CNugteren/CLBlast

Health v2 · maintenance only

Adoption not part of the score

1186 stars · 212 forks observed · 2026-08-28

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

CLBlast is a lightweight, tunable OpenCL BLAS library written in C++11 that implements basic linear algebra subprograms for vectors and matrices. It targets a wide variety of OpenCL devices from different vendors, including desktop, laptop, and embedded GPUs, and includes tuners for optimizing performance on specific hardware.

Use cases

• run BLAS matrix operations on non-NVIDIA GPUs via OpenCL
• accelerate matrix multiplication (GEMM) on embedded or Intel GPUs
• tune BLAS kernels for a specific OpenCL device or matrix shape
• use half-precision fp16 linear algebra routines
• replace clBLAS with a modern, customizable C++ BLAS library

When to choose

• you need BLAS on OpenCL devices from any vendor, not just CUDA GPUs
• you want to inspect or customize the BLAS kernels yourself
• you run on exotic or older OpenCL 1.1 hardware and can run the tuners
• you prefer a C++ API (with a C API also available) over clBLAS's C API

When to avoid

• you run exclusively on NVIDIA CUDA-enabled GPUs where cuBLAS's assembly-tuned kernels are faster
• you need out-of-the-box peak performance without running the tuning step on untuned devices

Facets

library · maturity stable

math gpu-computing benchmarking machine-learning gpu-computing performance cpp cross-platform windows blas opencl gemm matrix-multiplication linear-algebra tunable-kernels fp16 algorithms linux macos gpu

Member repositories

For agents

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