NVIDIA CUDA

E41922

NVIDIA CUDA is a parallel computing platform and programming model that enables developers to use NVIDIA GPUs for general-purpose high-performance computing.


Statements (65)
Predicate Object
instanceOf parallel computing platform
programming model
abbreviationOf Compute Unified Device Architecture
compatibleWith Linux
Windows
macOS (historical support)
developer NVIDIA
documentationURL https://docs.nvidia.com/cuda/
ecosystem NVIDIA GPU Computing Ecosystem
enables general-purpose computing on GPUs
high-performance computing
hasComponent CUDA Events
CUDA Graphs
CUDA Occupancy Calculator
CUDA Streams
includes CUDA Driver API
CUDA Math Library
CUDA Profiler tools
CUDA Runtime API
CUDA Toolkit
NCCL
Thrust
cuBLAS
cuDNN
cuFFT
cuRAND
cuSPARSE
nvcc compiler
introducedBy NVIDIA in 2006
license proprietary
marketedAs Compute Unified Device Architecture
programmingModel SIMT
requires NVIDIA GPU with CUDA capability
supports GPU-accelerated computing
GPUDirect
asynchronous execution
constant memory on GPU
dynamic parallelism
events
heterogeneous computing
massively parallel kernels
multi-GPU programming
pinned host memory
shared memory on GPU
streams
texture memory on GPU
unified memory
supportsLanguage CUDA C
CUDA C++
CUDA Fortran
Fortran via PGI/NVHPC compilers
Java via CUDA bindings
MATLAB via CUDA integration
Python via CUDA bindings
targetHardware NVIDIA GPU
toolkitDownloadURL https://developer.nvidia.com/cuda-downloads
usedFor computational finance
computer vision
data analytics
deep learning
image processing
machine learning
numerical simulation
scientific computing
signal processing

Referenced by (26)
Subject (surface form when different) Predicate
NVIDIA AI Enterprise ("NVIDIA CUDA libraries")
NVIDIA CUDA ("CUDA Runtime API")
NVIDIA CUDA ("CUDA Toolkit")
NVIDIA CUDA ("cuFFT")
NVIDIA CUDA ("CUDA Math Library")
includes
NVIDIA CUDA ("CUDA Graphs")
NVIDIA CUDA ("CUDA Streams")
NVIDIA CUDA ("CUDA Events")
NVIDIA CUDA ("CUDA Occupancy Calculator")
hasComponent
Clang ("CUDA C++")
NVIDIA CUDA ("CUDA C")
NVIDIA CUDA ("CUDA C++")
supportsLanguage
CuPy ("CUDA")
torchvision ("CUDA")
supports
NVIDIA DRIVE ("NVIDIA GPU technology")
basedOn
NVIDIA Corporation
brand
OpenACC ("CUDA")
competesWith
NVIDIA CUDA ("NVIDIA GPU Computing Ecosystem")
ecosystem
Theano ("CUDA")
implementationLanguage
NVIDIA DGX
includesSoftware
NVIDIA Corporation ("NVIDIA CUDA parallel computing platform")
knownFor
NVIDIA Corporation ("CUDA parallel computing platform")
notableTechnology
Jensen Huang ("CUDA parallel computing platform")
notableWork
TensorFlow ("CUDA")
programmingLanguage
PyTorch ("NVIDIA GPUs")
supportsHardware
Microsoft Cognitive Toolkit ("CUDA")
supportsStandard

Please wait…