work graphs spec: inconsequential variable name change to match DXC i…

…mplementation

d3d/WorkGraphs.md

@@ -4327,7 +4327,7 @@ enum class MEMORY_TYPE_FLAG : uint {
     ALL_MEMORY               = 0x0000000f,
 };
 
-enum class SEMANTIC_FLAG
+enum class BARRIER_SEMANTIC_FLAG
 {
     GROUP_SYNC  = 0x00000001,
     GROUP_SCOPE = 0x00000002,
@@ -4345,7 +4345,7 @@ void Barrier(Object o, uint SemanticFlags)
 
 ```
 
-> Note for users of the June 2023 Work Graphs preview release:  In that release, the function was `Barrier(uint MemoryTypeFlags, uint AccessFlags, uint SyncFlags)`, where `AccessFlags` could be `DEVICE|GROUP_VISIBLE` and `SyncFlags` could be `GROUP_SYNC`.  All that has changed is the last two parameters have been merged into `SEMANTIC_FLAG`.
+> Note for users of the June 2023 Work Graphs preview release:  In that release, the function was `Barrier(uint MemoryTypeFlags, uint AccessFlags, uint SyncFlags)`, where `AccessFlags` could be `DEVICE|GROUP_VISIBLE` and `SyncFlags` could be `GROUP_SYNC`.  All that has changed is the last two parameters have been merged into `BARRIER_SEMANTIC_FLAG`.
 
 As a basis, D3D's shader memory consistency model is relaxed, as generally understood in existing architectures and literature. Loosely, this means the program author and/or compiler are responsible for identifying all memory and thread synchronization points via some appropriately expressive labeling.
 
@@ -4357,7 +4357,7 @@ As a basis, D3D's shader memory consistency model is relaxed, as generally under
 |`NODE_OUTPUT_MEMORY`|Output record memory.  Not needed to make data visible to [OutputComplete()](#outputcomplete).  Rather, this is for the case of threads in a group communicating with each other through writes/reads in allocated node output memory while it is being prepared, before calling `OutputComplete()`.|
 |`ALL_MEMORY`|Shorthand for all memoy type flags combined.|
 
-|`SEMANTIC_FLAG`|Definition|
+|`BARRIER_SEMANTIC_FLAG`|Definition|
 |-|-|
 |`GROUP_SYNC`|Synchronize threads in group to the current location in the shader.|
 |`GROUP_SCOPE`|Scope for memory visibility is current thread group.|
@@ -6156,5 +6156,5 @@ v0.40|5/24/2023|<li>Fixed typos in DDI defines: [D3D12DDI_NODE_IO_KIND_0108](#d3
 v0.41|5/26/2023|<li>In DXIL [Node input and output metadata table](#node-input-and-output-metadata-table), `NodeMaxRecords` and `NodeRecordType` entries were reversed from what the compiler implemented, so swapped the entries in the spec to match the code.</li><li>In [Example of creating node input and output handles](#example-of-creating-node-input-and-output-handles) and [Example DXIL metadata diagram](#example-dxil-metadata-diagram) fixed metadata encoding values, several of which were stale.</li><li>In [Node input](#node-input) fixed examples that used old GetInputRecordCount() to be .count().</li><li>In DXIL [NodeIOFlags and NodeIOKind encoding](#nodeioflags-and-nodeiokind-encoding), there was a typo `GroupNodeOutputRecord` fixed to `GroupNodeOutputRecords` and `ThreadNodeOutputRecord` fixed to `ThreadNodeOutputRecords`.  Similar typo in the DDI header under [D3D12DDI_NODE_IO_KIND_0108](#d3d12ddi_node_io_kind_0108).  This will require new headers and driver recompile but fortunately not a binary breaking change.</li>
 v0.42|6/12/2023|<li>In [Shader function attributes](#shader-function-attributes) section removed stale text under `[NodeShareInputOf()]` that said that nodes sharing input need to have the same node type and dispatch grid size - these constraints were never actually needed.</li>
 v0.43|6/25/2023|<li>In [Shader target](#shader-target) and [Shader function attributes](#shader-function-attributes) sections fixed the case sensitivity to match expected final compiler behavior: For attributes like `Shader("node")` and `NodeLaunch("mode")` (`"mode"` is `"coalescing"` etc.), the attribute name is case insensitive, while the string is case sensitive.  And for these two attributes, the strings also must be lowercase.  The current preview compiler is incorrectly case insensitive for the string in `NodeLaunch("mode")`.  Updated samples to match.</li><li>Linked to [Thread visibility in wave operations](#thread-visibility-in-wave-operations) from the launch mode sections such as [Coalescing launch nodes](#coalescing-launch-nodes) for discoverability.</li>
