General-purpose computing on graphics processing units

General-purpose computing on graphics processing units (GPGPU) uses GPUs to perform workloads beyond graphics rendering, exploiting their ability to process many data elements in parallel.

Connect

SIMT: Single instruction, multiple threads is an execution model in which groups of threads execute the same instruction across different data. GPU programming commonly organizes parallel work as many threads following this model.
Deep learning: Deep learning is machine learning based on neural networks with multiple layers of learned representations. Large neural-network operations map efficiently to GPU parallel arithmetic.
Central processing unit: A central processing unit is a general-purpose processor designed to handle a wide range of sequential and parallel tasks. CPUs often suit branch-heavy, latency-sensitive work better than throughput-oriented GPU execution.
Parallel computing: Parallel computing performs multiple computations at the same time using multiple processing elements. GPGPU is a hardware-specific approach to parallel computing.
NVIDIA: NVIDIA is a technology company that designs GPUs, processors, and software platforms for computing. Its GPUs and CUDA platform helped establish a large GPGPU ecosystem.
CUDA: CUDA is NVIDIA’s parallel computing platform and programming model for GPUs. It provides a widely used way to write general-purpose programs for NVIDIA GPUs.
Molecular dynamics: Molecular dynamics simulates atomic motion by numerically integrating equations of motion over time. Many particle-force calculations can run concurrently on GPUs.
Field-programmable gate array: A field-programmable gate array is an integrated circuit whose logic can be configured after manufacturing. FPGAs offer a reconfigurable alternative when fixed GPU execution patterns are a poor fit.
Data parallelism: Data parallelism applies the same operation concurrently to different elements of a dataset. It is the central workload pattern that GPU execution is designed to exploit.
Heterogeneous computing: Heterogeneous computing combines different processor types to execute workloads suited to their capabilities. GPGPU made CPU–GPU cooperation a common design approach.