CUPTI :: CUPTI Documentation

6. Data Structures

Here are the data structures with brief descriptions:

BufferInfo: BufferInfo will be stored in the file for every buffer i.e for every call of UtilDumpPcSamplingBufferInFile() API
CUPTI::PcSamplingUtil::CUptiUtil_GetBufferInfoParams: Params for CuptiUtilGetBufferInfo
CUPTI::PcSamplingUtil::CUptiUtil_GetHeaderDataParams: Params for CuptiUtilGetHeaderData
CUPTI::PcSamplingUtil::CUptiUtil_GetPcSampDataParams: Params for CuptiUtilGetPcSampData
CUPTI::PcSamplingUtil::CUptiUtil_MergePcSampDataParams: Params for CuptiUtilMergePcSampData
CUPTI::PcSamplingUtil::CUptiUtil_PutPcSampDataParams: Params for CuptiUtilPutPcSampData
CUpti_Activity: The base activity record
CUpti_ActivityAPI: The activity record for a driver or runtime API invocation
CUpti_ActivityAutoBoostState: Device auto boost state structure
CUpti_ActivityBranch: The activity record for source level result branch. (deprecated)
CUpti_ActivityBranch2: The activity record for source level result branch
CUpti_ActivityCdpKernel: The activity record for CDP (CUDA Dynamic Parallelism) kernel
CUpti_ActivityContext: The activity record for a context
CUpti_ActivityCudaEvent: The activity record for CUDA event
CUpti_ActivityDevice: The activity record for a device. (deprecated)
CUpti_ActivityDevice2: The activity record for a device. (deprecated)
CUpti_ActivityDevice3: The activity record for a device. (CUDA 7.0 onwards)
CUpti_ActivityDevice4: The activity record for a device. (CUDA 11.6 onwards)
CUpti_ActivityDevice5: The activity record for a device. (CUDA 11.6 onwards)
CUpti_ActivityDeviceAttribute: The activity record for a device attribute
CUpti_ActivityEnvironment: The activity record for CUPTI environmental data
CUpti_ActivityEvent: The activity record for a CUPTI event
CUpti_ActivityEventInstance: The activity record for a CUPTI event with instance information
CUpti_ActivityExternalCorrelation: The activity record for correlation with external records
CUpti_ActivityFunction: The activity record for global/device functions
CUpti_ActivityGlobalAccess: The activity record for source-level global access. (deprecated)
CUpti_ActivityGlobalAccess2: The activity record for source-level global access. (deprecated in CUDA 9.0)
CUpti_ActivityGlobalAccess3: The activity record for source-level global access
CUpti_ActivityGraphTrace: The activity record for trace of graph execution
CUpti_ActivityInstantaneousEvent: The activity record for an instantaneous CUPTI event
CUpti_ActivityInstantaneousEventInstance: The activity record for an instantaneous CUPTI event with event domain instance information
CUpti_ActivityInstantaneousMetric: The activity record for an instantaneous CUPTI metric
CUpti_ActivityInstantaneousMetricInstance: The instantaneous activity record for a CUPTI metric with instance information
CUpti_ActivityInstructionCorrelation: The activity record for source-level sass/source line-by-line correlation
CUpti_ActivityInstructionExecution: The activity record for source-level instruction execution
CUpti_ActivityJit: The activity record for JIT operations. This activity represents the JIT operations (compile, load, store) of a CUmodule from the Compute Cache. Gives the exact hashed path of where the cached module is loaded from, or where the module will be stored after Just-In-Time (JIT) compilation
CUpti_ActivityKernel: The activity record for kernel. (deprecated)
CUpti_ActivityKernel2: The activity record for kernel. (deprecated)
CUpti_ActivityKernel3: The activity record for a kernel (CUDA 6.5(with sm_52 support) onwards). (deprecated in CUDA 9.0)
CUpti_ActivityKernel4: The activity record for a kernel (CUDA 9.0(with sm_70 support) onwards). (deprecated in CUDA 11.0)
CUpti_ActivityKernel5: The activity record for a kernel (CUDA 11.0(with sm_80 support) onwards). (deprecated in CUDA 11.2) This activity record represents a kernel execution (CUPTI_ACTIVITY_KIND_KERNEL and CUPTI_ACTIVITY_KIND_CONCURRENT_KERNEL) but is no longer generated by CUPTI. Kernel activities are now reported using the CUpti_ActivityKernel9 activity record
CUpti_ActivityKernel6: The activity record for kernel. (deprecated in CUDA 11.6)
CUpti_ActivityKernel7: The activity record for kernel. (deprecated in CUDA 11.8)
CUpti_ActivityKernel8: The activity record for kernel
CUpti_ActivityKernel9: The activity record for kernel
CUpti_ActivityMarker: The activity record providing a marker which is an instantaneous point in time. (deprecated in CUDA 8.0)
CUpti_ActivityMarker2: The activity record providing a marker which is an instantaneous point in time
CUpti_ActivityMarkerData: The activity record providing detailed information for a marker
CUpti_ActivityMemcpy: The activity record for memory copies. (deprecated)
CUpti_ActivityMemcpy3: The activity record for memory copies. (deprecated in CUDA 11.1)
CUpti_ActivityMemcpy4: The activity record for memory copies. (deprecated in CUDA 11.6)
CUpti_ActivityMemcpy5: The activity record for memory copies
CUpti_ActivityMemcpyPtoP: The activity record for peer-to-peer memory copies
CUpti_ActivityMemcpyPtoP2: The activity record for peer-to-peer memory copies. (deprecated in CUDA 11.1)
CUpti_ActivityMemcpyPtoP3: The activity record for peer-to-peer memory copies. (deprecated in CUDA 11.6)
CUpti_ActivityMemcpyPtoP4: The activity record for peer-to-peer memory copies
CUpti_ActivityMemory: The activity record for memory
CUpti_ActivityMemory2: The activity record for memory
CUpti_ActivityMemory3: The activity record for memory
CUpti_ActivityMemory3::CUpti_ActivityMemory3::PACKED_ALIGNMENT
CUpti_ActivityMemoryPool: The activity record for memory pool
CUpti_ActivityMemoryPool2: The activity record for memory pool
CUpti_ActivityMemset: The activity record for memset. (deprecated)
CUpti_ActivityMemset2: The activity record for memset. (deprecated in CUDA 11.1)
CUpti_ActivityMemset3: The activity record for memset. (deprecated in CUDA 11.6)
CUpti_ActivityMemset4: The activity record for memset
CUpti_ActivityMetric: The activity record for a CUPTI metric
CUpti_ActivityMetricInstance: The activity record for a CUPTI metric with instance information
CUpti_ActivityModule: The activity record for a CUDA module
CUpti_ActivityName: The activity record providing a name
CUpti_ActivityNvLink: NVLink information. (deprecated in CUDA 9.0)
CUpti_ActivityNvLink2: NVLink information. (deprecated in CUDA 10.0)
CUpti_ActivityNvLink3: NVLink information
CUpti_ActivityNvLink4: NVLink information
CUpti_ActivityObjectKindId: Identifiers for object kinds as specified by CUpti_ActivityObjectKind
CUpti_ActivityOpenAcc: The base activity record for OpenAcc records
CUpti_ActivityOpenAccData: The activity record for OpenACC data
CUpti_ActivityOpenAccLaunch: The activity record for OpenACC launch
CUpti_ActivityOpenAccOther: The activity record for OpenACC other
CUpti_ActivityOpenMp: The base activity record for OpenMp records
CUpti_ActivityOverhead: The activity record for CUPTI and driver overheads
CUpti_ActivityPcie: PCI devices information required to construct topology
CUpti_ActivityPCSampling: The activity record for PC sampling. (deprecated in CUDA 8.0)
CUpti_ActivityPCSampling2: The activity record for PC sampling. (deprecated in CUDA 9.0)
CUpti_ActivityPCSampling3: The activity record for PC sampling
CUpti_ActivityPCSamplingConfig: PC sampling configuration structure
CUpti_ActivityPCSamplingRecordInfo: The activity record for record status for PC sampling
CUpti_ActivityPreemption: The activity record for a preemption of a CDP kernel
CUpti_ActivitySharedAccess: The activity record for source-level shared access
CUpti_ActivitySourceLocator: The activity record for source locator
CUpti_ActivityStream: The activity record for CUDA stream
CUpti_ActivitySynchronization: The activity record for synchronization management
CUpti_ActivityUnifiedMemoryCounter: The activity record for Unified Memory counters (deprecated in CUDA 7.0)
CUpti_ActivityUnifiedMemoryCounter2: The activity record for Unified Memory counters (CUDA 7.0 and beyond)
CUpti_ActivityUnifiedMemoryCounterConfig: Unified Memory counters configuration structure
CUpti_CallbackData: Data passed into a runtime or driver API callback function
CUpti_EventGroupSet: A set of event groups
CUpti_EventGroupSets: A set of event group sets
CUpti_GetCubinCrcParams: Params for cuptiGetCubinCrc
CUpti_GetSassToSourceCorrelationParams: Params for cuptiGetSassToSourceCorrelation
CUpti_GraphData: CUDA graphs data passed into a resource callback function
CUpti_MetricValue: A metric value
CUpti_ModuleResourceData: Module data passed into a resource callback function
CUpti_NvtxData: Data passed into a NVTX callback function
CUpti_PCSamplingConfigurationInfo: PC sampling configuration information structure
CUpti_PCSamplingConfigurationInfoParams: PC sampling configuration structure
CUpti_PCSamplingData: Collected PC Sampling data
CUpti_PCSamplingDisableParams: Params for cuptiPCSamplingDisable
CUpti_PCSamplingEnableParams: Params for cuptiPCSamplingEnable
CUpti_PCSamplingGetDataParams: Params for cuptiPCSamplingEnable
CUpti_PCSamplingGetNumStallReasonsParams: Params for cuptiPCSamplingGetNumStallReasons
CUpti_PCSamplingGetStallReasonsParams: Params for cuptiPCSamplingGetStallReasons
CUpti_PCSamplingPCData: PC Sampling data
CUpti_PCSamplingStallReason: PC Sampling stall reasons
CUpti_PCSamplingStartParams: Params for cuptiPCSamplingStart
CUpti_PCSamplingStopParams: Params for cuptiPCSamplingStop
CUpti_Profiler_BeginPass_Params: Params for cuptiProfilerBeginPass
CUpti_Profiler_BeginSession_Params: Params for cuptiProfilerBeginSession
CUpti_Profiler_CounterDataImage_CalculateScratchBufferSize_Params: Params for cuptiProfilerCounterDataImageCalculateScratchBufferSize
CUpti_Profiler_CounterDataImage_CalculateSize_Params: Params for cuptiProfilerCounterDataImageCalculateSize
CUpti_Profiler_CounterDataImage_Initialize_Params: Params for cuptiProfilerCounterDataImageInitialize
CUpti_Profiler_CounterDataImage_InitializeScratchBuffer_Params: Params for cuptiProfilerCounterDataImageInitializeScratchBuffer
CUpti_Profiler_CounterDataImageOptions: Input parameter to define the counterDataImage
CUpti_Profiler_DeInitialize_Params: Default parameter for cuptiProfilerDeInitialize
CUpti_Profiler_DeviceSupported_Params: Params for cuptiProfilerDeviceSupported
CUpti_Profiler_DisableProfiling_Params: Params for cuptiProfilerDisableProfiling
CUpti_Profiler_EnableProfiling_Params: Params for cuptiProfilerEnableProfiling
CUpti_Profiler_EndPass_Params: Params for cuptiProfilerEndPass
CUpti_Profiler_EndSession_Params: Params for cuptiProfilerEndSession
CUpti_Profiler_FlushCounterData_Params: Params for cuptiProfilerFlushCounterData
CUpti_Profiler_GetCounterAvailability_Params: Params for cuptiProfilerGetCounterAvailability
CUpti_Profiler_Initialize_Params: Default parameter for cuptiProfilerInitialize
CUpti_Profiler_IsPassCollected_Params: Params for cuptiProfilerIsPassCollected
CUpti_Profiler_SetConfig_Params: Params for cuptiProfilerSetConfig
CUpti_Profiler_UnsetConfig_Params: Params for cuptiProfilerUnsetConfig
CUpti_ResourceData: Data passed into a resource callback function
CUpti_StateData: Data passed into a State callback function
CUpti_SynchronizeData: Data passed into a synchronize callback function
Header: Header info will be stored in file
NV::Cupti::Checkpoint::CUpti_Checkpoint: Configuration and handle for a CUPTI Checkpoint
PcSamplingStallReasons: All available stall reasons name and respective indexes will be stored in it

6.1. BufferInfo Struct Reference

[CUPTI PC Sampling Utility API]

Public Variables

uint64_t bufferByteSize
uint64_t numSelectedStallReasons
size_t numStallReasons
uint64_t recordCount

Variables

uint64_t BufferInfo::bufferByteSize [inherited]: Buffer size in Bytes.
uint64_t BufferInfo::numSelectedStallReasons [inherited]: Total number of stall reasons in single record.
size_t BufferInfo::numStallReasons [inherited]: Count of all stall reasons supported on the GPU
uint64_t BufferInfo::recordCount [inherited]: Total number of PC records.

6.2. CUPTI::PcSamplingUtil::CUptiUtil_GetBufferInfoParams Struct Reference

[CUPTI PC Sampling Utility API]

Public Variables

struct BufferInfo bufferInfoData
std::ifstream * fileHandler
size_t size

Variables

struct BufferInfo CUPTI::PcSamplingUtil::CUptiUtil_GetBufferInfoParams::bufferInfoData [inherited]: Buffer Info.
std::ifstream * CUPTI::PcSamplingUtil::CUptiUtil_GetBufferInfoParams::fileHandler [inherited]: File handle.
size_t CUPTI::PcSamplingUtil::CUptiUtil_GetBufferInfoParams::size [inherited]: Size of the data structure i.e. CUpti_PCSamplingDisableParamsSize CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are available in the structure. Used to preserve backward compatibility.

6.3. CUPTI::PcSamplingUtil::CUptiUtil_GetHeaderDataParams Struct Reference

[CUPTI PC Sampling Utility API]

Public Variables

std::ifstream * fileHandler
struct Header headerInfo
size_t size

Variables

std::ifstream * CUPTI::PcSamplingUtil::CUptiUtil_GetHeaderDataParams::fileHandler [inherited]: File handle.
struct Header CUPTI::PcSamplingUtil::CUptiUtil_GetHeaderDataParams::headerInfo [inherited]: Header Info.
size_t CUPTI::PcSamplingUtil::CUptiUtil_GetHeaderDataParams::size [inherited]: Size of the data structure i.e. CUpti_PCSamplingDisableParamsSize CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are available in the structure. Used to preserve backward compatibility.

6.4. CUPTI::PcSamplingUtil::CUptiUtil_GetPcSampDataParams Struct Reference

[CUPTI PC Sampling Utility API]

Public Variables

PcSamplingBufferType bufferType
std::ifstream * fileHandler
size_t numAttributes
BufferInfo * pBufferInfoData
CUpti_PCSamplingConfigurationInfo * pPCSamplingConfigurationInfo
PcSamplingStallReasons * pPcSamplingStallReasons
void * pSamplingData
size_t size

Variables

PcSamplingBufferType CUPTI::PcSamplingUtil::CUptiUtil_GetPcSampDataParams::bufferType [inherited]: Type of buffer to store in file
std::ifstream * CUPTI::PcSamplingUtil::CUptiUtil_GetPcSampDataParams::fileHandler [inherited]: File handle.
size_t CUPTI::PcSamplingUtil::CUptiUtil_GetPcSampDataParams::numAttributes [inherited]: Number of configuration attributes
BufferInfo * CUPTI::PcSamplingUtil::CUptiUtil_GetPcSampDataParams::pBufferInfoData [inherited]: Pointer to collected buffer info using CuptiUtilGetBufferInfo
CUpti_PCSamplingConfigurationInfo * CUPTI::PcSamplingUtil::CUptiUtil_GetPcSampDataParams::pPCSamplingConfigurationInfo [inherited]: Refer CUpti_PCSamplingConfigurationInfo
PcSamplingStallReasons * CUPTI::PcSamplingUtil::CUptiUtil_GetPcSampDataParams::pPcSamplingStallReasons [inherited]: Refer PcSamplingStallReasons. For stallReasons field of PcSamplingStallReasons it is expected to allocate memory for each string element of array.
void * CUPTI::PcSamplingUtil::CUptiUtil_GetPcSampDataParams::pSamplingData [inherited]: Pointer to allocated memory to store retrieved data from file.
size_t CUPTI::PcSamplingUtil::CUptiUtil_GetPcSampDataParams::size [inherited]: Size of the data structure i.e. CUpti_PCSamplingDisableParamsSize CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are available in the structure. Used to preserve backward compatibility.

6.5. CUPTI::PcSamplingUtil::CUptiUtil_MergePcSampDataParams Struct Reference

[CUPTI PC Sampling Utility API]

Public Variables

* MergedPcSampDataBuffers
CUpti_PCSamplingData * PcSampDataBuffer
size_t * numMergedBuffer
size_t numberOfBuffers
size_t size

Variables

* CUPTI::PcSamplingUtil::CUptiUtil_MergePcSampDataParams::MergedPcSampDataBuffers [inherited]: Pointer to array of merged buffers as per the range id.
CUpti_PCSamplingData * CUPTI::PcSamplingUtil::CUptiUtil_MergePcSampDataParams::PcSampDataBuffer [inherited]: Pointer to array of buffers to merge
size_t * CUPTI::PcSamplingUtil::CUptiUtil_MergePcSampDataParams::numMergedBuffer [inherited]: Number of merged buffers.
size_t CUPTI::PcSamplingUtil::CUptiUtil_MergePcSampDataParams::numberOfBuffers [inherited]: Number of buffers to merge.
size_t CUPTI::PcSamplingUtil::CUptiUtil_MergePcSampDataParams::size [inherited]: Size of the data structure i.e. CUpti_PCSamplingDisableParamsSize CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are available in the structure. Used to preserve backward compatibility.

