CH-Abhinav/JNum: High-performance pure Java 25 & Kotlin numerical computing and tensor library

JNum is a high-performance, pure Java 25 and Kotlin numerical computing and tensor library powered by Project Panama's Foreign Function and Memory API and Vector API. It delivers native BLAS performance without native C++ build toolchains or JNI boundaries, offering hardware SIMD vector lanes and a 4-tier GEMM engine. Benchmarks show it significantly outperforms legacy Java tensor libraries and competes closely with C-based NumPy. Open article →