Init (Deprecated)

This API has been deprecated and will be removed in later versions. Do not use this API. Use the InitV2 API for initialization.

Product	Supported
Atlas 350 Accelerator Card	√
Atlas A3 training product / Atlas A3 inference product	√
Atlas A2 training product / Atlas A2 inference product	√
Atlas 200I/500 A2 inference product	x
Atlas inference product AI Core	x
Atlas inference product Vector Core	x
Atlas training product	x

Initializes the HCCL client. By default, this API works on all cores. You can also specify a core by setting GetBlockIdx before calling this API.

      
           __aicore__ inline void Init(GM_ADDR context, __gm__ void *initTiling = nullptr)

**Table 1** API parameters
Parameter	Input/Output	Description
context	Input	Communication context, including information such as rankDim and rankID.
initTiling	Input	(Optional) Address for initializing the Mc2InitTiling in the communicator. Mc2InitTiling is computed on the host. For details, see Table 1 Mc2InitTiling parameters. This parameter is passed by the framework to the kernel function for use. For details about the complete example, see 8.13.1.2-Example.

None

If the initTiling parameter is passed when this API is called, this API must be used together with the SetCcTiling API.
An API with the optional parameter initTiling and another API without initTiling cannot be called at the same time in a program. You are advised to call an API with initTiling.
If initTiling is passed when this API is called, you must use the standard C++ syntax to define the development mode of the TilingData structure. For details, see Using the Standard C++ Syntax to Define the Tiling Structure.
This API does not support the initialization of multiple HCCL objects using the same context. Each HCCL object must obtain its own communication context.

Parent topic: HCCL Kernel APIs