Skip to content

Pull requests: intel/sycl-tla

New pull request New
42 Open 479 Closed
42 Open 479 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

Add python API for flash-attn
#558 opened Oct 13, 2025 by YangKai0616 Loading…
Rewrite mma unit tests
#557 opened Oct 13, 2025 by yuanhang-dev Loading…
7
[WIP] Support different dtype for LinearCombination
#556 opened Oct 13, 2025 by joyalbin • Draft
Skip alignment check for sourceless epilogues
#555 opened Oct 13, 2025 by nsingh-habana • Draft
Gemm Universal unit tests for MainloopIntelW8A8 API
#554 opened Oct 10, 2025 by rishi-yadav • Draft
Use newer version of mma_atom and copy_atom in CollectiveEpilogue for 00_bmg_gemm test
#553 opened Oct 10, 2025 by anamikac-intel Loading…
[CI][WIP] Fix coverity workflow
#552 opened Oct 9, 2025 by anupren • Draft
[PYTORCHDGQ-6865] Added support for RoPE on chunk prefill [WIP]
#551 opened Oct 9, 2025 by pralay-das • Draft
Tiled MMA tests using XE_DPAS_TT
#550 opened Oct 8, 2025 by rishi-yadav Loading…
1
[CuTe] [Xe] Fix make_block_2d_copy_* for batched tensors
#549 opened Oct 7, 2025 by petercad Loading…
First version of SDPA Fwd - No need to review
#548 opened Oct 6, 2025 by cfgfung Loading…
2
Re-implement FlashAttention with new Xe atoms
#547 opened Oct 4, 2025 by petercad Loading…
15
upload 2nd version of sdpa backward
#546 opened Oct 3, 2025 by yuankuns Loading…
[Reorder] add optimized e8m0 -> f32 upconversion
#544 opened Oct 1, 2025 by petercad Loading…
Support of FP8 Chunk Prefill kernel
#542 opened Oct 1, 2025 by adityachatter Loading…
Support nullptr value of argument ptr_C for xe_array_epilogue
#541 opened Sep 29, 2025 by sanchitintel Loading…
Use newer version of mma_atom and copy_atom in 00_bmg_gemm
#540 opened Sep 29, 2025 by anamikac-intel Loading…
37
Attention sink support
#533 opened Sep 25, 2025 by kareemshaik80 Loading…
18
Add dimension check to prevent out-of-bounds access in example 05_bmg_gemm_with_epilogue_splitk
#529 opened Sep 23, 2025 by ClarkChin08 Loading…
2
Model-specific tests for flash attention decode
#526 opened Sep 22, 2025 by rishi-yadav Loading…
[PYTORCHDGQ-7000] Added support for Rotary Embedding in flash_attention
#523 opened Sep 19, 2025 by pralay-das Loading…
12
MoEGEMM as an extension of GroupGEMM
#520 opened Sep 18, 2025 by sanchitintel • Draft
15
Add a new tile scheduler for varlen prefill to avoid launching empty work groups
#516 opened Sep 18, 2025 by carsonwang Loading…
Also use column-major B matrix in the example 00_bmg_gemm.cpp
#510 opened Sep 13, 2025 by sanchitintel Loading…
Remove redundant code from GroupGEMM implementation
#508 opened Sep 12, 2025 by sanchitintel Loading…
ProTip! Add no:assignee to see everything that’s not assigned.