 2. If ClusterSize is greater than the declared SubGroupSize, executing this instruction results in undefined behavior.

Capability:
GroupNonUniformArithmetic, GroupNonUniformClustered, GroupNonUniformPartitionedNV

Missing before version 1.3.

6 + variable

358

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
Value

Optional
<id>
ClusterSize

OpGroupNonUniformBitwiseAnd

A bitwise and group operation of all Value operands contributed by active invocations in the group.

Result Type must be a scalar or vector of integer type.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is ~0. If Operation is ClusteredReduce, ClusterSize must be specified.

The type of Value must be the same as Result Type.

ClusterSize is the size of cluster to use. ClusterSize must be a scalar of integer type, whose Signedness operand is 0. ClusterSize must come from a constant instruction. ClusterSize must be at least 1, and must be a power of 2. If ClusterSize is greater than the declared SubGroupSize, executing this instruction results in undefined behavior.

Capability:
GroupNonUniformArithmetic, GroupNonUniformClustered, GroupNonUniformPartitionedNV

Missing before version 1.3.

6 + variable

359

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
Value

Optional
<id>
ClusterSize

OpGroupNonUniformBitwiseOr

A bitwise or group operation of all Value operands contributed by active invocations in the group.

Result Type must be a scalar or vector of integer type.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is 0. If Operation is ClusteredReduce, ClusterSize must be specified.

The type of Value must be the same as Result Type.

ClusterSize is the size of cluster to use. ClusterSize must be a scalar of integer type, whose Signedness operand is 0. ClusterSize must come from a constant instruction. ClusterSize must be at least 1, and must be a power of 2. If ClusterSize is greater than the declared SubGroupSize, executing this instruction results in undefined behavior.

Capability:
GroupNonUniformArithmetic, GroupNonUniformClustered, GroupNonUniformPartitionedNV

Missing before version 1.3.

6 + variable

360

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
Value

Optional
<id>
ClusterSize

OpGroupNonUniformBitwiseXor

A bitwise xor group operation of all Value operands contributed by active invocations in the group.

Result Type must be a scalar or vector of integer type.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is 0. If Operation is ClusteredReduce, ClusterSize must be specified.

The type of Value must be the same as Result Type.

ClusterSize is the size of cluster to use. ClusterSize must be a scalar of integer type, whose Signedness operand is 0. ClusterSize must come from a constant instruction. ClusterSize must be at least 1, and must be a power of 2. If ClusterSize is greater than the declared SubGroupSize, executing this instruction results in undefined behavior.

Capability:
GroupNonUniformArithmetic, GroupNonUniformClustered, GroupNonUniformPartitionedNV

Missing before version 1.3.

6 + variable

361

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
Value

Optional
<id>
ClusterSize

OpGroupNonUniformLogicalAnd

A logical and group operation of all Value operands contributed by active invocations in the group.

Result Type must be a scalar or vector of Boolean type.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is ~0. If Operation is ClusteredReduce, ClusterSize must be specified.

The type of Value must be the same as Result Type.

ClusterSize is the size of cluster to use. ClusterSize must be a scalar of integer type, whose Signedness operand is 0. ClusterSize must come from a constant instruction. ClusterSize must be at least 1, and must be a power of 2. If ClusterSize is greater than the declared SubGroupSize, executing this instruction results in undefined behavior.

Capability:
GroupNonUniformArithmetic, GroupNonUniformClustered, GroupNonUniformPartitionedNV

Missing before version 1.3.

6 + variable

362

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
Value

Optional
<id>
ClusterSize

OpGroupNonUniformLogicalOr

A logical or group operation of all Value operands contributed by active invocations in the group.

Result Type must be a scalar or vector of Boolean type.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is 0. If Operation is ClusteredReduce, ClusterSize must be specified.

The type of Value must be the same as Result Type.

ClusterSize is the size of cluster to use. ClusterSize must be a scalar of integer type, whose Signedness operand is 0. ClusterSize must come from a constant instruction. ClusterSize must be at least 1, and must be a power of 2. If ClusterSize is greater than the declared SubGroupSize, executing this instruction results in undefined behavior.

Capability:
GroupNonUniformArithmetic, GroupNonUniformClustered, GroupNonUniformPartitionedNV

Missing before version 1.3.

