6.14. CUpti_ActivityDevice5

struct CUpti_ActivityDevice5

The activity record for a device.

(CUDA 11.6 onwards)

This activity record represents information about a GPU device (CUPTI_ACTIVITY_KIND_DEVICE).

Public Members

CUpti_ActivityKind kind: The activity record kind, must be CUPTI_ACTIVITY_KIND_DEVICE.

CUpti_ActivityFlag flags

The flags associated with the device.

See also

CUpti_ActivityFlag

uint64_t globalMemoryBandwidth: The global memory bandwidth available on the device, in kBytes/sec.

uint64_t globalMemorySize: The amount of global memory on the device, in bytes.

uint32_t constantMemorySize: The amount of constant memory on the device, in bytes.

uint32_t l2CacheSize: The size of the L2 cache on the device, in bytes.

uint32_t numThreadsPerWarp: The number of threads per warp on the device.

uint32_t coreClockRate: The core clock rate of the device, in kHz.

uint32_t numMemcpyEngines: Number of memory copy engines on the device.

uint32_t numMultiprocessors: Number of multiprocessors on the device.

uint32_t maxIPC: The maximum “instructions per cycle” possible on each device multiprocessor.

uint32_t maxWarpsPerMultiprocessor: Maximum number of warps that can be present on a multiprocessor at any given time.

uint32_t maxBlocksPerMultiprocessor: Maximum number of blocks that can be present on a multiprocessor at any given time.

uint32_t maxSharedMemoryPerMultiprocessor: Maximum amount of shared memory available per multiprocessor, in bytes.

uint32_t maxRegistersPerMultiprocessor: Maximum number of 32-bit registers available per multiprocessor.

uint32_t maxRegistersPerBlock: Maximum number of registers that can be allocated to a block.

uint32_t maxSharedMemoryPerBlock: Maximum amount of shared memory that can be assigned to a block, in bytes.

uint32_t maxThreadsPerBlock: Maximum number of threads allowed in a block.

uint32_t maxBlockDimX: Maximum allowed X dimension for a block.

uint32_t maxBlockDimY: Maximum allowed Y dimension for a block.

uint32_t maxBlockDimZ: Maximum allowed Z dimension for a block.

uint32_t maxGridDimX: Maximum allowed X dimension for a grid.

uint32_t maxGridDimY: Maximum allowed Y dimension for a grid.

uint32_t maxGridDimZ: Maximum allowed Z dimension for a grid.

uint32_t computeCapabilityMajor: Compute capability for the device, major number.

uint32_t computeCapabilityMinor: Compute capability for the device, minor number.

uint32_t id: The device ID.

uint32_t eccEnabled: ECC enabled flag for device.

CUuuid uuid

The device UUID.

This value is the globally unique immutable alphanumeric identifier of the device.

const char *name

The device name.

This name is shared across all activity records representing instances of the device, and so should not be modified.

uint8_t isCudaVisible

Flag to indicate whether the device is visible to CUDA.

Users can set the device visibility using CUDA_VISIBLE_DEVICES environment

uint8_t isMigEnabled: MIG enabled flag for device.

uint32_t gpuInstanceId

GPU Instance id for MIG enabled devices.

If mig mode is disabled value is set to UINT32_MAX

uint32_t computeInstanceId

Compute Instance id for MIG enabled devices.

If mig mode is disabled value is set to UINT32_MAX

CUuuid migUuid

The MIG UUID.

This value is the globally unique immutable alphanumeric identifier of the device.

uint32_t isNumaNode: Numa (Non-uniform memory access) information for device GPU is a NUMA node or not.

uint32_t numaId: Numa (Non-uniform memory access) information for device NUMA node ID of the GPU memory if GPU is not a NUMA node, it returns invalidNumaId.