kokkos/kokkos

Health v2 · maintenance only

Flags: no_license

Adoption not part of the score

2649 stars · 515 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Kokkos Core is a C++ programming model providing abstractions for parallel execution and memory management to write performance-portable applications across HPC platforms. It supports CUDA, HIP, SYCL, HPX, OpenMP, and C++ threads as backends and is a Linux Foundation / HPSF project.

Use cases

• write c++ code that runs on both gpus and cpus without rewriting
• port a cuda application to amd gpus
• parallelize scientific simulation code for supercomputers
• manage multi-level memory hierarchies in hpc applications
• abstract away vendor-specific gpu programming models
• prepare code for exascale computing platforms

When to choose

• you need a single c++ codebase targeting nvidia, amd, and intel gpus plus multicore cpus
• you are developing large-scale scientific or engineering simulation software
• you want a mature, production-level, community-supported performance portability layer

When to avoid

• you target only a single gpu vendor and can use native toolkits like CUDA directly
• your project is small-scale and does not need hardware portability
• you need a language other than C++

Facets

library · maturity stable

concurrency gpu-computing math simulation sdk gpu-computing performance cross-platform programming-languages cpp windows cross-platform performance-portability hpc parallel-computing memory-abstraction kokkos cuda hip sycl openmp exascale high-performance-computing algorithms gpu linux macos

Member repositories

For agents

markdown · JSON · MCP: product_card(name="kokkos/kokkos")