[Bugfix][Build] Grouped MoE build should respect GPU_ARCHS

[simondanielsson]

Motivation

Fixes #4071.

FlyDSL Grouped MoE kernel (gfx1250 only) was always built, irrespective of GPU_ARCHS. This slows down build time for say gfx950-only builds.

Technical Details

Don't build kernels whose gfx column in the csv is not part of GPU_ARCHS. When GPU_ARCHS is not set, we always build the kernels (as expected).

At the moment all rows in the csv have gfx1250 so technically we could skip parsing the entire file if GPU_ARCHS is not gfx1250, but to avoid future bugs we make a decision based on the gfx value in every row.

Test Plan

Build AITER for say gfx950 on this branch and main. Measure build time.

PREBUILD_KERNELS=1 GPU_ARCHS=gfx950 \
        python3 setup.py bdist_wheel --dist-dir=/aiter/dist;

Test Result

• main: 1h48m build time (192 CPUs on MI350 node)
• this branch: X build time (192 CPUs on MI350 node)

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: 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: GPT-OSS-120B, DeepSeek-R1-0528, Kimi-K2.5
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 4072 --add-label <label>