CopyToNpuSync

Function Usage

Copies data from the host to the device synchronously. The data copy is completed in blocking mode and will not be placed in the executor task queue. The device memory allocated in this API needs to be freed by the caller.

Prototype

const aclTensor *CopyToNpuSync(const aclTensor *src, aclOpExecutor *executor)

Parameters

Returns

aclTensor that points to the data copied to the device. If the task fails to be created, nullptr is returned.

Constraints

The input parameter pointer must not be null.

Example

1
2
3
4
5
6

// Initialize a tensor on the host and copy it to the device (dst is a tensor on the device).
void Func(aclOpExecutor *executor) {
    int64_t myArray[10];
    auto src = executor->ConvertToTensor(myArray, 10, DT_INT64);
    auto dst = CopyToNpuSync(src, executor);
}