ROCm 7.2.1 Runtime Bundle v7.2.1

ROCm 7.2.1 Runtime Libraries Bundle

This release provides a standalone bundle of ROCm 7.2.1 runtime libraries for use with llama.cpp and other HIP-enabled applications.

What's Included

Core Runtime Libraries

• libamdhip64.so* - HIP runtime
• libhsa-runtime64.so* - HSA runtime
• libhipblas.so* - HIP BLAS library
• libhipblaslt.so* - HIP BLAS LT library
• librocblas.so* - ROCm BLAS library
• librocsparse.so* - ROCm sparse linear algebra
• Additional supporting libraries (comgr, hsakmt, drm)

GPU Kernel Files

• rocblas/ directory with GPU kernels
• hipblaslt/ directory with GPU kernels

Helper Scripts

• setup-env.sh - Environment configuration script
• README.md - Detailed usage instructions

System Requirements

• AMD GPU with ROCm support
• Linux with AMDGPU driver
• No system ROCm installation required

Compatible With

• llama.cpp binaries built for ROCm 7.2.1
• Other applications compiled against ROCm 7.2.1

Quick Start

# Download and extract
tar -xzf rocm-7.2.1-runtime-libs.tar.gz

# Configure environment
source rocm-7.2.1-runtime/setup-env.sh

# Run your HIP-enabled application
./llama-cli --version

Using with llama.cpp

1. Download llama.cpp ROCm binaries from: https://github.com/ggml-org/llama.cpp/releases
2. Download this ROCm runtime bundle
3. Extract both archives
4. Source the setup-env.sh script from this bundle
5. Run llama.cpp binaries

See the included README.md for detailed instructions and troubleshooting.

Verification

SHA256 Checksum:

$(cat rocm-7.2.1-runtime-libs.tar.gz.sha256)

Verify with:

sha256sum -c rocm-7.2.1-runtime-libs.tar.gz.sha256

License

ROCm libraries are licensed under MIT and Apache 2.0 licenses by AMD.

---

Note: This is an unofficial redistribution of ROCm runtime libraries for convenience. For official ROCm releases, visit: https://github.com/RadeonOpenCompute/ROCm

ROCm 7.2 Runtime Bundle v7.2

ROCm 7.2 Runtime Libraries Bundle

This release provides a standalone bundle of ROCm 7.2 runtime libraries for use with llama.cpp and other HIP-enabled applications.

What's Included

Core Runtime Libraries

• libamdhip64.so* - HIP runtime
• libhsa-runtime64.so* - HSA runtime
• libhipblas.so* - HIP BLAS library
• libhipblaslt.so* - HIP BLAS LT library
• librocblas.so* - ROCm BLAS library
• librocsparse.so* - ROCm sparse linear algebra
• Additional supporting libraries (comgr, hsakmt, drm)

GPU Kernel Files

• rocblas/ directory with GPU kernels
• hipblaslt/ directory with GPU kernels

Helper Scripts

• setup-env.sh - Environment configuration script
• README.md - Detailed usage instructions

System Requirements

• AMD GPU with ROCm support
• Linux with AMDGPU driver
• No system ROCm installation required

Compatible With

• llama.cpp binaries built for ROCm 7.2
• Other applications compiled against ROCm 7.2

Quick Start

# Download and extract
tar -xzf rocm-7.2-runtime-libs.tar.gz

# Configure environment
source rocm-7.2-runtime/setup-env.sh

# Run your HIP-enabled application
./llama-cli --version

Using with llama.cpp

1. Download llama.cpp ROCm binaries from: https://github.com/ggml-org/llama.cpp/releases
2. Download this ROCm runtime bundle
3. Extract both archives
4. Source the setup-env.sh script from this bundle
5. Run llama.cpp binaries

See the included README.md for detailed instructions and troubleshooting.

Verification

SHA256 Checksum:

$(cat rocm-7.2-runtime-libs.tar.gz.sha256)

Verify with:

sha256sum -c rocm-7.2-runtime-libs.tar.gz.sha256

License

ROCm libraries are licensed under MIT and Apache 2.0 licenses by AMD.

---

Note: This is an unofficial redistribution of ROCm runtime libraries for convenience. For official ROCm releases, visit: https://github.com/RadeonOpenCompute/ROCm