[FlyDSL MOE] update a8w4 moe

[Zzz9990]

Motivation

Technical Details

Test Plan

Test Result

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 (only when aiter/ops/triton/** or related paths are changed)

Extended tests (opt-in via labels):

Label	Tests
ci:triton-355	Run Triton tests on MI355 in addition to MI325
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 2726 --add-label <label>

[github-actions]

Remaining comments which cannot be posted as a review comment to avoid GitHub Rate Limit

ruff

---

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

aiter/aiter/ops/flydsl/kernels/mixed_moe_gemm_2stage.py

Line 2914 in 7f3d055

size_w = (

---

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

aiter/aiter/ops/flydsl/kernels/mixed_moe_gemm_2stage.py

Line 3051 in 7f3d055

w_elem = T.f16 if is_f16_b else (T.i8 if is_int8 else T.f8)

---

⚠️ [ruff] <F841> _{reported by reviewdog 🐶}
Local variable vec1_f16 is assigned to but never used

aiter/aiter/ops/flydsl/kernels/mixed_moe_gemm_2stage.py

Line 3058 in 7f3d055

vec1_f16 = T.vec(1, f16)

---

⚠️ [ruff] <F841> _{reported by reviewdog 🐶}
Local variable vec2_f16 is assigned to but never used

aiter/aiter/ops/flydsl/kernels/mixed_moe_gemm_2stage.py

Line 3059 in 7f3d055

vec2_f16 = T.vec(2, f16)

---

⚠️ [ruff] <F841> _{reported by reviewdog 🐶}
Local variable vec4_f16 is assigned to but never used

aiter/aiter/ops/flydsl/kernels/mixed_moe_gemm_2stage.py

Line 3060 in 7f3d055

vec4_f16 = T.vec(4, f16)

---

⚠️ [ruff] <F841> _{reported by reviewdog 🐶}
Local variable vec8_x is assigned to but never used

aiter/aiter/ops/flydsl/kernels/mixed_moe_gemm_2stage.py

Line 3064 in 7f3d055

vec8_x = T.vec(vec8_elems, x_elem)

---

⚠️ [ruff] <F841> _{reported by reviewdog 🐶}
Local variable vec1_i64 is assigned to but never used

aiter/aiter/ops/flydsl/kernels/mixed_moe_gemm_2stage.py

Line 3066 in 7f3d055

vec1_i64 = T.vec(1, i64)

---

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

aiter/aiter/ops/flydsl/kernels/mixed_moe_gemm_2stage.py

Line 3081 in 7f3d055

layout_x = fx.make_layout((m_i32_v, k_i32_v), stride=(k_i32_v, 1))

---

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

aiter/aiter/ops/flydsl/kernels/mixed_moe_gemm_2stage.py

Line 3095 in 7f3d055

layout_b = b_layout.layout_b

---

⚠️ [ruff] <F841> _{reported by reviewdog 🐶}
Local variable c_k0 is assigned to but never used

aiter/aiter/ops/flydsl/kernels/mixed_moe_gemm_2stage.py

Line 3096 in 7f3d055

c_k0 = _div_pow2(k_in * arith.constant(int(a_elem_bytes), index=True), 64)

---

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

aiter/aiter/ops/flydsl/kernels/mixed_moe_gemm_2stage.py

Line 3189 in 7f3d055

c_a_pack = arith.constant(int(a_elem_vec_pack), index=True)

---

⚠️ [ruff] <F841> _{reported by reviewdog 🐶}
Local variable c32 is assigned to but never used

aiter/aiter/ops/flydsl/kernels/mixed_moe_gemm_2stage.py

Line 3261 in 7f3d055

c32 = arith.constant(32, index=True)

---

⚠️ [ruff] <F841> _{reported by reviewdog 🐶}
Local variable layout_x_div4 is assigned to but never used

aiter/aiter/ops/flydsl/kernels/mixed_moe_gemm_2stage.py

Line 3430 in 7f3d055

layout_x_div4 = fx.make_layout(

---

⚠️ [ruff] <F841> _{reported by reviewdog 🐶}
Local variable vec4_x is assigned to but never used

aiter/aiter/ops/flydsl/kernels/mixed_moe_gemm_2stage.py

Line 3459 in 7f3d055

vec4_x = T.vec(4, x_elem)

---

⚠️ [ruff] <F841> _{reported by reviewdog 🐶}
Local variable c_n0 is assigned to but never used

aiter/aiter/ops/flydsl/kernels/mixed_moe_gemm_2stage.py

Line 3560 in 7f3d055

c_n0 = _div_pow2(c_n_total, 16)

---

⚠️ [ruff] <F841> _{reported by reviewdog 🐶}
Local variable c_n0_static is assigned to but never used

aiter/aiter/ops/flydsl/kernels/mixed_moe_gemm_2stage.py

Line 3561 in 7f3d055

c_n0_static = experts * model_dim // 16

---

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

aiter/csrc/ck_gemm_moe_2stages_codegen/gemm_moe_tune.py

Line 2893 in 7f3d055

scale_tmp = torch.ones(