Module cuda 

CUDA profiler: wraps ncu, nsys, and CUPTI. See spec sections 4.1.1 (ncu), 4.1.2 (nsys), 4.1.3 (CUPTI).

Structs

NcuProfiler

Wraps ncu CLI for kernel-level profiling.

NsysKernelStat

Parsed nsys stats output — one entry per kernel.

NsysProfiler

Wraps nsys CLI for system-wide timeline profiling.

Enums

NcuSection

ncu metric sections — lazily collect only what’s requested.

Functions

ncu_metrics_to_profile

Build a FullProfile from ncu metrics.

parse_ncu_csv

Parse ncu CSV output into a metric name → value map. ncu CSV format: “ID”,“Metric Name”,“Metric Unit”,“Metric Value”

profile_binary

Profile an arbitrary binary via nsys.

profile_cublas

Profile cuBLAS operations.

profile_kernel

Profile a CUDA PTX kernel via ncu.

profile_python

Profile a Python script via nsys + perf stat.

run_trace

Run cgp trace — system-wide timeline via nsys.