6 + variable

363

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
Value

Optional
<id>
ClusterSize

OpGroupNonUniformLogicalXor

A logical xor group operation of all Value operands contributed by active invocations in the group.

Result Type must be a scalar or vector of Boolean type.

Execution must be Workgroup or Subgroup Scope.

The identity I for Operation is 0. If Operation is ClusteredReduce, ClusterSize must be specified.

The type of Value must be the same as Result Type.

ClusterSize is the size of cluster to use. ClusterSize must be a scalar of integer type, whose Signedness operand is 0. ClusterSize must come from a constant instruction. ClusterSize must be at least 1, and must be a power of 2. If ClusterSize is greater than the declared SubGroupSize, executing this instruction results in undefined behavior.

Capability:
GroupNonUniformArithmetic, GroupNonUniformClustered, GroupNonUniformPartitionedNV

Missing before version 1.3.

6 + variable

364

<id>
Result Type

Result <id>

Scope <id>
Execution

Group Operation
Operation

<id>
Value

Optional
<id>
ClusterSize

OpGroupNonUniformQuadBroadcast

Return the Value of the invocation within the quad whose SubgroupLocalInvocationId % 4 is equal to Index.

Result Type must be a scalar or vector of floating-point type, integer type, or Boolean type.

Execution must be Workgroup or Subgroup Scope.

The type of Value must be the same as Result Type.

Index must be a scalar of integer type, whose Signedness operand is 0.

Index must come from a constant instruction.

If the value of Index is greater or equal to 4, an undefined result is returned.

Capability:
GroupNonUniformQuad

Missing before version 1.3.

6

365

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Value

<id>
Index

OpGroupNonUniformQuadSwap

Swap the Value of the invocation within the quad with another invocation in the quad using Direction.

Result Type must be a scalar or vector of floating-point type, integer type, or Boolean type.

Execution must be Workgroup or Subgroup Scope.

The type of Value must be the same as Result Type.

Direction is the kind of swap to perform.

Direction must be a scalar of integer type, whose Signedness operand is 0.

Direction must come from a constant instruction.

The value of Direction is evaluated such that:
0 indicates a horizontal swap within the quad.
1 indicates a vertical swap within the quad.
2 indicates a diagonal swap within the quad.

Capability:
GroupNonUniformQuad

Missing before version 1.3.

6

366

<id>
Result Type

Result <id>

Scope <id>
Execution

<id>
Value

<id>
Direction

OpGroupNonUniformPartitionNV

TBD

Capability:
GroupNonUniformPartitionedNV

Reserved.

4

5296

<id>
Result Type

Result <id>

<id>
Value

4. Appendix A: Changes

4.1. Changes from Version 0.99, Revision 31

  • Added the PushConstant Storage Class.

  • Added OpIAddCarry, OpISubBorrow, OpUMulExtended, and OpSMulExtended.

  • Added OpInBoundsPtrAccessChain.

  • Added the Decoration NoContraction to prevent combining multiple operations into a single operation (bug 14396).

  • Added sparse texturing (14486):

    • Added OpImageSparse… for accessing images that might not be resident.

    • Added MinLod functionality for accessing images with a minimum level of detail.

  • Added back the Alignment Decoration, for the Kernel capability (14505).

  • Added a NonTemporal Memory Access (14566).

  • Structured control flow changes:

    • Changed structured loops to have a structured continue Continue Target in OpLoopMerge (14422).

    • Added rules for how "fall through" works with OpSwitch (13579).

    • Added definitions for what is "inside" a structured control-flow construct (14422).

  • Added SubpassData Dim to support input targets written by a previous subpass as an output target (14304). This is also a Decoration and a Capability, and can be used by some image ops to read the input target.

  • Added OpTypeForwardPointer to establish the Storage Class of a forward reference to a pointer type (13822).

  • Improved Debuggability

    • Changed OpLine to not have a target <id>, but instead be placed immediately preceding the instruction(s) it is annotating (13905).

    • Added OpNoLine to terminate the affect of OpLine (13905).

    • Changed OpSource to include the source code:

      • Allow multiple occurrences.

      • Be mixed in with the OpString instructions.

      • Optionally consume an OpString result to say which file it is annotating.

      • Optionally include the source text corresponding to that OpString.

      • Included adding OpSourceContinued for source text that is too long for a single instruction.

  • Added a large number of Capabilities for subsetting functionality (14520, 14453), including 8-bit integer support for OpenCL kernels.

