aclplatformDevInfo

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
typedef enum aclplatformDevInfo {
    ACL_PLATFORM_AICORE_CNT        = 0,
    ACL_PLATFORM_AICORE_UB_SIZE    = 1,
    ACL_PLATFORM_CUBE_CORE_CNT     = 2,
    ACL_PLATFORM_VECTOR_CORE_CNT   = 3,
    ACL_PLATFORM_L2_SIZE           = 4,
    ACL_PLATFORM_MEMORY_SIZE       = 5,
    ACL_PLATFORM_CUBE_FREQ         = 6,
    ACL_PLATFORM_VEC_FREQ          = 7,
    ACL_PLATFORM_BT_SIZE           = 8,
    ACL_PLATFORM_L0_A_SIZE         = 9,
    ACL_PLATFORM_L0_B_SIZE         = 10,
    ACL_PLATFORM_L0_C_SIZE         = 11,
    ACL_PLATFORM_L1_SIZE           = 12,
    ACL_PLATFORM_SOC_VERSION       = 13,
    ACL_PLATFORM_AIC_VERSION       = 14,
    ACL_PLATFORM_NPU_ARCH          = 15,
    ACL_PLATFORM_MEMORY_TYPE       = 16,
} aclplatformDevInfo;

Enumeration

Description

ACL_PLATFORM_AICORE_CNT

Number of AI Cores.

ACL_PLATFORM_AICORE_UB_SIZE

Unified Buffer (UB) size of a single AI Core, in bytes.

ACL_PLATFORM_CUBE_CORE_CNT

Number of Cube Cores.

ACL_PLATFORM_VECTOR_CORE_CNT

Number of Vector Cores.

ACL_PLATFORM_L2_SIZE

Size of the L2 cache, in bytes.

ACL_PLATFORM_MEMORY_SIZE

Device memory size, in bytes.

ACL_PLATFORM_CUBE_FREQ

Operating frequency of the Cube Core, in MHz.

ACL_PLATFORM_VEC_FREQ

Operating frequency of the Vector Core, in MHz.

ACL_PLATFORM_BT_SIZE

Size of the Base Tile, in bytes.

ACL_PLATFORM_L0_A_SIZE

Size of the L0-A buffer, in bytes. It is used as the input buffer for the left matrix in Cube Core matrix operations.

ACL_PLATFORM_L0_B_SIZE

Size of the L0-B buffer, in bytes. It is used as the input buffer for the right matrix in Cube Core matrix operations.

ACL_PLATFORM_L0_C_SIZE

Size of the L0-C buffer, in bytes. It stores the output results of Cube Core matrix operations.

ACL_PLATFORM_L1_SIZE

Size of the L1 buffer, in bytes.

ACL_PLATFORM_SOC_VERSION

AI processor model.

ACL_PLATFORM_AIC_VERSION

AI Core version.

ACL_PLATFORM_NPU_ARCH

NPU architecture version. The returned value corresponds to the integer value of the aclplatformNpuArch enumeration.

The mapping between product models and NPU architecture versions is as follows:

  • Ascend 950PR/Ascend 950DT: 3510
  • Atlas A3 training products/Atlas A3 inference products: 2201
  • Atlas A2 training products/Atlas A2 inference products: 2201
  • Atlas 200I/500 A2 inference products: 3002
  • Atlas inference products: 2002
  • Atlas training products: 1001

ACL_PLATFORM_MEMORY_TYPE

Device memory type. The returned value is the integer value of the aclplatformLocalMemType enumeration.