Setting Up ROCm on Rocky Linux 10

AMD ROCm (Radeon Open Compute) is an open-source platform for GPU computing. This guide walks through installing ROCm on Rocky Linux 10.

Prerequisites

Rocky Linux 10 installed
AMD GPU (check compatibility at ROCm compatibility list)
Root or sudo access
Stable internet connection

System Requirements

# Check your GPU
lspci | grep -i amd

# Verify kernel version (5.x or higher recommended)
uname -r

Installation Steps

1. Add ROCm Repository

# Add the ROCm repository
sudo tee /etc/yum.repos.d/rocm.repo <<EOF
[ROCm]
name=ROCm
baseurl=https://repo.radeon.com/rocm/rhel9/rpm
enabled=1
gpgcheck=1
gpgkey=https://repo.radeon.com/rocm/rocm.gpg.key
EOF

2. Install ROCm

# Update package cache
sudo dnf clean all
sudo dnf update

# Install ROCm
sudo dnf install rocm-hip-sdk rocm-opencl-sdk

3. Configure User Permissions

# Add user to video and render groups
sudo usermod -a -G video,render $USER

# Log out and back in for changes to take effect

4. Verify Installation

# Check ROCm version
rocminfo

# Test with rocm-smi
rocm-smi

# Verify HIP installation
hipconfig

Post-Installation

Set Environment Variables

Add to ~/.bashrc:

export PATH=$PATH:/opt/rocm/bin
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/opt/rocm/lib

Test GPU Compute

# Clone ROCm examples
git clone https://github.com/ROCm/HIP-Examples.git
cd HIP-Examples/vectorAdd

# Compile and run
make
./vectorAdd

Troubleshooting

GPU Not Detected

# Check if amdgpu driver is loaded
lsmod | grep amdgpu

# If not, load it manually
sudo modprobe amdgpu

Permission Denied Errors

Ensure you’re in the correct groups:

groups $USER
# Should show: video render

Next Steps

Install PyTorch with ROCm support
Set up TensorFlow for AMD GPUs
Explore ROCm libraries (rocBLAS, rocFFT, etc.)

Resources

Last updated: Feb 20, 2025