Release Notes

The CUPTI Python Release Notes include new features, important bug fixes, known issues, supported platforms, and supported GPUs.

Updates in CUPTI Python Releases

Updates in CUPTI Python 13.5.0

Refer CUPTI C 13.5.0 release notes.

New Features

• Added support for multi-subscriber mode in the Activity and Callback APIs.

  • Each subscriber created with cupti.cupti.subscribe() or cupti.cupti.subscribe_v2() receives only the callbacks and user data that it registered.

  • New per-subscriber Activity APIs:

    • cupti.cupti.activity_register_callbacks_v2()

    • cupti.cupti.activity_enable_v2()

    • cupti.cupti.activity_enable_and_dump_v2()

    • cupti.cupti.activity_disable_v2()

    • cupti.cupti.activity_enable_runtime_api_v2()

    • cupti.cupti.activity_enable_driver_api_v2()

    • cupti.cupti.activity_get_attribute_v2()

    • cupti.cupti.activity_set_attribute_v2()

    • cupti.cupti.activity_get_num_dropped_records_v2()

    • cupti.cupti.activity_push_external_correlation_id_v2()

    • cupti.cupti.activity_pop_external_correlation_id_v2()

    • cupti.cupti.get_timestamp_v2()

  • Added cupti.cupti.activity_get_enabled_kinds() and cupti.cupti.get_enabled_callbacks() to query the activity kinds and callback IDs enabled for a subscriber.

  • Added the sample CuptiMultiSubscriberVectorAdd.py.

  • Updated the CUPTI Python samples to use the per-subscriber (v2) Activity APIs.

• Added the --hes option to enable hardware event tracing in the CuptiVectorAddDrv.py and cupyprof.py samples.

• Added support for API cupti.cupti.get_version().

• Added support for logical endpoint resource callbacks: cupti.cupti.CallbackIdResource.LOGICAL_ENDPOINT_CREATED and cupti.cupti.CallbackIdResource.LOGICAL_ENDPOINT_DESTROYED, with the new cupti.cupti.LogicalEndpointResourceData and cupti.cupti.LogicalEndpointType.

• New activity class: cupti.cupti.ActivityCudaEvent3.

• Added the following function parameter classes for CUDA Runtime APIs. Some members of these classes are not exposed, see Known Limitations.

  • cupti.cupti.CudaMalloc3DV3020Params

  • cupti.cupti.CudaMalloc3DArrayV3020Params

  • cupti.cupti.CudaMallocMipmappedArrayV5000Params

  • cupti.cupti.CudaMemset3DV3020Params

  • cupti.cupti.CudaMemset3DPtdsV7000Params

  • cupti.cupti.CudaMemset3DAsyncV3020Params

  • cupti.cupti.CudaMemset3DAsyncPtszV7000Params

  • cupti.cupti.CudaMemAdviseV12020Params

  • cupti.cupti.CudaMemPrefetchAsyncV12020Params

  • cupti.cupti.CudaMemPrefetchAsyncPtszV12020Params

Deprecated features

• Deprecated activity class: cupti.cupti.ActivityCudaEvent2.

• Support for Python 3.10 is deprecated and will be dropped in a future release.

Updates in CUPTI Python 13.4.0

Refer CUPTI C 13.4.0 release notes.

New Features

• Added support for function parameter structures for CUDA Runtime, CUDA Driver and NVTX APIs. These allow for convenient consumption of CallbackData.function_params and NvtxData.function_params in CUPTI callbacks. See Function Parameter Classes for CUDA and NVTX APIs.

• Added support for GPU clock control. See Clock Control.

• Added support for API cupti.cupti.get_cuda_event_id().

• New activity classes: cupti.cupti.ActivityGreenContext3, cupti.cupti.ActivityKernel13, cupti.cupti.ActivityMemcpy7, cupti.cupti.ActivityMemset5.

API Changes

• Renamed classes and enums to follow PEP 8 naming conventions. The old names remain available as aliases for backwards compatibility.

