2. Hardware and Software Requirements

AMD uProf supports the following specifications. For a detailed list of supported processors and operating systems, refer to the AMD uProf Release Notes available at: AMD uProf product page.

2.1. Processors

AMD uProf supports the following processors:

• AMD Zen based CPU and APU Processors

• AMD Instinct™ MI200 and MI300 accelerators (for GPU kernel profiling and tracing)

2.2. Operating Systems

AMD uProf supports the 64-bit versions of the following operating systems:

• Microsoft®

  • Windows® 10 (up to 22H2)

  • Windows® 11 (up to 25H2)

  • Windows Server® 2019

  • Windows Server® 2022

  • Windows Server® 2025

• Linux

  • Ubuntu® 22.04 and later

  • RHEL® 8.6 and later

  • SLES, openSUSE® Leap 15.5*, and Debian 12

  • RHEL based distros - Rocky Linux 9.3*, Alma Linux 9.4

• FreeBSD® 13, FreeBSD® 14, FreeBSD® 15

For OS support on AMD EPYC™ processors, refer to AMD EPYC™ Processors Minimum Operating System (OS) Versions.

2.3. Compilers and Application Environment

AMD uProf supports the following application environments:

• Languages

  • Native languages: C, C++, Fortran, and Assembly

  • Non-native languages: Java, C# and Python (only with Hotspots profile on Linux)

• Programs compiled with

  • Microsoft® compilers, GNU compilers, and LLVM

  • AMD Optimizing C/C++ and Fortran Compilers (AOCC)

  • Intel Compilers (ICC)

• Parallelism

  • OpenMP

  • MPI

• Debug info formats: PDB, COFF, DWARF, and STABS

• Applications compiled with and without optimization or debug information

• Single-process, multi-process, single-thread, and multi-threaded applications

• Dynamically linked/loaded libraries

2.4. Virtualization Support

AMD uProf supports virtualized environments. The following virtualized environments are supported:

• VMware ESXi

• Linux KVM

• Citrix Xen

• Microsoft Hyper-V

2.5. Container Support

AMD uProf Power Profiler can be used for system wide profiling and for application profiling inside the docker environment. This support is only on Linux platforms. To run the Power Profiler in the docker environment, begin by installing the Power Profiler driver inside the container. By default, docker containers have their own isolated PID namespace, which means they do not see processes outside their own container. If your application needs visibility into processes outside the container (such as host processes), you can run the container in the host PID namespace using the --pid=host option when starting the container.

$ sudo docker run --pid=host <image>

AMD uProf CPU Profiler can be used for analysis of applications running inside the Docker container environments. This is supported only on Linux platforms. Choose one of the following approaches for application analysis:

• Run AMD uProf inside the Docker container to analyze the application. CAP_SYS_ADMIN permission (docker run --cap-add=CAP_SYS_ADMIN) is required to enable profiling. Both CLI and GUI based profiling and analysis are supported in this mode.

• Run AMD uProf CLI outside the Docker container to profile and analyze the target application running in the container:

  • Attach uProf CLI to the containerized process using the option --pid, during collection. Alternatively, collect the system-wide data and filter by PID during report generation.

  • During report generation, provide the path to the binary and source code (--bin-path and --src-path) of the profiled application running in the container. AMD uProf GUI doesn’t support profiling and analysis in this mode.