1  include::meta/VK_EXT_shader_subgroup_vote.txt[]
  2  
  3  *Last Modified Date*::
  4      2016-11-28
  5  *IP Status*::
  6      No known IP claims.
  7  *Interactions and External Dependencies*::
  8    - This extension requires the
  9      https://www.khronos.org/registry/spir-v/extensions/KHR/SPV_KHR_subgroup_vote.html[`SPV_KHR_subgroup_vote`]
 10      SPIR-V extension.
 11    - This extension requires the
 12      https://www.khronos.org/registry/OpenGL/extensions/ARB/ARB_shader_group_vote.txt[`GL_ARB_shader_group_vote`]
 13      extension for GLSL source languages.
 14  *Contributors*::
 15    - Neil Henning, Codeplay
 16    - Daniel Koch, NVIDIA Corporation
 17  
 18  This extension adds support for the following SPIR-V extension in Vulkan:
 19  
 20    * `SPV_KHR_subgroup_vote`
 21  
 22  This extension provides new SPIR-V instructions:
 23  
 24    * code:OpSubgroupAllKHR,
 25    * code:OpSubgroupAnyKHR, and
 26    * code:OpSubgroupAllEqualKHR.
 27  
 28  to compute the composite of a set of boolean conditions across a group of
 29  shader invocations that are running concurrently (a _subgroup_).
 30  These composite results may be used to execute shaders more efficiently on a
 31  slink:VkPhysicalDevice.
 32  
 33  When using GLSL source-based shader languages, the following shader
 34  functions from GL_ARB_shader_group_vote can map to these SPIR-V
 35  instructions:
 36  
 37    * code:anyInvocationARB() -> code:OpSubgroupAnyKHR,
 38    * code:allInvocationsARB() -> code:OpSubgroupAllKHR, and
 39    * code:allInvocationsEqualARB() -> code:OpSubgroupAllEqualKHR.
 40  
 41  The subgroup across which the boolean conditions are evaluated is
 42  implementation-dependent, and this extension provides no guarantee over how
 43  individual shader invocations are assigned to subgroups.
 44  In particular, a subgroup has no necessary relationship with the compute
 45  shader _local workgroup_ -- any pair of shader invocations in a compute
 46  local workgroup may execute in different subgroups as used by these
 47  instructions.
 48  
 49  Compute shaders operate on an explicitly specified group of threads (a local
 50  workgroup), but many implementations will also group non-compute shader
 51  invocations and execute them concurrently.
 52  When executing code like
 53  
 54  [source,c++]
 55  ----------------------------------------
 56  if (condition) {
 57    result = do_fast_path();
 58  } else {
 59    result = do_general_path();
 60  }
 61  ----------------------------------------
 62  
 63  where code:condition diverges between invocations, an implementation might
 64  first execute code:do_fast_path() for the invocations where code:condition
 65  is true and leave the other invocations dormant.
 66  Once code:do_fast_path() returns, it might call code:do_general_path() for
 67  invocations where code:condition is code:false and leave the other
 68  invocations dormant.
 69  In this case, the shader executes *both* the fast and the general path and
 70  might be better off just using the general path for all invocations.
 71  
 72  This extension provides the ability to avoid divergent execution by
 73  evaluating a condition across an entire subgroup using code like:
 74  
 75  [source,c++]
 76  ----------------------------------------
 77  if (allInvocationsARB(condition)) {
 78    result = do_fast_path();
 79  } else {
 80    result = do_general_path();
 81  }
 82  ----------------------------------------
 83  
 84  The built-in function code:allInvocationsARB() will return the same value
 85  for all invocations in the group, so the group will either execute
 86  code:do_fast_path() or code:do_general_path(), but never both.
 87  For example, shader code might want to evaluate a complex function
 88  iteratively by starting with an approximation of the result and then
 89  refining the approximation.
 90  Some input values may require a small number of iterations to generate an
 91  accurate result (code:do_fast_path) while others require a larger number
 92  (code:do_general_path).
 93  In another example, shader code might want to evaluate a complex function
 94  (code:do_general_path) that can be greatly simplified when assuming a
 95  specific value for one of its inputs (code:do_fast_path).
 96  
 97  === New Object Types
 98  
 99  None.
100  
101  === New Enum Constants
102  
103  None.
104  
105  === New Enums
106  
107  None.
108  
109  === New Structures
110  
111  None.
112  
113  === New Functions
114  
115  None.
116  
117  === New Built-In Variables
118  
119  None.
120  
121  === New SPIR-V Capabilities
122  
123    * <<spirvenv-capabilities-table-subgroupvote,SubgroupVoteKHR>>
124  
125  === Issues
126  
127  None.
128  
129  === Version History
130  
131    * Revision 1, 2016-11-28 (Daniel Koch)
132      - Initial draft