Here are some GPU tools and libraries commonly used and (partially) developed at NLeSC.
- Kernel Tuner: Software development tool for the creation of highly-optimized and tuned GPU applications.
- Kernel Launcher: Using C++ magic to launch/capture CUDA kernels and tune them with Kernel Tuner
- Kernel Float: header-only library to use vector and low-precision floating-point types (16 bit, 8 bit) in GPU code
- CUDA wrappers: C++ wrapper for the Nvidia/HIP C libraries (e.g. CUDA driver, nvrtc, hiprtc, cuFFT, hipFFT, etc.)
- KMM: A lightweight C++ middleware for accelerated GPU computing
- PowerSensor 3: custom-built device that measures the instantaneous power consumption of GPUs and other devices at a high time resolution.
- nvmetrics: library designed to facilitate the measurement of GPU metrics using NVIDIA CUPTI (CUDA Profiling Tools Interface)
- KT dashboard: Monitor, analyze, and visualize an active or completed auto-tuning run of Kernel Tuner
- ccglib: CUDA library provides complex GEMM kernels for complex data.
- PMT: PMT is a high-level software library capable of collecting power consumption measurements on various hardware.