  • Added VertexIndex and InstanceIndex BuiltIn Decorations (14255).

  • Added GenericPointer capability that allows the ability to use the Generic Storage Class (14287).

  • Added IndependentForwardProgress Execution Mode (14271).

  • Added OpAtomicFlagClear and OpAtomicFlagTestAndSet instructions (14315).

  • Changed OpEntryPoint to take a list of Input and Output <id> for declaring the entry point’s interface.

  • Fixed internal bugs

    • 14411 Added missing documentation for mad_sat OpenCL extended instructions (enums existed, just the documentation was missing)

    • 14241 Removed shader capability requirement from OpImageQueryLevels and OpImageQuerySamples.

    • 14241 Removed unneeded OpImageQueryDim instruction.

    • 14241 Filled in TBD section for OpAtomicCompareExchangeWeek

    • 14366 All OpSampledImage must appear before uses of sampled images (and still in the first block of the entry point).

    • 14450 DeviceEnqueue capability is required for OpTypeQueue and OpTypeDeviceEvent

    • 14363 OpTypePipe is opaque - moved packet size and alignment to opcodes

    • 14367 Float16Buffer capability clarified

    • 14241 Clarified how OpSampledImage can be used

    • 14402 Clarified OpTypeImage encodings for OpenCL extended instructions

    • 14569 Removed mention of non-existent OpFunctionDecl

    • 14372 Clarified usage of OpGenericPtrMemSemantics

    • 13801 Clarified the SpecId Decoration is just for constants

    • 14447 Changed literal values of Memory Semantic enums to match OpenCL/C++11 atomics, and made the Memory Semantic None and Relaxed be aliases

    • 14637 Removed subgroup scope from OpGroupAsyncCopy and OpGroupWaitEvents

4.2. Changes from Version 0.99, Revision 32

  • Added UnormInt101010_2 to the Image Channel Data Type table.

  • Added place holder for C++11 atomic Consume Memory Semantics along with an explicit AcquireRelease memory semantic.

  • Fixed internal bugs:

    • 14690 OpSwitch literal width (and hence number of operands) is determined by the type of Selector, and be rigorous about how sub-32-bit literals are stored.

    • 14485 The client API owns the semantics of built-ins that only have "pass through" semantics WRT SPIR-V.

    • 14862 Removed the IndependentForwardProgress Execution Mode.

  • Fixed public bugs:

4.3. Changes from Version 1.00, Revision 1

  • Adjusted Capabilities:

    • Split geometry-stream functionality into its own GeometryStreams capability (14873).

    • Have InputAttachmentIndex to depend on InputAttachment instead of Shader (14797).

    • Merge AdvancedFormats and StorageImageExtendedFormats into just StorageImageExtendedFormats (14824).

    • Require StorageImageReadWithoutFormat and StorageImageWriteWithoutFormat to read and write storage images with an Unknown Image Format.

    • Removed the ImageSRGBWrite capability.

  • Clarifications

    • RelaxedPrecision Decoration can be applied to OpFunction (14662).

  • Fixed internal bugs:

    • 14797 The literal argument was missing for the InputAttachmentIndex Decoration.

    • 14547 Remove the FragColor BuiltIn, so that no implicit broadcast is implied.

    • 13292 Make statements about "Volatile" be more consistent with the memory model specification (non-functional change).

    • 14948 Remove image-"Query" overloading on image/sampled-image type and "fetch" on non-sampled images, by adding the OpImage instruction to get the image from a sampled image.

    • 14949 Make consistent placement between OpSource and OpSourceExtension in the logical layout of a module.

    • 14865 Merge WorkgroupLinearId with LocalInvocationId BuiltIn Decorations.

    • 14806 Include 3D images for OpImageQuerySize.

    • 14325 Removed the Smooth Decoration.

    • 12771 Make the version word formatted as: "0 | Major Number | Minor Number | 0" in the physical layout.

    • 15035 Allow OpTypeImage to use a Depth operand of 2 for not indicating a depth or non-depth image.

    • 15009 Split the OpenCL Source Language into two: OpenCL_C and OpenCL_CPP.

    • 14683 OpSampledImage instructions can only be the consuming block, for scalars, and directly consumed by an image lookup or query instruction.

    • 14325 mutual exclusion validation rules of Execution Modes and Decorations

    • 15112 add definitions for invocation, dynamically uniform, and uniform control flow.

