GPU Glossary
GPU Glossary
The GPU Glossary stands out as a highly structured and comprehensive reference that organizes complex GPU concepts into clear categories spanning hardware, device software, host software, and performance metrics. It provides concise definitions for foundational architectural components such as streaming multiprocessors, tensor cores, memory hierarchies, and cache structures, making it useful for both beginners and advanced practitioners. The inclusion of CUDA-related programming model terms-like warps, thread blocks, kernels, and PTX-helps bridge the gap between hardware understanding and practical software development. Its coverage of performance concepts, including bottlenecks, arithmetic intensity, occupancy, and warp divergence, adds depth for users seeking to optimize GPU workloads. The glossary’s breadth is impressive, capturing everything from low-level assembly to high-level profiling tools, which makes it a strong one-stop resource for GPU terminology. The layout is straightforward and easy to navigate, allowing readers to quickly jump between categories without feeling overwhelmed. Overall, the page succeeds as a well-organized, technically rich glossary that supports anyone working with GPU computing or CUDA-based workflows.
Source: modal.com/gpu-glossary