Building and Deploying Custom hipBLASLt Libraries on AMD Instinct GPUs

2026-06-18 · Source: AMD ROCm Blogs · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Cloud Computing & IT Infrastructure · Depth: Advanced, long

Summary

A comprehensive workflow for building and deploying custom "hipBLASLt" libraries on AMD Instinct GPUs is outlined, crucial for optimizing General Matrix Multiply (GEMM) operations in generative AI workloads. It addresses scenarios where standard ROCm binaries in "/opt/rocm" are insufficient, such as testing bug fixes, validating TensileLite-tuned kernels, or evaluating new architectures like the AMD Instinct MI300X ("gfx942"). The process involves preparing Ubuntu/Debian or RHEL/RPM build environments, fetching the "rocm-libraries" source at a specific commit ("1784d40186"), and compiling "hipBLASLt" for target architectures using "install.sh -a". Deployment strategies include generating ".deb" or ".rpm" packages for cluster-wide distribution or utilizing "LD_LIBRARY_PATH" for isolated, root-privilege-free testing, ensuring operational flexibility and avoiding dependency conflicts.

Key takeaway

For MLOps Engineers deploying generative AI workloads on AMD Instinct GPUs, especially those requiring specific "hipBLASLt" versions or custom TensileLite kernels, you should adopt this workflow. Building architecture-specific packages or using "LD_LIBRARY_PATH" for runtime selection provides granular control over GEMM performance and avoids system-wide dependency conflicts. This ensures your deployments are reproducible, traceable, and optimized for target hardware like the MI300X ("gfx942"), significantly improving end-to-end latency and throughput.

Key insights

Custom "hipBLASLt" builds offer precise GPU optimization and flexible deployment for AMD Instinct GPUs, avoiding system-wide conflicts.

Principles

• Avoid "sudo make install" in multi-user environments.
• Target specific GPU architectures to optimize builds.
• Document source revision and build logs for traceability.

Method

Prepare OS environment, clone "rocm-libraries" at commit "1784d40186", build "hipBLASLt" for specific architectures like "gfx942" using "install.sh -a", then package as ".deb"/".rpm" or use "LD_LIBRARY_PATH" for runtime selection.

In practice

• Use "LD_LIBRARY_PATH" for local testing.
• Generate ".deb"/".rpm" for cluster deployment.
• Verify loaded library with "ldd" and "hipblaslt-bench".

Topics

• hipBLASLt
• AMD Instinct GPUs
• GPU Optimization
• Generative AI
• ROCm
• Software Deployment

Code references

• ROCm/rocm-libraries
• ROCm/rocm-libraries

Best for: Machine Learning Engineer, AI Engineer, MLOps Engineer

Related on AIssential

• Customizing Kernels with hipBLASLt TensileLite GEMM Tuning - Advanced User Guide — TensileLite Tuning generates custom GEMM kernels for AMD GPUs, significantly outperforming existing solutions.
• Accelerating ComfyUI Workflows on AMD Instinct™ MI355X GPUs with ROCm — AMD Instinct MI355X GPUs significantly accelerate ComfyUI generative AI workflows over NVIDIA B200.
• Deal between AMD and Meta could bring greater innovation to GenAI market — Strategic "chips-for-stock" partnerships are diversifying AI compute supply and challenging single-vendor dominance.
• Taking on CUDA With ROCm: ‘One Step After Another’ — AMD's ROCm software stack is rapidly maturing, leveraging open-source tools and community engagement to challenge Nvidia's CUDA dominance.
• Meta 🤝 AMD Today we’re announcing a multi-year agreement with @AMD to integrate their latest Instinct GPUs into our global infrastructure. With approximately 6GW of planned data center capacity dedicated to this deployment, we’re scaling our compute ca - x.com — Meta is massively scaling AI compute with AMD Instinct GPUs for advanced model development.
• GPU Cloud Deployment Without Leaving Your IDE — Audry Hsu, RunPod — RunPod's Flash SDK enables direct GPU cloud deployment and rapid iteration of AI inference functions from a local IDE.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AMD ROCm Blogs.