6.6. CUPTI::PcSamplingUtil::CUptiUtil_PutPcSampDataParams Struct Reference

[CUPTI PC Sampling Utility API]

Public Variables

PcSamplingBufferType bufferType
const char * fileName
size_t numAttributes
CUpti_PCSamplingConfigurationInfo * pPCSamplingConfigurationInfo
PcSamplingStallReasons * pPcSamplingStallReasons
void * pSamplingData
size_t size

Variables

PcSamplingBufferType CUPTI::PcSamplingUtil::CUptiUtil_PutPcSampDataParams::bufferType [inherited]: Type of buffer to store in file
const char * CUPTI::PcSamplingUtil::CUptiUtil_PutPcSampDataParams::fileName [inherited]: File name to store buffer into it.
size_t CUPTI::PcSamplingUtil::CUptiUtil_PutPcSampDataParams::numAttributes [inherited]: Number of configured attributes
CUpti_PCSamplingConfigurationInfo * CUPTI::PcSamplingUtil::CUptiUtil_PutPcSampDataParams::pPCSamplingConfigurationInfo [inherited]: Refer CUpti_PCSamplingConfigurationInfo It is expected to provide configuration details of at least CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_STALL_REASON attribute.
PcSamplingStallReasons * CUPTI::PcSamplingUtil::CUptiUtil_PutPcSampDataParams::pPcSamplingStallReasons [inherited]: Refer PcSamplingStallReasons.
void * CUPTI::PcSamplingUtil::CUptiUtil_PutPcSampDataParams::pSamplingData [inherited]: PC sampling buffer.
size_t CUPTI::PcSamplingUtil::CUptiUtil_PutPcSampDataParams::size [inherited]: Size of the data structure i.e. CUpti_PCSamplingDisableParamsSize CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are available in the structure. Used to preserve backward compatibility.

6.7. CUpti_Activity Struct Reference

[CUPTI Activity API]

The activity API uses a CUpti_Activity as a generic representation for any activity. The 'kind' field is used to determine the specific activity kind, and from that the CUpti_Activity object can be cast to the specific activity record type appropriate for that kind.

Note that all activity record types are padded and aligned to ensure that each member of the record is naturally aligned.

See also:

CUpti_ActivityKind

Public Variables

CUpti_ActivityKind kind

Variables

CUpti_ActivityKind CUpti_Activity::kind [inherited]: The kind of this activity.

6.8. CUpti_ActivityAPI Struct Reference

[CUPTI Activity API]

This activity record represents an invocation of a driver or runtime API (CUPTI_ACTIVITY_KIND_DRIVER and CUPTI_ACTIVITY_KIND_RUNTIME).

Public Variables

CUpti_CallbackId cbid
uint32_t correlationId
uint64_t end
CUpti_ActivityKind kind
uint32_t processId
uint32_t returnValue
uint64_t start
uint32_t threadId

Variables

CUpti_CallbackId CUpti_ActivityAPI::cbid [inherited]: The ID of the driver or runtime function.
uint32_t CUpti_ActivityAPI::correlationId [inherited]: The correlation ID of the driver or runtime CUDA function. Each function invocation is assigned a unique correlation ID that is identical to the correlation ID in the memcpy, memset, or kernel activity record that is associated with this function.
uint64_t CUpti_ActivityAPI::end [inherited]: The end timestamp for the function, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the function.
CUpti_ActivityKind CUpti_ActivityAPI::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_DRIVER, CUPTI_ACTIVITY_KIND_RUNTIME, or CUPTI_ACTIVITY_KIND_INTERNAL_LAUNCH_API.
uint32_t CUpti_ActivityAPI::processId [inherited]: The ID of the process where the driver or runtime CUDA function is executing.
uint32_t CUpti_ActivityAPI::returnValue [inherited]: The return value for the function. For a CUDA driver function with will be a CUresult value, and for a CUDA runtime function this will be a cudaError_t value.
uint64_t CUpti_ActivityAPI::start [inherited]: The start timestamp for the function, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the function.
uint32_t CUpti_ActivityAPI::threadId [inherited]: The ID of the thread where the driver or runtime CUDA function is executing.

6.9. CUpti_ActivityAutoBoostState Struct Reference

[CUPTI Activity API]

This structure defines auto boost state for a device. See function cuptiGetAutoBoostState

Public Variables

uint32_t enabled
uint32_t pid

Variables

uint32_t CUpti_ActivityAutoBoostState::enabled [inherited]: Returned auto boost state. 1 is returned in case auto boost is enabled, 0 otherwise
uint32_t CUpti_ActivityAutoBoostState::pid [inherited]: Id of process that has set the current boost state. The value will be CUPTI_AUTO_BOOST_INVALID_CLIENT_PID if the user does not have the permission to query process ids or there is an error in querying the process id.

6.10. CUpti_ActivityBranch Struct Reference

[CUPTI Activity API]

This activity record the locations of the branches in the source (CUPTI_ACTIVITY_KIND_BRANCH). Branch activities are now reported using the CUpti_ActivityBranch2 activity record.

Public Variables

uint32_t correlationId
uint32_t diverged
uint32_t executed
CUpti_ActivityKind kind
uint32_t pcOffset
uint32_t sourceLocatorId
uint64_t threadsExecuted

Variables

uint32_t CUpti_ActivityBranch::correlationId [inherited]: The correlation ID of the kernel to which this result is associated.
uint32_t CUpti_ActivityBranch::diverged [inherited]: Number of times this branch diverged
uint32_t CUpti_ActivityBranch::executed [inherited]: The number of times this instruction was executed per warp. It will be incremented regardless of predicate or condition code.
CUpti_ActivityKind CUpti_ActivityBranch::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_BRANCH.
uint32_t CUpti_ActivityBranch::pcOffset [inherited]: The pc offset for the branch.
uint32_t CUpti_ActivityBranch::sourceLocatorId [inherited]: The ID for source locator.
uint64_t CUpti_ActivityBranch::threadsExecuted [inherited]: This increments each time when this instruction is executed by number of threads that executed this instruction

6.11. CUpti_ActivityBranch2 Struct Reference

[CUPTI Activity API]

This activity record the locations of the branches in the source (CUPTI_ACTIVITY_KIND_BRANCH).

Public Variables

uint32_t correlationId
uint32_t diverged
uint32_t executed
uint32_t functionId
CUpti_ActivityKind kind
uint32_t pad
uint32_t pcOffset
uint32_t sourceLocatorId
uint64_t threadsExecuted

Variables

uint32_t CUpti_ActivityBranch2::correlationId [inherited]: The correlation ID of the kernel to which this result is associated.
uint32_t CUpti_ActivityBranch2::diverged [inherited]: Number of times this branch diverged
uint32_t CUpti_ActivityBranch2::executed [inherited]: The number of times this instruction was executed per warp. It will be incremented regardless of predicate or condition code.
uint32_t CUpti_ActivityBranch2::functionId [inherited]: Correlation ID with global/device function name
CUpti_ActivityKind CUpti_ActivityBranch2::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_BRANCH.
uint32_t CUpti_ActivityBranch2::pad [inherited]: Undefined. Reserved for internal use.
uint32_t CUpti_ActivityBranch2::pcOffset [inherited]: The pc offset for the branch.
uint32_t CUpti_ActivityBranch2::sourceLocatorId [inherited]: The ID for source locator.
uint64_t CUpti_ActivityBranch2::threadsExecuted [inherited]: This increments each time when this instruction is executed by number of threads that executed this instruction

6.12. CUpti_ActivityCdpKernel Struct Reference

[CUPTI Activity API]

This activity record represents a CDP kernel execution.

Public Variables

int32_t blockX
int32_t blockY
int32_t blockZ
uint64_t completed
uint32_t contextId
uint32_t correlationId
uint32_t deviceId
int32_t dynamicSharedMemory
uint64_t end
uint8_t executed
int64_t gridId
int32_t gridX
int32_t gridY
int32_t gridZ
CUpti_ActivityKind kind
uint32_t localMemoryPerThread
uint32_t localMemoryTotal
const char * name
uint32_t parentBlockX
uint32_t parentBlockY
uint32_t parentBlockZ
int64_t parentGridId
uint64_t queued
uint16_t registersPerThread
uint8_t requested
uint8_t sharedMemoryConfig
uint64_t start
int32_t staticSharedMemory
uint32_t streamId
uint64_t submitted

Variables

int32_t CUpti_ActivityCdpKernel::blockX [inherited]: The X-dimension block size for the kernel.
int32_t CUpti_ActivityCdpKernel::blockY [inherited]: The Y-dimension block size for the kernel.
int32_t CUpti_ActivityCdpKernel::blockZ [inherited]: The Z-dimension grid size for the kernel.
uint64_t CUpti_ActivityCdpKernel::completed [inherited]: The timestamp when kernel is marked as completed, in ns. A value of CUPTI_TIMESTAMP_UNKNOWN indicates that the completion time is unknown.
uint32_t CUpti_ActivityCdpKernel::contextId [inherited]: The ID of the context where the kernel is executing.
uint32_t CUpti_ActivityCdpKernel::correlationId [inherited]: The correlation ID of the kernel. Each kernel execution is assigned a unique correlation ID that is identical to the correlation ID in the driver API activity record that launched the kernel.
uint32_t CUpti_ActivityCdpKernel::deviceId [inherited]: The ID of the device where the kernel is executing.
int32_t CUpti_ActivityCdpKernel::dynamicSharedMemory [inherited]: The dynamic shared memory reserved for the kernel, in bytes.
uint64_t CUpti_ActivityCdpKernel::end [inherited]: The end timestamp for the kernel execution, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the kernel.
uint8_t CUpti_ActivityCdpKernel::executed [inherited]: The cache configuration used for the kernel. The value is one of the CUfunc_cache enumeration values from cuda.h.
int64_t CUpti_ActivityCdpKernel::gridId [inherited]: The grid ID of the kernel. Each kernel execution is assigned a unique grid ID.
int32_t CUpti_ActivityCdpKernel::gridX [inherited]: The X-dimension grid size for the kernel.
int32_t CUpti_ActivityCdpKernel::gridY [inherited]: The Y-dimension grid size for the kernel.
int32_t CUpti_ActivityCdpKernel::gridZ [inherited]: The Z-dimension grid size for the kernel.
CUpti_ActivityKind CUpti_ActivityCdpKernel::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_CDP_KERNEL
uint32_t CUpti_ActivityCdpKernel::localMemoryPerThread [inherited]: The amount of local memory reserved for each thread, in bytes.
uint32_t CUpti_ActivityCdpKernel::localMemoryTotal [inherited]: The total amount of local memory reserved for the kernel, in bytes.
const char * CUpti_ActivityCdpKernel::name [inherited]: The name of the kernel. This name is shared across all activity records representing the same kernel, and so should not be modified.
uint32_t CUpti_ActivityCdpKernel::parentBlockX [inherited]: The X-dimension of the parent block.
uint32_t CUpti_ActivityCdpKernel::parentBlockY [inherited]: The Y-dimension of the parent block.
uint32_t CUpti_ActivityCdpKernel::parentBlockZ [inherited]: The Z-dimension of the parent block.
int64_t CUpti_ActivityCdpKernel::parentGridId [inherited]: The grid ID of the parent kernel.
uint64_t CUpti_ActivityCdpKernel::queued [inherited]: The timestamp when kernel is queued up, in ns. A value of CUPTI_TIMESTAMP_UNKNOWN indicates that the queued time is unknown.
uint16_t CUpti_ActivityCdpKernel::registersPerThread [inherited]: The number of registers required for each thread executing the kernel.
uint8_t CUpti_ActivityCdpKernel::requested [inherited]: The cache configuration requested by the kernel. The value is one of the CUfunc_cache enumeration values from cuda.h.
uint8_t CUpti_ActivityCdpKernel::sharedMemoryConfig [inherited]: The shared memory configuration used for the kernel. The value is one of the CUsharedconfig enumeration values from cuda.h.
uint64_t CUpti_ActivityCdpKernel::start [inherited]: The start timestamp for the kernel execution, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the kernel.
int32_t CUpti_ActivityCdpKernel::staticSharedMemory [inherited]: The static shared memory allocated for the kernel, in bytes.
uint32_t CUpti_ActivityCdpKernel::streamId [inherited]: The ID of the stream where the kernel is executing.
uint64_t CUpti_ActivityCdpKernel::submitted [inherited]: The timestamp when kernel is submitted to the gpu, in ns. A value of CUPTI_TIMESTAMP_UNKNOWN indicates that the submission time is unknown.

6.13. CUpti_ActivityContext Struct Reference

[CUPTI Activity API]

This activity record represents information about a context (CUPTI_ACTIVITY_KIND_CONTEXT).

Public Variables

uint16_t computeApiKind
uint32_t contextId
uint32_t deviceId
CUpti_ActivityKind kind
uint16_t nullStreamId

Variables

uint16_t CUpti_ActivityContext::computeApiKind [inherited]: The compute API kind.

See also:

CUpti_ActivityComputeApiKind
uint32_t CUpti_ActivityContext::contextId [inherited]: The context ID.
uint32_t CUpti_ActivityContext::deviceId [inherited]: The device ID.
CUpti_ActivityKind CUpti_ActivityContext::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_CONTEXT.
uint16_t CUpti_ActivityContext::nullStreamId [inherited]: The ID for the NULL stream in this context

6.14. CUpti_ActivityCudaEvent Struct Reference

[CUPTI Activity API]

This activity is used to track recorded events. (CUPTI_ACTIVITY_KIND_CUDA_EVENT).

Public Variables

uint32_t contextId
uint32_t correlationId
uint32_t eventId
CUpti_ActivityKind kind
uint32_t pad
uint32_t streamId

Variables

uint32_t CUpti_ActivityCudaEvent::contextId [inherited]: The ID of the context where the event was recorded.
uint32_t CUpti_ActivityCudaEvent::correlationId [inherited]: The correlation ID of the API to which this result is associated.
uint32_t CUpti_ActivityCudaEvent::eventId [inherited]: A unique event ID to identify the event record.
CUpti_ActivityKind CUpti_ActivityCudaEvent::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_CUDA_EVENT.
uint32_t CUpti_ActivityCudaEvent::pad [inherited]: Undefined. Reserved for internal use.
uint32_t CUpti_ActivityCudaEvent::streamId [inherited]: The compute stream where the event was recorded.

6.15. CUpti_ActivityDevice Struct Reference

[CUPTI Activity API]

This activity record represents information about a GPU device (CUPTI_ACTIVITY_KIND_DEVICE). Device activity is now reported using the CUpti_ActivityDevice5 activity record.

Public Variables

uint32_t computeCapabilityMajor
uint32_t computeCapabilityMinor
uint32_t constantMemorySize
uint32_t coreClockRate
CUpti_ActivityFlag flags
uint64_t globalMemoryBandwidth
uint64_t globalMemorySize
uint32_t id
CUpti_ActivityKind kind
uint32_t l2CacheSize
uint32_t maxBlockDimX
uint32_t maxBlockDimY
uint32_t maxBlockDimZ
uint32_t maxBlocksPerMultiprocessor
uint32_t maxGridDimX
uint32_t maxGridDimY
uint32_t maxGridDimZ
uint32_t maxIPC
uint32_t maxRegistersPerBlock
uint32_t maxSharedMemoryPerBlock
uint32_t maxThreadsPerBlock
uint32_t maxWarpsPerMultiprocessor
const char * name
uint32_t numMemcpyEngines
uint32_t numMultiprocessors
uint32_t numThreadsPerWarp

Variables

uint32_t CUpti_ActivityDevice::computeCapabilityMajor [inherited]: Compute capability for the device, major number.
uint32_t CUpti_ActivityDevice::computeCapabilityMinor [inherited]: Compute capability for the device, minor number.
uint32_t CUpti_ActivityDevice::constantMemorySize [inherited]: The amount of constant memory on the device, in bytes.
uint32_t CUpti_ActivityDevice::coreClockRate [inherited]: The core clock rate of the device, in kHz.
CUpti_ActivityFlag CUpti_ActivityDevice::flags [inherited]: The flags associated with the device.

See also:

CUpti_ActivityFlag
uint64_t CUpti_ActivityDevice::globalMemoryBandwidth [inherited]: The global memory bandwidth available on the device, in kBytes/sec.
uint64_t CUpti_ActivityDevice::globalMemorySize [inherited]: The amount of global memory on the device, in bytes.
uint32_t CUpti_ActivityDevice::id [inherited]: The device ID.
CUpti_ActivityKind CUpti_ActivityDevice::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_DEVICE.
uint32_t CUpti_ActivityDevice::l2CacheSize [inherited]: The size of the L2 cache on the device, in bytes.
uint32_t CUpti_ActivityDevice::maxBlockDimX [inherited]: Maximum allowed X dimension for a block.
uint32_t CUpti_ActivityDevice::maxBlockDimY [inherited]: Maximum allowed Y dimension for a block.
uint32_t CUpti_ActivityDevice::maxBlockDimZ [inherited]: Maximum allowed Z dimension for a block.
uint32_t CUpti_ActivityDevice::maxBlocksPerMultiprocessor [inherited]: Maximum number of blocks that can be present on a multiprocessor at any given time.
uint32_t CUpti_ActivityDevice::maxGridDimX [inherited]: Maximum allowed X dimension for a grid.
uint32_t CUpti_ActivityDevice::maxGridDimY [inherited]: Maximum allowed Y dimension for a grid.
uint32_t CUpti_ActivityDevice::maxGridDimZ [inherited]: Maximum allowed Z dimension for a grid.
uint32_t CUpti_ActivityDevice::maxIPC [inherited]: The maximum "instructions per cycle" possible on each device multiprocessor.
uint32_t CUpti_ActivityDevice::maxRegistersPerBlock [inherited]: Maximum number of registers that can be allocated to a block.
uint32_t CUpti_ActivityDevice::maxSharedMemoryPerBlock [inherited]: Maximum amount of shared memory that can be assigned to a block, in bytes.
uint32_t CUpti_ActivityDevice::maxThreadsPerBlock [inherited]: Maximum number of threads allowed in a block.
uint32_t CUpti_ActivityDevice::maxWarpsPerMultiprocessor [inherited]: Maximum number of warps that can be present on a multiprocessor at any given time.
const char * CUpti_ActivityDevice::name [inherited]: The device name. This name is shared across all activity records representing instances of the device, and so should not be modified.
uint32_t CUpti_ActivityDevice::numMemcpyEngines [inherited]: Number of memory copy engines on the device.
uint32_t CUpti_ActivityDevice::numMultiprocessors [inherited]: Number of multiprocessors on the device.
uint32_t CUpti_ActivityDevice::numThreadsPerWarp [inherited]: The number of threads per warp on the device.

