Skip to content

[SPIRV] Use scalar rules for BDA alignment - #8576

Open
Chris B (llvm-beanz) wants to merge 2 commits into
microsoft:mainfrom
llvm-beanz:8572
Open

[SPIRV] Use scalar rules for BDA alignment#8576
Chris B (llvm-beanz) wants to merge 2 commits into
microsoft:mainfrom
llvm-beanz:8572

Conversation

@llvm-beanz

Copy link
Copy Markdown
Collaborator

Previously the vk::RawBufferLoad and vk::RawBufferStore instructions used a default alignment of 4 unless the user explicitly specified a different value. This change instead makes the default be based on the minimum legal value for the type being loaded so that users don't need to explicitly override it when loading arbitrary data types.

Fixes #8572

Assisted by Claude Opus 4.8

Previously the `vk::RawBufferLoad` and `vk::RawBufferStore`
instructions used a default alignment of 4 unless the user explicitly
specified a different value. This change instead makes the default be
based on the minimum legal value for the type being loaded so that
users don't need to explicitly override it when loading arbitrary data
types.

Fixes microsoft#8572

Assisted by Claude Opus 4.8
@github-actions

github-actions Bot commented Jun 20, 2026

Copy link
Copy Markdown
Contributor

✅ With the latest revision this PR passed the C/C++ code formatter.

@dnovillo Diego Novillo (dnovillo) added the spirv Work related to SPIR-V label Jul 10, 2026
@damyanp

Copy link
Copy Markdown
Member

[Auto-generated note from Damyan Pepper (@damyanp)]

Did you consider adding a release note?

Per our CONTRIBUTING guidance, user-visible bug fixes and new features usually need an entry in docs/ReleaseNotes.md. If this PR is part of a larger multi-PR effort, a single shared release note across that effort is totally fine.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

spirv Work related to SPIR-V

Projects

Status: New

Development

Successfully merging this pull request may close these issues.

[SPIR-V] Improperly aligned OpLoad/OpStore for structures with 64-bit scalar types

4 participants