-v0.44|9/7/2023|<li>Added [Producer - consumer dataflow through UAVs](#producer---consumer-dataflow-through-uavs), summarizing the rules for how to make sure data written to UAVs in a producer node ends up visible to a consumer.</li><li> Added example code for a couple of barrier scenarios:<ul><li>[Dispatch grid writing to UAV for consumer node to read](#dispatch-grid-writing-to-uav-for-consumer-node-to-read)</li><li>[Single thread writing to UAV for consumer node to read](#single-thread-writing-to-uav-for-consumer-node-to-read)</li></ul></li><li>Joined the ACCESS_FLAGS and SYNC_FLAGS field in [Barrier](#barrier) into a single flags field: SEMANTIC_FLAGS for simplicity.  Matching change to DXIL in [Lowering barrier](#lowering-barrier).</li><li>In [Node input declaration](#node-input-declaration) and [Node input atttributes](#node-input-attributes), added a `globallycoherent` option to `RWDispatchNodeInputRecord` for cases loads/stores will be used on the input record for cross-group communication.  Matching change to DXIL in [NodeIOFlags and NodeIOKind encoding](#nodeioflags-and-nodeiokind-encoding)</li><li>Added [Quad and derivative operation semantics](#quad-and-derivative-operation-semantics) and [NonUniformResourceIndex semantics](#nonuniformresourceindex-semantics) sections to clarify how these operate in work graphs.</li><li>Added [Support for WaveSize shader function attribute](#support-for-wavesize-shader-function-attribute) section to clarify that the HLSL `WaveSize(N)` shader function attribute works for shaders node shaders of all launch types.</li><li>Refined behavior for [AddToStateObject()](#addtostateobject) with more specifics and more flexibility than the initial proposal.  Will need more refinement with feedback.</li><li>Along with the [AddToStateObject()](#addtostateobject) refinement, changed various work graphs creation DDIs so that nodes identify their connections to other nodes with lists of pointers rather than by index.  This way when additions are made, existing node definitons can be updated in-place by appending pointers to lists when a node points to a newly added node, or is targetted by a newly added node.  The addition lists all nodes, with new ones at the start of the list. All affected existing node definitions from previous DDIs are modified in-place to point to each other appropriately.  Each node definition includes a versionAdded number so that with a single DDI memory representation of the full work graph, drivers can see what the structure at any given version looked like if needed.  See [D3D12DDI_WORK_GRAPH_DESC_0108](#d3d12ddi_work_graph_desc_0108)(added flags, version, and changed node arrays to lists), [D3D12DDI_WORK_GRAPH_FLAGS_0108](#d3d12ddi_work_graph_flags_0108)(new addition flag), [D3D12DDI_NODE_LIST_ENTRY_0108](#d3d12ddi_node_list_entry_0108)(new, used for lists of nodes), [D3D12DDI_NODE_0108](#d3d12ddi_node_0108)(updated with `VersionAdded` instead of `NodeIndex`), [D3D12DDI_NODE_OUTPUT_0108](#d3d12ddi_node_output_0108)(updated with node list instead of array), [D3D12DDI_BROADCASTING_LAUNCH_NODE_PROPERTIES_0108](#d3d12ddi_broadcasting_launch_node_properties_0108)(updated with input node list instead of arrays), [D3D12DDI_COALESCING_LAUNCH_NODE_PROPERTIES_0108](#d3d12ddi_coalescing_launch_node_properties_0108)(updated with input node list instead of arrays), [D3D12DDI_THREAD_LAUNCH_NODE_PROPERTIES_0108](#d3d12ddi_thread_launch_node_properties_0108)(updated with input node list instead of arrays)</li><li>In [Node output attributes](#node-output-attributes) added `[UnboundedSparseNodes]` option, which is equivalent to setting `[AllowSparseNodes]` and `[NodeArraySize(0xffffffff)]`.  In DXIL [Node input and output metadata table](#node-input-and-output-metadata-table), noted that for `NodeOutputArraySize`, `0xffffffff` means unbounded array size and `NodeAllowSparseNodes` must be `true`.</li><li>Removed changes to allow multiple conflicting `shader` attributes on entry points. Replaced **Repurposing plain compute shader code** with [Reusing shader entries](#reusing-shader-entries). Removed references to default or implicit node inputs throughout.</li><li>In [D3D12_WORK_GRAPHS_TIER](#d3d12_work_graphs_tier) added `D3D12_WORK_GRAPHS_TIER_1_0` (and same at DDI) to get ready for first full release.</li>