6.16. CUpti_ActivityDevice2 Struct Reference

[CUPTI Activity API]

This activity record represents information about a GPU device (CUPTI_ACTIVITY_KIND_DEVICE). Device activity is now reported using the CUpti_ActivityDevice5 activity record.

Public Variables

uint32_t computeCapabilityMajor
uint32_t computeCapabilityMinor
uint32_t constantMemorySize
uint32_t coreClockRate
uint32_t eccEnabled
CUpti_ActivityFlag flags
uint64_t globalMemoryBandwidth
uint64_t globalMemorySize
uint32_t id
CUpti_ActivityKind kind
uint32_t l2CacheSize
uint32_t maxBlockDimX
uint32_t maxBlockDimY
uint32_t maxBlockDimZ
uint32_t maxBlocksPerMultiprocessor
uint32_t maxGridDimX
uint32_t maxGridDimY
uint32_t maxGridDimZ
uint32_t maxIPC
uint32_t maxRegistersPerBlock
uint32_t maxRegistersPerMultiprocessor
uint32_t maxSharedMemoryPerBlock
uint32_t maxSharedMemoryPerMultiprocessor
uint32_t maxThreadsPerBlock
uint32_t maxWarpsPerMultiprocessor
const char * name
uint32_t numMemcpyEngines
uint32_t numMultiprocessors
uint32_t numThreadsPerWarp
uint32_t pad
CUuuid uuid

Variables

uint32_t CUpti_ActivityDevice2::computeCapabilityMajor [inherited]: Compute capability for the device, major number.
uint32_t CUpti_ActivityDevice2::computeCapabilityMinor [inherited]: Compute capability for the device, minor number.
uint32_t CUpti_ActivityDevice2::constantMemorySize [inherited]: The amount of constant memory on the device, in bytes.
uint32_t CUpti_ActivityDevice2::coreClockRate [inherited]: The core clock rate of the device, in kHz.
uint32_t CUpti_ActivityDevice2::eccEnabled [inherited]: ECC enabled flag for device
CUpti_ActivityFlag CUpti_ActivityDevice2::flags [inherited]: The flags associated with the device.

See also:

CUpti_ActivityFlag
uint64_t CUpti_ActivityDevice2::globalMemoryBandwidth [inherited]: The global memory bandwidth available on the device, in kBytes/sec.
uint64_t CUpti_ActivityDevice2::globalMemorySize [inherited]: The amount of global memory on the device, in bytes.
uint32_t CUpti_ActivityDevice2::id [inherited]: The device ID.
CUpti_ActivityKind CUpti_ActivityDevice2::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_DEVICE.
uint32_t CUpti_ActivityDevice2::l2CacheSize [inherited]: The size of the L2 cache on the device, in bytes.
uint32_t CUpti_ActivityDevice2::maxBlockDimX [inherited]: Maximum allowed X dimension for a block.
uint32_t CUpti_ActivityDevice2::maxBlockDimY [inherited]: Maximum allowed Y dimension for a block.
uint32_t CUpti_ActivityDevice2::maxBlockDimZ [inherited]: Maximum allowed Z dimension for a block.
uint32_t CUpti_ActivityDevice2::maxBlocksPerMultiprocessor [inherited]: Maximum number of blocks that can be present on a multiprocessor at any given time.
uint32_t CUpti_ActivityDevice2::maxGridDimX [inherited]: Maximum allowed X dimension for a grid.
uint32_t CUpti_ActivityDevice2::maxGridDimY [inherited]: Maximum allowed Y dimension for a grid.
uint32_t CUpti_ActivityDevice2::maxGridDimZ [inherited]: Maximum allowed Z dimension for a grid.
uint32_t CUpti_ActivityDevice2::maxIPC [inherited]: The maximum "instructions per cycle" possible on each device multiprocessor.
uint32_t CUpti_ActivityDevice2::maxRegistersPerBlock [inherited]: Maximum number of registers that can be allocated to a block.
uint32_t CUpti_ActivityDevice2::maxRegistersPerMultiprocessor [inherited]: Maximum number of 32-bit registers available per multiprocessor.
uint32_t CUpti_ActivityDevice2::maxSharedMemoryPerBlock [inherited]: Maximum amount of shared memory that can be assigned to a block, in bytes.
uint32_t CUpti_ActivityDevice2::maxSharedMemoryPerMultiprocessor [inherited]: Maximum amount of shared memory available per multiprocessor, in bytes.
uint32_t CUpti_ActivityDevice2::maxThreadsPerBlock [inherited]: Maximum number of threads allowed in a block.
uint32_t CUpti_ActivityDevice2::maxWarpsPerMultiprocessor [inherited]: Maximum number of warps that can be present on a multiprocessor at any given time.
const char * CUpti_ActivityDevice2::name [inherited]: The device name. This name is shared across all activity records representing instances of the device, and so should not be modified.
uint32_t CUpti_ActivityDevice2::numMemcpyEngines [inherited]: Number of memory copy engines on the device.
uint32_t CUpti_ActivityDevice2::numMultiprocessors [inherited]: Number of multiprocessors on the device.
uint32_t CUpti_ActivityDevice2::numThreadsPerWarp [inherited]: The number of threads per warp on the device.
uint32_t CUpti_ActivityDevice2::pad [inherited]: Undefined. Reserved for internal use.
CUuuid CUpti_ActivityDevice2::uuid [inherited]: The device UUID. This value is the globally unique immutable alphanumeric identifier of the device.

6.17. CUpti_ActivityDevice3 Struct Reference

[CUPTI Activity API]

This activity record represents information about a GPU device (CUPTI_ACTIVITY_KIND_DEVICE). Device activity is now reported using the CUpti_ActivityDevice5 activity record.

Public Variables

uint32_t computeCapabilityMajor
uint32_t computeCapabilityMinor
uint32_t constantMemorySize
uint32_t coreClockRate
uint32_t eccEnabled
CUpti_ActivityFlag flags
uint64_t globalMemoryBandwidth
uint64_t globalMemorySize
uint32_t id
uint8_t isCudaVisible
CUpti_ActivityKind kind
uint32_t l2CacheSize
uint32_t maxBlockDimX
uint32_t maxBlockDimY
uint32_t maxBlockDimZ
uint32_t maxBlocksPerMultiprocessor
uint32_t maxGridDimX
uint32_t maxGridDimY
uint32_t maxGridDimZ
uint32_t maxIPC
uint32_t maxRegistersPerBlock
uint32_t maxRegistersPerMultiprocessor
uint32_t maxSharedMemoryPerBlock
uint32_t maxSharedMemoryPerMultiprocessor
uint32_t maxThreadsPerBlock
uint32_t maxWarpsPerMultiprocessor
const char * name
uint32_t numMemcpyEngines
uint32_t numMultiprocessors
uint32_t numThreadsPerWarp
uint32_t pad
CUuuid uuid

Variables

uint32_t CUpti_ActivityDevice3::computeCapabilityMajor [inherited]: Compute capability for the device, major number.
uint32_t CUpti_ActivityDevice3::computeCapabilityMinor [inherited]: Compute capability for the device, minor number.
uint32_t CUpti_ActivityDevice3::constantMemorySize [inherited]: The amount of constant memory on the device, in bytes.
uint32_t CUpti_ActivityDevice3::coreClockRate [inherited]: The core clock rate of the device, in kHz.
uint32_t CUpti_ActivityDevice3::eccEnabled [inherited]: ECC enabled flag for device
CUpti_ActivityFlag CUpti_ActivityDevice3::flags [inherited]: The flags associated with the device.

See also:

CUpti_ActivityFlag
uint64_t CUpti_ActivityDevice3::globalMemoryBandwidth [inherited]: The global memory bandwidth available on the device, in kBytes/sec.
uint64_t CUpti_ActivityDevice3::globalMemorySize [inherited]: The amount of global memory on the device, in bytes.
uint32_t CUpti_ActivityDevice3::id [inherited]: The device ID.
uint8_t CUpti_ActivityDevice3::isCudaVisible [inherited]: Flag to indicate whether the device is visible to CUDA. Users can set the device visibility using CUDA_VISIBLE_DEVICES environment
CUpti_ActivityKind CUpti_ActivityDevice3::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_DEVICE.
uint32_t CUpti_ActivityDevice3::l2CacheSize [inherited]: The size of the L2 cache on the device, in bytes.
uint32_t CUpti_ActivityDevice3::maxBlockDimX [inherited]: Maximum allowed X dimension for a block.
uint32_t CUpti_ActivityDevice3::maxBlockDimY [inherited]: Maximum allowed Y dimension for a block.
uint32_t CUpti_ActivityDevice3::maxBlockDimZ [inherited]: Maximum allowed Z dimension for a block.
uint32_t CUpti_ActivityDevice3::maxBlocksPerMultiprocessor [inherited]: Maximum number of blocks that can be present on a multiprocessor at any given time.
uint32_t CUpti_ActivityDevice3::maxGridDimX [inherited]: Maximum allowed X dimension for a grid.
uint32_t CUpti_ActivityDevice3::maxGridDimY [inherited]: Maximum allowed Y dimension for a grid.
uint32_t CUpti_ActivityDevice3::maxGridDimZ [inherited]: Maximum allowed Z dimension for a grid.
uint32_t CUpti_ActivityDevice3::maxIPC [inherited]: The maximum "instructions per cycle" possible on each device multiprocessor.
uint32_t CUpti_ActivityDevice3::maxRegistersPerBlock [inherited]: Maximum number of registers that can be allocated to a block.
uint32_t CUpti_ActivityDevice3::maxRegistersPerMultiprocessor [inherited]: Maximum number of 32-bit registers available per multiprocessor.
uint32_t CUpti_ActivityDevice3::maxSharedMemoryPerBlock [inherited]: Maximum amount of shared memory that can be assigned to a block, in bytes.
uint32_t CUpti_ActivityDevice3::maxSharedMemoryPerMultiprocessor [inherited]: Maximum amount of shared memory available per multiprocessor, in bytes.
uint32_t CUpti_ActivityDevice3::maxThreadsPerBlock [inherited]: Maximum number of threads allowed in a block.
uint32_t CUpti_ActivityDevice3::maxWarpsPerMultiprocessor [inherited]: Maximum number of warps that can be present on a multiprocessor at any given time.
const char * CUpti_ActivityDevice3::name [inherited]: The device name. This name is shared across all activity records representing instances of the device, and so should not be modified.
uint32_t CUpti_ActivityDevice3::numMemcpyEngines [inherited]: Number of memory copy engines on the device.
uint32_t CUpti_ActivityDevice3::numMultiprocessors [inherited]: Number of multiprocessors on the device.
uint32_t CUpti_ActivityDevice3::numThreadsPerWarp [inherited]: The number of threads per warp on the device.
uint32_t CUpti_ActivityDevice3::pad [inherited]: Undefined. Reserved for internal use.
CUuuid CUpti_ActivityDevice3::uuid [inherited]: The device UUID. This value is the globally unique immutable alphanumeric identifier of the device.

6.18. CUpti_ActivityDevice4 Struct Reference

[CUPTI Activity API]

This activity record represents information about a GPU device (CUPTI_ACTIVITY_KIND_DEVICE). Device activity is now reported using the CUpti_ActivityDevice5 activity record.

Public Variables

uint32_t computeCapabilityMajor
uint32_t computeCapabilityMinor
uint32_t computeInstanceId
uint32_t constantMemorySize
uint32_t coreClockRate
uint32_t eccEnabled
CUpti_ActivityFlag flags
uint64_t globalMemoryBandwidth
uint64_t globalMemorySize
uint32_t gpuInstanceId
uint32_t id
uint8_t isCudaVisible
uint8_t isMigEnabled
CUpti_ActivityKind kind
uint32_t l2CacheSize
uint32_t maxBlockDimX
uint32_t maxBlockDimY
uint32_t maxBlockDimZ
uint32_t maxBlocksPerMultiprocessor
uint32_t maxGridDimX
uint32_t maxGridDimY
uint32_t maxGridDimZ
uint32_t maxIPC
uint32_t maxRegistersPerBlock
uint32_t maxRegistersPerMultiprocessor
uint32_t maxSharedMemoryPerBlock
uint32_t maxSharedMemoryPerMultiprocessor
uint32_t maxThreadsPerBlock
uint32_t maxWarpsPerMultiprocessor
CUuuid migUuid
const char * name
uint32_t numMemcpyEngines
uint32_t numMultiprocessors
uint32_t numThreadsPerWarp
uint32_t pad
CUuuid uuid

Variables

uint32_t CUpti_ActivityDevice4::computeCapabilityMajor [inherited]: Compute capability for the device, major number.
uint32_t CUpti_ActivityDevice4::computeCapabilityMinor [inherited]: Compute capability for the device, minor number.
uint32_t CUpti_ActivityDevice4::computeInstanceId [inherited]: Compute Instance id for MIG enabled devices. If mig mode is disabled value is set to UINT32_MAX
uint32_t CUpti_ActivityDevice4::constantMemorySize [inherited]: The amount of constant memory on the device, in bytes.
uint32_t CUpti_ActivityDevice4::coreClockRate [inherited]: The core clock rate of the device, in kHz.
uint32_t CUpti_ActivityDevice4::eccEnabled [inherited]: ECC enabled flag for device
CUpti_ActivityFlag CUpti_ActivityDevice4::flags [inherited]: The flags associated with the device.

See also:

CUpti_ActivityFlag
uint64_t CUpti_ActivityDevice4::globalMemoryBandwidth [inherited]: The global memory bandwidth available on the device, in kBytes/sec.
uint64_t CUpti_ActivityDevice4::globalMemorySize [inherited]: The amount of global memory on the device, in bytes.
uint32_t CUpti_ActivityDevice4::gpuInstanceId [inherited]: GPU Instance id for MIG enabled devices. If mig mode is disabled value is set to UINT32_MAX
uint32_t CUpti_ActivityDevice4::id [inherited]: The device ID.
uint8_t CUpti_ActivityDevice4::isCudaVisible [inherited]: Flag to indicate whether the device is visible to CUDA. Users can set the device visibility using CUDA_VISIBLE_DEVICES environment
uint8_t CUpti_ActivityDevice4::isMigEnabled [inherited]: MIG enabled flag for device
CUpti_ActivityKind CUpti_ActivityDevice4::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_DEVICE.
uint32_t CUpti_ActivityDevice4::l2CacheSize [inherited]: The size of the L2 cache on the device, in bytes.
uint32_t CUpti_ActivityDevice4::maxBlockDimX [inherited]: Maximum allowed X dimension for a block.
uint32_t CUpti_ActivityDevice4::maxBlockDimY [inherited]: Maximum allowed Y dimension for a block.
uint32_t CUpti_ActivityDevice4::maxBlockDimZ [inherited]: Maximum allowed Z dimension for a block.
uint32_t CUpti_ActivityDevice4::maxBlocksPerMultiprocessor [inherited]: Maximum number of blocks that can be present on a multiprocessor at any given time.
uint32_t CUpti_ActivityDevice4::maxGridDimX [inherited]: Maximum allowed X dimension for a grid.
uint32_t CUpti_ActivityDevice4::maxGridDimY [inherited]: Maximum allowed Y dimension for a grid.
uint32_t CUpti_ActivityDevice4::maxGridDimZ [inherited]: Maximum allowed Z dimension for a grid.
uint32_t CUpti_ActivityDevice4::maxIPC [inherited]: The maximum "instructions per cycle" possible on each device multiprocessor.
uint32_t CUpti_ActivityDevice4::maxRegistersPerBlock [inherited]: Maximum number of registers that can be allocated to a block.
uint32_t CUpti_ActivityDevice4::maxRegistersPerMultiprocessor [inherited]: Maximum number of 32-bit registers available per multiprocessor.
uint32_t CUpti_ActivityDevice4::maxSharedMemoryPerBlock [inherited]: Maximum amount of shared memory that can be assigned to a block, in bytes.
uint32_t CUpti_ActivityDevice4::maxSharedMemoryPerMultiprocessor [inherited]: Maximum amount of shared memory available per multiprocessor, in bytes.
uint32_t CUpti_ActivityDevice4::maxThreadsPerBlock [inherited]: Maximum number of threads allowed in a block.
uint32_t CUpti_ActivityDevice4::maxWarpsPerMultiprocessor [inherited]: Maximum number of warps that can be present on a multiprocessor at any given time.
CUuuid CUpti_ActivityDevice4::migUuid [inherited]: The MIG UUID. This value is the globally unique immutable alphanumeric identifier of the device.
const char * CUpti_ActivityDevice4::name [inherited]: The device name. This name is shared across all activity records representing instances of the device, and so should not be modified.
uint32_t CUpti_ActivityDevice4::numMemcpyEngines [inherited]: Number of memory copy engines on the device.
uint32_t CUpti_ActivityDevice4::numMultiprocessors [inherited]: Number of multiprocessors on the device.
uint32_t CUpti_ActivityDevice4::numThreadsPerWarp [inherited]: The number of threads per warp on the device.
uint32_t CUpti_ActivityDevice4::pad [inherited]: Undefined. Reserved for internal use.
CUuuid CUpti_ActivityDevice4::uuid [inherited]: The device UUID. This value is the globally unique immutable alphanumeric identifier of the device.

