Add support for bf16 for afoneapi/afcuda
Nobody has claimed this yet.
Assessment
- Difficulty
- 5/5
- Estimated time
- Over a week
- Newbie friendliness
- 30/100
- Issue type
- Feature
- Clarity
- Mostly clear
- Activity status
- Stale
- Tech stack
- cpp
- Domain
- hpc, performance
Research direction
Start with include/af/defines.h, then read the linked CUDA bf16 header and the issue's requirements for NVIDIA GPUs and fourth-generation Intel CPUs. Determine how the new datatype must be represented for AVX512BF16/AMX and CUDA; done means bf16 is available for the requested targets without adding a new API.
Written by the indexing model from the issue text.
Description
Add suppor for bf16 datatype
https://github.com/arrayfire/arrayfire/blob/master/include/af/defines.h
for :
-
afcuda/nvidia gpus
https://sources.debian.org/src/nvidia-cuda-toolkit/11.2.2-3%2Bdeb11u3/arm64/cuda_cudart/targets/sbsa-linux/include/cuda_bf16.hpp/#L126 -
4th gen intel CPUs (SaphirreRapids) with avx512bf16 / AMX
Description
-
Additional information about the feature you would like to add
bf16 datatype -
What problem are you trying to solve?
computation at bf16 precision -
(Optional) API of new function
No new API, just a new type in defines.h -
(Optional)Are there other libraries that implement this feature?
pytorch, numpy, oneDNN, ...
- Dominant language
- C++
- Stars
- 4.9k
- Forks
- 555
- Avg merge
- 1h 24m
- Merged PRs (30d)
- 1
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from arrayfire/arrayfire
-
bug
Difficulty 2/5 1-3 hours Newbie friendliness 74/100
-
bug
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
-
bug
Difficulty 2/5 1-3 hours Newbie friendliness 76/100
-
bug
Difficulty 4/5 3-5 days Newbie friendliness 45/100
-
Difficulty 5/5 Over a week Newbie friendliness 25/100
All issues in arrayfire/arrayfire
Similar issues
-
Difficulty 2/5 1-3 hours Newbie friendliness 70/100
-
Difficulty 2/5 1-3 hours Newbie friendliness 65/100
duckdb/duckdb-wasm#2258 ·
-
Difficulty 2/5 1-3 hours Newbie friendliness 75/100
objectionary/eo-graphs#75 ·
-
Coarray integration tests carry no LABELS, so run_tests.py silently skips them under every backend Opencoarray
Difficulty 2/5 1-3 hours Newbie friendliness 70/100
-
Difficulty 2/5 1-3 hours Newbie friendliness 75/100
FISCO-BCOS/FISCO-BCOS#5642 ·