Skip to content

Pull requests: ROCm/aiter

New pull request New
436 Open 3,143 Closed
436 Open 3,143 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

remove kv cache assert for old arch for upstream compatibility
#3942 opened Jun 26, 2026 by ganyi1996ppo Contributor Loading…
1 task
Feat/flydsl mxfp4 gemm
#3941 opened Jun 26, 2026 by lizamd Loading…
2 of 4 tasks
1
[Triton] Add fused_gemm_a16w16_split_cat
#3940 opened Jun 25, 2026 by rbrugaro-amd Contributor Loading…
1
Map top-left map to bottom-right for self-attn
#3939 opened Jun 25, 2026 by Micky774 Contributor Loading…
1 task
2
gate custom all-reduce on XGMI topology
#3938 opened Jun 25, 2026 by skysnow2001 Contributor Loading…
1 task done
1
Gluon MXFP4 Fuse Reduce Quant
#3937 opened Jun 25, 2026 by amd-jrosas Loading…
1 task done
1
Spatial Attention: XCD-aware spatial workgroup mapping for MHA and GQA (SWIZZLE=1)
#3936 opened Jun 25, 2026 by mc186 Loading…
1
[DO NOT MERGE] [TESTING CI] ci:all
#3935 opened Jun 25, 2026 by Boss2002n Contributor Loading…
2
[test] test_topk_plain: parametrize sweep to fix collection-time OOM
#3934 opened Jun 25, 2026 by JohnQinAMD Contributor Loading…
1 task
4
edit aiter_opus_plus.h using opus api instead of asm code
#3932 opened Jun 25, 2026 by junhaha666 Contributor Loading…
1 task
6
fix(quick_all_reduce): make flag sync CUDA-graph-safe
#3928 opened Jun 25, 2026 by Jasen2201 Contributor Loading…
3
Feat/gfx942 flydsl mxfp4 moe
#3926 opened Jun 25, 2026 by msaffari-amd • Draft
1 task
1
feat: support flydsl all2all
#3924 opened Jun 25, 2026 by JiaoliangYu • Draft
1 task
1
change default pa reduce kernel from cxx to flydsl ci:atom
#3923 opened Jun 25, 2026 by Bernard-Liu Contributor Loading…
3
[OPUS][ATOM] gfx1250 (wave32/WMMA) NoPE-fp8/RoPE-bf16 sparse prefill (separate module)
#3922 opened Jun 25, 2026 by carlushuang Collaborator Loading…
1
feat: update A8W8 MLA kernels to global-load ckv variant on gfx950
#3921 opened Jun 25, 2026 by fangche123 Contributor Loading…
1 task
2
[Bugfix] base_tuner: gfx-aware tuned CSV handling and column-safe row…
#3920 opened Jun 25, 2026 by yzhou103 Contributor Loading…
1 task
7
feat(gfx1151): allow gfx1151 in cpp_itfs JIT arch validation
#3919 opened Jun 25, 2026 by carlushuang Collaborator Loading…
1
Fix aot deadlock
#3918 opened Jun 25, 2026 by zhiding512 Contributor Loading…
1 task
6
feat(gfx1151): add INT8 W8A8 GEMM default config
#3917 opened Jun 25, 2026 by carlushuang Collaborator Loading…
1
CI: run vLLM DSv4 and MiniMax M3 on MI350X ci:vllm
#3916 opened Jun 25, 2026 by gyohuangxin Member Loading…
5
perf(unified_attention): use 8 warps for gfx1151 3D decode
#3915 opened Jun 25, 2026 by carlushuang Collaborator Loading…
1
[Triton] Optimize MoE
#3914 opened Jun 25, 2026 by vgokhale Contributor Loading…
1
flydsl: gfx942 FP8 MQA logits indexer kernel (+ Triton FN/FNUZ fix)
#3913 opened Jun 25, 2026 by akii96 Contributor Loading…
4
configs(fmoe): add expert=256/topk=8 MXFP4 tuned entries for DeepSeek-V3/R1 (gfx950)
#3911 opened Jun 24, 2026 by Rohan138 Contributor • Draft
1
ProTip! Updated in the last three days: updated:>2026-06-22.