  Old Name	New Name
  driver_api_trace_cbid	DriverApiTraceCbid
  runtime_api_trace_cbid	RuntimeApiTraceCbid
  Driver_api_trace_cbid	DriverApiTraceCbid
  Runtime_api_trace_cbid	RuntimeApiTraceCbid
  Nvtx_api_trace_cbid	NvtxApiTraceCbid
  PmSampling_TriggerMode	PmSamplingTriggerMode
  PmSampling_DecodeStopReason	PmSamplingDecodeStopReason
  PmSampling_HardwareBuffer_AppendMode	PmSamplingHardwareBufferAppendMode
  Device_GetChipName_Params	DeviceGetChipNameParams
  PmSampling_CounterDataImage_Initialize_Params	PmSamplingCounterDataImageInitializeParams
  PmSampling_CounterData_GetSampleInfo_Params	PmSamplingCounterDataGetSampleInfoParams
  PmSampling_DecodeData_Params	PmSamplingDecodeDataParams
  PmSampling_Disable_Params	PmSamplingDisableParams
  PmSampling_Enable_Params	PmSamplingEnableParams
  PmSampling_GetCounterAvailability_Params	PmSamplingGetCounterAvailabilityParams
  PmSampling_GetCounterDataInfo_Params	PmSamplingGetCounterDataInfoParams
  PmSampling_GetCounterDataSize_Params	PmSamplingGetCounterDataSizeParams
  PmSampling_SetConfig_Params	PmSamplingSetConfigParams
  PmSampling_Start_Params	PmSamplingStartParams
  PmSampling_Stop_Params	PmSamplingStopParams
  Profiler_DeInitialize_Params	ProfilerDeInitializeParams
  Profiler_Host_ConfigAddMetrics_Params	ProfilerHostConfigAddMetricsParams
  Profiler_Host_Deinitialize_Params	ProfilerHostDeinitializeParams
  Profiler_Host_EvaluateToGpuValues_Params	ProfilerHostEvaluateToGpuValuesParams
  Profiler_Host_GetBaseMetrics_Params	ProfilerHostGetBaseMetricsParams
  Profiler_Host_GetConfigImageSize_Params	ProfilerHostGetConfigImageSizeParams
  Profiler_Host_GetConfigImage_Params	ProfilerHostGetConfigImageParams
  Profiler_Host_GetMaxNumHardwareMetricsPerPass_Params	ProfilerHostGetMaxNumHardwareMetricsPerPassParams
  Profiler_Host_GetMetricProperties_Params	ProfilerHostGetMetricPropertiesParams
  Profiler_Host_GetMetricsInSinglePassSet_Params	ProfilerHostGetMetricsInSinglePassSetParams
  Profiler_Host_GetNumOfPasses_Params	ProfilerHostGetNumOfPassesParams
  Profiler_Host_GetRangeName_Params	ProfilerHostGetRangeNameParams
  Profiler_Host_GetSinglePassSets_Params	ProfilerHostGetSinglePassSetsParams
  Profiler_Host_GetSubMetrics_Params	ProfilerHostGetSubMetricsParams
  Profiler_Host_GetSupportedChips_Params	ProfilerHostGetSupportedChipsParams
  Profiler_Host_Initialize_Params	ProfilerHostInitializeParams
  Profiler_Initialize_Params	ProfilerInitializeParams

• In cupti.pm_sampling, HardwareBuffer_AppendMode has been renamed to HardwareBufferAppendMode. The old name remains available as an alias.

Deprecated features

• Deprecated activity classes: cupti.cupti.ActivityGreenContext2, cupti.cupti.ActivityKernel12, cupti.cupti.ActivityMemcpy6, cupti.cupti.ActivityMemset4.

Updates in CUPTI Python 13.3.1

Refer CUPTI C 13.3.1 release notes.

Backwards Compatibility Fixes

• Restored activity classes that were previously dropped, now marked as deprecated for backwards compatibility.