  • Renames

4.4. Changes from Version 1.00, Revision 2

  • Updated example at the end of Section 1 to conform to the KHR_vulkan_glsl extension and treat OpTypeBool as an abstract type.

  • Adjusted Capabilities:

    • MatrixStride depends on Matrix (15234).

    • Sample, SampleId, SamplePosition, and SampleMask depend on SampleRateShading (15234).

    • ClipDistance and CullDistance BuiltIns depend on, respectively, ClipDistance and CullDistance (1407, 15234).

    • ViewportIndex depends on MultiViewport (15234).

    • AtomicCounterMemory should be the AtomicStorage (15234).

    • Float16 has no dependencies (15234).

    • Offset Decoration should only be for Shader (15268).

    • Generic Storage Class is supposed to need the GenericPointer Capability (14287).

    • Remove capability restriction on the BuiltIn Decoration (15248).

  • Fixed internal bugs:

  • Improved accuracy of the opcode word count in each instruction regarding which operands are optional. For sampling operations with explicit LOD, this included not marking the required LOD operands as optional.

  • Clarified that when NonWritable, NonReadable, Volatile, and Coherent Decorations are applied to the Uniform storage class, the BufferBlock decoration must be present.

  • Fixed external bugs:

4.5. Changes from Version 1.00, Revision 3

  • Added definition of derivative group, and use it to say when derivatives are well defined.

4.6. Changes from Version 1.00, Revision 4

4.7. Changes from Version 1.00, Revision 5

  • Khronos SPIR-V issue #27: Removed Shader dependency from SampledBuffer and Sampled1D Capabilities.

  • Khronos SPIR-V issue #56: Clarify that the meaning of "read-only" in the Storage Classes includes not allowing initializers.

  • Khronos SPIR-V issue #57: Clarify "modulo" means "remainder" in OpFMod's description.

  • Khronos SPIR-V issue #60: OpControlBarrier synchronizes Output variables when used in tessellation-control shader.

  • Public SPIRV-Headers issue #1: Remove the Shader capability requirement from the Input Storage Class.

  • Public SPIRV-Headers issue #10: Don’t say the (u [, v] [, w], q) has four components, as it can be closed up when the optional ones are missing. Seen in the projective image instructions.

  • Public SPIRV-Headers issues #12 and #13 and Khronos SPIR-V issue #65: Allow OpVariable as an initializer for another OpVariable instruction or the Base of an OpSpecConstantOp with an AccessChain opcode.

  • Public SPIRV-Headers issues #14: add Max enumerants of 0x7FFFFFFF to each of the non-mask enums in the C-based header files.

4.8. Changes from Version 1.00, Revision 6

  • Khronos SPIR-V issue #63: Be clear that OpUndef can be used in sequence 9 (and is preferred to be) of the Logical Layout and can be part of partially-defined OpConstantComposite.

  • Khronos SPIR-V issue #70: Don’t explicitly require operand truncation for integer operations when operating at RelaxedPrecision.

  • Khronos SPIR-V issue #76: Include OpINotEqual in the list of allowed instructions for OpSpecConstantOp.

  • Khronos SPIR-V issue #79: Remove implication that OpImageQueryLod should have a component for the array index.

  • Public SPIRV-Headers issue #17: Decorations Noperspective, Flat, Patch, Centroid, and Sample can apply to a top-level member that is itself a structure, so don’t disallow it through restrictions to numeric types.

4.9. Changes from Version 1.00, Revision 7

4.10. Changes from Version 1.00, Revision 8

  • Add SPV_KHR_subgroup_vote tokens.

  • Typo: Change "without a sampler" to "with a sampler" for the description of the SampledBuffer Capability.

  • Khronos SPIR-V issue #61: Clarification of packet size and alignment on all instructions that use the Pipes Capability.

  • Khronos SPIR-V issue #99: Use "invalid" language to replace any "compile-time error" language.

  • Khronos SPIR-V issue #55: Distinguish between branch instructions and termination instructions.

  • Khronos SPIR-V issue #94: Add missing OpSubgroupReadInvocationKHR enumerant.

  • Khronos SPIR-V issue #114: Header blocks strictly dominate their merge blocks.

  • Khronos SPIR-V issue #119: OpSpecConstantOp allows OpUndef where allowed by its opcode.

