data: port complete MoE + xllm layer call chains from upstream repos

MoE call chain from ds_vllm (vllm-project/vllm latest):
  ex_engine/moe/ — 20 files, 8736 lines
  - modular_kernel.py (1630 lines) — base classes for modular MoE
  - experts/fused_batched_moe.py (972 lines) — NaiveBatchedExperts
  - prepare_finalize/batched.py (171 lines) — token grouping by expert
  - topk_weight_and_reduce.py (176 lines) — scatter-add finalize
  - fused_moe.py (1740 lines) — main fused_moe dispatch
  - config.py (1407 lines) — FusedMoEQuantConfig
  - activation.py, utils.py, layer.py, etc.

xllm layer code (jd-opensource/xllm):
  ex_engine/xllm_layers/ — 39 files, 5859 lines
  - ilu/fused_moe.cpp (797 lines) — production ixformer 7-step MoE pipeline
  - ilu/attention.cpp (189 lines) — paged_attention + flash_attn bridge
  - npu_torch/qwen3_gated_delta_net_base.cpp (576 lines) — GDN reference
  - common/rms_norm.cpp, rotary_embedding.cpp, activation.cpp, dense_mlp.cpp

xllm ILU kernels — synced 10 files to upstream (diffs from prior edits)

These are reference implementations, NOT hand-written.
Source repos: vllm-project/vllm, jd-opensource/xllm

ex_engine/moe/activation.py

@@ -122,32 +122,17 @@ def apply_moe_activation(
 
     # Activations with gated multiplication (gate × activation(up))
     if activation == MoEActivation.SILU:
-        # BI-V100: torch.ops._C.silu_and_mul not available
-        # Use corex_attn_head_rms_norm pattern: try C++ first, fallback to PyTorch
-        d = output.size(-1)
-        gate = input[..., :d]
-        up = input[..., d:]
-        output.copy_(F.silu(gate) * up)
+        torch.ops._C.silu_and_mul(output, input)
     elif activation == MoEActivation.GELU:
-        d = output.size(-1)
-        gate = input[..., :d]
-        up = input[..., d:]
-        output.copy_(F.gelu(gate) * up)
+        torch.ops._C.gelu_and_mul(output, input)
     elif activation == MoEActivation.GELU_TANH:
-        d = output.size(-1)
-        gate = input[..., :d]
-        up = input[..., d:]
-        output.copy_(F.gelu(gate, approximate="tanh") * up)
+        torch.ops._C.gelu_tanh_and_mul(output, input)
     elif activation == MoEActivation.SWIGLUOAI:
-        d = output.size(-1)
-        gate = input[..., :d]
-        up = input[..., d:]
-        output.copy_(F.silu(gate) * up)
+        torch.ops._C.swigluoai_and_mul(output, input)
     elif activation == MoEActivation.SWIGLUSTEP:
-        d = output.size(-1)
-        gate = input[..., :d]
-        up = input[..., d:]
-        output.copy_(F.silu(gate) * up)
+        from vllm.model_executor.layers.activation import swiglustep_and_mul_triton
+
+        swiglustep_and_mul_triton(output, input)
 
     # Activations without gated multiplication
     elif activation == MoEActivation.SILU_NO_MUL: