[FLYDSL] [TRITON] Attention backward mxfp8 gfx950

[lburzawa]

Motivation

Support mxfp8 attention backward in FlyDSL on gfx950.

Technical Details

• Main attn bwd kernel in FlyDSL
• Bwd preprocess kernel in Triton
• Quant kernels in Triton

Test Plan

Correctness tests for each kernel.

Test Result

Tests pass.

Submission Checklist

• Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.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:triton-300x	Run an additional Triton test job on MI300X in PRs; main branch always runs both MI35X and MI300X
ci:sglang	SGLang integration tests
ci:atom	ATOM benchmark (DeepSeek-R1 + GPT-OSS)
ci:vllm	vLLM benchmark
ci:all	All of the above

Add labels via the sidebar or gh pr edit 3094 --add-label <label>

[github-actions]

⚠️ [ruff] <F841> _{reported by reviewdog 🐶}
Local variable bytes_per_thread_qo_scale is assigned to but never used

Suggested change

	bytes_per_thread_qo_scale = max(1, bytes_per_tile_qo_scale // total_threads)
	max(1, bytes_per_tile_qo_scale // total_threads)

[github-actions]

⚠️ [ruff] <F841> _{reported by reviewdog 🐶}
Local variable data is assigned to but never used

Suggested change

	data = func(*args, **kwargs)
	func(*args, **kwargs)