AMD also offers readily available libraries to use with ROCm for application-specific purposes. Some of these libraries are listed here:
- MIOpen: Similar to cuDNN on NVIDIA CUDA, AMD ROCm has the MIOpen library for deep neural networks can be used for TensorFlow and PyTorch deep learning frameworks.
- rocFFT: A software library for computing FFT written in HIP. The library can be used with both NVIDIA and AMD GPU devices.
- rocBLAS: Implementation of basic linear algebra subroutines for compute intensive operations implemented in the HIP programming language and optimized for AMD's GPUs.
- Open Compute Math Library (OCML): Contains 250+ optimized math functions for single, double, and half precision.
- rocSPARSE: Accelerated ...