6.19. CUpti_ActivityDevice5 Struct Reference

[CUPTI Activity API]

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

Public Variables

uint32_t computeCapabilityMajor
uint32_t computeCapabilityMinor
uint32_t computeInstanceId
uint32_t constantMemorySize
uint32_t coreClockRate
uint32_t eccEnabled
CUpti_ActivityFlag flags
uint64_t globalMemoryBandwidth
uint64_t globalMemorySize
uint32_t gpuInstanceId
uint32_t id
uint8_t isCudaVisible
uint8_t isMigEnabled
uint32_t isNumaNode
CUpti_ActivityKind kind
uint32_t l2CacheSize
uint32_t maxBlockDimX
uint32_t maxBlockDimY
uint32_t maxBlockDimZ
uint32_t maxBlocksPerMultiprocessor
uint32_t maxGridDimX
uint32_t maxGridDimY
uint32_t maxGridDimZ
uint32_t maxIPC
uint32_t maxRegistersPerBlock
uint32_t maxRegistersPerMultiprocessor
uint32_t maxSharedMemoryPerBlock
uint32_t maxSharedMemoryPerMultiprocessor
uint32_t maxThreadsPerBlock
uint32_t maxWarpsPerMultiprocessor
CUuuid migUuid
const char * name
uint32_t numMemcpyEngines
uint32_t numMultiprocessors
uint32_t numThreadsPerWarp
uint32_t numaId
uint32_t pad
CUuuid uuid

Variables

uint32_t CUpti_ActivityDevice5::computeCapabilityMajor [inherited]: Compute capability for the device, major number.
uint32_t CUpti_ActivityDevice5::computeCapabilityMinor [inherited]: Compute capability for the device, minor number.
uint32_t CUpti_ActivityDevice5::computeInstanceId [inherited]: Compute Instance id for MIG enabled devices. If mig mode is disabled value is set to UINT32_MAX
uint32_t CUpti_ActivityDevice5::constantMemorySize [inherited]: The amount of constant memory on the device, in bytes.
uint32_t CUpti_ActivityDevice5::coreClockRate [inherited]: The core clock rate of the device, in kHz.
uint32_t CUpti_ActivityDevice5::eccEnabled [inherited]: ECC enabled flag for device
CUpti_ActivityFlag CUpti_ActivityDevice5::flags [inherited]: The flags associated with the device.

See also:

CUpti_ActivityFlag
uint64_t CUpti_ActivityDevice5::globalMemoryBandwidth [inherited]: The global memory bandwidth available on the device, in kBytes/sec.
uint64_t CUpti_ActivityDevice5::globalMemorySize [inherited]: The amount of global memory on the device, in bytes.
uint32_t CUpti_ActivityDevice5::gpuInstanceId [inherited]: GPU Instance id for MIG enabled devices. If mig mode is disabled value is set to UINT32_MAX
uint32_t CUpti_ActivityDevice5::id [inherited]: The device ID.
uint8_t CUpti_ActivityDevice5::isCudaVisible [inherited]: Flag to indicate whether the device is visible to CUDA. Users can set the device visibility using CUDA_VISIBLE_DEVICES environment
uint8_t CUpti_ActivityDevice5::isMigEnabled [inherited]: MIG enabled flag for device
uint32_t CUpti_ActivityDevice5::isNumaNode [inherited]: Numa (Non-uniform memory access) information for device GPU is a NUMA node or not
CUpti_ActivityKind CUpti_ActivityDevice5::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_DEVICE.
uint32_t CUpti_ActivityDevice5::l2CacheSize [inherited]: The size of the L2 cache on the device, in bytes.
uint32_t CUpti_ActivityDevice5::maxBlockDimX [inherited]: Maximum allowed X dimension for a block.
uint32_t CUpti_ActivityDevice5::maxBlockDimY [inherited]: Maximum allowed Y dimension for a block.
uint32_t CUpti_ActivityDevice5::maxBlockDimZ [inherited]: Maximum allowed Z dimension for a block.
uint32_t CUpti_ActivityDevice5::maxBlocksPerMultiprocessor [inherited]: Maximum number of blocks that can be present on a multiprocessor at any given time.
uint32_t CUpti_ActivityDevice5::maxGridDimX [inherited]: Maximum allowed X dimension for a grid.
uint32_t CUpti_ActivityDevice5::maxGridDimY [inherited]: Maximum allowed Y dimension for a grid.
uint32_t CUpti_ActivityDevice5::maxGridDimZ [inherited]: Maximum allowed Z dimension for a grid.
uint32_t CUpti_ActivityDevice5::maxIPC [inherited]: The maximum "instructions per cycle" possible on each device multiprocessor.
uint32_t CUpti_ActivityDevice5::maxRegistersPerBlock [inherited]: Maximum number of registers that can be allocated to a block.
uint32_t CUpti_ActivityDevice5::maxRegistersPerMultiprocessor [inherited]: Maximum number of 32-bit registers available per multiprocessor.
uint32_t CUpti_ActivityDevice5::maxSharedMemoryPerBlock [inherited]: Maximum amount of shared memory that can be assigned to a block, in bytes.
uint32_t CUpti_ActivityDevice5::maxSharedMemoryPerMultiprocessor [inherited]: Maximum amount of shared memory available per multiprocessor, in bytes.
uint32_t CUpti_ActivityDevice5::maxThreadsPerBlock [inherited]: Maximum number of threads allowed in a block.
uint32_t CUpti_ActivityDevice5::maxWarpsPerMultiprocessor [inherited]: Maximum number of warps that can be present on a multiprocessor at any given time.
CUuuid CUpti_ActivityDevice5::migUuid [inherited]: The MIG UUID. This value is the globally unique immutable alphanumeric identifier of the device.
const char * CUpti_ActivityDevice5::name [inherited]: The device name. This name is shared across all activity records representing instances of the device, and so should not be modified.
uint32_t CUpti_ActivityDevice5::numMemcpyEngines [inherited]: Number of memory copy engines on the device.
uint32_t CUpti_ActivityDevice5::numMultiprocessors [inherited]: Number of multiprocessors on the device.
uint32_t CUpti_ActivityDevice5::numThreadsPerWarp [inherited]: The number of threads per warp on the device.
uint32_t CUpti_ActivityDevice5::numaId [inherited]: 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
uint32_t CUpti_ActivityDevice5::pad [inherited]: Undefined. Reserved for internal use.
CUuuid CUpti_ActivityDevice5::uuid [inherited]: The device UUID. This value is the globally unique immutable alphanumeric identifier of the device.

6.20. CUpti_ActivityDeviceAttribute Struct Reference

[CUPTI Activity API]

This activity record represents information about a GPU device: either a CUpti_DeviceAttribute or CUdevice_attribute value (CUPTI_ACTIVITY_KIND_DEVICE_ATTRIBUTE).

Public Variables

CUpti_ActivityDeviceAttribute::@23 attribute
uint32_t deviceId
CUpti_ActivityFlag flags
CUpti_ActivityKind kind
CUpti_ActivityDeviceAttribute::@24 value

Variables

CUpti_ActivityDeviceAttribute::@23 CUpti_ActivityDeviceAttribute::attribute [inherited]: The attribute, either a CUpti_DeviceAttribute or CUdevice_attribute. Flag CUPTI_ACTIVITY_FLAG_DEVICE_ATTRIBUTE_CUDEVICE is used to indicate what kind of attribute this is. If CUPTI_ACTIVITY_FLAG_DEVICE_ATTRIBUTE_CUDEVICE is 1 then CUdevice_attribute field is value, otherwise CUpti_DeviceAttribute field is valid.
uint32_t CUpti_ActivityDeviceAttribute::deviceId [inherited]: The ID of the device that this attribute applies to.
CUpti_ActivityFlag CUpti_ActivityDeviceAttribute::flags [inherited]: The flags associated with the device.

See also:

CUpti_ActivityFlag
CUpti_ActivityKind CUpti_ActivityDeviceAttribute::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_DEVICE_ATTRIBUTE.
CUpti_ActivityDeviceAttribute::@24 CUpti_ActivityDeviceAttribute::value [inherited]: The value for the attribute. See CUpti_DeviceAttribute and CUdevice_attribute for the type of the value for a given attribute.

6.21. CUpti_ActivityEnvironment Struct Reference

[CUPTI Activity API]

This activity record provides CUPTI environmental data, include power, clocks, and thermals. This information is sampled at various rates and returned in this activity record. The consumer of the record needs to check the environmentKind field to figure out what kind of environmental record this is.

Public Variables

CUpti_EnvironmentClocksThrottleReason clocksThrottleReasons
CUpti_ActivityEnvironment::@25::@29 cooling
uint32_t deviceId
CUpti_ActivityEnvironmentKind environmentKind
uint32_t fanSpeed
uint32_t gpuTemperature
CUpti_ActivityKind kind
uint32_t memoryClock
uint32_t pcieLinkGen
uint32_t pcieLinkWidth
CUpti_ActivityEnvironment::@25::@28 power
uint32_t power
uint32_t powerLimit
uint32_t smClock
CUpti_ActivityEnvironment::@25::@26 speed
CUpti_ActivityEnvironment::@25::@27 temperature
uint64_t timestamp

Variables

CUpti_EnvironmentClocksThrottleReason CUpti_ActivityEnvironment::clocksThrottleReasons [inherited]: The clocks throttle reasons.
CUpti_ActivityEnvironment::@25::@29 CUpti_ActivityEnvironment::cooling [inherited]: Data returned for CUPTI_ACTIVITY_ENVIRONMENT_COOLING environment kind.
uint32_t CUpti_ActivityEnvironment::deviceId [inherited]: The ID of the device
CUpti_ActivityEnvironmentKind CUpti_ActivityEnvironment::environmentKind [inherited]: The kind of data reported in this record.
uint32_t CUpti_ActivityEnvironment::fanSpeed [inherited]: The fan speed as percentage of maximum.
uint32_t CUpti_ActivityEnvironment::gpuTemperature [inherited]: The GPU temperature in degrees C.
CUpti_ActivityKind CUpti_ActivityEnvironment::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_ENVIRONMENT.
uint32_t CUpti_ActivityEnvironment::memoryClock [inherited]: The memory frequency in MHz
uint32_t CUpti_ActivityEnvironment::pcieLinkGen [inherited]: The PCIe link generation.
uint32_t CUpti_ActivityEnvironment::pcieLinkWidth [inherited]: The PCIe link width.
CUpti_ActivityEnvironment::@25::@28 CUpti_ActivityEnvironment::power [inherited]: Data returned for CUPTI_ACTIVITY_ENVIRONMENT_POWER environment kind.
uint32_t CUpti_ActivityEnvironment::power [inherited]: The power in milliwatts consumed by GPU and associated circuitry.
uint32_t CUpti_ActivityEnvironment::powerLimit [inherited]: The power in milliwatts that will trigger power management algorithm.
uint32_t CUpti_ActivityEnvironment::smClock [inherited]: The SM frequency in MHz
CUpti_ActivityEnvironment::@25::@26 CUpti_ActivityEnvironment::speed [inherited]: Data returned for CUPTI_ACTIVITY_ENVIRONMENT_SPEED environment kind.
CUpti_ActivityEnvironment::@25::@27 CUpti_ActivityEnvironment::temperature [inherited]: Data returned for CUPTI_ACTIVITY_ENVIRONMENT_TEMPERATURE environment kind.
uint64_t CUpti_ActivityEnvironment::timestamp [inherited]: The timestamp when this sample was retrieved, in ns. A value of 0 indicates that timestamp information could not be collected for the marker.

6.22. CUpti_ActivityEvent Struct Reference

[CUPTI Activity API]

This activity record represents a CUPTI event value (CUPTI_ACTIVITY_KIND_EVENT). This activity record kind is not produced by the activity API but is included for completeness and ease-of-use. Profile frameworks built on top of CUPTI that collect event data may choose to use this type to store the collected event data.

Public Variables

uint32_t correlationId
CUpti_EventDomainID domain
CUpti_EventID id
CUpti_ActivityKind kind
uint64_t value

Variables

uint32_t CUpti_ActivityEvent::correlationId [inherited]: The correlation ID of the event. Use of this ID is user-defined, but typically this ID value will equal the correlation ID of the kernel for which the event was gathered.
CUpti_EventDomainID CUpti_ActivityEvent::domain [inherited]: The event domain ID.
CUpti_EventID CUpti_ActivityEvent::id [inherited]: The event ID.
CUpti_ActivityKind CUpti_ActivityEvent::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_EVENT.
uint64_t CUpti_ActivityEvent::value [inherited]: The event value.

6.23. CUpti_ActivityEventInstance Struct Reference

[CUPTI Activity API]

This activity record represents the a CUPTI event value for a specific event domain instance (CUPTI_ACTIVITY_KIND_EVENT_INSTANCE). This activity record kind is not produced by the activity API but is included for completeness and ease-of-use. Profile frameworks built on top of CUPTI that collect event data may choose to use this type to store the collected event data. This activity record should be used when event domain instance information needs to be associated with the event.

Public Variables

uint32_t correlationId
CUpti_EventDomainID domain
CUpti_EventID id
uint32_t instance
CUpti_ActivityKind kind
uint32_t pad
uint64_t value

Variables

uint32_t CUpti_ActivityEventInstance::correlationId [inherited]: The correlation ID of the event. Use of this ID is user-defined, but typically this ID value will equal the correlation ID of the kernel for which the event was gathered.
CUpti_EventDomainID CUpti_ActivityEventInstance::domain [inherited]: The event domain ID.
CUpti_EventID CUpti_ActivityEventInstance::id [inherited]: The event ID.
uint32_t CUpti_ActivityEventInstance::instance [inherited]: The event domain instance.
CUpti_ActivityKind CUpti_ActivityEventInstance::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_EVENT_INSTANCE.
uint32_t CUpti_ActivityEventInstance::pad [inherited]: Undefined. Reserved for internal use.
uint64_t CUpti_ActivityEventInstance::value [inherited]: The event value.

6.24. CUpti_ActivityExternalCorrelation Struct Reference

[CUPTI Activity API]

This activity record correlates native CUDA records (e.g. CUDA Driver API, kernels, memcpys, ...) with records from external APIs such as OpenACC. (CUPTI_ACTIVITY_KIND_EXTERNAL_CORRELATION).

See also:

CUpti_ActivityKind

Public Variables

uint32_t correlationId
uint64_t externalId
CUpti_ExternalCorrelationKind externalKind
CUpti_ActivityKind kind
uint32_t reserved

Variables

uint32_t CUpti_ActivityExternalCorrelation::correlationId [inherited]: The correlation ID of the associated CUDA driver or runtime API record.
uint64_t CUpti_ActivityExternalCorrelation::externalId [inherited]: The correlation ID of the associated non-CUDA API record. The exact field in the associated external record depends on that record's activity kind (

See also:

externalKind).
CUpti_ExternalCorrelationKind CUpti_ActivityExternalCorrelation::externalKind [inherited]: The kind of external API this record correlated to.
CUpti_ActivityKind CUpti_ActivityExternalCorrelation::kind [inherited]: The kind of this activity.
uint32_t CUpti_ActivityExternalCorrelation::reserved [inherited]: Undefined. Reserved for internal use.

6.25. CUpti_ActivityFunction Struct Reference

[CUPTI Activity API]

This activity records function name and corresponding module information. (CUPTI_ACTIVITY_KIND_FUNCTION).

Public Variables

uint32_t contextId
uint32_t functionIndex
uint32_t id
CUpti_ActivityKind kind
uint32_t moduleId
const char * name

Variables

uint32_t CUpti_ActivityFunction::contextId [inherited]: The ID of the context where the function is launched.
uint32_t CUpti_ActivityFunction::functionIndex [inherited]: The function's unique symbol index in the module.
uint32_t CUpti_ActivityFunction::id [inherited]: ID to uniquely identify the record
CUpti_ActivityKind CUpti_ActivityFunction::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_FUNCTION.
uint32_t CUpti_ActivityFunction::moduleId [inherited]: The module ID in which this global/device function is present.
const char * CUpti_ActivityFunction::name [inherited]: The name of the function. This name is shared across all activity records representing the same kernel, and so should not be modified.

6.26. CUpti_ActivityGlobalAccess Struct Reference

[CUPTI Activity API]

This activity records the locations of the global accesses in the source (CUPTI_ACTIVITY_KIND_GLOBAL_ACCESS). Global access activities are now reported using the CUpti_ActivityGlobalAccess3 activity record.

Public Variables

uint32_t correlationId
uint32_t executed
CUpti_ActivityFlag flags
CUpti_ActivityKind kind
uint64_t l2_transactions
uint32_t pcOffset
uint32_t sourceLocatorId
uint64_t threadsExecuted

Variables

uint32_t CUpti_ActivityGlobalAccess::correlationId [inherited]: The correlation ID of the kernel to which this result is associated.
uint32_t CUpti_ActivityGlobalAccess::executed [inherited]: The number of times this instruction was executed per warp. It will be incremented when at least one of thread among warp is active with predicate and condition code evaluating to true.
CUpti_ActivityFlag CUpti_ActivityGlobalAccess::flags [inherited]: The properties of this global access.
CUpti_ActivityKind CUpti_ActivityGlobalAccess::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_GLOBAL_ACCESS.
uint64_t CUpti_ActivityGlobalAccess::l2_transactions [inherited]: The total number of 32 bytes transactions to L2 cache generated by this access
uint32_t CUpti_ActivityGlobalAccess::pcOffset [inherited]: The pc offset for the access.
uint32_t CUpti_ActivityGlobalAccess::sourceLocatorId [inherited]: The ID for source locator.
uint64_t CUpti_ActivityGlobalAccess::threadsExecuted [inherited]: This increments each time when this instruction is executed by number of threads that executed this instruction with predicate and condition code evaluating to true.

