[Kernel] add custom moe ops for prefill (#4194)

### What this PR does / why we need it?
1.Add the implementation of normal Aclnn operators: MoeCombineNormal,
MoeDispatchNormal, NotifyDispatch，and DispatchLayout.

- MoeCombineNormal: Implements the combine logic within MoE operations.
- MoeDispatchNormal: Implements the dispatch logic within MoE
operations.
- NotifyDispatch: Exchanges topk_idx information among different ranks
to calculate the device memory required for the dispatch stage.
- DispatchLayout: Used to calculate information related to the device
memory layout for the dispatch stage.

2.Provide PyTorch interfaces for normal operators—get_dispatch_layout,
dispatch_prefill, and combine_prefill—to be used for MoE communication
during the prefill stage in vLLM.

- get_dispatch_layout: Calculates information related to the device
memory layout for the dispatch operator, and is called before
dispatch_prefill.
- dispatch_prefill: Initiates the dispatch operation.
- combine_prefill: Initiates the combine operation.

### Does this PR introduce _any_ user-facing change?
No

### How was this patch tested?
The functionality has already been validated using the local Qwen model.
Test cases will be added after support for multi-NPU use cases in the CI
pipeline is finalized.


- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c

Signed-off-by: shiro-zzzz <zhangdianhao@huawei.com>

csrc/notify_dispatch/op_host/notify_dispatch.cpp

@@ -0,0 +1,60 @@
+#include "register/op_def_registry.h"
+
+namespace ops {
+class NotifyDispatch : public OpDef {
+public:
+    explicit NotifyDispatch(const char *name) : OpDef(name)
+    {
+        this->Input("sendData")
+            .ParamType(REQUIRED)
+            .DataType({ge::DT_FLOAT16, ge::DT_FLOAT, ge::DT_INT32})
+            .Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND})
+            .UnknownShapeFormat({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
+        this->Input("tokenPerExpertData")
+            .ParamType(REQUIRED)
+            .DataType({ge::DT_FLOAT16, ge::DT_FLOAT, ge::DT_INT32})
+            .Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND})
+            .UnknownShapeFormat({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
+        this->Output("sendDataOffset")
+            .ParamType(REQUIRED)
+            .DataType({ge::DT_FLOAT16, ge::DT_FLOAT, ge::DT_INT32})
+            .Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND})
+            .UnknownShapeFormat({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
+        this->Output("recvData")
+            .ParamType(REQUIRED)
+            .DataType({ge::DT_FLOAT16, ge::DT_FLOAT, ge::DT_INT32})
+            .Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND})
+            .UnknownShapeFormat({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
+
+        this->Attr("sendCount").Int();
+        this->Attr("num_tokens").Int();
+        this->Attr("comm_group").String();
+        this->Attr("rank_size").Int();
+        this->Attr("rank_id").Int();
+        this->Attr("local_rank_size").Int();
+        this->Attr("local_rank_id").Int();
+
+        OpAICoreConfig aicore_config_base;
+        aicore_config_base.DynamicCompileStaticFlag(true)
+            .DynamicFormatFlag(true)
+            .DynamicRankSupportFlag(true)
+            .DynamicShapeSupportFlag(true)
+            .NeedCheckSupportFlag(false)
+            .PrecisionReduceFlag(true)
+            .ExtendCfgInfo("aclnnSupport.value", "support_aclnn")
+            .ExtendCfgInfo("multiKernelSupportDynamicGraph.value", "multi_kernel");
+
+        OpAICoreConfig aicore_config_A2 = aicore_config_base;
+        aicore_config_A2.ExtendCfgInfo("jitCompile.flag", "static_false");
+
+        OpAICoreConfig aicore_config = aicore_config_base;
+        aicore_config.ExtendCfgInfo("jitCompile.flag", "static_true");
+
+        this->AICore().AddConfig("ascend910_93", aicore_config);
+        this->AICore().AddConfig("ascend910b", aicore_config_A2);
+        this->MC2().HcclGroup("comm_group");
+    }
+};
+
+OP_ADD(NotifyDispatch);
+}  // namespace ops