Some Eclipse Foundation services are deprecated, or will be soon. Please ensure you've read this important communication.

Bug 292408

Summary: [LTTng] Statistics View - Support for counters at Experiment level
Product: z_Archived Reporter: Alvaro Sanchez-Leon <alvaro.sanchez-leon>
Component: LinuxToolsAssignee: Francois Chouinard <fchouinard>
Status: CLOSED WONTFIX QA Contact: Francois Chouinard <fchouinard>
Severity: enhancement    
Priority: P3 CC: mathieu.denis, patrick.tasse
Version: unspecified   
Target Milestone: ---   
Hardware: PC   
OS: Linux   
Whiteboard:
Bug Depends on:    
Bug Blocks: 301631    
Attachments:
Description Flags
Adding of a trace set folder
none
Correct the calculation of cumulative cpu time none

Description Alvaro Sanchez-Leon CLA 2009-10-15 10:28:35 EDT
The statistics view is missing to reflect the counters at experiment level which are the cumulative values of all traces within the same experiment.
This to provide feature parity with c-library.
Comment 1 Francois Chouinard CLA 2010-06-06 11:23:22 EDT
Postponed to 0.6.1
Comment 2 Mathieu Denis CLA 2011-08-01 14:09:34 EDT
Created attachment 200665 [details]
Adding of a trace set folder
Comment 3 Mathieu Denis CLA 2011-08-01 14:14:06 EDT
Here is a patch to add a trace set folder that displays the same kind of statistics about Lttng traces. It counts the total number of events in case there is multiple traces in one experiment.

Also, the total is always displayed either at the top or at the bottom of the level column depending on how it is sorted.
Comment 4 Mathieu Denis CLA 2011-08-01 17:09:07 EDT
Created attachment 200680 [details]
Correct the calculation of cumulative cpu time

The children node of the trace were showing incorrect statistics about cumulative cpu time.

This patch corrects the calculation of this statistic so that the trace level cumulative cpu time is coherent with the sum of the cumulative cpu time of all the CPU and other levels of children node.
Comment 5 Mathieu Denis CLA 2011-08-01 17:19:03 EDT
About the cumulative CPU time and the elapsed time, we suspect that there is an overlap of the eventType timestamp so that the calculation of the elapsed time may appear falsely boosted by counting multiple times the same part of range.


-----
CPU 1 (event occuring on a single CPU)
-----

event 1  |                                  |    
   begin |----------------------------------| end


event 2      |                      |
       begin |----------------------| end


overlap      |----------------------|
Comment 6 Mathieu Denis CLA 2011-08-23 14:19:50 EDT
The bug from the comment #4 has been transferred to a new one since this is not part of the enhancement : https://bugs.eclipse.org/bugs/show_bug.cgi?id=355551

A new patch will be created based on the master branch.
Comment 7 Patrick Tasse CLA 2013-05-23 13:23:27 EDT
This bug was for feature parity with Legacy LTTng which is being removed in Linux Tools 2.0.

Statistics for an experiment are displayed individually per trace into separate tree roots.