Performance Profiling on AMD GPUs - Part 6: Advanced Thread Trace (ATT) - The Microscope for Your Application
AMD documents Advanced Thread Trace, a profiling feature that records wavefront instruction execution for single kernels. The guide explains how to use ATT to diagnose specific performance bottlenecks within GPU applications. It serves as the sixth installment in a series covering ROCm profiling tools, providing developers with deeper visibility into hardware-level behavior.
AMD published the sixth installment in a series detailing profiling techniques for their graphics hardware. This specific entry focuses on the Advanced Thread Trace capability within the ROCm software ecosystem. The documentation explains how developers can enable this tool to capture detailed execution logs. It provides instructions for analyzing instruction-level activity within a single specific compute kernel. Earlier parts of the series identified which kernels consume the most resources. This new guide moves beyond identifying bottlenecks to explaining the root causes of specific behavior. It allows engineers to observe exactly how hardware components execute instructions over time. This visibility helps teams optimize code by revealing subtle performance issues at the hardware level. The provided text does not state whether ATT is available on all supported GPU models. It offers no quantitative data regarding the overhead associated with enabling trace recording. The document avoids claiming that this single tool solves all performance problems. It remains unclear if third-party analysis tools can directly consume the generated trace files.