[solinzby1]

Motivation

Technical Details

Test Plan

Test Result

Submission Checklist

• Look over the contributing guidelines at https://github.com/ROCm/TheRock/blob/main/GOVERNANCE.md#pull-requests.

[github-actions]

🏷️ CI Guide

Runs automatically on every PR:

• ✅ Pre-checks (submodule verification, code formatting)
• ✅ Aiter op tests (gfx942 + gfx950)
• ✅ Triton tests on MI35X (only when aiter/ops/triton/** or related paths are changed)

Extended tests (opt-in via labels):

Label	Tests
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 gfx1201 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 6603 --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.

[Copilot]

🟡 Changes recommended

Tuner defaults can write to the wrong configuration tables, and mxscale benchmarking still selects an incompatible operator.

2 open findings

• Restore default tune file redirection after changing ARG_DEFAULTS · New
• Initialize mxscale mode and switch blockscale workflow defaults · New

What changed in this PR

Extends gfx950 FlyDSL MXFP8 GEMM support to raw 32×32 and 128×128 block scales, with tuning against batched GEMM at batch size one.

Changes:

• Adds coarse-scale handling to both MXFP8 kernel pipelines.
• Unifies backend selection and extends tuning and ahead-of-time compilation.
• Updates numerical tests and removes obsolete scale-shuffle helpers.

File	Description
op_tests/​tuning_tests/​test_csv_validation.py	Enforces one tuning row per shape and block.
op_tests/​test_gemm_group32_interface.py	Tests both MXFP8 pipelines with coarse scales.
op_tests/​test_gemm_a8w8_blockscale.py	Passes raw scales to GEMM.
csrc/​gemm_a8w8_mxfp8_bpreshuffle/​gemm_a8w8_mxfp8_bpreshuffle_tune.py	Adds coarse-scale tuning against batch-one GEMM.
csrc/​ck_gemm_a8w8_blockscale/​gemm_a8w8_blockscale_tune.py	Removes the legacy FlyDSL tuning path.
aiter/​ops/​shuffle.py	Removes obsolete blockscale shuffle helpers.
aiter/​ops/​gemm_op_a8w8.py	Dispatches raw scales by tuned backend.
aiter/​ops/​flydsl/​mxscale_preshuffle_kernels.py	Routes supported raw-scale layouts.
aiter/​ops/​flydsl/​kernels/​mxfp8_gemm_8wave.py	Stages coarse scales in the eight-wave kernel.
aiter/​ops/​flydsl/​kernels/​gemm/​mxscale_preshuffle.py	Adds coarse-scale loading and packing.
aiter/​ops/​flydsl/​gemm_tune/​flydsl_gemm_mxscale_preshuffle_common.py	Adds scale-specific candidates and memory estimates.
aiter/​ops/​flydsl/​gemm_a8w8_mxfp8_8wave.py	Validates and forwards scale-block settings.
aiter/​configs/​a8w8_blockscale_mxscale_bpreshuffle_untuned_gemm.csv	Removes backend identity from input keys.
aiter/​aot/​flydsl/​gemm.py	Compiles kernels for the new scale contracts.

🧠 Review effort: Balanced

---

💡 Configure MCP servers for context-aware, tailored reviews. Learn more in the docs.