All issues

Search results

Create Triton kernel backup for paged_mxfp4_mqa_logits operator

Status: Open.

#6021 In ROCm/aiter;

· brunomazzottiamd opened on Sep 30, 2026

[Feature]: Create a Perf CI

Triton/Gluon

Status: Open.

Feature

#6014 In ROCm/aiter;

· Boss2002n opened on Sep 30, 2026

[Kimik3] Accumulate the latent-MoE up-projection in aiter a16w16 atomic

Status: Open.

#6012 In ROCm/aiter;

· nsusanto opened on Sep 30, 2026

[Kimik3] Fuse Kimi-K3 MLA chunked-context gather on AITER

Status: Open.

#6011 In ROCm/aiter;

· nsusanto opened on Sep 30, 2026

[KimiK3] Fuse shared-expert down-proj into the latent-MoE tail GEMM

Status: Open.

#6010 In ROCm/aiter;

· nsusanto opened on Sep 30, 2026

move gluon kernels into proper Dirs

Triton/Gluon

Status: Open.

Task

#6008 In ROCm/aiter;

· Boss2002n opened on Sep 30, 2026

[gfx950][Triton/Gluon] Unified attention: scf.for i64/i32 loop-carried type mismatch breaks CUDA-graph capture

bugSomething isn't workingSomething isn't working

Triton/Gluon

Status: Open.

#5990 In ROCm/aiter;

· kangwangamd opened on Sep 30, 2026

[Issue]: gfx950 FlyDSL BF16 GEMM (gemm_a16w16) reduces split-K / slice-K partial sums in BF16: ~6% of outputs deviate ≥2 ULP from an FP32-accumulation reference at small M

Status: Open.

#5976 In ROCm/aiter;

· irvineoy opened on Sep 30, 2026

[Issue]: AITER_TRITON_ONLY leaves aiter.dtypes unbound, silently disabling aiter.ops.triton.quant

Status: Open.

#5959 In ROCm/aiter;

· Ragua1 opened on Sep 29, 2026

[Bug] gfx950 OPUS persistent BF16 kid 300 output depends on partial-K row padding

Status: Open.

#5954 In ROCm/aiter;

· harsh-nod opened on Sep 29, 2026

gdr_decode_packed_bf16 accepts NaN fixed scale and returns NaN output

Status: Open.

#5951 In ROCm/aiter;

· harsh-nod opened on Sep 29, 2026

[Feature]: Tune fp8 unified attention for Gemma4 prefill shapes [gfx942]

Status: Open.

#5947 In ROCm/aiter;

· simondanielsson opened on Sep 29, 2026