6.27. CUpti_ActivityGlobalAccess2 Struct Reference

[CUPTI Activity API]

Public Variables

uint32_t correlationId
uint32_t executed
CUpti_ActivityFlag flags
uint32_t functionId
CUpti_ActivityKind kind
uint64_t l2_transactions
uint32_t pad
uint32_t pcOffset
uint32_t sourceLocatorId
uint64_t theoreticalL2Transactions
uint64_t threadsExecuted

Variables

uint32_t CUpti_ActivityGlobalAccess2::correlationId [inherited]: The correlation ID of the kernel to which this result is associated.
uint32_t CUpti_ActivityGlobalAccess2::executed [inherited]: The number of times this instruction was executed per warp. It will be incremented when at least one of thread among warp is active with predicate and condition code evaluating to true.
CUpti_ActivityFlag CUpti_ActivityGlobalAccess2::flags [inherited]: The properties of this global access.
uint32_t CUpti_ActivityGlobalAccess2::functionId [inherited]: Correlation ID with global/device function name
CUpti_ActivityKind CUpti_ActivityGlobalAccess2::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_GLOBAL_ACCESS.
uint64_t CUpti_ActivityGlobalAccess2::l2_transactions [inherited]: The total number of 32 bytes transactions to L2 cache generated by this access
uint32_t CUpti_ActivityGlobalAccess2::pad [inherited]: Undefined. Reserved for internal use.
uint32_t CUpti_ActivityGlobalAccess2::pcOffset [inherited]: The pc offset for the access.
uint32_t CUpti_ActivityGlobalAccess2::sourceLocatorId [inherited]: The ID for source locator.
uint64_t CUpti_ActivityGlobalAccess2::theoreticalL2Transactions [inherited]: The minimum number of L2 transactions possible based on the access pattern.
uint64_t CUpti_ActivityGlobalAccess2::threadsExecuted [inherited]: This increments each time when this instruction is executed by number of threads that executed this instruction with predicate and condition code evaluating to true.

6.28. CUpti_ActivityGlobalAccess3 Struct Reference

[CUPTI Activity API]

This activity records the locations of the global accesses in the source (CUPTI_ACTIVITY_KIND_GLOBAL_ACCESS).

Public Variables

uint32_t correlationId
uint32_t executed
CUpti_ActivityFlag flags
uint32_t functionId
CUpti_ActivityKind kind
uint64_t l2_transactions
uint64_t pcOffset
uint32_t sourceLocatorId
uint64_t theoreticalL2Transactions
uint64_t threadsExecuted

Variables

uint32_t CUpti_ActivityGlobalAccess3::correlationId [inherited]: The correlation ID of the kernel to which this result is associated.
uint32_t CUpti_ActivityGlobalAccess3::executed [inherited]: The number of times this instruction was executed per warp. It will be incremented when at least one of thread among warp is active with predicate and condition code evaluating to true.
CUpti_ActivityFlag CUpti_ActivityGlobalAccess3::flags [inherited]: The properties of this global access.
uint32_t CUpti_ActivityGlobalAccess3::functionId [inherited]: Correlation ID with global/device function name
CUpti_ActivityKind CUpti_ActivityGlobalAccess3::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_GLOBAL_ACCESS.
uint64_t CUpti_ActivityGlobalAccess3::l2_transactions [inherited]: The total number of 32 bytes transactions to L2 cache generated by this access
uint64_t CUpti_ActivityGlobalAccess3::pcOffset [inherited]: The pc offset for the access.
uint32_t CUpti_ActivityGlobalAccess3::sourceLocatorId [inherited]: The ID for source locator.
uint64_t CUpti_ActivityGlobalAccess3::theoreticalL2Transactions [inherited]: The minimum number of L2 transactions possible based on the access pattern.
uint64_t CUpti_ActivityGlobalAccess3::threadsExecuted [inherited]: This increments each time when this instruction is executed by number of threads that executed this instruction with predicate and condition code evaluating to true.

6.29. CUpti_ActivityGraphTrace Struct Reference

[CUPTI Activity API]

This activity record represents execution for a graph without giving visibility about the execution of its nodes. This is intended to reduce overheads in tracing each node. The activity kind is CUPTI_ACTIVITY_KIND_GRAPH_TRACE

Public Variables

uint32_t contextId
uint32_t correlationId
uint32_t deviceId
uint64_t end
uint32_t graphId
CUpti_ActivityKind kind
void * reserved
uint64_t start
uint32_t streamId

Variables

uint32_t CUpti_ActivityGraphTrace::contextId [inherited]: The ID of the context where the graph is being launched.
uint32_t CUpti_ActivityGraphTrace::correlationId [inherited]: The correlation ID of the graph launch. Each graph launch is assigned a unique correlation ID that is identical to the correlation ID in the driver API activity record that launched the graph.
uint32_t CUpti_ActivityGraphTrace::deviceId [inherited]: The ID of the device where the graph execution is occurring.
uint64_t CUpti_ActivityGraphTrace::end [inherited]: The end timestamp for the graph execution, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the graph.
uint32_t CUpti_ActivityGraphTrace::graphId [inherited]: The unique ID of the graph that is launched.
CUpti_ActivityKind CUpti_ActivityGraphTrace::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_GRAPH_TRACE
void * CUpti_ActivityGraphTrace::reserved [inherited]: This field is reserved for internal use
uint64_t CUpti_ActivityGraphTrace::start [inherited]: The start timestamp for the graph execution, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the graph.
uint32_t CUpti_ActivityGraphTrace::streamId [inherited]: The ID of the stream where the graph is being launched.

6.30. CUpti_ActivityInstantaneousEvent Struct Reference

[CUPTI Activity API]

This activity record represents a CUPTI event value (CUPTI_ACTIVITY_KIND_EVENT) sampled at a particular instant. This activity record kind is not produced by the activity API but is included for completeness and ease-of-use. Profiler frameworks built on top of CUPTI that collect event data at a particular time may choose to use this type to store the collected event data.

Public Variables

uint32_t deviceId
CUpti_EventID id
CUpti_ActivityKind kind
uint32_t reserved
uint64_t timestamp
uint64_t value

Variables

uint32_t CUpti_ActivityInstantaneousEvent::deviceId [inherited]: The device id
CUpti_EventID CUpti_ActivityInstantaneousEvent::id [inherited]: The event ID.
CUpti_ActivityKind CUpti_ActivityInstantaneousEvent::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_INSTANTANEOUS_EVENT.
uint32_t CUpti_ActivityInstantaneousEvent::reserved [inherited]: Undefined. reserved for internal use
uint64_t CUpti_ActivityInstantaneousEvent::timestamp [inherited]: The timestamp at which event is sampled
uint64_t CUpti_ActivityInstantaneousEvent::value [inherited]: The event value.

6.31. CUpti_ActivityInstantaneousEventInstance Struct Reference

[CUPTI Activity API]

This activity record represents the a CUPTI event value for a specific event domain instance (CUPTI_ACTIVITY_KIND_EVENT_INSTANCE) sampled at a particular instant. This activity record kind is not produced by the activity API but is included for completeness and ease-of-use. Profiler frameworks built on top of CUPTI that collect event data may choose to use this type to store the collected event data. This activity record should be used when event domain instance information needs to be associated with the event.

Public Variables

uint32_t deviceId
CUpti_EventID id
uint8_t instance
CUpti_ActivityKind kind
uint8_t pad[3]
uint64_t timestamp
uint64_t value

Variables

uint32_t CUpti_ActivityInstantaneousEventInstance::deviceId [inherited]: The device id
CUpti_EventID CUpti_ActivityInstantaneousEventInstance::id [inherited]: The event ID.
uint8_t CUpti_ActivityInstantaneousEventInstance::instance [inherited]: The event domain instance
CUpti_ActivityKind CUpti_ActivityInstantaneousEventInstance::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_INSTANTANEOUS_EVENT_INSTANCE.
uint8_t CUpti_ActivityInstantaneousEventInstance::pad[3] [inherited]: Undefined. reserved for internal use
uint64_t CUpti_ActivityInstantaneousEventInstance::timestamp [inherited]: The timestamp at which event is sampled
uint64_t CUpti_ActivityInstantaneousEventInstance::value [inherited]: The event value.

6.32. CUpti_ActivityInstantaneousMetric Struct Reference

[CUPTI Activity API]

This activity record represents the collection of a CUPTI metric value (CUPTI_ACTIVITY_KIND_METRIC) at a particular instance. This activity record kind is not produced by the activity API but is included for completeness and ease-of-use. Profiler frameworks built on top of CUPTI that collect metric data may choose to use this type to store the collected metric data.

Public Variables

uint32_t deviceId
uint8_t flags
CUpti_MetricID id
CUpti_ActivityKind kind
uint8_t pad[3]
uint64_t timestamp
union CUpti_MetricValue value

Variables

uint32_t CUpti_ActivityInstantaneousMetric::deviceId [inherited]: The device id
uint8_t CUpti_ActivityInstantaneousMetric::flags [inherited]: The properties of this metric.

See also:

CUpti_ActivityFlag
CUpti_MetricID CUpti_ActivityInstantaneousMetric::id [inherited]: The metric ID.
CUpti_ActivityKind CUpti_ActivityInstantaneousMetric::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_INSTANTANEOUS_METRIC.
uint8_t CUpti_ActivityInstantaneousMetric::pad[3] [inherited]: Undefined. reserved for internal use
uint64_t CUpti_ActivityInstantaneousMetric::timestamp [inherited]: The timestamp at which metric is sampled
union CUpti_MetricValue CUpti_ActivityInstantaneousMetric::value [inherited]: The metric value.

6.33. CUpti_ActivityInstantaneousMetricInstance Struct Reference

[CUPTI Activity API]

This activity record represents a CUPTI metric value for a specific metric domain instance (CUPTI_ACTIVITY_KIND_METRIC_INSTANCE) sampled at a particular time. This activity record kind is not produced by the activity API but is included for completeness and ease-of-use. Profiler frameworks built on top of CUPTI that collect metric data may choose to use this type to store the collected metric data. This activity record should be used when metric domain instance information needs to be associated with the metric.

Public Variables

uint32_t deviceId
uint8_t flags
CUpti_MetricID id
uint8_t instance
CUpti_ActivityKind kind
uint8_t pad[2]
uint64_t timestamp
union CUpti_MetricValue value

Variables

uint32_t CUpti_ActivityInstantaneousMetricInstance::deviceId [inherited]: The device id
uint8_t CUpti_ActivityInstantaneousMetricInstance::flags [inherited]: The properties of this metric.

See also:

CUpti_ActivityFlag
CUpti_MetricID CUpti_ActivityInstantaneousMetricInstance::id [inherited]: The metric ID.
uint8_t CUpti_ActivityInstantaneousMetricInstance::instance [inherited]: The metric domain instance
CUpti_ActivityKind CUpti_ActivityInstantaneousMetricInstance::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_INSTANTANEOUS_METRIC_INSTANCE.
uint8_t CUpti_ActivityInstantaneousMetricInstance::pad[2] [inherited]: Undefined. reserved for internal use
uint64_t CUpti_ActivityInstantaneousMetricInstance::timestamp [inherited]: The timestamp at which metric is sampled
union CUpti_MetricValue CUpti_ActivityInstantaneousMetricInstance::value [inherited]: The metric value.

6.34. CUpti_ActivityInstructionCorrelation Struct Reference

[CUPTI Activity API]

This activity records source level sass/source correlation information. (CUPTI_ACTIVITY_KIND_INSTRUCTION_CORRELATION).

Public Variables

CUpti_ActivityFlag flags
uint32_t functionId
CUpti_ActivityKind kind
uint32_t pad
uint32_t pcOffset
uint32_t sourceLocatorId

Variables

CUpti_ActivityFlag CUpti_ActivityInstructionCorrelation::flags [inherited]: The properties of this instruction.
uint32_t CUpti_ActivityInstructionCorrelation::functionId [inherited]: Correlation ID with global/device function name
CUpti_ActivityKind CUpti_ActivityInstructionCorrelation::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_INSTRUCTION_CORRELATION.
uint32_t CUpti_ActivityInstructionCorrelation::pad [inherited]: Undefined. Reserved for internal use.
uint32_t CUpti_ActivityInstructionCorrelation::pcOffset [inherited]: The pc offset for the instruction.
uint32_t CUpti_ActivityInstructionCorrelation::sourceLocatorId [inherited]: The ID for source locator.

6.35. CUpti_ActivityInstructionExecution Struct Reference

[CUPTI Activity API]

This activity records result for source level instruction execution. (CUPTI_ACTIVITY_KIND_INSTRUCTION_EXECUTION).

Public Variables

uint32_t correlationId
uint32_t executed
CUpti_ActivityFlag flags
uint32_t functionId
CUpti_ActivityKind kind
uint64_t notPredOffThreadsExecuted
uint32_t pad
uint32_t pcOffset
uint32_t sourceLocatorId
uint64_t threadsExecuted

Variables

uint32_t CUpti_ActivityInstructionExecution::correlationId [inherited]: The correlation ID of the kernel to which this result is associated.
uint32_t CUpti_ActivityInstructionExecution::executed [inherited]: The number of times this instruction was executed per warp. It will be incremented regardless of predicate or condition code.
CUpti_ActivityFlag CUpti_ActivityInstructionExecution::flags [inherited]: The properties of this instruction execution.
uint32_t CUpti_ActivityInstructionExecution::functionId [inherited]: Correlation ID with global/device function name
CUpti_ActivityKind CUpti_ActivityInstructionExecution::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_INSTRUCTION_EXECUTION.
uint64_t CUpti_ActivityInstructionExecution::notPredOffThreadsExecuted [inherited]: This increments each time when this instruction is executed by number of threads that executed this instruction with predicate and condition code evaluating to true.
uint32_t CUpti_ActivityInstructionExecution::pad [inherited]: Undefined. Reserved for internal use.
uint32_t CUpti_ActivityInstructionExecution::pcOffset [inherited]: The pc offset for the instruction.
uint32_t CUpti_ActivityInstructionExecution::sourceLocatorId [inherited]: The ID for source locator.
uint64_t CUpti_ActivityInstructionExecution::threadsExecuted [inherited]: This increments each time when this instruction is executed by number of threads that executed this instruction, regardless of predicate or condition code.

6.36. CUpti_ActivityJit Struct Reference

[CUPTI Activity API]

Public Variables

const char * cachePath
uint64_t cacheSize
uint32_t correlationId
uint32_t deviceId
uint64_t end
CUpti_ActivityJitEntryType jitEntryType
uint64_t jitOperationCorrelationId
CUpti_ActivityJitOperationType jitOperationType
CUpti_ActivityKind kind
uint32_t padding
uint64_t start

Variables

const char * CUpti_ActivityJit::cachePath [inherited]: The path where the fat binary is cached.
uint64_t CUpti_ActivityJit::cacheSize [inherited]: The size of compute cache.
uint32_t CUpti_ActivityJit::correlationId [inherited]: The correlation ID of the JIT operation to which records belong to. Each JIT operation is assigned a unique correlation ID that is identical to the correlation ID in the driver or runtime API activity record that launched the JIT operation.
uint32_t CUpti_ActivityJit::deviceId [inherited]: The device ID.
uint64_t CUpti_ActivityJit::end [inherited]: The end timestamp for the JIT operation, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the JIT operation.
CUpti_ActivityJitEntryType CUpti_ActivityJit::jitEntryType [inherited]: The JIT entry type.
uint64_t CUpti_ActivityJit::jitOperationCorrelationId [inherited]: The correlation ID to correlate JIT compilation, load and store operations. Each JIT compilation unit is assigned a unique correlation ID at the time of the JIT compilation. This correlation id can be used to find the matching JIT cache load/store records.
CUpti_ActivityJitOperationType CUpti_ActivityJit::jitOperationType [inherited]: The JIT operation type.
CUpti_ActivityKind CUpti_ActivityJit::kind [inherited]: The activity record kind must be CUPTI_ACTIVITY_KIND_JIT.
uint32_t CUpti_ActivityJit::padding [inherited]: Internal use.
uint64_t CUpti_ActivityJit::start [inherited]: The start timestamp for the JIT operation, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the JIT operation.

6.37. CUpti_ActivityKernel Struct Reference

[CUPTI Activity API]

This activity record represents a kernel execution (CUPTI_ACTIVITY_KIND_KERNEL and CUPTI_ACTIVITY_KIND_CONCURRENT_KERNEL) but is no longer generated by CUPTI. Kernel activities are now reported using the CUpti_ActivityKernel9 activity record.

