Skip to content

feat: add Ascend950 TileLang and causal convolution support. - #2073

Merged
zhang-minchao merged 2 commits into
xLLM-AI:mainfrom
wuyifan2017:codex/a5-tilelang-causal-conv
Jul 31, 2026
Merged

feat: add Ascend950 TileLang and causal convolution support.#2073
zhang-minchao merged 2 commits into
xLLM-AI:mainfrom
wuyifan2017:codex/a5-tilelang-causal-conv

Conversation

@wuyifan2017

@wuyifan2017 wuyifan2017 commented Jul 29, 2026

Copy link
Copy Markdown
Contributor

Description

  • Resolve Ascend950 TileLang builds to dav-c310 and use the CANN-compatible A5 BiSheng flags.
  • Pass device-specific compile flags through variant workers while preserving the existing A2/A3 --npu-arch path.
  • Add the FP16 bs1_d6144_w4_silu0_f16 causal-conv variant used by Qwen3.5 decode.

Related Issues

N/A

Change Type

  • Bug fix
  • New feature
  • Performance improvement
  • Refactor
  • Documentation
  • Test
  • Build or CI

Pull Request Checklist

Thank you for contributing to xLLM. Before requesting review, please make sure the following items are complete.

PR Title and Commit Messages

  • The PR title and each commit message follow the xLLM commit format: <type>: <subject>.

Allowed types: feat, bugfix, docs, test, refactor, chore, style, revert, perf, model, build, release.
The subject should use clear English, start with a verb, include at least 4 words, and end with ..

Pre-commit Checks

  • I have installed pre-commit by running pip install pre-commit or an equivalent command.
  • I have installed the hooks with pre-commit install.
  • I have run pre-commit run --all-files and fixed any reported issues.

If you are unsure how to set up pre-commit, see the pre-commit documentation.

Self Review

  • I have self-reviewed the code according to .agents/skills/code-review/references/custom-code-style.md, especially code written or assisted by AI.
  • I have rebased this PR onto the latest main branch.

Build and Test Coverage

  • Tests have been added or updated as needed.
  • CUDA: python setup.py build test has passed on a CUDA machine.
  • NPU: python setup.py build test has passed on an NPU machine.
  • MLU: python setup.py build test has passed on an MLU machine.

Reviewer Notes

  • A focused Ascend950 AOT build passed for all five causal_conv1d variants, including bs1_d6144_w4_silu0_f16; generated manifests and object files report a5 / dav-c310.
  • A2/A3 flag assertions confirm that both retained platforms still use the existing --npu-arch route.
  • Python bytecode compilation and pre-commit run --all-files passed.

@wuyifan2017
wuyifan2017 force-pushed the codex/a5-tilelang-causal-conv branch from 8cb3c02 to 5900a50 Compare July 30, 2026 02:38
@zhang-minchao
zhang-minchao merged commit 44fec5f into xLLM-AI:main Jul 31, 2026
3 of 14 checks passed
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.

3 participants