4.11. Changes from Version 1.00, Revision 9

  • Khronos Vulkan issue #652: Remove statements about matrix offsets and padding. These are described correctly in the Vulkan API specifications.

  • Khronos SPIR-V issue #113: Remove the "By Default" statements in FP Rounding Mode. These should be properly specified by the client API.

  • Add extension enumerants for

    • SPV_KHR_16bit_storage

    • SPV_KHR_device_group

    • SPV_KHR_multiview

    • SPV_NV_sample_mask_override_coverage

    • SPV_NV_geometry_shader_passthrough

    • SPV_NV_viewport_array2

    • SPV_NV_stereo_view_rendering

    • SPV_NVX_multiview_per_view_attributes

4.12. Changes from Version 1.00, Revision 10

4.13. Changes from Version 1.00, Revision 11

  • Public issue #2: Disallow the Cube dimension from use with the Offset, ConstOffset, and ConstOffset image operands.

  • Public issue #48: OpConvertPtrToU only returns a scalar, not a vector.

  • Khronos SPIR-V issue #130: Be more clear which masks are literal and which are not.

  • Khronos SPIR-V issue #154: Clarify only one of the listed Capabilities needs to be declared to use a feature that lists multiple capabilities. The non-declared capabilities need not be supported by the underlying implementation.

  • Khronos SPIR-V issue #174: OpImageDrefGather and OpImageSparseDrefGather return vectors, not scalars.

  • Khronos SPIR-V issue #182: The SampleMask built in does not depend on SampleRateShading, only Shader.

  • Khronos SPIR-V issue #183: OpQuantizeToF16 with too-small magnitude can result in either +0 or -0.

  • Khronos SPIR-V issue #203: OpImageTexelPointer has 3 components for cube arrays, not 4.

  • Khronos SPIR-V issue #217: Clearer language for OpArrayLength.

  • Khronos SPIR-V issue #213: Image Operand LoD is not used by query operations.

  • Khronos SPIR-V issue #223: OpPhi has exactly one parent operand per parent block.

  • Khronos SPIR-V issue #212: In the Validation Rules, make clear a pointer can be an operand in an extended instruction set.

  • Add extension enumerants for

    • SPV_AMD_shader_ballot

    • SPV_KHR_post_depth_coverage

    • SPV_AMD_shader_explicit_vertex_parameter

    • SPV_EXT_shader_stencil_export

    • SPV_INTEL_subgroups

4.14. Changes from Version 1.00

4.15. Changes from Version 1.1, Revision 1

  • Incorporated bug fixes from Revision 6 of Version 1.00 (see section 4.7. Changes from Version 1.00, Revision 5).

4.16. Changes from Version 1.1, Revision 2

  • Incorporated bug fixes from Revision 7 of Version 1.00 (see section 4.8. Changes from Version 1.00, Revision 6).

4.17. Changes from Version 1.1, Revision 3

  • Incorporated bug fixes from Revision 8 of Version 1.00 (see section 4.9. Changes from Version 1.00, Revision 7).

4.18. Changes from Version 1.1, Revision 4

  • Incorporated bug fixes from Revision 9 of Version 1.00 (see section 4.10. Changes from Version 1.00, Revision 8).

4.19. Changes from Version 1.1, Revision 5

  • Incorporated changes from Revision 10 of Version 1.00 (see section 4.11. Changes from Version 1.00, Revision 9).

4.20. Changes from Version 1.1, Revision 6

  • Incorporated changes from Revision 11 of Version 1.00 (see section 4.12. Changes from Version 1.00, Revision 10).

4.21. Changes from Version 1.1, Revision 7

  • Incorporated changes from Revision 12 of Version 1.00 (see section 4.13. Changes from Version 1.00, Revision 11).

  • State where all OpModuleProcessed belong, in the logical layout.

4.22. Changes from Version 1.1

  • Moved version number to SPIR-V 1.2

  • New functionality:

4.23. Changes from Version 1.2, Revision 1

  • Incorporated changes from Revision 12 of Version 1.00 (see section 4.13. Changes from Version 1.00, Revision 11).

  • Incorporated changes from Revision 8 of Version 1.1 (see section 4.21. Changes from Version 1.1, Revision 7).

4.24. Changes from Version 1.2, Revision 2

  • Combine the 1.0, 1.1, and 1.2 specifications, making a unified specification. The previous 1.0, 1.1, and 1.2 specifications are replaced with this one unified specification.

