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/cuda_compat.h

@@ -0,0 +1,76 @@
+#pragma once
+
+#ifdef USE_ROCM
+  #include <hip/hip_runtime.h>
+#endif
+
+#ifdef USE_ROCM
+struct Utils {
+  static __host__ int get_warp_size() {
+    static bool is_cached = false;
+    static int result;
+
+    if (!is_cached) {
+      int device_id;
+      cudaDeviceProp deviceProp;
+      cudaGetDevice(&device_id);
+      cudaGetDeviceProperties(&deviceProp, device_id);
+
+      result = deviceProp.warpSize;
+      is_cached = true;
+    }
+
+    return result;
+  }
+
+  static __device__ constexpr int get_warp_size() {
+  #ifdef __GFX9__
+    return 64;
+  #else
+    return 32;
+  #endif
+  }
+};
+
+  #define WARP_SIZE Utils::get_warp_size()
+#else
+  #define WARP_SIZE 32
+#endif
+
+#ifndef USE_ROCM
+  #define VLLM_LDG(arg) __ldg(arg)
+#else
+  #define VLLM_LDG(arg) *(arg)
+#endif
+
+#ifndef USE_ROCM
+  #define VLLM_SHFL_XOR_SYNC(var, lane_mask) \
+    __shfl_xor_sync(uint32_t(-1), var, lane_mask)
+  #define VLLM_SHFL_XOR_SYNC_WIDTH(var, lane_mask, width) \
+    __shfl_xor_sync(uint32_t(-1), var, lane_mask, width)
+#else
+  #define VLLM_SHFL_XOR_SYNC(var, lane_mask) __shfl_xor(var, lane_mask)
+  #define VLLM_SHFL_XOR_SYNC_WIDTH(var, lane_mask, width) \
+    __shfl_xor(var, lane_mask, width)
+#endif
+
+#ifndef USE_ROCM
+  #define VLLM_SHFL_SYNC(var, src_lane) __shfl_sync(uint32_t(-1), var, src_lane)
+#else
+  #define VLLM_SHFL_SYNC(var, src_lane) __shfl(var, src_lane)
+#endif
+
+#ifndef USE_ROCM
+  #define VLLM_SHFL_DOWN_SYNC(var, lane_delta) \
+    __shfl_down_sync(uint32_t(-1), var, lane_delta)
+#else
+  #define VLLM_SHFL_DOWN_SYNC(var, lane_delta) __shfl_down(var, lane_delta)
+#endif
+
+#ifndef USE_ROCM
+  #define VLLM_DevFuncAttribute_SET_MaxDynamicSharedMemorySize(FUNC, VAL) \
+    cudaFuncSetAttribute(FUNC, cudaFuncAttributeMaxDynamicSharedMemorySize, VAL)
+#else
+  #define VLLM_DevFuncAttribute_SET_MaxDynamicSharedMemorySize(FUNC, VAL) \
+    hipFuncSetAttribute(FUNC, hipFuncAttributeMaxDynamicSharedMemorySize, VAL)
+#endif