CuPerf
2026-01-13·1 min·80 words
C++23 CLI tool for NVIDIA GPU benchmarking with extensible architecture.
Filed underproject · cuda · gpu · benchmarking
C++23 CLI tool for NVIDIA GPU benchmarking (memory, compute, tensor cores) with extensible architecture and JSON/CSV output.
Features
- Memory bandwidth, compute throughput, and tensor core benchmarking
- Kernel launch overhead and reduction performance measurements
- Multiple data types (FP32, FP16, BF16, INT8, FP4)
- Extensible architecture
- Multiple output formats (console, JSON, CSV)
Technologies
CUDA, C++23, Parallel Computing, Profiling
Links
Status
Active Development - 2025 -- Present