ROCm/MIOpen

Health v2 · maintenance only

Flags: no_license

Adoption not part of the score

1195 stars · 273 forks observed · 2026-09-27

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

MIOpen is AMD's open-source library providing high-performance GPU kernels for machine learning primitives (convolutions, normalization, activation, etc.) on the ROCm platform, comparable to cuDNN for NVIDIA GPUs. This repository has been retired and development has moved into the ROCm/rocm-libraries super-repo.

Use cases

• run deep learning primitives on AMD GPUs
• cuDNN equivalent for ROCm
• accelerate convolution and normalization kernels on AMD hardware
• build ML frameworks on top of ROCm
• optimize GPU kernel performance for training and inference

When to choose

• you need optimized DNN primitives on AMD GPUs with ROCm
• you are building a deep learning framework targeting AMD hardware
• you need HIP-based GPU kernels for convolutions, pooling, or batch norm

When to avoid

• you use NVIDIA GPUs (use cuDNN instead)
• you need the actively developed version (use ROCm/rocm-libraries)
• you rely on the OpenCL backend, which is deprecated

Facets

library · maturity maintenance

machine-learning deep-learning gpu-computing sdk machine-learning deep-learning gpu-computing cpp python amd rocm miopen gpu-kernels deprecated hip dnn-primitives convolution blas linux gpu

Member repositories

For agents

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