4.25. Changes from Version 1.2, Revision 3

Fixed Khronos-internal issues:

4.26. Changes from Version 1.2

  • Moved version number to SPIR-V 1.3

  • New functionality:

  • Incorporated the following extensions:

    • SPV_KHR_shader_draw_parameters

    • SPV_KHR_16bit_storage

    • SPV_KHR_device_group

    • SPV_KHR_multiview

    • SPV_KHR_storage_buffer_storage_class

    • SPV_KHR_variable_pointers

  • Reserved symbols for

    • SPV_GOOGLE_decorate_string

    • SPV_GOOGLE_hlsl_functionality1

    • SPV_AMD_gpu_shader_half_float_fetch

  • Added deprecation model.

4.27. Changes from Version 1.3, Revision 1

  • Fixed Issues:

    • Public SPIRV-Headers PR #73: Add missing fields for some NVIDIA-specific tokens.

    • Khronos SPIR-V Issue #202: Shader Validation: Be clear that arrays of blocks set by the client API cannot have an ArrayStride.

    • Khronos SPIR-V Issue #210: Clarify the Result Type of OpSampledImage.

    • Khronos SPIR-V Issue #211: State that Derivative instructions only work on 32-bit width components.

    • Khronos SPIR-V Issue #239: Clarify OpImageFetch is for an image whose Sampled operand is 1.

    • Khronos SPIR-V Issue #256: OpAtomicCompareExchange does not store if comparison fails.

    • Khronos SPIR-V Issue #269: Be more clear which bits are mutually exclusive for memory semantics.

    • Khronos SPIR-V Issue #278: Delete OpTypeRuntimeArray restriction on storage classes, as this is already covered by the client API.

    • Khronos SPIR-V Issue #279:

      • Add section expository section 2.8.1 "Unsigned Versus Signed Integers".

      • As expected, OpUConvert can have vector Result Type.

    • Khronos SPIR-V Issue #280: OpImageQuerySizeLod and OpImageQueryLevels can be limited by the client API.

    • Khronos SPIR-V Issue #285: Remove Kernel as a capability implicitly declared by Int8.

    • Khronos SPIR-V Issue #290: Clarify implicit declaration of capabilities, in part by changing the column heading to *Implicitly Declares".

    • Khronos SPIR-V Issues #295: Explicitly say blocks cannot be nested in blocks, in the validation section. (This was already indirectly required.)

    • Khronos SPIR-V Issue #299: Add the ImageGatherExtended capability to ConstOffsets in the image operands section.

    • Khronos SPIR-V Issues #303 and #304: OpGroupNonUniformBallotBitExtract documentation: add Result Type and fix Index parameter.

    • Khronos SPIR-V Issue #310: Remove instruction word count from the Limits table, as it is already intrinsically limited.

    • Khronos SPIR-V Issue #313: Move the FPRoundingMode-decoration validation rule to the shader validation section (not a universal rule). Also, include the StorageBuffer storage class in this rule.

4.28. Changes from Version 1.3, Revision 2

  • New enumarents:

    • For SPV_KHR_8bit_storage

  • Fixed Issues:

    • Add definition of Memory Object Declaration.

    • Khronos SPIR-V Issue #275: Clarify the meaning of Aliased and Restrict in the Aliasing section.

    • Khronos SPIR-V Issue #315: Be more specific about where many decorations are allowed, particularly for OpFunctionParameter. Includes being clear that the BuiltIn decoration does not apply to OpFunctionParamater.

    • Khronos SPIR-V Issue #348: Clarify remainder descriptions in OpFRem, OpFMod, OpSRem, and OpSMod.

    • Khronos SPIR-V Issue #342: State the DepthReplacing execution-mode behavior more specifically.

    • Khronos SPIR-V Issue #341: More specific wording for depth-hint execution modes DepthGreater, DepthLess, and DepthUnchanged.

    • Khronos SPIR-V Issues #276 and #311: Take more care with unreachable blocks in structured control flow and how to branch into a construct.

    • Khronos SPIR-V Issue #320: Include OpExecutionModeId in the logical layout.

    • Khronos SPIR-V Issue #238: Fix description of OpImageQuerySize to correct Sampled Type → Sampled and list the correct set of dimensions.

