https://lesitedefrancois.be/en/tags/parallel-computing/Recent content in Parallel-Computing on Le Site de FrançoisHugo -- gohugo.ioen© 2026 Françoishttps://lesitedefrancois.be/en/ai/cuda/Mon, 01 Jan 0001 00:00:00 +0000https://lesitedefrancois.be/en/ai/cuda/<p><strong>CUDA (Compute Unified Device Architecture)</strong> is NVIDIA&rsquo;s software platform for general-purpose computing on GPUs. Its objective is to give developers a familiar C/C++ (and Fortran, Python bindings) programming model with explicit control over <strong>kernels</strong> (functions that run on the device), <strong>streams</strong> (ordered queues of work), and <strong>memory spaces</strong> (host, device, unified). CUDA sits above the GPU driver and below frameworks such as cuDNN, NCCL, and higher-level ML stacks; it is the layer that makes it practical to implement custom operators, HPC solvers, and inference engines that are not covered by a closed library. For AI, virtually every major training and inference stack ultimately depends on CUDA (or a CUDA-compatible runtime) on NVIDIA hardware.</p>