Examples - Examples - 57368

uProf User Guide
Document_ID
57368
Release_Date
2025-06-09
Revision
5.1 English
Note:

The following example commands are given with the assumption that the AMDuProfSys binary is located at the CWD (current working directory) -- <Installed Directory>/bin/AMDPerf. Linux-valid commands are given here. To run on Windows, the appropriate executable path should be used (.\AMDuProfSys.exe , if located in CWD). Modify these according to your relative path when using.

Table 1. Example table
Task Description Command
Display all help ./AMDuProfSys --help-all
Default metrics (core, L3 and DF) system-wide collection for 100 s. AMDuProfSys -o default -a -d 100
Monitor the entire system to collect core events defined in config files for 50 seconds and generate the profiled metrics report. ./AMDuProfSys --config core -a -d 50
Launch the workload /tmp/scimark2 with core affinity set to core 0 and monitor that core and generate profile report. ./AMDuProfSys --config core -C 0 —affinity 0 / tmp/scimark2
Monitor the entire system to collect raw data over 20 s and generate metric reports in a new folder /tmp/all for core, L3, DF, UMC and HSMP profile types. ./AMDuProfSys --config core,l3,df,umc,hsmp -o / tmp/all -a -d 20
Collect any user defined custom metrics from command line. ./AMDuProfSys --metrics core/ BrMisPredExTime="(0x4300C3)/(0x4300C2)",core/ ratio="((BrMisPredExTime * 0x430076)/0x4300C0)"-d 20
Time series profile data for core metrics (core 0-5) using Linux Perf with a logging interval of 1000 ms and set affinity of the workload (/tmp/scimark2) to core 0. ./AMDuProfSys --config core -C 0-5 -I 1000 --use-linux-perf -T -o output --affinity 0 /tmp/scimark2
Collect using custom config files and generate report in two steps. ./AMDuProfSys collect --config data/0x17_0x3/ configs/core/core_config.yaml -C 0 -o sci_perf taskset -c 0 scimark2
To generate a report file sci_perf.csv containing computed metrics.
./AMDuProfSys report -i sci_perf/sci_perf.ses -o all_events
Collect multiple metric-types using predefined config files and generate report (in 2 steps)
  1. Collect core, l3, df data for CPUs 0-10

    ./AMDuProfSys collect --config core,l3,df -C 0-10 -o sci_perf taskset -c 0 scimark2
    Note: C / -a option is applicable only for core counters.
  2. To generate a report file sci_perf.csv containing computed metrics:

    ./AMDuProfSys report -i sci_perf/sci_perf.ses -o all_events
Collect core events with multiplexing interval set to 32 ms ./AMDuProfSys --config core -a --mux-interval 32 -d 10