CUPTI
CUPTI is NVIDIA’s interface for building tools that observe CUDA activity, API calls and supported performance information.
How it works
A profiler can use callbacks and activity records to relate host submission to device execution. Records require correlation and timestamp interpretation. Collection can add overhead or lose records when buffers and consumers cannot keep up.
Work through an example
A timeline with a long CPU API duration and a later GPU kernel start contains two different intervals. Keep their boundaries separate when assigning end-to-end latency.
Where the term can mislead
Instrumentation changes the workload. Compare against an uninstrumented baseline and inspect collection errors before trusting a trace.
Connect the idea
Take it into practice
Continue with the existing concept, apply it to an interview problem, or follow a course lesson. Access requirements are shown before you open a page.
New to GPUs? Start with the free execution-model concept.
Sources and scope
Examples here are teaching calculations, not measured GPU performance. Consult the linked documentation for the exact device, instruction and tool version. Source review: 25 October 2026.