Function: memcpy2d_async

Applicable Products

Product

Supported (√/x)

Ascend 950PR/Ascend 950DT

√

Atlas A3 training products/Atlas A3 inference products

√

Atlas A2 training products/Atlas A2 inference products

√

Atlas training products

√

Atlas inference products

√

Atlas 200I/500 A2 inference products

√

Description

Implements asynchronous memory copy, which is mainly used for matrix data replication. This is an asynchronous API.

The Atlas inference accelerator module does not support this API.

Prototype

  • C Prototype
    1
    aclError aclrtMemcpy2dAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, aclrtMemcpyKind kind, aclrtStream stream)
    
  • Python Function
    1
    ret = acl.rt.memcpy2d_async(dst, dpitch, src, spitch, width, height, kind, stream)
    

Parameters

Parameter

Description

dst

Int, pointer address of the destination memory address.

dpitch

Int, address distance between two adjacent columns of vectors in the destination memory.

src

Int, pointer address of the source memory address.

spitch

Int, address distance between two adjacent columns of vectors in the source memory.

width

Int, matrix width to be copied.

It must be less than or equal to dpitch and spitch. The maximum value of width is 5000000.

height

Int, matrix height to be copied.

The maximum height can be set to 5242880 (5 x 1024 x 1024). Otherwise, a failure is returned.

kind

Int, memory copy type.

ACL_MEMCPY_HOST_TO_HOST = 0 // Copy memory within the host.
ACL_MEMCPY_HOST_TO_DEVICE = 1 // Copy memory from the host to the device.
ACL_MEMCPY_DEVICE_TO_HOST = 2 // Copy memory from the device to the host.
ACL_MEMCPY_DEVICE_TO_DEVICE = 3 // Copy memory within a device.
ACL_MEMCPY_DEFAULT = 4 // The system determines the copy direction based on the source and destination memory addresses.

stream

Int, pointer address of the stream.

Return Values

Return Value

Description

ret

Int, error code. 0 on success; else, failure.

Restrictions

  • This API is asynchronous. The API call delivers a task rather than executes a task. After calling this API, you need to call a synchronization API, such as acl.rt.synchronize_stream, to ensure that the memory copy task is complete.
  • This API supports only the ACL_MEMCPY_HOST_TO_DEVICE, ACL_MEMCPY_DEVICE_TO_HOST, and ACL_MEMCPY_DEVICE_TO_DEVICE memory copy kinds. Different models support different memory copy kinds. For unsupported memory copy kinds, the API returns ACL_ERROR_INVALID_PARAM.
    • The following models support the ACL_MEMCPY_HOST_TO_DEVICE and ACL_MEMCPY_DEVICE_TO_HOST memory copy kinds:

      Ascend 950PR/Ascend 950DT

      Atlas A3 training products/Atlas A3 inference products

      Atlas A2 training products/Atlas A2 inference products

      Atlas 200I/500 A2 inference products

      Atlas inference products

      Atlas training products

    • The following models support the ACL_MEMCPY_DEVICE_TO_DEVICE memory copy kind:

      Ascend 950PR/Ascend 950DT

      Atlas A3 training products/Atlas A3 inference products

      Atlas A2 training products/Atlas A2 inference products

  • Atlas inference products in Open Ctrl CPU form: This API is not supported. Atlas inference accelerator module: This API is not supported.
  • Atlas 200I/500 A2 inference products in Ascend RC form: This API is not supported.