Skip to content

Create Highlevel Bindings for FP8 Datatype#788

Open
isVoid wants to merge 7 commits into
NVIDIA:mainfrom
isVoid:numba-fp8-highlevel
Open

Create Highlevel Bindings for FP8 Datatype#788
isVoid wants to merge 7 commits into
NVIDIA:mainfrom
isVoid:numba-fp8-highlevel

Conversation

@isVoid

@isVoid isVoid commented Feb 11, 2026

Copy link
Copy Markdown
Contributor

This PR adds high level exposure of fp8 data type to Numba-CUDA.

Supported features include:

  • FP8 constructors from existing data type (elementwise and packed)
  • Conversion intrinsics that provide finer control of saturation type

Supported FP8 variants (element wise):

  • fp8_[e5m2, e4m3, e8m0]

Supported packed FP8 variants:

  • fp8[x2, x4]_[e5m2, e4m3, e8m0]

This PR also adds tests for packed type bindings introduced in #686.

closes #200

@greptile-apps

greptile-apps Bot commented Feb 11, 2026

Copy link
Copy Markdown
Contributor

Automatic reviews are disabled for this repository.

@copy-pr-bot

copy-pr-bot Bot commented Feb 11, 2026

Copy link
Copy Markdown

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

Comment thread numba_cuda/numba/cuda/fp8.py Outdated
@isVoid

isVoid commented Feb 12, 2026

Copy link
Copy Markdown
Contributor Author

/ok to test 894b79d

@gmarkall gmarkall added the 3 - Ready for Review Ready for review by team label Feb 23, 2026
@isVoid

isVoid commented Feb 23, 2026

Copy link
Copy Markdown
Contributor Author

/ok to test 203c3a8

@isVoid

isVoid commented Feb 24, 2026

Copy link
Copy Markdown
Contributor Author

/ok to test 028dfe0

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

Labels

3 - Ready for Review Ready for review by team

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[FEA] Support fp8 data types

2 participants