Skip to content

Pull requests: AMD-AGI/Primus-Turbo

New pull request New
23 Open 332 Closed
23 Open 332 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

[Attention][Parallelism] Ulysses context-parallel Varlen attention support
#377 opened Jun 10, 2026 by paulpak58 Contributor Loading…
5 of 12 tasks
opt: mxfp4 and mxfp8 dequantize kernel
#376 opened Jun 10, 2026 by RuibinCheung Collaborator Loading…
5 of 12 tasks
12
[Op][Normalization] Add zero_centered_gamma to RMSNorm
#375 opened Jun 10, 2026 by paulpak58 Contributor Loading…
5 of 12 tasks
2
feat: support build on gfx1250
#374 opened Jun 9, 2026 by RuibinCheung Collaborator Loading…
6 of 12 tasks
14
feat: add mxfp8 grouped quantize api
#373 opened Jun 9, 2026 by RuibinCheung Collaborator Loading…
4 of 12 tasks
11
opt(gemm): add AITER MXFP4 preshuffle fast path
#366 opened Jun 6, 2026 by jasainio Contributor Loading…
7 of 12 tasks
14
feat(quantize): enable stochastic rounding on MXFP4 gradients
#365 opened Jun 6, 2026 by jasainio Contributor Loading…
6 of 12 tasks
13
feat(quantize): add fused FP8 quantization kernels with amax+scale and cast+transpose
#364 opened Jun 6, 2026 by jasainio Contributor Loading…
6 of 12 tasks
4
[feat] flydsl based fp8 per tensor gemm
#356 opened Jun 3, 2026 by kyle-256 Collaborator Loading…
7 of 12 tasks
44
feat: triton_grouped_gemm: add work-stealing variant with ws_mode API
#353 opened Jun 2, 2026 by wenchenvincent Loading…
4 of 12 tasks
21
[feat] Add mxfp8 triton grouped gemm support
#349 opened May 30, 2026 by kyle-256 Collaborator Loading…
7 of 12 tasks
13
feat: ck_grouped_gemm: add work-stealing variant with ws_mode API
#348 opened May 27, 2026 by wenchenvincent Loading…
4 of 12 tasks
19
feat: update gemm tensorwise default backend on gfx950
#347 opened May 27, 2026 by RuibinCheung Collaborator Loading…
5 of 12 tasks
4
chore: remove ck tensorwise pytest skip
#334 opened May 9, 2026 by RuibinCheung Collaborator Loading…
4 of 12 tasks
2
[WIP] [Feature] Add Turbo MXFP8 Grouped GEMM (gfx950) for MoE
#330 opened May 7, 2026 by kyle-256 Collaborator Loading…
6 of 12 tasks
67
feat: add more activation func
#329 opened May 7, 2026 by RuibinCheung Collaborator Loading…
8 of 9 tasks
42
opt(gemm): add hipBLASLt algorithm cache and thread-local workspace
#321 opened Apr 30, 2026 by jasainio Contributor Loading…
6 of 12 tasks
5
Refactor: moe dispatch combine autotune
#312 opened Apr 24, 2026 by zhenhuang12 Collaborator Loading…
7 of 12 tasks
54
feat: enable hybrid FP8 dtypes on Triton grouped GEMM backends
#288 opened Apr 15, 2026 by sarthak-amd • Draft
2
perf: optimize hipBLASLt grouped GEMM with algo tuning, enable grouped_gemm autotune hipblaslt support
#284 opened Apr 14, 2026 by kyle-256 Collaborator Loading…
49
feat: Online tuning for hipblaslt gemm
#277 opened Apr 10, 2026 by Z-Y00 Loading…
5
feat(benchmark): per-model/GPU batch sizes and vocab projection for GEMM bench
#265 opened Mar 31, 2026 by Z-Y00 Loading…
1
refactor: reorganize moe ops and kernels
#243 opened Mar 5, 2026 by zhenhuang12 Collaborator Loading…
1
ProTip! Follow long discussions with comments:>50.