6 dépôts
Enabling GPU-accelerated implementations of FFT, matrix multiply, and solver operations via dedicated math libraries.
Distinct from GPU Acceleration Libraries: Distinct from general GPU Acceleration Libraries: focuses on enabling specific math library backends (cuFFTDx, cuBLASDx, cuSolverDx) rather than general GPU offloading.
Explore 6 awesome GitHub repositories matching devops & infrastructure · Math Library Accelerators. Refine with filters or upvote what's useful.
Provides optimized libraries for mixed-precision matrix operations and HPC workloads on AMD GPUs.
Warp is a Python framework that JIT-compiles Python functions into CUDA kernels for GPU-accelerated parallel computation, with built-in automatic differentiation and multi-framework array interoperability. At its core, it provides a GPU kernel compilation system that enables writing and executing custom GPU kernels directly from Python, while supporting automatic gradient computation through those kernels for integration with machine learning pipelines. The framework also includes tile-based cooperative computing, where thread blocks partition into tiles for shared-memory and tensor-core opera
Switches on GPU-accelerated implementations for FFT, matrix multiply, and solver operations using cuFFTDx, cuBLASDx, and cuSolverDx.
Provides Pythonic APIs and low-level bindings to NVIDIA's CPU and GPU math libraries.
TileLang is a Python-embedded domain-specific language compiler that JIT-compiles and autotunes GPU kernels. It uses a tile-based DSL, automatic software pipelining, and parallel autotuning to generate optimized GPU kernels at runtime. It supports tensor core operations with Pythonic syntax, automatic memory management, and thread mapping. The compiler searches over tile sizes, thread counts, and scheduling policies, compiling and benchmarking candidates in parallel to find the fastest kernel. It also caches compiled binaries and tuning results to disk for reuse across sessions. TileLang inc
Provides accelerated implementations of common math functions on GPU and CPU.
ndarray est une bibliothèque de tableaux multidimensionnels pour Rust qui sert de framework d'algèbre linéaire et d'outil de calcul scientifique. Elle fournit l'infrastructure de base pour créer et manipuler des tableaux n-dimensionnels, fonctionnant à la fois comme un processeur de tableaux parallèle et une boîte à outils pour l'analyse de données numériques. La bibliothèque se distingue en fournissant un découpage (slicing) et des vues mémoire efficaces, permettant le partage de données sans copie. Elle tire parti de bibliothèques mathématiques backend optimisées pour la multiplication de matrices à haute vitesse et distribue les itérations mathématiques lourdes sur plusieurs threads CPU pour accélérer le traitement. Le projet couvre un large éventail d'opérations mathématiques, notamment l'arithmétique élément par élément, l'agrégation de données basée sur les axes et les calculs de produit scalaire. Elle inclut également des utilitaires complets pour la manipulation de tableaux tels que le remodelage, l'aplatissement, l'empilement et la génération de grilles de coordonnées, ainsi qu'une prise en charge de la génération de tableaux aléatoires et de la sérialisation.
Leverages optimized backend math libraries for high-speed floating-point matrix multiplication.
oneDNN is a library for deep learning acceleration that provides optimized building blocks for neural network training and inference. It manages tensor computation across CPU and GPU hardware, enabling the execution of high-performance primitives for model training and neural network inference optimization. The project distinguishes itself through hardware-specific kernel optimization and the use of just-in-time compilation to target specific processor instruction sets. It supports quantized neural network execution using both static and dynamic quantization to reduce memory usage and increas
Links to vendor-specific BLAS libraries to accelerate general matrix multiplication operations on CPUs.