  • cupti.cupti.ActivitySynchronization, cupti.cupti.ActivityCudaEvent, cupti.cupti.ActivityUnifiedMemoryCounter2, cupti.cupti.ActivityMemcpy5 (dropped in 12.8.0)

  • cupti.cupti.ActivityKernel9, cupti.cupti.ActivityMemoryPool2 (dropped in 13.0.0)

  • cupti.cupti.ActivityKernel10, cupti.cupti.ActivityContext3, cupti.cupti.ActivityMarkerData, cupti.cupti.ActivityDevice5 (dropped in 13.2.0)

  • cupti.cupti.ActivityKernel11, cupti.cupti.ActivityGreenContext (dropped in 13.3.0)

• Moving forward, CUPTI Python will be consistent with CUPTI C and also include types moved to cupti_activity_deprecated.h .

Updates in CUPTI Python 13.3.0

Refer CUPTI C 13.3.0 release notes.

New Features

• Added support for CUPTI’s PM Sampling API: 1:1 bindings and a higher-level pythonic API in cupti.pm_sampling.

• Added support for CUPTI’s Profiler Host API: 1:1 bindings and a higher-level pythonic API in cupti.profiler_host.

• CUPTI Python now uses cuda-pathfinder to locate and load libcupti.so. For more details, see Locating the CUPTI Library.

• New activity classes: cupti.cupti.ActivityKernel12, cupti.cupti.ActivityGreenContext2.

• Added support for API cupti.cupti.is_tracing_session_running().

Deprecated and dropped features

• Dropped activity classes: cupti.cupti.ActivityKernel11, cupti.cupti.ActivityGreenContext.

Updates in CUPTI Python 13.2.0

Refer CUPTI C 13.2.0 release notes.

New Features

• The cupti.cupti.CallbackDomain.NVTX callback domain is now supported.

• The cupti.cupti.CallbackData.function_params field of cupti.cupti.CallbackData is now supported.

• New Activity kinds and classes:

  • cupti.cupti.ActivityKind.GRAPH_HOST_NODE with cupti.cupti.ActivityGraphHostNode

  • cupti.cupti.ActivityKind.COMPUTE_ENGINE_CTX_SWITCH with cupti.cupti.ActivityComputeEngineCtxSwitch

  • cupti.cupti.ActivityKind.GREEN_CONTEXT with cupti.cupti.ActivityGreenContext

  • cupti.cupti.ActivityKind.HOST_LAUNCH with cupti.cupti.ActivityHostLaunch

• New activity classes: cupti.cupti.ActivityKernel11, cupti.cupti.ActivityContext4, cupti.cupti.ActivityMarkerData2, cupti.cupti.ActivityDevice6

• Support for Python 3.14 has been added.

Deprecated and dropped features

• Dropped activity classes: cupti.cupti.ActivityKernel10, cupti.cupti.ActivityContext3, cupti.cupti.ActivityMarkerData, cupti.cupti.ActivityDevice5

• Support for Python 3.9 has been dropped.

Updates in CUPTI Python 13.0.1

Refer CUPTI C 13.0.1 release notes.

New Features

• Relaxed nvidia-cuda-cupti dependency: cupti-python now supports nvidia-cuda-cupti versions >=13.0.0 and <14.

• Relaxed cuda-bindings dependency: cupti-python now supports cuda-bindings versions >=13.0.0 and <14.

Updates in CUPTI Python 13.0.0

Refer CUPTI C 13.0.0 release notes.

New Features

• New Activity Kind cupti.cupti.ActivityKind.ROTATION and its corresponding activity class cupti.cupti.ActivityConfidentialComputeRotation has been added.

• Support for Activity Kind cupti.cupti.ActivityKind.RUNTIME and Callback Domain cupti.cupti.CallbackDomain.RUNTIME_API have been added.

• New activity classes cupti.cupti.ActivityKernel10 and cupti.cupti.ActivityMemoryPool3 have been introduced.

• Support for APIs cupti.cupti.subscribe_v2(), cupti.cupti.activity_enable_cuda_event_device_timestamps() and cupti.cupti.activity_enable_all_sync_records() has been added.

• Added cupti.pyi stub file to improve IDE support, including type hints and auto-completion.

• Support for Python 3.13 has been added.

• Support for Linux (aarch64 sbsa) architecture has been added.

