Gpu Glossary

Browse 16 gpu terms defined in plain English, from the cultural dictionary of computing.

16 Gpu Terms

Compute Shader
A Compute Shader is a development concept used to shape how software is built, executed, or maintained in real systems. Developers rely on it to make code...
CUDA
NVIDIA's programming platform and runtime for accelerating parallel workloads on GPUs. It influences how models are trained, evaluated, or served, and it can...
CUDA Core
A GPU execution unit that performs many floating-point or integer operations in parallel. It influences how models are trained, evaluated, or served, and it...
Data Parallelism
A form of parallelization where the same operation is applied simultaneously to different elements of a dataset, distributing data across multiple processors....
Flash Attention
An optimized attention implementation that reduces memory traffic and accelerates transformer training or inference. It influences how models are trained,...
Fragment Shader
A Fragment Shader is a development concept used to shape how software is built, executed, or maintained in real systems. Developers rely on it to make code...
PCIe
Peripheral Component Interconnect Express — a high-speed serial expansion bus standard used to connect GPUs, NVMe SSDs, network cards, and other peripherals to...
Pixel Buffer
A contiguous block of memory that stores raw pixel data (color values for each pixel in a grid) representing a frame or texture. In OpenGL it's called a Pixel...
Rasterization
The process of converting vector graphics or 3D geometry (triangles, primitives) into a grid of pixels for display on a screen. The GPU's rasterization stage...
Render Pipeline
The sequence of stages that transforms 3D scene data into a 2D image on screen — vertex processing, primitive assembly, rasterization, fragment shading, and...
Shader
A small program that runs on the GPU to control how vertices, pixels, or compute workloads are processed, written in languages like GLSL, HLSL, or WGSL.
Texture
A bitmap image mapped onto the surface of a 3D model or 2D sprite to give it visual detail — stored in GPU memory and sampled by shaders during rendering using...
Torch
Short for PyTorch — the open-source deep learning framework built by Meta that provides GPU-accelerated tensor computation and automatic differentiation,...
Triton Kernel
A custom GPU compute kernel written with Triton to optimize operations common in machine learning workloads.
Vertex Shader
A GPU program that runs once per vertex in the graphics pipeline, transforming vertex positions from model space through view and projection matrices into clip...
WebGPU
A modern browser API for GPU-accelerated graphics and general-purpose GPU computation, succeeding WebGL. Provides lower-level GPU access similar to...

Related Topics