[JiaoliangYu]

Change

• Adds aiter/ops/triton/configs/gfx1250/gluon/gemm/gemm_a16w16/GEMM-A16W16-N=32768-K=512.json: DeepSeek-V3/R1 kv_b_proj (128 heads x (128 + 128), kv_lora_rank 512) at attention TP1.
• M <= 128 keeps the DEFAULT entries; M <= 256 uses compute_bound 256x128x64; larger M uses compute_bound 256x256x64 (4 warps, 4 buffers).
• Config only, no code change.

Results

MI455X (gfx1250), single GPU, BF16:

• M=11776: 500.7 -> 255.7 us.
• Faster than DEFAULT for every M from 256 to 16384; outputs bit-identical.
• DeepSeek-R1 MLPerf offline prefill step median 248.2 -> 229.5 ms.

Relation to #6397

• Use the FlyDSL BF16 GEMM by default for large untuned gfx1250 shapes #6397 (open) sends large untuned gfx1250 shapes to the FlyDSL BF16 GEMM by default. With this file the N=32768, K=512 shape counts as tuned and stays on gluon.
• We have not compared this config with Use the FlyDSL BF16 GEMM by default for large untuned gfx1250 shapes #6397's FlyDSL path for this shape.

Checks

• JSON parses; same buckets as DEFAULT; M <= 128 entries equal DEFAULT; compute_bound entries use the key set other files in the directory use.

[github-actions]

🏷️ CI Guide

Runs automatically on every PR:

• ✅ Pre-checks (submodule verification, code formatting)
• ✅ Aiter standard tests on MI35X
• ✅ Triton tests on MI35X (only when aiter/ops/triton/** or related paths are changed)

Extended tests (opt-in via labels):

Label	Tests
ci:mi300x	Run Aiter standard tests on MI300X in PRs; main branch always runs them
ci:gfx1250-ffm-triton	Run the five-shard gfx1250 FFM Triton test suite
ci:triton-300x	Run an additional Triton test job on MI300X in PRs (added automatically when gfx942 configs change); main branch always runs both MI35X and MI300X
ci:triton-355	Run the full Triton test suite on MI35X, not only the tests the change affects
ci:windows	Windows JIT op smoke test on the Radeon R9700 ROCm runner
ci:multi-gpu	Aiter multi-GPU tests on the 8-GPU runner
ci:sglang	SGLang integration tests: DeepSeek-R1-MXFP4 accuracy, Qwen 3.5 accuracy
ci:atom	ATOM benchmark: DeepSeek-R1-0528, GPT-OSS-120B
ci:atom_full	ATOM accuracy suite for PR and main models from ATOM models_accuracy.json
ci:vllm	vLLM benchmark: DeepSeek-V4-Pro, MiniMax-M3-MXFP4
ci:all	All standard extended tests (excludes ci:atom_full)

Only add ci:atom_full for FlyDSL or Triton upgrades.
Add labels via the sidebar or gh pr edit 7962 --add-label <label>

One backend per PR:
A PR changes one kernel backend: [Triton/Gluon] (Triton and Gluon count as one), [HIP], [ASM], [CK], [OPUS] or [FlyDSL]. If the title ends up with two backend tags, split the PR -- as stacked pull requests when one part cannot merge without the other.

PR title tags & labels:
Component tags ([Triton/Gluon], [HIP], [CK], [ASM], ...) are added to the PR title and as PR labels automatically from the changed files and re-synced on every push — change-type tags like [fix]/[Perf], op tags like [MLA], and human labels (ci:*) are left untouched. Add the no-auto-title label to stop the title rewrites; labels stay in sync either way.

Reviews:
A reviewer is usually assigned within one working day after the PR is ready for review (see Code Review Process). If nobody has responded for about a week, mention the area contact listed in CONTACTS.md. Want to help review? See REVIEWERS.md.