Public Variables

int32_t blockX
int32_t blockY
int32_t blockZ
uint8_t cacheConfigExecuted
uint8_t cacheConfigRequested
uint32_t contextId
uint32_t correlationId
uint32_t deviceId
int32_t dynamicSharedMemory
uint64_t end
int32_t gridX
int32_t gridY
int32_t gridZ
CUpti_ActivityKind kind
uint32_t localMemoryPerThread
uint32_t localMemoryTotal
const char * name
uint32_t pad
uint16_t registersPerThread
void * reserved0
uint32_t runtimeCorrelationId
uint64_t start
int32_t staticSharedMemory
uint32_t streamId

Variables

int32_t CUpti_ActivityKernel::blockX [inherited]: The X-dimension block size for the kernel.
int32_t CUpti_ActivityKernel::blockY [inherited]: The Y-dimension block size for the kernel.
int32_t CUpti_ActivityKernel::blockZ [inherited]: The Z-dimension grid size for the kernel.
uint8_t CUpti_ActivityKernel::cacheConfigExecuted [inherited]: The cache configuration used for the kernel. The value is one of the CUfunc_cache enumeration values from cuda.h.
uint8_t CUpti_ActivityKernel::cacheConfigRequested [inherited]: The cache configuration requested by the kernel. The value is one of the CUfunc_cache enumeration values from cuda.h.
uint32_t CUpti_ActivityKernel::contextId [inherited]: The ID of the context where the kernel is executing.
uint32_t CUpti_ActivityKernel::correlationId [inherited]: The correlation ID of the kernel. Each kernel execution is assigned a unique correlation ID that is identical to the correlation ID in the driver API activity record that launched the kernel.
uint32_t CUpti_ActivityKernel::deviceId [inherited]: The ID of the device where the kernel is executing.
int32_t CUpti_ActivityKernel::dynamicSharedMemory [inherited]: The dynamic shared memory reserved for the kernel, in bytes.
uint64_t CUpti_ActivityKernel::end [inherited]: The end timestamp for the kernel execution, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the kernel.
int32_t CUpti_ActivityKernel::gridX [inherited]: The X-dimension grid size for the kernel.
int32_t CUpti_ActivityKernel::gridY [inherited]: The Y-dimension grid size for the kernel.
int32_t CUpti_ActivityKernel::gridZ [inherited]: The Z-dimension grid size for the kernel.
CUpti_ActivityKind CUpti_ActivityKernel::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_KERNEL or CUPTI_ACTIVITY_KIND_CONCURRENT_KERNEL.
uint32_t CUpti_ActivityKernel::localMemoryPerThread [inherited]: The amount of local memory reserved for each thread, in bytes.
uint32_t CUpti_ActivityKernel::localMemoryTotal [inherited]: The total amount of local memory reserved for the kernel, in bytes.
const char * CUpti_ActivityKernel::name [inherited]: The name of the kernel. This name is shared across all activity records representing the same kernel, and so should not be modified.
uint32_t CUpti_ActivityKernel::pad [inherited]: Undefined. Reserved for internal use.
uint16_t CUpti_ActivityKernel::registersPerThread [inherited]: The number of registers required for each thread executing the kernel.
void * CUpti_ActivityKernel::reserved0 [inherited]: Undefined. Reserved for internal use.
uint32_t CUpti_ActivityKernel::runtimeCorrelationId [inherited]: The runtime correlation ID of the kernel. Each kernel execution is assigned a unique runtime correlation ID that is identical to the correlation ID in the runtime API activity record that launched the kernel.
uint64_t CUpti_ActivityKernel::start [inherited]: The start timestamp for the kernel execution, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the kernel.
int32_t CUpti_ActivityKernel::staticSharedMemory [inherited]: The static shared memory allocated for the kernel, in bytes.
uint32_t CUpti_ActivityKernel::streamId [inherited]: The ID of the stream where the kernel is executing.

6.38. CUpti_ActivityKernel2 Struct Reference

[CUPTI Activity API]

Public Variables

int32_t blockX
int32_t blockY
int32_t blockZ
uint64_t completed
uint32_t contextId
uint32_t correlationId
uint32_t deviceId
int32_t dynamicSharedMemory
uint64_t end
uint8_t executed
int64_t gridId
int32_t gridX
int32_t gridY
int32_t gridZ
CUpti_ActivityKind kind
uint32_t localMemoryPerThread
uint32_t localMemoryTotal
const char * name
uint16_t registersPerThread
uint8_t requested
void * reserved0
uint8_t sharedMemoryConfig
uint64_t start
int32_t staticSharedMemory
uint32_t streamId

Variables

int32_t CUpti_ActivityKernel2::blockX [inherited]: The X-dimension block size for the kernel.
int32_t CUpti_ActivityKernel2::blockY [inherited]: The Y-dimension block size for the kernel.
int32_t CUpti_ActivityKernel2::blockZ [inherited]: The Z-dimension grid size for the kernel.
uint64_t CUpti_ActivityKernel2::completed [inherited]: The completed timestamp for the kernel execution, in ns. It represents the completion of all it's child kernels and the kernel itself. A value of CUPTI_TIMESTAMP_UNKNOWN indicates that the completion time is unknown.
uint32_t CUpti_ActivityKernel2::contextId [inherited]: The ID of the context where the kernel is executing.
uint32_t CUpti_ActivityKernel2::correlationId [inherited]: The correlation ID of the kernel. Each kernel execution is assigned a unique correlation ID that is identical to the correlation ID in the driver or runtime API activity record that launched the kernel.
uint32_t CUpti_ActivityKernel2::deviceId [inherited]: The ID of the device where the kernel is executing.
int32_t CUpti_ActivityKernel2::dynamicSharedMemory [inherited]: The dynamic shared memory reserved for the kernel, in bytes.
uint64_t CUpti_ActivityKernel2::end [inherited]: The end timestamp for the kernel execution, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the kernel.
uint8_t CUpti_ActivityKernel2::executed [inherited]: The cache configuration used for the kernel. The value is one of the CUfunc_cache enumeration values from cuda.h.
int64_t CUpti_ActivityKernel2::gridId [inherited]: The grid ID of the kernel. Each kernel is assigned a unique grid ID at runtime.
int32_t CUpti_ActivityKernel2::gridX [inherited]: The X-dimension grid size for the kernel.
int32_t CUpti_ActivityKernel2::gridY [inherited]: The Y-dimension grid size for the kernel.
int32_t CUpti_ActivityKernel2::gridZ [inherited]: The Z-dimension grid size for the kernel.
CUpti_ActivityKind CUpti_ActivityKernel2::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_KERNEL or CUPTI_ACTIVITY_KIND_CONCURRENT_KERNEL.
uint32_t CUpti_ActivityKernel2::localMemoryPerThread [inherited]: The amount of local memory reserved for each thread, in bytes.
uint32_t CUpti_ActivityKernel2::localMemoryTotal [inherited]: The total amount of local memory reserved for the kernel, in bytes.
const char * CUpti_ActivityKernel2::name [inherited]: The name of the kernel. This name is shared across all activity records representing the same kernel, and so should not be modified.
uint16_t CUpti_ActivityKernel2::registersPerThread [inherited]: The number of registers required for each thread executing the kernel.
uint8_t CUpti_ActivityKernel2::requested [inherited]: The cache configuration requested by the kernel. The value is one of the CUfunc_cache enumeration values from cuda.h.
void * CUpti_ActivityKernel2::reserved0 [inherited]: Undefined. Reserved for internal use.
uint8_t CUpti_ActivityKernel2::sharedMemoryConfig [inherited]: The shared memory configuration used for the kernel. The value is one of the CUsharedconfig enumeration values from cuda.h.
uint64_t CUpti_ActivityKernel2::start [inherited]: The start timestamp for the kernel execution, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the kernel.
int32_t CUpti_ActivityKernel2::staticSharedMemory [inherited]: The static shared memory allocated for the kernel, in bytes.
uint32_t CUpti_ActivityKernel2::streamId [inherited]: The ID of the stream where the kernel is executing.

6.39. CUpti_ActivityKernel3 Struct Reference

[CUPTI Activity API]

This activity record represents a kernel execution (CUPTI_ACTIVITY_KIND_KERNEL and CUPTI_ACTIVITY_KIND_CONCURRENT_KERNEL). Kernel activities are now reported using the CUpti_ActivityKernel9 activity record.

Public Variables

int32_t blockX
int32_t blockY
int32_t blockZ
uint64_t completed
uint32_t contextId
uint32_t correlationId
uint32_t deviceId
int32_t dynamicSharedMemory
uint64_t end
uint8_t executed
int64_t gridId
int32_t gridX
int32_t gridY
int32_t gridZ
CUpti_ActivityKind kind
uint32_t localMemoryPerThread
uint32_t localMemoryTotal
const char * name
CUpti_ActivityPartitionedGlobalCacheConfig partitionedGlobalCacheExecuted
CUpti_ActivityPartitionedGlobalCacheConfig partitionedGlobalCacheRequested
uint16_t registersPerThread
uint8_t requested
void * reserved0
uint8_t sharedMemoryConfig
uint64_t start
int32_t staticSharedMemory
uint32_t streamId

Variables

int32_t CUpti_ActivityKernel3::blockX [inherited]: The X-dimension block size for the kernel.
int32_t CUpti_ActivityKernel3::blockY [inherited]: The Y-dimension block size for the kernel.
int32_t CUpti_ActivityKernel3::blockZ [inherited]: The Z-dimension grid size for the kernel.
uint64_t CUpti_ActivityKernel3::completed [inherited]: The completed timestamp for the kernel execution, in ns. It represents the completion of all it's child kernels and the kernel itself. A value of CUPTI_TIMESTAMP_UNKNOWN indicates that the completion time is unknown.
uint32_t CUpti_ActivityKernel3::contextId [inherited]: The ID of the context where the kernel is executing.
uint32_t CUpti_ActivityKernel3::correlationId [inherited]: The correlation ID of the kernel. Each kernel execution is assigned a unique correlation ID that is identical to the correlation ID in the driver or runtime API activity record that launched the kernel.
uint32_t CUpti_ActivityKernel3::deviceId [inherited]: The ID of the device where the kernel is executing.
int32_t CUpti_ActivityKernel3::dynamicSharedMemory [inherited]: The dynamic shared memory reserved for the kernel, in bytes.
uint64_t CUpti_ActivityKernel3::end [inherited]: The end timestamp for the kernel execution, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the kernel.
uint8_t CUpti_ActivityKernel3::executed [inherited]: The cache configuration used for the kernel. The value is one of the CUfunc_cache enumeration values from cuda.h.
int64_t CUpti_ActivityKernel3::gridId [inherited]: The grid ID of the kernel. Each kernel is assigned a unique grid ID at runtime.
int32_t CUpti_ActivityKernel3::gridX [inherited]: The X-dimension grid size for the kernel.
int32_t CUpti_ActivityKernel3::gridY [inherited]: The Y-dimension grid size for the kernel.
int32_t CUpti_ActivityKernel3::gridZ [inherited]: The Z-dimension grid size for the kernel.
CUpti_ActivityKind CUpti_ActivityKernel3::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_KERNEL or CUPTI_ACTIVITY_KIND_CONCURRENT_KERNEL.
uint32_t CUpti_ActivityKernel3::localMemoryPerThread [inherited]: The amount of local memory reserved for each thread, in bytes.
uint32_t CUpti_ActivityKernel3::localMemoryTotal [inherited]: The total amount of local memory reserved for the kernel, in bytes.
const char * CUpti_ActivityKernel3::name [inherited]: The name of the kernel. This name is shared across all activity records representing the same kernel, and so should not be modified.
CUpti_ActivityPartitionedGlobalCacheConfig CUpti_ActivityKernel3::partitionedGlobalCacheExecuted [inherited]: The partitioned global caching executed for the kernel. Partitioned global caching is required to enable caching on certain chips, such as devices with compute capability 5.2. Partitioned global caching can be automatically disabled if the occupancy requirement of the launch cannot support caching.
CUpti_ActivityPartitionedGlobalCacheConfig CUpti_ActivityKernel3::partitionedGlobalCacheRequested [inherited]: The partitioned global caching requested for the kernel. Partitioned global caching is required to enable caching on certain chips, such as devices with compute capability 5.2.
uint16_t CUpti_ActivityKernel3::registersPerThread [inherited]: The number of registers required for each thread executing the kernel.
uint8_t CUpti_ActivityKernel3::requested [inherited]: The cache configuration requested by the kernel. The value is one of the CUfunc_cache enumeration values from cuda.h.
void * CUpti_ActivityKernel3::reserved0 [inherited]: Undefined. Reserved for internal use.
uint8_t CUpti_ActivityKernel3::sharedMemoryConfig [inherited]: The shared memory configuration used for the kernel. The value is one of the CUsharedconfig enumeration values from cuda.h.
uint64_t CUpti_ActivityKernel3::start [inherited]: The start timestamp for the kernel execution, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the kernel.
int32_t CUpti_ActivityKernel3::staticSharedMemory [inherited]: The static shared memory allocated for the kernel, in bytes.
uint32_t CUpti_ActivityKernel3::streamId [inherited]: The ID of the stream where the kernel is executing.

6.40. CUpti_ActivityKernel4 Struct Reference

[CUPTI Activity API]

Public Variables

int32_t blockX
int32_t blockY
int32_t blockZ
CUpti_ActivityKernel4::@9 cacheConfig
uint64_t completed
uint32_t contextId
uint32_t correlationId
uint32_t deviceId
int32_t dynamicSharedMemory
uint64_t end
uint8_t executed
int64_t gridId
int32_t gridX
int32_t gridY
int32_t gridZ
uint8_t isSharedMemoryCarveoutRequested
CUpti_ActivityKind kind
uint8_t launchType
uint32_t localMemoryPerThread
uint32_t localMemoryTotal
const char * name
uint8_t padding
CUpti_ActivityPartitionedGlobalCacheConfig partitionedGlobalCacheExecuted
CUpti_ActivityPartitionedGlobalCacheConfig partitionedGlobalCacheRequested
uint64_t queued
uint16_t registersPerThread
uint8_t requested
void * reserved0
uint8_t sharedMemoryCarveoutRequested
uint8_t sharedMemoryConfig
uint32_t sharedMemoryExecuted
uint64_t start
int32_t staticSharedMemory
uint32_t streamId
uint64_t submitted

Variables

int32_t CUpti_ActivityKernel4::blockX [inherited]: The X-dimension block size for the kernel.
int32_t CUpti_ActivityKernel4::blockY [inherited]: The Y-dimension block size for the kernel.
int32_t CUpti_ActivityKernel4::blockZ [inherited]: The Z-dimension grid size for the kernel.
CUpti_ActivityKernel4::@9 CUpti_ActivityKernel4::cacheConfig [inherited]: For devices with compute capability 7.0+ cacheConfig values are not updated in case field isSharedMemoryCarveoutRequested is set
uint64_t CUpti_ActivityKernel4::completed [inherited]: The completed timestamp for the kernel execution, in ns. It represents the completion of all it's child kernels and the kernel itself. A value of CUPTI_TIMESTAMP_UNKNOWN indicates that the completion time is unknown.
uint32_t CUpti_ActivityKernel4::contextId [inherited]: The ID of the context where the kernel is executing.
uint32_t CUpti_ActivityKernel4::correlationId [inherited]: The correlation ID of the kernel. Each kernel execution is assigned a unique correlation ID that is identical to the correlation ID in the driver or runtime API activity record that launched the kernel.
uint32_t CUpti_ActivityKernel4::deviceId [inherited]: The ID of the device where the kernel is executing.
int32_t CUpti_ActivityKernel4::dynamicSharedMemory [inherited]: The dynamic shared memory reserved for the kernel, in bytes.
uint64_t CUpti_ActivityKernel4::end [inherited]: The end timestamp for the kernel execution, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the kernel.
uint8_t CUpti_ActivityKernel4::executed [inherited]: The cache configuration used for the kernel. The value is one of the CUfunc_cache enumeration values from cuda.h.
int64_t CUpti_ActivityKernel4::gridId [inherited]: The grid ID of the kernel. Each kernel is assigned a unique grid ID at runtime.
int32_t CUpti_ActivityKernel4::gridX [inherited]: The X-dimension grid size for the kernel.
int32_t CUpti_ActivityKernel4::gridY [inherited]: The Y-dimension grid size for the kernel.
int32_t CUpti_ActivityKernel4::gridZ [inherited]: The Z-dimension grid size for the kernel.
uint8_t CUpti_ActivityKernel4::isSharedMemoryCarveoutRequested [inherited]: This indicates if CU_FUNC_ATTRIBUTE_PREFERRED_SHARED_MEMORY_CARVEOUT was updated for the kernel launch
CUpti_ActivityKind CUpti_ActivityKernel4::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_KERNEL or CUPTI_ACTIVITY_KIND_CONCURRENT_KERNEL.
uint8_t CUpti_ActivityKernel4::launchType [inherited]: The indicates if the kernel was executed via a regular launch or via a single/multi device cooperative launch.

See also:

CUpti_ActivityLaunchType
uint32_t CUpti_ActivityKernel4::localMemoryPerThread [inherited]: The amount of local memory reserved for each thread, in bytes.
uint32_t CUpti_ActivityKernel4::localMemoryTotal [inherited]: The total amount of local memory reserved for the kernel, in bytes.
const char * CUpti_ActivityKernel4::name [inherited]: The name of the kernel. This name is shared across all activity records representing the same kernel, and so should not be modified.
uint8_t CUpti_ActivityKernel4::padding [inherited]: Undefined. Reserved for internal use.
CUpti_ActivityPartitionedGlobalCacheConfig CUpti_ActivityKernel4::partitionedGlobalCacheExecuted [inherited]: The partitioned global caching executed for the kernel. Partitioned global caching is required to enable caching on certain chips, such as devices with compute capability 5.2. Partitioned global caching can be automatically disabled if the occupancy requirement of the launch cannot support caching.
CUpti_ActivityPartitionedGlobalCacheConfig CUpti_ActivityKernel4::partitionedGlobalCacheRequested [inherited]: The partitioned global caching requested for the kernel. Partitioned global caching is required to enable caching on certain chips, such as devices with compute capability 5.2.
uint64_t CUpti_ActivityKernel4::queued [inherited]: The timestamp when the kernel is queued up in the command buffer, in ns. A value of CUPTI_TIMESTAMP_UNKNOWN indicates that the queued time could not be collected for the kernel. This timestamp is not collected by default. Use API cuptiActivityEnableLatencyTimestamps() to enable collection.

Command buffer is a buffer written by CUDA driver to send commands like kernel launch, memory copy etc to the GPU. All launches of CUDA kernels are asynchrnous with respect to the host, the host requests the launch by writing commands into the command buffer, then returns without checking the GPU's progress.
uint16_t CUpti_ActivityKernel4::registersPerThread [inherited]: The number of registers required for each thread executing the kernel.
uint8_t CUpti_ActivityKernel4::requested [inherited]: The cache configuration requested by the kernel. The value is one of the CUfunc_cache enumeration values from cuda.h.
void * CUpti_ActivityKernel4::reserved0 [inherited]: Undefined. Reserved for internal use.
uint8_t CUpti_ActivityKernel4::sharedMemoryCarveoutRequested [inherited]: Shared memory carveout value requested for the function in percentage of the total resource. The value will be updated only if field isSharedMemoryCarveoutRequested is set.
uint8_t CUpti_ActivityKernel4::sharedMemoryConfig [inherited]: The shared memory configuration used for the kernel. The value is one of the CUsharedconfig enumeration values from cuda.h.
uint32_t CUpti_ActivityKernel4::sharedMemoryExecuted [inherited]: Shared memory size set by the driver.
uint64_t CUpti_ActivityKernel4::start [inherited]: The start timestamp for the kernel execution, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the kernel.
int32_t CUpti_ActivityKernel4::staticSharedMemory [inherited]: The static shared memory allocated for the kernel, in bytes.
uint32_t CUpti_ActivityKernel4::streamId [inherited]: The ID of the stream where the kernel is executing.
uint64_t CUpti_ActivityKernel4::submitted [inherited]: The timestamp when the command buffer containing the kernel launch is submitted to the GPU, in ns. A value of CUPTI_TIMESTAMP_UNKNOWN indicates that the submitted time could not be collected for the kernel. This timestamp is not collected by default. Use API cuptiActivityEnableLatencyTimestamps() to enable collection.

6.41. CUpti_ActivityKernel5 Struct Reference

[CUPTI Activity API]

Public Variables

int32_t blockX
int32_t blockY
int32_t blockZ
CUpti_ActivityKernel5::@11 cacheConfig
uint64_t completed
uint32_t contextId
uint32_t correlationId
uint32_t deviceId
int32_t dynamicSharedMemory
uint64_t end
uint8_t executed
uint32_t graphId
uint64_t graphNodeId
int64_t gridId
int32_t gridX
int32_t gridY
int32_t gridZ
uint8_t isSharedMemoryCarveoutRequested
CUpti_ActivityKind kind
uint8_t launchType
uint32_t localMemoryPerThread
uint32_t localMemoryTotal
const char * name
uint8_t padding
CUpti_ActivityPartitionedGlobalCacheConfig partitionedGlobalCacheExecuted
CUpti_ActivityPartitionedGlobalCacheConfig partitionedGlobalCacheRequested
uint64_t queued
uint16_t registersPerThread
uint8_t requested
void * reserved0
uint8_t sharedMemoryCarveoutRequested
uint8_t sharedMemoryConfig
uint32_t sharedMemoryExecuted
CUpti_FuncShmemLimitConfig shmemLimitConfig
uint64_t start
int32_t staticSharedMemory
uint32_t streamId
uint64_t submitted

Variables

int32_t CUpti_ActivityKernel5::blockX [inherited]: The X-dimension block size for the kernel.
int32_t CUpti_ActivityKernel5::blockY [inherited]: The Y-dimension block size for the kernel.
int32_t CUpti_ActivityKernel5::blockZ [inherited]: The Z-dimension grid size for the kernel.
CUpti_ActivityKernel5::@11 CUpti_ActivityKernel5::cacheConfig [inherited]: For devices with compute capability 7.0+ cacheConfig values are not updated in case field isSharedMemoryCarveoutRequested is set
uint64_t CUpti_ActivityKernel5::completed [inherited]: The completed timestamp for the kernel execution, in ns. It represents the completion of all it's child kernels and the kernel itself. A value of CUPTI_TIMESTAMP_UNKNOWN indicates that the completion time is unknown.
uint32_t CUpti_ActivityKernel5::contextId [inherited]: The ID of the context where the kernel is executing.
uint32_t CUpti_ActivityKernel5::correlationId [inherited]: The correlation ID of the kernel. Each kernel execution is assigned a unique correlation ID that is identical to the correlation ID in the driver or runtime API activity record that launched the kernel.
uint32_t CUpti_ActivityKernel5::deviceId [inherited]: The ID of the device where the kernel is executing.
int32_t CUpti_ActivityKernel5::dynamicSharedMemory [inherited]: The dynamic shared memory reserved for the kernel, in bytes.
uint64_t CUpti_ActivityKernel5::end [inherited]: The end timestamp for the kernel execution, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the kernel.
uint8_t CUpti_ActivityKernel5::executed [inherited]: The cache configuration used for the kernel. The value is one of the CUfunc_cache enumeration values from cuda.h.
uint32_t CUpti_ActivityKernel5::graphId [inherited]: The unique ID of the graph that launched this kernel through graph launch APIs. This field will be 0 if the kernel is not launched through graph launch APIs.
uint64_t CUpti_ActivityKernel5::graphNodeId [inherited]: The unique ID of the graph node that launched this kernel through graph launch APIs. This field will be 0 if the kernel is not launched through graph launch APIs.
int64_t CUpti_ActivityKernel5::gridId [inherited]: The grid ID of the kernel. Each kernel is assigned a unique grid ID at runtime.
int32_t CUpti_ActivityKernel5::gridX [inherited]: The X-dimension grid size for the kernel.
int32_t CUpti_ActivityKernel5::gridY [inherited]: The Y-dimension grid size for the kernel.
int32_t CUpti_ActivityKernel5::gridZ [inherited]: The Z-dimension grid size for the kernel.
uint8_t CUpti_ActivityKernel5::isSharedMemoryCarveoutRequested [inherited]: This indicates if CU_FUNC_ATTRIBUTE_PREFERRED_SHARED_MEMORY_CARVEOUT was updated for the kernel launch
CUpti_ActivityKind CUpti_ActivityKernel5::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_KERNEL or CUPTI_ACTIVITY_KIND_CONCURRENT_KERNEL.
uint8_t CUpti_ActivityKernel5::launchType [inherited]: The indicates if the kernel was executed via a regular launch or via a single/multi device cooperative launch.

See also:

CUpti_ActivityLaunchType
uint32_t CUpti_ActivityKernel5::localMemoryPerThread [inherited]: The amount of local memory reserved for each thread, in bytes.
uint32_t CUpti_ActivityKernel5::localMemoryTotal [inherited]: The total amount of local memory reserved for the kernel, in bytes.
const char * CUpti_ActivityKernel5::name [inherited]: The name of the kernel. This name is shared across all activity records representing the same kernel, and so should not be modified.
uint8_t CUpti_ActivityKernel5::padding [inherited]: Undefined. Reserved for internal use.
CUpti_ActivityPartitionedGlobalCacheConfig CUpti_ActivityKernel5::partitionedGlobalCacheExecuted [inherited]: The partitioned global caching executed for the kernel. Partitioned global caching is required to enable caching on certain chips, such as devices with compute capability 5.2. Partitioned global caching can be automatically disabled if the occupancy requirement of the launch cannot support caching.
CUpti_ActivityPartitionedGlobalCacheConfig CUpti_ActivityKernel5::partitionedGlobalCacheRequested [inherited]: The partitioned global caching requested for the kernel. Partitioned global caching is required to enable caching on certain chips, such as devices with compute capability 5.2.
uint64_t CUpti_ActivityKernel5::queued [inherited]: The timestamp when the kernel is queued up in the command buffer, in ns. A value of CUPTI_TIMESTAMP_UNKNOWN indicates that the queued time could not be collected for the kernel. This timestamp is not collected by default. Use API cuptiActivityEnableLatencyTimestamps() to enable collection.

Command buffer is a buffer written by CUDA driver to send commands like kernel launch, memory copy etc to the GPU. All launches of CUDA kernels are asynchrnous with respect to the host, the host requests the launch by writing commands into the command buffer, then returns without checking the GPU's progress.
uint16_t CUpti_ActivityKernel5::registersPerThread [inherited]: The number of registers required for each thread executing the kernel.
uint8_t CUpti_ActivityKernel5::requested [inherited]: The cache configuration requested by the kernel. The value is one of the CUfunc_cache enumeration values from cuda.h.
void * CUpti_ActivityKernel5::reserved0 [inherited]: Undefined. Reserved for internal use.
uint8_t CUpti_ActivityKernel5::sharedMemoryCarveoutRequested [inherited]: Shared memory carveout value requested for the function in percentage of the total resource. The value will be updated only if field isSharedMemoryCarveoutRequested is set.
uint8_t CUpti_ActivityKernel5::sharedMemoryConfig [inherited]: The shared memory configuration used for the kernel. The value is one of the CUsharedconfig enumeration values from cuda.h.
uint32_t CUpti_ActivityKernel5::sharedMemoryExecuted [inherited]: Shared memory size set by the driver.
CUpti_FuncShmemLimitConfig CUpti_ActivityKernel5::shmemLimitConfig [inherited]: The shared memory limit config for the kernel. This field shows whether user has opted for a higher per block limit of dynamic shared memory.
uint64_t CUpti_ActivityKernel5::start [inherited]: The start timestamp for the kernel execution, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the kernel.
int32_t CUpti_ActivityKernel5::staticSharedMemory [inherited]: The static shared memory allocated for the kernel, in bytes.
uint32_t CUpti_ActivityKernel5::streamId [inherited]: The ID of the stream where the kernel is executing.
uint64_t CUpti_ActivityKernel5::submitted [inherited]: The timestamp when the command buffer containing the kernel launch is submitted to the GPU, in ns. A value of CUPTI_TIMESTAMP_UNKNOWN indicates that the submitted time could not be collected for the kernel. This timestamp is not collected by default. Use API cuptiActivityEnableLatencyTimestamps() to enable collection.

6.42. CUpti_ActivityKernel6 Struct Reference

[CUPTI Activity API]

Public Variables

int32_t blockX
int32_t blockY
int32_t blockZ
CUpti_ActivityKernel6::@13 cacheConfig
uint64_t completed
uint32_t contextId
uint32_t correlationId
uint32_t deviceId
int32_t dynamicSharedMemory
uint64_t end
uint8_t executed
uint32_t graphId
uint64_t graphNodeId
int64_t gridId
int32_t gridX
int32_t gridY
int32_t gridZ
uint8_t isSharedMemoryCarveoutRequested
CUpti_ActivityKind kind
uint8_t launchType
uint32_t localMemoryPerThread
uint32_t localMemoryTotal
const char * name
CUaccessPolicyWindow * pAccessPolicyWindow
uint8_t padding
CUpti_ActivityPartitionedGlobalCacheConfig partitionedGlobalCacheExecuted
CUpti_ActivityPartitionedGlobalCacheConfig partitionedGlobalCacheRequested
uint64_t queued
uint16_t registersPerThread
uint8_t requested
void * reserved0
uint8_t sharedMemoryCarveoutRequested
uint8_t sharedMemoryConfig
uint32_t sharedMemoryExecuted
CUpti_FuncShmemLimitConfig shmemLimitConfig
uint64_t start
int32_t staticSharedMemory
uint32_t streamId
uint64_t submitted

Variables

int32_t CUpti_ActivityKernel6::blockX [inherited]: The X-dimension block size for the kernel.
int32_t CUpti_ActivityKernel6::blockY [inherited]: The Y-dimension block size for the kernel.
int32_t CUpti_ActivityKernel6::blockZ [inherited]: The Z-dimension grid size for the kernel.
CUpti_ActivityKernel6::@13 CUpti_ActivityKernel6::cacheConfig [inherited]: For devices with compute capability 7.0+ cacheConfig values are not updated in case field isSharedMemoryCarveoutRequested is set
uint64_t CUpti_ActivityKernel6::completed [inherited]: The completed timestamp for the kernel execution, in ns. It represents the completion of all it's child kernels and the kernel itself. A value of CUPTI_TIMESTAMP_UNKNOWN indicates that the completion time is unknown.
uint32_t CUpti_ActivityKernel6::contextId [inherited]: The ID of the context where the kernel is executing.
uint32_t CUpti_ActivityKernel6::correlationId [inherited]: The correlation ID of the kernel. Each kernel execution is assigned a unique correlation ID that is identical to the correlation ID in the driver or runtime API activity record that launched the kernel.
uint32_t CUpti_ActivityKernel6::deviceId [inherited]: The ID of the device where the kernel is executing.
int32_t CUpti_ActivityKernel6::dynamicSharedMemory [inherited]: The dynamic shared memory reserved for the kernel, in bytes.
uint64_t CUpti_ActivityKernel6::end [inherited]: The end timestamp for the kernel execution, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the kernel.
uint8_t CUpti_ActivityKernel6::executed [inherited]: The cache configuration used for the kernel. The value is one of the CUfunc_cache enumeration values from cuda.h.
uint32_t CUpti_ActivityKernel6::graphId [inherited]: The unique ID of the graph that launched this kernel through graph launch APIs. This field will be 0 if the kernel is not launched through graph launch APIs.
uint64_t CUpti_ActivityKernel6::graphNodeId [inherited]: The unique ID of the graph node that launched this kernel through graph launch APIs. This field will be 0 if the kernel is not launched through graph launch APIs.
int64_t CUpti_ActivityKernel6::gridId [inherited]: The grid ID of the kernel. Each kernel is assigned a unique grid ID at runtime.
int32_t CUpti_ActivityKernel6::gridX [inherited]: The X-dimension grid size for the kernel.
int32_t CUpti_ActivityKernel6::gridY [inherited]: The Y-dimension grid size for the kernel.
int32_t CUpti_ActivityKernel6::gridZ [inherited]: The Z-dimension grid size for the kernel.
uint8_t CUpti_ActivityKernel6::isSharedMemoryCarveoutRequested [inherited]: This indicates if CU_FUNC_ATTRIBUTE_PREFERRED_SHARED_MEMORY_CARVEOUT was updated for the kernel launch
CUpti_ActivityKind CUpti_ActivityKernel6::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_KERNEL or CUPTI_ACTIVITY_KIND_CONCURRENT_KERNEL.
uint8_t CUpti_ActivityKernel6::launchType [inherited]: The indicates if the kernel was executed via a regular launch or via a single/multi device cooperative launch.

See also:

CUpti_ActivityLaunchType
uint32_t CUpti_ActivityKernel6::localMemoryPerThread [inherited]: The amount of local memory reserved for each thread, in bytes.
uint32_t CUpti_ActivityKernel6::localMemoryTotal [inherited]: The total amount of local memory reserved for the kernel, in bytes.
const char * CUpti_ActivityKernel6::name [inherited]: The name of the kernel. This name is shared across all activity records representing the same kernel, and so should not be modified.
CUaccessPolicyWindow * CUpti_ActivityKernel6::pAccessPolicyWindow [inherited]: The pointer to the access policy window. The structure CUaccessPolicyWindow is defined in cuda.h.
uint8_t CUpti_ActivityKernel6::padding [inherited]: Undefined. Reserved for internal use.
CUpti_ActivityPartitionedGlobalCacheConfig CUpti_ActivityKernel6::partitionedGlobalCacheExecuted [inherited]: The partitioned global caching executed for the kernel. Partitioned global caching is required to enable caching on certain chips, such as devices with compute capability 5.2. Partitioned global caching can be automatically disabled if the occupancy requirement of the launch cannot support caching.
CUpti_ActivityPartitionedGlobalCacheConfig CUpti_ActivityKernel6::partitionedGlobalCacheRequested [inherited]: The partitioned global caching requested for the kernel. Partitioned global caching is required to enable caching on certain chips, such as devices with compute capability 5.2.
uint64_t CUpti_ActivityKernel6::queued [inherited]: The timestamp when the kernel is queued up in the command buffer, in ns. A value of CUPTI_TIMESTAMP_UNKNOWN indicates that the queued time could not be collected for the kernel. This timestamp is not collected by default. Use API cuptiActivityEnableLatencyTimestamps() to enable collection.

Command buffer is a buffer written by CUDA driver to send commands like kernel launch, memory copy etc to the GPU. All launches of CUDA kernels are asynchrnous with respect to the host, the host requests the launch by writing commands into the command buffer, then returns without checking the GPU's progress.
uint16_t CUpti_ActivityKernel6::registersPerThread [inherited]: The number of registers required for each thread executing the kernel.
uint8_t CUpti_ActivityKernel6::requested [inherited]: The cache configuration requested by the kernel. The value is one of the CUfunc_cache enumeration values from cuda.h.
void * CUpti_ActivityKernel6::reserved0 [inherited]: Undefined. Reserved for internal use.
uint8_t CUpti_ActivityKernel6::sharedMemoryCarveoutRequested [inherited]: Shared memory carveout value requested for the function in percentage of the total resource. The value will be updated only if field isSharedMemoryCarveoutRequested is set.
uint8_t CUpti_ActivityKernel6::sharedMemoryConfig [inherited]: The shared memory configuration used for the kernel. The value is one of the CUsharedconfig enumeration values from cuda.h.
uint32_t CUpti_ActivityKernel6::sharedMemoryExecuted [inherited]: Shared memory size set by the driver.
CUpti_FuncShmemLimitConfig CUpti_ActivityKernel6::shmemLimitConfig [inherited]: The shared memory limit config for the kernel. This field shows whether user has opted for a higher per block limit of dynamic shared memory.
uint64_t CUpti_ActivityKernel6::start [inherited]: The start timestamp for the kernel execution, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the kernel.
int32_t CUpti_ActivityKernel6::staticSharedMemory [inherited]: The static shared memory allocated for the kernel, in bytes.
uint32_t CUpti_ActivityKernel6::streamId [inherited]: The ID of the stream where the kernel is executing.
uint64_t CUpti_ActivityKernel6::submitted [inherited]: The timestamp when the command buffer containing the kernel launch is submitted to the GPU, in ns. A value of CUPTI_TIMESTAMP_UNKNOWN indicates that the submitted time could not be collected for the kernel. This timestamp is not collected by default. Use API cuptiActivityEnableLatencyTimestamps() to enable collection.

6.43. CUpti_ActivityKernel7 Struct Reference

[CUPTI Activity API]

Public Variables

int32_t blockX
int32_t blockY
int32_t blockZ
CUpti_ActivityKernel7::@15 cacheConfig
uint32_t channelID
CUpti_ChannelType channelType
uint64_t completed
uint32_t contextId
uint32_t correlationId
uint32_t deviceId
int32_t dynamicSharedMemory
uint64_t end
uint8_t executed
uint32_t graphId
uint64_t graphNodeId
int64_t gridId
int32_t gridX
int32_t gridY
int32_t gridZ
uint8_t isSharedMemoryCarveoutRequested
CUpti_ActivityKind kind
uint8_t launchType
uint32_t localMemoryPerThread
uint32_t localMemoryTotal
const char * name
CUaccessPolicyWindow * pAccessPolicyWindow
uint8_t padding
CUpti_ActivityPartitionedGlobalCacheConfig partitionedGlobalCacheExecuted
CUpti_ActivityPartitionedGlobalCacheConfig partitionedGlobalCacheRequested
uint64_t queued
uint16_t registersPerThread
uint8_t requested
void * reserved0
uint8_t sharedMemoryCarveoutRequested
uint8_t sharedMemoryConfig
uint32_t sharedMemoryExecuted
CUpti_FuncShmemLimitConfig shmemLimitConfig
uint64_t start
int32_t staticSharedMemory
uint32_t streamId
uint64_t submitted

Variables

int32_t CUpti_ActivityKernel7::blockX [inherited]: The X-dimension block size for the kernel.
int32_t CUpti_ActivityKernel7::blockY [inherited]: The Y-dimension block size for the kernel.
int32_t CUpti_ActivityKernel7::blockZ [inherited]: The Z-dimension grid size for the kernel.
CUpti_ActivityKernel7::@15 CUpti_ActivityKernel7::cacheConfig [inherited]: For devices with compute capability 7.0+ cacheConfig values are not updated in case field isSharedMemoryCarveoutRequested is set
uint32_t CUpti_ActivityKernel7::channelID [inherited]: The ID of the HW channel on which the kernel is launched.
CUpti_ChannelType CUpti_ActivityKernel7::channelType [inherited]: The type of the channel
uint64_t CUpti_ActivityKernel7::completed [inherited]: The completed timestamp for the kernel execution, in ns. It represents the completion of all it's child kernels and the kernel itself. A value of CUPTI_TIMESTAMP_UNKNOWN indicates that the completion time is unknown.
uint32_t CUpti_ActivityKernel7::contextId [inherited]: The ID of the context where the kernel is executing.
uint32_t CUpti_ActivityKernel7::correlationId [inherited]: The correlation ID of the kernel. Each kernel execution is assigned a unique correlation ID that is identical to the correlation ID in the driver or runtime API activity record that launched the kernel.
uint32_t CUpti_ActivityKernel7::deviceId [inherited]: The ID of the device where the kernel is executing.
int32_t CUpti_ActivityKernel7::dynamicSharedMemory [inherited]: The dynamic shared memory reserved for the kernel, in bytes.
uint64_t CUpti_ActivityKernel7::end [inherited]: The end timestamp for the kernel execution, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the kernel.
uint8_t CUpti_ActivityKernel7::executed [inherited]: The cache configuration used for the kernel. The value is one of the CUfunc_cache enumeration values from cuda.h.
uint32_t CUpti_ActivityKernel7::graphId [inherited]: The unique ID of the graph that launched this kernel through graph launch APIs. This field will be 0 if the kernel is not launched through graph launch APIs.
uint64_t CUpti_ActivityKernel7::graphNodeId [inherited]: The unique ID of the graph node that launched this kernel through graph launch APIs. This field will be 0 if the kernel is not launched through graph launch APIs.
int64_t CUpti_ActivityKernel7::gridId [inherited]: The grid ID of the kernel. Each kernel is assigned a unique grid ID at runtime.
int32_t CUpti_ActivityKernel7::gridX [inherited]: The X-dimension grid size for the kernel.
int32_t CUpti_ActivityKernel7::gridY [inherited]: The Y-dimension grid size for the kernel.
int32_t CUpti_ActivityKernel7::gridZ [inherited]: The Z-dimension grid size for the kernel.
uint8_t CUpti_ActivityKernel7::isSharedMemoryCarveoutRequested [inherited]: This indicates if CU_FUNC_ATTRIBUTE_PREFERRED_SHARED_MEMORY_CARVEOUT was updated for the kernel launch
CUpti_ActivityKind CUpti_ActivityKernel7::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_KERNEL or CUPTI_ACTIVITY_KIND_CONCURRENT_KERNEL.
uint8_t CUpti_ActivityKernel7::launchType [inherited]: The indicates if the kernel was executed via a regular launch or via a single/multi device cooperative launch.

See also:

CUpti_ActivityLaunchType
uint32_t CUpti_ActivityKernel7::localMemoryPerThread [inherited]: The amount of local memory reserved for each thread, in bytes.
uint32_t CUpti_ActivityKernel7::localMemoryTotal [inherited]: The total amount of local memory reserved for the kernel, in bytes.
const char * CUpti_ActivityKernel7::name [inherited]: The name of the kernel. This name is shared across all activity records representing the same kernel, and so should not be modified.
CUaccessPolicyWindow * CUpti_ActivityKernel7::pAccessPolicyWindow [inherited]: The pointer to the access policy window. The structure CUaccessPolicyWindow is defined in cuda.h.
uint8_t CUpti_ActivityKernel7::padding [inherited]: Undefined. Reserved for internal use.
CUpti_ActivityPartitionedGlobalCacheConfig CUpti_ActivityKernel7::partitionedGlobalCacheExecuted [inherited]: The partitioned global caching executed for the kernel. Partitioned global caching is required to enable caching on certain chips, such as devices with compute capability 5.2. Partitioned global caching can be automatically disabled if the occupancy requirement of the launch cannot support caching.
CUpti_ActivityPartitionedGlobalCacheConfig CUpti_ActivityKernel7::partitionedGlobalCacheRequested [inherited]: The partitioned global caching requested for the kernel. Partitioned global caching is required to enable caching on certain chips, such as devices with compute capability 5.2.
uint64_t CUpti_ActivityKernel7::queued [inherited]: The timestamp when the kernel is queued up in the command buffer, in ns. A value of CUPTI_TIMESTAMP_UNKNOWN indicates that the queued time could not be collected for the kernel. This timestamp is not collected by default. Use API cuptiActivityEnableLatencyTimestamps() to enable collection.

Command buffer is a buffer written by CUDA driver to send commands like kernel launch, memory copy etc to the GPU. All launches of CUDA kernels are asynchrnous with respect to the host, the host requests the launch by writing commands into the command buffer, then returns without checking the GPU's progress.
uint16_t CUpti_ActivityKernel7::registersPerThread [inherited]: The number of registers required for each thread executing the kernel.
uint8_t CUpti_ActivityKernel7::requested [inherited]: The cache configuration requested by the kernel. The value is one of the CUfunc_cache enumeration values from cuda.h.
void * CUpti_ActivityKernel7::reserved0 [inherited]: Undefined. Reserved for internal use.
uint8_t CUpti_ActivityKernel7::sharedMemoryCarveoutRequested [inherited]: Shared memory carveout value requested for the function in percentage of the total resource. The value will be updated only if field isSharedMemoryCarveoutRequested is set.
uint8_t CUpti_ActivityKernel7::sharedMemoryConfig [inherited]: The shared memory configuration used for the kernel. The value is one of the CUsharedconfig enumeration values from cuda.h.
uint32_t CUpti_ActivityKernel7::sharedMemoryExecuted [inherited]: Shared memory size set by the driver.
CUpti_FuncShmemLimitConfig CUpti_ActivityKernel7::shmemLimitConfig [inherited]: The shared memory limit config for the kernel. This field shows whether user has opted for a higher per block limit of dynamic shared memory.
uint64_t CUpti_ActivityKernel7::start [inherited]: The start timestamp for the kernel execution, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the kernel.
int32_t CUpti_ActivityKernel7::staticSharedMemory [inherited]: The static shared memory allocated for the kernel, in bytes.
uint32_t CUpti_ActivityKernel7::streamId [inherited]: The ID of the stream where the kernel is executing.
uint64_t CUpti_ActivityKernel7::submitted [inherited]: The timestamp when the command buffer containing the kernel launch is submitted to the GPU, in ns. A value of CUPTI_TIMESTAMP_UNKNOWN indicates that the submitted time could not be collected for the kernel. This timestamp is not collected by default. Use API cuptiActivityEnableLatencyTimestamps() to enable collection.

6.44. CUpti_ActivityKernel8 Struct Reference

[CUPTI Activity API]

This activity record represents a kernel execution (CUPTI_ACTIVITY_KIND_KERNEL and CUPTI_ACTIVITY_KIND_CONCURRENT_KERNEL)

Public Variables

int32_t blockX
int32_t blockY
int32_t blockZ
CUpti_ActivityKernel8::@17 cacheConfig
uint32_t channelID
CUpti_ChannelType channelType
uint32_t clusterSchedulingPolicy
uint32_t clusterX
uint32_t clusterY
uint32_t clusterZ
uint64_t completed
uint32_t contextId
uint32_t correlationId
uint32_t deviceId
int32_t dynamicSharedMemory
uint64_t end
uint8_t executed
uint32_t graphId
uint64_t graphNodeId
int64_t gridId
int32_t gridX
int32_t gridY
int32_t gridZ
uint8_t isSharedMemoryCarveoutRequested
CUpti_ActivityKind kind
uint8_t launchType
uint32_t localMemoryPerThread
uint32_t localMemoryTotal
uint64_t localMemoryTotal_v2
const char * name
CUaccessPolicyWindow * pAccessPolicyWindow
uint8_t padding
CUpti_ActivityPartitionedGlobalCacheConfig partitionedGlobalCacheExecuted
CUpti_ActivityPartitionedGlobalCacheConfig partitionedGlobalCacheRequested
uint64_t queued
uint16_t registersPerThread
uint8_t requested
void * reserved0
uint8_t sharedMemoryCarveoutRequested
uint8_t sharedMemoryConfig
uint32_t sharedMemoryExecuted
CUpti_FuncShmemLimitConfig shmemLimitConfig
uint64_t start
int32_t staticSharedMemory
uint32_t streamId
uint64_t submitted

Variables

int32_t CUpti_ActivityKernel8::blockX [inherited]: The X-dimension block size for the kernel.
int32_t CUpti_ActivityKernel8::blockY [inherited]: The Y-dimension block size for the kernel.
int32_t CUpti_ActivityKernel8::blockZ [inherited]: The Z-dimension grid size for the kernel.
CUpti_ActivityKernel8::@17 CUpti_ActivityKernel8::cacheConfig [inherited]: For devices with compute capability 7.0+ cacheConfig values are not updated in case field isSharedMemoryCarveoutRequested is set
uint32_t CUpti_ActivityKernel8::channelID [inherited]: The ID of the HW channel on which the kernel is launched.
CUpti_ChannelType CUpti_ActivityKernel8::channelType [inherited]: The type of the channel
uint32_t CUpti_ActivityKernel8::clusterSchedulingPolicy [inherited]: The cluster scheduling policy for the kernel. Refer CUclusterSchedulingPolicy Field is valid for devices with compute capability 9.0 and higher
uint32_t CUpti_ActivityKernel8::clusterX [inherited]: The X-dimension cluster size for the kernel. Field is valid for devices with compute capability 9.0 and higher
uint32_t CUpti_ActivityKernel8::clusterY [inherited]: The Y-dimension cluster size for the kernel. Field is valid for devices with compute capability 9.0 and higher
uint32_t CUpti_ActivityKernel8::clusterZ [inherited]: The Z-dimension cluster size for the kernel. Field is valid for devices with compute capability 9.0 and higher
uint64_t CUpti_ActivityKernel8::completed [inherited]: The completed timestamp for the kernel execution, in ns. It represents the completion of all it's child kernels and the kernel itself. A value of CUPTI_TIMESTAMP_UNKNOWN indicates that the completion time is unknown.
uint32_t CUpti_ActivityKernel8::contextId [inherited]: The ID of the context where the kernel is executing.
uint32_t CUpti_ActivityKernel8::correlationId [inherited]: The correlation ID of the kernel. Each kernel execution is assigned a unique correlation ID that is identical to the correlation ID in the driver or runtime API activity record that launched the kernel.
uint32_t CUpti_ActivityKernel8::deviceId [inherited]: The ID of the device where the kernel is executing.
int32_t CUpti_ActivityKernel8::dynamicSharedMemory [inherited]: The dynamic shared memory reserved for the kernel, in bytes.
uint64_t CUpti_ActivityKernel8::end [inherited]: The end timestamp for the kernel execution, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the kernel.
uint8_t CUpti_ActivityKernel8::executed [inherited]: The cache configuration used for the kernel. The value is one of the CUfunc_cache enumeration values from cuda.h.
uint32_t CUpti_ActivityKernel8::graphId [inherited]: The unique ID of the graph that launched this kernel through graph launch APIs. This field will be 0 if the kernel is not launched through graph launch APIs.
uint64_t CUpti_ActivityKernel8::graphNodeId [inherited]: The unique ID of the graph node that launched this kernel through graph launch APIs. This field will be 0 if the kernel is not launched through graph launch APIs.
int64_t CUpti_ActivityKernel8::gridId [inherited]: The grid ID of the kernel. Each kernel is assigned a unique grid ID at runtime.
int32_t CUpti_ActivityKernel8::gridX [inherited]: The X-dimension grid size for the kernel.
int32_t CUpti_ActivityKernel8::gridY [inherited]: The Y-dimension grid size for the kernel.
int32_t CUpti_ActivityKernel8::gridZ [inherited]: The Z-dimension grid size for the kernel.
uint8_t CUpti_ActivityKernel8::isSharedMemoryCarveoutRequested [inherited]: This indicates if CU_FUNC_ATTRIBUTE_PREFERRED_SHARED_MEMORY_CARVEOUT was updated for the kernel launch
CUpti_ActivityKind CUpti_ActivityKernel8::kind [inherited]: The activity record kind, must be CUPTI_ACTIVITY_KIND_KERNEL or CUPTI_ACTIVITY_KIND_CONCURRENT_KERNEL.
uint8_t CUpti_ActivityKernel8::launchType [inherited]: The indicates if the kernel was executed via a regular launch or via a single/multi device cooperative launch.

See also:

CUpti_ActivityLaunchType
uint32_t CUpti_ActivityKernel8::localMemoryPerThread [inherited]: The amount of local memory reserved for each thread, in bytes.
uint32_t CUpti_ActivityKernel8::localMemoryTotal [inherited]: The total amount of local memory reserved for the kernel, in bytes (deprecated in CUDA 11.8). Refer field localMemoryTotal_v2
uint64_t CUpti_ActivityKernel8::localMemoryTotal_v2 [inherited]: The total amount of local memory reserved for the kernel, in bytes.
const char * CUpti_ActivityKernel8::name [inherited]: The name of the kernel. This name is shared across all activity records representing the same kernel, and so should not be modified.
CUaccessPolicyWindow * CUpti_ActivityKernel8::pAccessPolicyWindow [inherited]: The pointer to the access policy window. The structure CUaccessPolicyWindow is defined in cuda.h.
uint8_t CUpti_ActivityKernel8::padding [inherited]: Undefined. Reserved for internal use.
CUpti_ActivityPartitionedGlobalCacheConfig CUpti_ActivityKernel8::partitionedGlobalCacheExecuted [inherited]: The partitioned global caching executed for the kernel. Partitioned global caching is required to enable caching on certain chips, such as devices with compute capability 5.2. Partitioned global caching can be automatically disabled if the occupancy requirement of the launch cannot support caching.
CUpti_ActivityPartitionedGlobalCacheConfig CUpti_ActivityKernel8::partitionedGlobalCacheRequested [inherited]: The partitioned global caching requested for the kernel. Partitioned global caching is required to enable caching on certain chips, such as devices with compute capability 5.2.
uint64_t CUpti_ActivityKernel8::queued [inherited]: The timestamp when the kernel is queued up in the command buffer, in ns. A value of CUPTI_TIMESTAMP_UNKNOWN indicates that the queued time could not be collected for the kernel. This timestamp is not collected by default. Use API cuptiActivityEnableLatencyTimestamps() to enable collection.

Command buffer is a buffer written by CUDA driver to send commands like kernel launch, memory copy etc to the GPU. All launches of CUDA kernels are asynchrnous with respect to the host, the host requests the launch by writing commands into the command buffer, then returns without checking the GPU's progress.
uint16_t CUpti_ActivityKernel8::registersPerThread [inherited]: The number of registers required for each thread executing the kernel.
uint8_t CUpti_ActivityKernel8::requested [inherited]: The cache configuration requested by the kernel. The value is one of the CUfunc_cache enumeration values from cuda.h.
void * CUpti_ActivityKernel8::reserved0 [inherited]: Undefined. Reserved for internal use.
uint8_t CUpti_ActivityKernel8::sharedMemoryCarveoutRequested [inherited]: Shared memory carveout value requested for the function in percentage of the total resource. The value will be updated only if field isSharedMemoryCarveoutRequested is set.
uint8_t CUpti_ActivityKernel8::sharedMemoryConfig [inherited]: The shared memory configuration used for the kernel. The value is one of the CUsharedconfig enumeration values from cuda.h.
uint32_t CUpti_ActivityKernel8::sharedMemoryExecuted [inherited]: Shared memory size set by the driver.
CUpti_FuncShmemLimitConfig CUpti_ActivityKernel8::shmemLimitConfig [inherited]: The shared memory limit config for the kernel. This field shows whether user has opted for a higher per block limit of dynamic shared memory.
uint64_t CUpti_ActivityKernel8::start [inherited]: The start timestamp for the kernel execution, in ns. A value of 0 for both the start and end timestamps indicates that timestamp information could not be collected for the kernel.
int32_t CUpti_ActivityKernel8::staticSharedMemory [inherited]: The static shared memory allocated for the kernel, in bytes.