    • Khronos SPIR-V Issue #346: Remove ordered rule for structures in the memory layout: Vulkan allows out-of-order Offset layouts.

    • Khronos SPIR-V Issue #322: Allow OpImageQuerySize to query the size of a NonReadable image.

    • Khronos SPIR-V Issue #244: Be more clear about the connections between dimensionalities and capabilities, and in refering to them from OpImageRead and OpImageWrite.

    • Khronos SPIR-V Issue #333: Be clear about overflow behavior for OpIAdd, OpISub, and OpIMul.

4.29. Changes from Version 1.3, Revision 3

  • Add enumerants for

    • SPV_KHR_vulkan_memory_model

  • Fixed Issues:

    • Typo: say OpMatrixTimesVector is Matrix X Vector.

    • Update on Khronos SPIR-V issue #244: Added Shader and Kernel capabilities to the 2D dimensionality.

    • Khronos SPIR-V Issue #317: Clarify that the Uniform decoration should apply only to objects, and that the dynamic instance of the object is the same, rather than at the consumer usage.

    • Khronos SPIR-V Issue #335: Clarify and correct when it is valid for pointers to be operands to OpFunctionCall. Corrections are believed to be consistent with existing front-end and back-end support.

    • Khronos SPIR-V Issue #344: don’t include inactive invocations in what makes the result of OpGroupNonUniformBallotBitExtract undefined.

4.30. Changes from Version 1.3, Revision 4

  • Add enumerants for

    • SPV_NV_fragment_shader_barycentric

    • SPV_NV_compute_shader_derivatives

    • SPV_NV_shader_image_footprint

    • SPV_NV_shading_rate

    • SPV_NV_mesh_shader

    • SPV_NVX_Raytracing

  • Formatting: Removed Enabling Extensions column and instead list the extensions in the Enabling Capabilities column.

4.31. Changes from Version 1.3, Revision 5

  • Reserve Tokens for:

    • SPV_KHR_no_integer_wrap_decoration

    • SPV_KHR_float_controls

  • Fixed Issues:

    • Khronos SPIR-V Issue #352: Remove from OpFunction the statement limiting the use its result. This does not result in any change in intent; it only avoids any past and potential future contradictions.

    • Khronos SPIR-V Issue #308: Don’t allow runtime-sized arrays to be loaded or copied by OpLoad or OpCopyMemory.

    • Include back-edge blocks in the list of blocks that can branch outside their own construct in the structured control-flow rules.

    • Khronos OpenGL API issue #77: Clarify the OriginUpperLeft and OriginLowerLeft execution modes apply only to FragCoord.

    • State the XfbStride and Stream restrictions in the Universal Validation Rules.

    • Khronos SPIR-V Issue #357: The Memory Operands of OpCopyMemory and OpCopyMemorySized applies to both Source and Target.

    • Khronos SPIR-V Issue #385: Be more clear what type <id> must be the same in OpCopyMemory.

    • Khronos SPIR-V Issue #359: OpAccessChain and OpPtrAccessChain do indexing with signed indexes, and OpPtrAccessChain is allowed to compute addresses of elements one past the end of an array.

    • Khronos SPIR-V Issue #367: General validation rules allow the Function storage class for atomic access, while the shader-specific validation rules do not.

    • Khronos SPIR-V Issue #382: In OpTypeFunction, disallow parameter types from being OpTypeVoid.

    • Khronos SPIR-V Issue #374: Built-in derocations can also apply to a constant instruction.

  • Editorial:

    • Make it more clear in OpVariable what Storage Classes must be the same.

    • Remove references to specific APIs, and instead generally refer only to "client API"s. Note that the previous lists of APIs was nonnormative.

    • State the FPRoundingMode decoration rule more clearly in the section listing Validation Rules for Shader Capabilities.

    • Don’t say "value preserving" in the Conversion instructions. These now convert the "value numerically".

    • State variable-pointer validation rules more clearly.

4.32. Changes from Version 1.3, Revision 6

  • Reserve Tokens for:

    • SPV_INTEL_media_block_io

    • SPV_NV_cooperative_matrix

    • SPV_INTEL_device_side_avc_motion_estimation, partially. See the SPV_INTEL_device_side_avc_motion_estimation extension specification for a full listing of tokens.

  • Fixed Issues:

    • Khronos SPIR-V Issue #406: Scope values must come from the table of scope values.

