Support dx12 raytracing opacity micromap trimming

[zongdu-arm]

Support dx12 raytracing opacity micromap trimming, track in #2868

DX12 AS trim metadata payloads now include serialized D3D12 build-input
structs to preserve OMM and OMM-array behavior:

• Add a new meta id kInitDx12AccelerationStructureCommand2.
• Encode and write full AS build inputs from state tracking and writer paths.
• Decode and route structured build inputs through DX12 consumers.
• Update DX12 utility and builder logic for OMM triangles and OMM-array
  input buffer entry handling.

[zongdu-arm]

Hi @locke-lunarg, @bradgrantham-lunarg ,
Could you review this PR?

[CLAassistant]

All committers have signed the CLA.

[locke-lunarg]

The CI failed during the replay and gfxrecon-info tests for D3D12RaytracingHelloWorld-priority-capture-1225-1235. They got a error [gfxrecon] ERROR - Failed to read init DX12 acceleration structure meta-data block header (frame 0 block 69).

I suspect the cause is the change to the trimming encoder, which would require generating new capture files for the tests to pass. Since we'd prefer not to update test assets, can we avoid modifying the encoder? or is this change necessary for the fix?

[locke-lunarg]

ParsedBlock& BlockParser::ParseMetaData(BlockBuffer& block_buffer)
{
    ...
    else if (meta_data_type == format::MetaDataType::kInitDx12AccelerationStructureCommand)
    ...

Here is where the decoder is changed. The new InitDx12AccelerationStructureCommandHeader doesn't match the old capture files. The value of header.inputs_size is bad, like 188978561025, to cause failure.

Not only new InitDx12AccelerationStructureCommandHeader, I feel that the new parameter of Dx12DecoderBase::DispatchInitDx12AccelerationStructureCommand might also cause failure for the old capture files.

[locke-lunarg]

I feel that we might need a new meta data for this new solution, but keep the original meta data and original solution to be compatible with the old capture files. @bradgrantham What do you think?

[zongdu-arm]

I feel that we might need a new meta data for this new solution, but keep the original meta data and original solution to be compatible with the old capture files. @bradgrantham What do you think?

The structure data has been changed, so the old gfxr replay failed and the new FF gfxr must be re-captured.
Whether to add a new meta block ID is up to you, I can modify it at any time.

[bradgrantham-lunarg]

It's important that existing captures continue to work, so this would need to add a new revision of that metadata with new encode/decode. (And decoder for original metadata needs to persist) Brad Grantham Director of Technology LunarG, Inc. - 3D Graphics Software Solutions ***@***.***

…

On Thu, May 14, 2026, 4:24 AM zongdu-arm ***@***.***> wrote: *zongdu-arm* left a comment (LunarG/gfxreconstruct#2934) <#2934 (comment)> I feel that we might need a new meta data for this new solution, but keep the original meta data and original solution to be compatible with the old capture files. @bradgrantham <https://github.com/bradgrantham> What do you think? The structure data has been changed, so the old gfxr replay failed and the new FF gfxr must be re-captured. Whether to add a new meta block ID is up to you, I can modify it at any time. — Reply to this email directly, view it on GitHub <#2934 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AMCLT76DWTTPDHQAA7ANZIL42UU4FAVCNFSM6AAAAACYUNAWMSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DINBWHA3TQOJVGQ> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[locke-lunarg]

I agree with @bradgrantham-lunarg . @zongdu-arm please also adds a warning for the old meta data to recommend the user to generate a new capture file. Thanks!