Deprecated and dropped features

• Support for activity classes cupti.cupti.ActivityKernel9 and cupti.cupti.ActivityMemoryPool2 has been removed.

Updates in CUPTI Python 12.8.0

Refer CUPTI C 12.8.0 release notes.

New Features

• New Activity Kind cupti.cupti.ActivityKind.MEM_DECOMPRESS and its corresponding activity class cupti.cupti.ActivityMemDecompress has been added.

• New Activity Kind cupti.cupti.ActivityKind.DEVICE_GRAPH_TRACE and its corresponding activity class cupti.cupti.ActivityDeviceGraphTrace has been added.

• New activity classes cupti.cupti.ActivityMemcpy6, cupti.cupti.ActivityUnifiedMemoryCounter3, cupti.cupti.ActivityCudaEvent2 and cupti.cupti.ActivitySynchronization2 have been introduced.

• Support for API cupti.cupti.activity_enable_hw_trace() has been added.

Deprecated and dropped features

• Support for activity classes cupti.cupti.ActivityMemcpy5, cupti.cupti.ActivityUnifiedMemoryCounter2, cupti.cupti.ActivityCudaEvent and cupti.cupti.ActivitySynchronization has been removed.

Updates in CUPTI Python 12.6.0

This was the first CUPTI Python release. Refer CUPTI C 12.6.0 release notes.

Limitations

Known Issues

1. Decoding of NVTX annotations with extended payloads (used by libraries like NCCL) is not supported by CUPTI Python.

Known Limitations

1. CUPTI Python cannot collect tracing data when the Python interpreter is tearing down or is not in a proper state. The callbacks registered by functions cupti.cupti.activity_register_callbacks(), cupti.cupti.activity_register_timestamp_callback(), cupti.cupti.subscribe(), and cupti.cupti.subscribe_v2() will not be dispatched in such cases. For example, CUDA API calls that occur in a class’s __del__ method will not be traced.

2. CUPTI Python does not support the following CUDA Runtime API function parameter class members:

   Python Class	Member
   cupti.cupti.CudaMalloc3DV3020Params	extent
   cupti.cupti.CudaMalloc3DArrayV3020Params	extent
   cupti.cupti.CudaMallocMipmappedArrayV5000Params	extent
   cupti.cupti.CudaMemset3DV3020Params	pitched_dev_ptr
   cupti.cupti.CudaMemset3DV3020Params	extent
   cupti.cupti.CudaMemset3DPtdsV7000Params	pitched_dev_ptr
   cupti.cupti.CudaMemset3DPtdsV7000Params	extent
   cupti.cupti.CudaMemset3DAsyncV3020Params	pitched_dev_ptr
   cupti.cupti.CudaMemset3DAsyncV3020Params	extent
   cupti.cupti.CudaMemset3DAsyncPtszV7000Params	pitched_dev_ptr
   cupti.cupti.CudaMemset3DAsyncPtszV7000Params	extent
   cupti.cupti.CudaMemAdviseV12020Params	location
   cupti.cupti.CudaMemPrefetchAsyncV12020Params	location
   cupti.cupti.CudaMemPrefetchAsyncPtszV12020Params	location

3. CUPTI Python does not support the following CUDA Driver API function parameter class members:

   Python Class	Member
   cupti.cupti.CuIpcOpenEventHandleParams	handle
   cupti.cupti.CuIpcOpenMemHandleParams	handle
   cupti.cupti.CuIpcOpenMemHandleV2Params	handle
   cupti.cupti.CuMemAdviseV2Params	location
   cupti.cupti.CuMemPrefetchAsyncV2Params	location
   cupti.cupti.CuMemPrefetchAsyncV2PtszParams	location

Support

• Information on supported python versions, platforms, and GPUs.

Python Support

• Python 3.10 - 3.14

  • Python 3.10 support is deprecated and will be dropped in a future release.

Platform Support

• Linux (x86_64)

• Linux (aarch64 sbsa)

GPU Support

• Turing and later GPU architectures, i.e. devices with compute capability 7.5 and higher.