    • Khronos SPIR-V Issue #419: Validation rules include AtomicCounter in the list of storage classes allowed for pointer operands to an OpFunctionCall.

    • Khronos SPIR-V Issue #325: OpPhi clarifications regarding parent dominance, in the instruction and the validation rules, and forward references in the Logical Layout section.

    • Khronos SPIR-V Issue #415: Remove the non-writable storage classes PushConstant and Input from the FPRoundingMode decoration shader validation rule.

    • Khronos SPIR-V Issue #404: Clarify when OpGroupNonUniformShuffleXor, OpGroupNonUniformShuffleUp, and OpGroupNonUniformShuffleDown are valid or result in undefined values.

    • Khronos SPIR-V Issue #393: Be more clear that OpConvertUToPtr and OpConvertPtrToU operate only on unsigned scalar integers.

    • Khronos SPIR-V Issue #416: Result are undefined for all Shift instructions for shifts amounts equal to the bit width of the operand.

    • Khronos SPIR-V Issue #399: Refine the definition of a variable pointer, particularly for function parameters receiving a variable pointer.

    • Khronos SPIR-V Issue #441: Clarify that atomic instruction’s Scope <id> must be a valid memory scope. More generally, all Scope <id> operands are now either Memory or Execution.

    • Khronos SPIR-V Issue #426: Be more direct about undefined behavior for non-uniform control flow in OpControlBarrier and the OpGroup… instructions that discuss this.

  • Deprecate

  • Editorial

4.33. Changes from Version 1.3, Revision 7

  • Fixed Issues:

    • Khronos SPIR-V Issue #371: Restrict intermediate object types to variable types allowed at global scope. See shader validation data rules.

    • Khronos SPIR-V Issue #408: (Re)allow the decorations Volatile, Coherent, NonWritable, and NonReadable on members of blocks. (Temporarily dropping this functionality was accidental/clerical; intent is that it has always been present.)

    • Khronos SPIR-V Issue #418: Add statements about undefinedness and how NaNs are mixed to OpGroupNonUniformFAdd, OpGroupNonUniformFMul, OpGroupNonUniformFMin, and OpGroupNonUniformFMax.

    • Khronos SPIR-V Issue #435: Expand the universal validation rule for variable pointers and matrices to also disallow pointing within a matrix.

    • Khronos SPIR-V Issue #447: Remove implication that OpPtrAccessChain obeys an ArrayStride decoration in storage classes laid out by the implementation.

    • Khronos SPIR-V Issue #450: Allow pointers to OpFunctionCall to be pointers to an element of an array of samplers or images. See the universal validation rules under the Logical addressing model without variable pointers.

    • Khronos SPIR-V Issue #452: OpGroupNonUniformAllEqual uses ordered compares for floating-point values.

    • Khronos SPIR-V Issue #454: Add OpExecutionModeId to the list of allowed forward references in the Logical Layout of a Module.

4.34. Changes from Version 1.3

  • New Functionality:

    • Public issue #35: OpEntryPoint must list all global variables in the interface. Additionally, duplication in the list is not allowed.

    • Khronos SPIR-V Issue #140: Generalize OpSelect to select between two objects.

    • Khronos SPIR-V Issue #156: Add OpUConvert to the list of required opcodes in OpSpecConstantOp.

    • Khronos SPIR-V Issue #345: Generalize the NonWritable decoration to include Private and Function storage classes. This helps identify lookup tables.

    • Khronos SPIR-V Issue #84: Add OpCopyLogical to copy similar but unequal types.

    • Khronos SPIR-V Issue #170: Add OpPtrEqual and OpPtrNotEqual to compare pointers.

    • Khronos SPIR-V Issue #362: Add OpPtrDiff to count the number of elements between two element pointers.

    • Khronos SPIR-V Issue #332: Add SignExtend and ZeroExtend image operands.

    • Khronos SPIR-V Issue #340: Add the UniformId decoration, which takes a Scope operand.

    • Khronos SPIR-V Issue #112: Add iteration-control loop controls.

    • Khronos SPIR-V Issue #366: Change Memory Access operands and the Memory Access section to now be Memory Operands and the Memory Operands section.

    • Khronos SPIR-V Issue #357: Allow OpCopyMemory and OpCopyMemorySized to have Memory Operands for both their Source and Target.

  • New Extensions Incorporated into SPIR-V 1.4:

  • Removed: