You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: README.md
+1-1Lines changed: 1 addition & 1 deletion
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -331,7 +331,7 @@ To illustrate how the TBS mode works let us consider the following example comma
331
331
...
332
332
333
333
334
-
This command provides the user with the number of instructions retired, last-level cache (LLC) misses and core energy consumption (in mJ) every second. The beginning of the command output shows the event-to-counter mapping for the various hardware events and virtual counters. The "Event counts" section in the output displays a table with the raw counts for the various events; each sample (one per second) is represented by a different row. Note that the sampling period is specified in seconds via the -T option; fractions of a second can be also specified (e.g, 0.3 for 300ms). If the user includes the -A switch in the command line, `pmctrack` will display the aggregate event count for the application's entire execution instead. At the end of the line, we specify the command to run the associated application we wish to monitor (e.g: ./mcf06).
334
+
This command provides the user with the number of instructions retired, last-level cache (LLC) misses and core energy consumption (in uJ) every second. The beginning of the command output shows the event-to-counter mapping for the various hardware events and virtual counters. The "Event counts" section in the output displays a table with the raw counts for the various events; each sample (one per second) is represented by a different row. Note that the sampling period is specified in seconds via the -T option; fractions of a second can be also specified (e.g, 0.3 for 300ms). If the user includes the -A switch in the command line, `pmctrack` will display the aggregate event count for the application's entire execution instead. At the end of the line, we specify the command to run the associated application we wish to monitor (e.g: ./mcf06).
335
335
336
336
In case a specific processor model does not integrate enough PMCs to monitor a given set of events at once, the user can turn to PMCTrack's event-multiplexing feature. This boils down to specifying several event sets by including multiple instances of the -c switch in the command line. In this case, the various events sets will be collected in a round-robin fashion and a new `expid` field in the output will indicate the event set a particular sample belongs to. In a similar vein, time-based sampling also supports multithreaded applications. In this case, samples from each thread in the application will be identified by a different value in the pid column.
0 commit comments