2025-08-21 - HLSL Working Group Minutes

• Discussion topics
  • CBuffer status.
    • Justin explained that the CBuffer feature is mostly working but has bugs related to the timing of CBuffer type creation, which can result in incorrect array offsets.
  • Design for simple matrix use status.
    • Steven noted that most of the simple matrix samples don’t rely on matrix orientation attributes which allows for a simpler implementation.
    • MSFT team has no plans to start on matrices this month, may begin planning in September.
    • Steven will keep us posted and reassess the timelines as we go.
  • Tracking Bugs vs Unimplemented Features in offload test suite
    • Sarah presented a proposal to use XFAIL to mark tests that are failing due to incomplete Clang implementation.
    • The proposed suggestion is to have all XFAILs refer to links to issues tracking either feature implementaiton or bugs.
    • Using XFAIL allows us to know when things start passing and enable tests.
    • Sarah will review existing tests to ensure proper XFAIL usage.
    • Chris has been doing recent cleanups to XFAILs to get the tests passing.
  • wg-hlsl website generation (see #330)
    • Damyan has a PR up to generate a website from the markdown.
    • Damyan has experimented with parsing marksdown files to extract status information vs using FrontMatter.
    • General consensus was that FrontMatter renders well enough and simplifies the tooling.
    • Chris raised concerns about validating the state information, and Damyan confirmed that it requires precise spelling and formatting.
    • Steven suggested a pre-submit check to verify the status information formatting, and Chris supported the idea of a linting step for FrontMatter.
    • Action item: Damyan will update the PR to stop using regexes to parse Markdown, and to provide clear diagnostics.