Skip to content

Pull requests: vllm-project/vllm-xpu-kernels

New pull request New
41 Open 386 Closed
41 Open 386 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

[MoE] Add a bf16xfp32 gemm kernel for router gemm
#465 opened Jul 16, 2026 by jjmiao1 Contributor Loading…
11
[XPU] Add & optimize silu_and_mul_with_clamp activation kernel
#461 opened Jul 15, 2026 by yangulei Loading…
5
[BugFix][MoE] Make xe2 grouped-GEMM tile scheduler cudagraph-capture-safe
#457 opened Jul 10, 2026 by AleksVidmantas Loading…
[Attn] Optimize the flash attn for head size 72 and corresponding test
#456 opened Jul 10, 2026 by jinyouzhi Loading…
3 of 4 tasks
17
upgrade to oneapi 2026.0
#451 opened Jul 7, 2026 by jikunshang Member Loading…
4 tasks
5
feat: add HND KV layout support
#449 opened Jul 3, 2026 by hlin99 Loading…
14
[Dont merge]Enale FusedMoe kernel V2
#448 opened Jul 3, 2026 by yihuaxu Contributor Loading…
10
[Add] add sycl implementation for dequantize_and_gather_k_cache
#447 opened Jul 3, 2026 by wangshier108 Loading…
8
[DNM][WIP]attn: convert boolean template params to runtime dispatch
#444 opened Jun 30, 2026 by jikunshang Member • Draft
4 tasks
6
[XPU][DeepSeekV4] Add fused_kv_compress_norm_rope_insert_indexer_mxfp…
#443 opened Jun 30, 2026 by nc-BobLee Loading…
4 tasks
7
fix low accuracy of Qwen3.5-27B/Qwen3.6-27B
#437 opened Jun 29, 2026 by ccrhx4 Loading…
8
Add xe2 block_fp8 grouped_gemm kernel
#436 opened Jun 25, 2026 by frost-intel Contributor Loading…
5
Add Sparse MLA prefill/decode kernel
#433 opened Jun 23, 2026 by wuxun-zhang Contributor Loading…
4 tasks done
16
[DNM]update umd 2622
#424 opened Jun 17, 2026 by jikunshang Member Loading…
4 tasks
2
[XPU][MiniMax-M3] Add and optimize fused qknorm/rope/kv-insert SYCL kernel for BMG
#422 opened Jun 16, 2026 by yangulei Loading…
9
[XPU][DeepSeekV4] Optimize _fused_kv_compress_norm_rope_insert_sparse_attn with SYCL
#419 opened Jun 16, 2026 by yingjie-han Loading…
5
add asym feature for chunk prefill and kv cache
#403 opened Jun 8, 2026 by tianyuan0211 Loading…
20
[GDN] Split gdn attention into conv1d and gated_delta_rule
#402 opened Jun 8, 2026 by HeJunyan Loading…
10
[XPU][MOE] Optimize XPU MoE activation by skipping invalid rows
#401 opened Jun 8, 2026 by ranzhejiang Loading…
13
[XPU][GDN] Use native recurrent algo for decode seq in mix batch
#394 opened Jun 4, 2026 by YangQun1 Contributor • Draft
4 tasks
[MoE] Allow reusable workspaces in XPU fused MoE
#392 opened Jun 3, 2026 by ehartford Loading…
1
15
[GDN] Accept graph-padded spec metadata for DFlash decode
#391 opened Jun 3, 2026 by ehartford Loading…
1
[ATTN] support HND format
#382 opened Jun 1, 2026 by YizhouZ Collaborator Loading…
8
[XPU][WAN] Flash attention optimizations for WAN 2.2 prefill
#381 opened May 30, 2026 by pengzhao-intel Collaborator • Draft
4 tasks done
8
[XPU][FMHA] Add reverse-odd-heads Q-tile scheduler for causal prefill
#363 opened May 21, 2026 by pengzhao-intel Collaborator Loading…
11
ProTip! Updated in the last three days: updated:>2026-07-15.