Skip to content

[Doc][SPIR-V] Move FP4 E2M1 to SPV_EXT_ocp_microscaling_types#22528

Merged
uditagarwal97 merged 1 commit into
intel:syclfrom
vmaksimo:fp4-ocp-microscaling-spec-update
Jul 6, 2026
Merged

[Doc][SPIR-V] Move FP4 E2M1 to SPV_EXT_ocp_microscaling_types#22528
uditagarwal97 merged 1 commit into
intel:syclfrom
vmaksimo:fp4-ocp-microscaling-spec-update

Conversation

@vmaksimo

@vmaksimo vmaksimo commented Jul 2, 2026

Copy link
Copy Markdown
Contributor

Represent the FP4 (E2M1) type with the multi-vendor SPV_EXT_ocp_microscaling_types extension (Float4E2M1EXT encoding) instead of the Intel-only SPV_INTEL_float4, for both plain values and cooperative-matrix component types.

Repoint mini_float_conversions_env.asciidoc to the EXT encoding and extension.

Narrow SPV_INTEL_float4 to a single role: it no longer defines an encoding or plain-type capability, and only adds Float4E2M1CooperativeMatrixINTEL to enable the Float4E2M1EXT type as a cooperative-matrix component type.

AI-assisted: Claude Opus 4.8 (commercial SaaS)

Represent the FP4 (E2M1) type with the multi-vendor
[SPV_EXT_ocp_microscaling_types](https://github.khronos.org/SPIRV-Registry/extensions/EXT/SPV_EXT_ocp_microscaling_types.html)
extension (Float4E2M1EXT encoding) instead of
the Intel-only SPV_INTEL_float4, for both plain values and cooperative-matrix
component types.

Repoint mini_float_conversions_env.asciidoc to the EXT encoding and extension.

Narrow SPV_INTEL_float4 to a single role: it no longer defines an encoding or
plain-type capability, and only adds Float4E2M1CooperativeMatrixINTEL to enable
the Float4E2M1EXT type as a cooperative-matrix component type.

AI-assisted: Claude Opus 4.8 (commercial SaaS)

@bashbaug bashbaug left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM, the move from the Intel-specific extension to the multi-vendor EXT extension is the right direction.

@YuriPlyakhin

Copy link
Copy Markdown
Contributor

@intel/llvm-gatekeepers , please, merge.

@uditagarwal97 uditagarwal97 merged commit cb9b7b7 into intel:sycl Jul 6, 2026
4 checks passed
Comment on lines 26 to 33
- Dmitry Sidorov, Intel +
- Victor Mustya, Intel +
- Ben Ashbaugh, Intel +
- Dounia Khaldi, Intel +
- Joe Garvey, Intel +
- Greg Lueck, Intel +
- Pawel Jurek, Intel +

@MrSidims MrSidims Jul 9, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

since https://github.com/intel/llvm/pull/22528/changes#diff-96030a078c5e0bac7ff1f01182144eafeed911397583aa68df47eed70ff283e3R147

Suggested change
- Dmitry Sidorov, Intel +
- Victor Mustya, Intel +
- Ben Ashbaugh, Intel +
- Dounia Khaldi, Intel +
- Joe Garvey, Intel +
- Greg Lueck, Intel +
- Pawel Jurek, Intel +
- Viktoria Maksimova, Intel +

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

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants