perf(qmoe): group prefill expert matmuls

[donghanbj]

Batch symmetric int4 Qwen3.5 expert projections into grid-z WMMA launches to eliminate per-expert dispatch overhead while preserving deterministic scatter order.

Summary

Related issue or design

Why

What

Test plan

Notes for reviewers

Checklist

• The change is focused, or links a design/series explaining its scope.
• Relevant tests were added or updated and the results are documented.
• User-facing or design documentation was updated when needed.
• Substantial AI assistance is disclosed, and I reviewed and understand the result.

[github-actions]

Thanks for opening a PR!

This project follows LLVM's incremental-development and AI-tool-use
guidance. See CONTRIBUTING.md
for the project workflow.

Before requesting review, please check that:

1. The change is focused. Substantial work links the relevant issue
   or design discussion.
2. The PR documents relevant test results and updates affected
   documentation.
3. If AI tools provided substantial assistance, the description
   explains what was assisted and how it was validated, and commit
   trailers identify the tool. The contributor has reviewed and
   understands the result.

Reviewers are assigned through
CODEOWNERS where ownership
is configured.

[github-actions]

L2 Accuracy Results (EP vs CPU)

Model	Combined L2	Total Elems	Skipped NaN/Inf
conv_test_hybrid	4.8668E-07	64	0
GroupQueryAttention_seq256	25.2366	2621440	0
MatMulNBits_o_seq128	259.906	368640	0
QMoE_seq128	34.9552	368640	0

Threshold: 0.01 | Run: 3444 - Commit: 3ed77fc

[github-actions]

MorphiZen EP Performance Results

Model	QPS	Session (s)	1st Infer (ms)	CPU%	Mem (MB)
full_model_seq128	7.52	6.14	369	3	1243
GroupQueryAttention_seq128	4499.78	1.79816	11	6	310
matmul_down_seq128	524.24	2.42	74	3	351

EPContext Export Performance

Model	QPS	Session (s)	1st Infer (ms)	CPU%	Mem (MB)
full_model_seq128	7.51	45.81	364	3	15590

EPContext Import Performance

Model	QPS	Session (s)	1st Infer (ms)	CPU%	Mem (MB)
full_model_seq128	7.53	9.78	363	3	15760

OGA Benchmark Results

Model	Warmup	Reps	Prompt Len	Gen Tokens	TTFT (ms)	TPS	Peak Mem (GB)	GPU Mem (GB)
gpt-oss-20b-webgpu-int4-rtn-block-32	1	5	128	128	177.5	78.5	1.33	13.54
Llama-3.1-8B-awq-g128-int4-asym-fp16-onnx-dml	1	5	128	128	343.7	40.6	1.22	6.43

OGA Wheel Smoke (Python benchmark_e2e.py)

Model	TTFT (ms)	TPS
Llama-3.1-8B-awq-g128-int4-asym-fp16-onnx-dml	197	39.7

Run: 3444 - Commit: 3ed77fc

[donghanbj]

/run-test models=all

[amd-yanjunz]

ROCm OGA CI finished with result: UNSTABLE.

• PR: perf(qmoe): group prefill expert matmuls #601
• Source comment: models=all
• TEST_MODE: performance,accuracy
• MODELZOO: LLM_ROCm_dynamic
• TEST_MODELS: (all)
• FORBID_MODELS: google-gemma-4-26B-A4B-it-FP16W4-qmoe,gemma-4-12B-it-fp16-int4-kquant,NVIDIA-Nemotron-3-Super-120B-A12B-BF16-onnx-fp16-int4-rtn-onnx
• NODE: (OGA job default)
• CI_OPTIONS: LLM_RUN_REPETITIONS=5 LLM_WARMUP_ITERATIONS=1 LLM_RUN_EPS=hip OGA_MODEL_BENCHMARK_MAX_LENGTH=0 LLM_TOKENS_MITRIX=128|128,2048|128
• Jenkins build: http://xsjvitisaijenkins:8080/job/daily_hip_ep_oga/1602/
• NTS report: http://xsjnts.xilinx.com/ROCm/hipdnn-detail?database=development&collection=daily_hip_ep_oga&id=6a6bdbc54747d50f10686cd7