Introduction to profiling tools for AMD hardware

2026-04-10 · Source: AMD ROCm Blogs · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning · Depth: Intermediate, long

Summary

AMD provides a suite of profiling tools designed to help developers optimize applications for AMD hardware, including "Zen" Core CPUs, RDNA™ GPUs, and CDNA™ accelerators. The article introduces key tools like rocprofiler-sdk, rocprofv3, rocprof-sys, rocprof-compute, Radeon™ GPU Profiler, and AMD uProf, detailing their specific capabilities and supported architectures/operating systems. It emphasizes that efficient application performance requires more than just benchmarking execution time; it necessitates understanding where a program spends its time and identifying bottlenecks. The post outlines a decision-making framework based on profiling objectives: identifying hot spots, assessing hardware utilization (e.g., through Roofline Analysis), and understanding the root causes of observed performance via hardware metrics. Several superseded tools are also noted, with their modern replacements highlighted.

Key takeaway

For AI Engineers and HPC developers optimizing applications on AMD hardware, you should select your profiling tool based on your specific objective and target architecture. Start by identifying performance hot spots with timeline tracing tools like rocprof-sys or rocprofv3. Then, use tools such as rocprof-compute for detailed kernel analysis or AMD uProf for broader system insights to diagnose the root causes of performance bottlenecks, ensuring efficient hardware utilization.

Key insights

AMD offers diverse profiling tools to optimize application performance across its CPU and GPU architectures.

Principles

• Benchmarking requires deep performance tuning.
• Profiling tools vary by architecture and OS.
• Identify hot spots before optimizing.

Method

To profile, first identify hot spots with timeline traces (rocprof-sys, rocprofv3). Then, assess hardware utilization via Roofline Analysis (rocprof-compute, AMD uProf). Finally, collect hardware metrics to understand performance causes (rocprofv3, rocprof-sys, rocprof-compute, AMD uProf).

In practice

• Use rocprofv3 for GPU hot spot analysis.
• Employ rocprof-sys for unified CPU/GPU traces.
• Apply rocprof-compute for Instinct™ GPU kernel analysis.

Topics

• ROCm Profiling
• AMD CDNA Architecture
• GPU Performance Analysis
• Roofline Analysis
• rocprof-sys

Code references

• ROCm/rocm-systems

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

Related on AIssential

• Performance Profiling on AMD GPUs - Part 4: Fortran OpenMP Offload Edition — Profiling Fortran OpenMP offload on AMD GPUs requires a systematic workflow to identify memory-bound kernels and optimize HSA_XNACK and workgroup sizes.
• AMD Device Metrics Exporter v1.4.2: Enhanced Observability, Deeper RAS Insights, and Smarter GPU Telemetry for Modern HPC & AI Clusters — DME v1.4.2 enhances GPU observability with process-level context, structured error reporting, and detailed performance constraint metrics.
• AI giveth and AI taketh CPU — AMD's AI strategy leverages heterogeneous computing and modular chiplet designs for performance, efficiency, and adaptability across diverse workloads.
• Profiling in PyTorch (Part 2): From nn.Linear to a Fused MLP — Effective PyTorch profiling reveals kernel fusion, CPU overheads, and memory traffic bottlenecks for performance optimization.
• 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.
• Your AI Setup Needs This — AMD Ryzen AI Processors integrate NPU, GPU, and CPU for enhanced on-device AI performance in business laptops.

Open in AIssential →

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