+v0.44|9/7/2023|<li>Added [Producer - consumer dataflow through UAVs](#producer---consumer-dataflow-through-uavs), summarizing the rules for how to make sure data written to UAVs in a producer node ends up visible to a consumer.</li><li> Added example code for a couple of barrier scenarios:<ul><li>[Dispatch grid writing to UAV for consumer node to read](#dispatch-grid-writing-to-uav-for-consumer-node-to-read)</li><li>[Single thread writing to UAV for consumer node to read](#single-thread-writing-to-uav-for-consumer-node-to-read)</li></ul></li><li>Joined the ACCESS_FLAGS and SYNC_FLAGS field in [Barrier](#barrier) into a single flags field: BARRIER_SEMANTIC_FLAGS for simplicity.  Matching change to DXIL in [Lowering barrier](#lowering-barrier).</li><li>In [Node input declaration](#node-input-declaration) and [Node input atttributes](#node-input-attributes), added a `globallycoherent` option to `RWDispatchNodeInputRecord` for cases loads/stores will be used on the input record for cross-group communication.  Matching change to DXIL in [NodeIOFlags and NodeIOKind encoding](#nodeioflags-and-nodeiokind-encoding)</li><li>Added [Quad and derivative operation semantics](#quad-and-derivative-operation-semantics) and [NonUniformResourceIndex semantics](#nonuniformresourceindex-semantics) sections to clarify how these operate in work graphs.</li><li>Added [Support for WaveSize shader function attribute](#support-for-wavesize-shader-function-attribute) section to clarify that the HLSL `WaveSize(N)` shader function attribute works for shaders node shaders of all launch types.</li><li>Refined behavior for [AddToStateObject()](#addtostateobject) with more specifics and more flexibility than the initial proposal.  Will need more refinement with feedback.</li><li>Along with the [AddToStateObject()](#addtostateobject) refinement, changed various work graphs creation DDIs so that nodes identify their connections to other nodes with lists of pointers rather than by index.  This way when additions are made, existing node definitons can be updated in-place by appending pointers to lists when a node points to a newly added node, or is targetted by a newly added node.  The addition lists all nodes, with new ones at the start of the list. All affected existing node definitions from previous DDIs are modified in-place to point to each other appropriately.  Each node definition includes a versionAdded number so that with a single DDI memory representation of the full work graph, drivers can see what the structure at any given version looked like if needed.  See [D3D12DDI_WORK_GRAPH_DESC_0108](#d3d12ddi_work_graph_desc_0108)(added flags, version, and changed node arrays to lists), [D3D12DDI_WORK_GRAPH_FLAGS_0108](#d3d12ddi_work_graph_flags_0108)(new addition flag), [D3D12DDI_NODE_LIST_ENTRY_0108](#d3d12ddi_node_list_entry_0108)(new, used for lists of nodes), [D3D12DDI_NODE_0108](#d3d12ddi_node_0108)(updated with `VersionAdded` instead of `NodeIndex`), [D3D12DDI_NODE_OUTPUT_0108](#d3d12ddi_node_output_0108)(updated with node list instead of array), [D3D12DDI_BROADCASTING_LAUNCH_NODE_PROPERTIES_0108](#d3d12ddi_broadcasting_launch_node_properties_0108)(updated with input node list instead of arrays), [D3D12DDI_COALESCING_LAUNCH_NODE_PROPERTIES_0108](#d3d12ddi_coalescing_launch_node_properties_0108)(updated with input node list instead of arrays), [D3D12DDI_THREAD_LAUNCH_NODE_PROPERTIES_0108](#d3d12ddi_thread_launch_node_properties_0108)(updated with input node list instead of arrays)</li><li>In [Node output attributes](#node-output-attributes) added `[UnboundedSparseNodes]` option, which is equivalent to setting `[AllowSparseNodes]` and `[NodeArraySize(0xffffffff)]`.  In DXIL [Node input and output metadata table](#node-input-and-output-metadata-table), noted that for `NodeOutputArraySize`, `0xffffffff` means unbounded array size and `NodeAllowSparseNodes` must be `true`.</li><li>Removed changes to allow multiple conflicting `shader` attributes on entry points. Replaced **Repurposing plain compute shader code** with [Reusing shader entries](#reusing-shader-entries). Removed references to default or implicit node inputs throughout.</li><li>In [D3D12_WORK_GRAPHS_TIER](#d3d12_work_graphs_tier) added `D3D12_WORK_GRAPHS_TIER_1_0` (and same at DDI) to get ready for first full release.</li>
 v0.45|9/22/2023|<li>Added new option for overriding node properties at the API, a new [D3D12_NODE_OVERRIDES_TYPE](#d3d12_node_overrides_type), [D3D12_NODE_OVERRIDES_TYPE_COMMON_COMPUTE](#d3d12_node_overrides_type).  This allows overriding common properties agnostic to launch mode, like renaming a node or its outputs, without having to know the launch mode of the node/shader.  Previously any override required the app to know the launch mode to pick the matching override type for the shader.</li><li>In [Interactions with other d3d12 features](#interactions-with-other-d3d12-features), noted that - Pipeline statistics report invocations of nodes against the corresponding shader type.  So compute-based node shader invocations increment `CSInvocations`.</li>