NVIDIA CUDA Profiling Tools Interface (CUPTI) - CUDA Toolkit
The NVIDIA® CUDA Profiling Tools Interface (CUPTI) is a dynamic library that enables the creation of profiling and tracing tools that target CUDA applications. CUPTI provides a set of APIs targeted at ISVs creating profilers and other performance optimization tools:
- the Activity API,
- the Callback API,
- the Event API,
- the Metric API, and
- the Profiler API
Using these CUPTI APIs, independent software developers can create profiling tools that provide low and deterministic profiling overhead on the target system, while giving insight into the CPU and GPU behavior of CUDA applications. Normally packaged with the CUDA Toolkit, NVIDIA occasionally uses this page to provide CUPTI improvements and bug fixes between toolkit releases.
- Trace CUDA API by registering callbacks for API calls of interest
- Full support for entry and exit points in the CUDA C Runtime (CUDART) and CUDA Driver
- GPU workload trace for the activities happening on the GPU, which includes kernel executions, memory operations (e.g., Host-to-Device memory copies) and memset operations.
- CUDA Unified Memory trace for transfers from host to device, device to host, device to device and page faults on CPU and GPU etc.
- Normalized timestamps for CPU and GPU trace
- Profile hardware and software event counters, including:
- Utilization metrics for various hardware units
- Instruction count and throughput
- Memory load/store events and throughput
- Cache hits/misses
- Branches and divergent branches
- Many more
- Enables automated bottleneck identification based on metrics such as instruction throughput, memory throughput, and more
- Range profiling to enable metric collection over concurrent kernel launches within a range
- Metrics attribution at the high-level source code and the executed assembly instructions.
- Device-wide sampling of the program counter (PC). The PC Sampling gives the number of samples for each source and assembly line with various stall reasons.
- CUPTI has made the following changes as part of the CUDA Toolkit 11.3 release,
- A new set of CUPTI APIs for PC sampling data collection are provided in the header file cupti_pcsampling.h which support continuous mode data collection without serializing kernel execution and have a lower runtime overhead. Along with these a utility library is provided in the header file cupti_pcsampling_util.h which has APIs for GPU assembly to CUDA-C source correlation and for reading and writing the PC sampling data from/to files. Refer to the section CUPTI PC Sampling API for more details.
- Overhead reduction for tracing of CUDA memcopies.
CUpti_PcieGenis extended to include PCIe Gen 5.
- The following functions are deprecated and will be removed in a future release:
APIs NVPW_MetricsContext_*from the header nvperf_host.h. It is recommended to use the struct
NVPW_MetricsEvaluatorand related APIs
NVPW_MetricsEvaluator_*instead. Profiling API samples have been updated to show how to use these APIs.
cuptiDeviceGetTimestampfrom the header cupti_events.h.
- Linux x86_64
- Windows x86_64
- Linux PowerPC
- Linux aarch64 SBSA
- DRIVE OS QNX aarch64
- DRIVE OS Linux aarch64
- Android aarch64
 available in the Embedded or Drive toolkits only
 does not support Profiling API data collection
Supported NVIDIA GPU architectures
- Activity and Callback APIs
- All architectures supported by CUDA Toolkit
- Event and Metric APIs
- Profiling APIs
- Ampere: A100 with Multi-Instance GPU, GA10x
- Volta: GV100, GV10B
- CUPTI can be found in the CUDA Toolkit 11.3 production release
- Please use the following drivers
- 465.87 (Windows)
- 465.19.01 (Linux)
To provide feedback, request additional features, or report issues, please use the Developer Forums.
When installing CUDA Toolkit 11.3 and specifying options, be sure to select CUDA > Development > Tools > CUPTI.