CUDA
also: CUDA Toolkit, CUDA Compute Capability
CUDA is NVIDIA's parallel computing platform that allows developers to use graphics processing units (GPUs) for general-purpose computing, significantly accelerating computationally intensive applications.
CUDA (Compute Unified Device Architecture) is a software framework developed by NVIDIA that enables developers to harness the parallel processing power of NVIDIA GPUs for tasks beyond graphics rendering. It allows code written in C, C++, Python, and other languages to execute on GPU hardware, achieving dramatic speedups for suitable workloads.
CUDA works by offloading parallel computation from the CPU to thousands of GPU cores operating simultaneously. For example, machine learning frameworks like TensorFlow and PyTorch use CUDA to accelerate neural network training, reducing what might take days on a CPU to hours on a GPU.
To use CUDA on Linux, you need the NVIDIA GPU driver installed along with the CUDA Toolkit, which provides libraries, compilers, and tools. You can check GPU availability with nvidia-smi to verify driver installation and monitor GPU usage.
CUDA is widely used in deep learning, scientific computing, data analytics, and video processing—any domain where massive parallelism can be exploited to speed up computation.