Serialize serializes the graph into a string. Each items of row is separated by comma (,), then each row is separated by semicolon (;). For example, a

Start register and patch periodically

Start health check and notify loop

Start starts the gRPC server, registers the device plugin with the Kubelet, and starts the device health checks

StartCollect starting periodic XPU information collection

Stop stops the gRPC server.

Topology() returns xpu topology information of the node.

WriteVxpuIdsConfig write vxpu ids assigned to the container to vxpu-ids.config

Beginning of GPU Architecture definitions

acquire acquires the limiting semaphore. Returns true if successfully acquired, false if the listener is closed or reach the max limit

allocates, but does NOT initialize. Use in-place new constructor, e.g. T* obj = pool.allocate(); ::new (static_cast<void*>(obj)) T();

backupName creates a backup name from the given name, inserting a timestamp between the filename and the extension, using the local time if requested

buildTopologyGraph builds topology graph for gpu. Currently, we get the GPU topology by parsing output of nvidia-smi output.

checkCommonParamValid 检查通用参数的有效性

CheckHealth performs health checks on a set of devices, writing to the 'unhealthy' channel with any unhealthy devices

compressLogFile compresses the given log file and removing src log file if successful.

todo: to be validated