ref(upstream): FULL TREE — Deep-Spark xllm (1470) + ds_vllm csrc/models (703)

Replaces cherry-picked upstream_ref with complete source trees.

xllm/ — Iluvatar official C++ inference engine (15MB, 1470 files)
  Complete: kernels → layers → models → runtime → scheduler → api
  Excluded: .git, binary images, third_party submodule checkouts

ds_vllm/ — Iluvatar official vllm fork (8MB, 703 files)
  Included: csrc/ (ALL CUDA kernels), fused_moe/, qwen3_5 model, _custom_ops
  Excluded: tests, benchmarks, docs, examples (not needed for reference)

Critical call chains now fully traceable:
  MoE: moe_topk_softmax_kernels.cuh → ixformer.h → fused_moe.cpp → layer
  GDN: qwen3_gated_delta_net_base.cpp → qwen3_5_gated_delta_net.cpp
  Attention: ixformer.h → xllm_paged_attention → attention.cpp

upstream_ref/ds_vllm/csrc/attention/attention_generic.cuh

@@ -0,0 +1,65 @@
+/*
+ * Adapted from
+ * https://github.com/NVIDIA/FasterTransformer/blob/release/v5.3_tag/src/fastertransformer/kernels/decoder_masked_multihead_attention_utils.h
+ * Copyright (c) 2023, The vLLM team.
+ * Copyright (c) 2020-2023, NVIDIA CORPORATION.  All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#pragma once
+
+#include <stdint.h>
+
+namespace vllm {
+
+// A vector type to store Q, K, V elements.
+template <typename T, int VEC_SIZE>
+struct Vec {};
+
+// A vector type to store FP32 accumulators.
+template <typename T>
+struct FloatVec {};
+
+// Template vector operations.
+template <typename Acc, typename A, typename B>
+inline __device__ Acc mul(A a, B b);
+
+template <typename T>
+inline __device__ float sum(T v);
+
+template <typename T>
+inline __device__ float dot(T a, T b) {
+  return sum(mul<T, T, T>(a, b));
+}
+
+template <typename A, typename T>
+inline __device__ float dot(T a, T b) {
+  return sum(mul<A, T, T>(a, b));
+}
+
+template <typename T>
+inline __device__ void zero(T& dst) {
+  constexpr int WORDS = sizeof(T) / 4;
+  union {
+    T raw;
+    uint32_t words[WORDS];
+  } tmp;
+
+#pragma unroll
+  for (int ii = 0; ii < WORDS; ++ii) {
+    tmp.words[ii] = 0u;
+  }
+  dst = tmp.raw;
+}
+
+}  // namespace vllm