[WIP] Allow duplication of OpTypeNodePayloadArrayAMDX #6470
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
jkwak-work
changed the title
jkwak-work
force-pushed
the
fix/workgraphcs_fix_blocking_issue_for_lionel
branch
from
2241997 to
f203597
Compare
|
This PR is not ready for review nor merge. The main problem is that it is not able to take "nodeID" in a String type. I need to get the nodeID from a new attribute |
|
/format |
|
🌈 Formatted, please merge the changes from this PR |
jkwak-work
force-pushed
the
fix/workgraphcs_fix_blocking_issue_for_lionel
branch
2 times, most recently
from
9b45d1b to
085fda4
Compare
|
/format |
|
🌈 Formatted, please merge the changes from this PR |
This commit is mainly to allow OpTypeNodePayloadArrayAMDX to be duplicated in SPIRV when using WorkGraphs. AMD spec requires that there should be dedicated type declaration with OpTypeNodePayloadArrayAMDX for each "node". And each of them should be decorated with PayloadNodeNameAMDX. https://docs.vulkan.org/features/latest/features/proposals/VK_AMDX_shader_enqueue.html This behavior is very different from how "types" are declared. Normally types are declared only once and they are shared across the whole program. But for the work-graphs, each "RecordType" belongs to each node and they need to be declared for each node. Consider the following HLSL example where a producer triggers two consumer nodes. Each node will be differentiated by "NodeID" attribute. ``` [Shader("node")] [NodeLaunch("broadcasting")] [NumThreads(4,1,1)] void myFancyNode(..., [NodeID("myNode1")] NodeOutput<MY_RECORD> myRecords1, [NodeID("myNode2")] NodeOutput<MY_RECORD> myRecords2, ...) { ThreadNodeOutputRecords<MY_RECORD> myRecord1 = myRecords1.GetThreadNodeOutputRecords(1); myRecord1.myData = 1; myRecord1.OutputCompelete(); ThreadNodeOutputRecords<MY_RECORD> myRecord2 = myRecords2.GetThreadNodeOutputRecords(1); myRecord2.myData = 1; myRecord2.OutputCompelete(); } ``` Even though both are using the same type, "MY_RECORD", each of them needs to be declared separately. The generated SPIR-V code for the HLSL above is something like the following, ``` %str_node1 = OpConstantStringAMDX "myNode1"; %str_node2 = OpConstantStringAMDX "myNode1"; %type_node1 = OpTypeNodePayloadArrayAMDX %type_myRecord; %type_node2 = OpTypeNodePayloadArrayAMDX %type_myRecord; OpDecorateId %type_node1 PayloadNodeNameAMDX %str_node1; OpDecorateId %type_node2 PayloadNodeNameAMDX %str_node2; ``` Note that `%type_node1` and `%type_node2` are equivalent, but they must be declared separately because each of them will be decorated with different nodeID string.
jkwak-work
force-pushed
the
fix/workgraphcs_fix_blocking_issue_for_lionel
branch
from
132a91e to
274c029
Compare
|
/format |
|
🌈 Formatted, please merge the changes from this PR |
Co-authored-by: slangbot <186143334+slangbot@users.noreply.github.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Closes #6391
This commit is mainly to allow OpTypeNodePayloadArrayAMDX to be duplicated in SPIRV when using WorkGraphs.
AMD spec requires that there should be dedicated type declaration with OpTypeNodePayloadArrayAMDX for each "node". And each of them should be decorated with PayloadNodeNameAMDX.
https://docs.vulkan.org/features/latest/features/proposals/VK_AMDX_shader_enqueue.html
This behavior is very different from how "types" are declared. Normally types are declared only once and they are shared across the whole program. But for the work-graphs, each "RecordType" belongs to each node and they need to be declared for each node.
Consider the following HLSL example where a producer triggers two consumer nodes. Each node will be differentiated by "NodeID" attribute.
Even though both are using the same type, "MY_RECORD", each of them needs to be declared separately. The generated SPIR-V code for the HLSL above is something like the following,
Note that
%type_node1and%type_node2are equivalent, but they must be declared separately because each of them will be decorated with different nodeID string.