Load Balance
The
Load Balance
tab displays the same data as the Flat Profile except that
it groups the data by function and not by process. The Load Balance
tab
compares the profiles of the same function across several processes. The top level
entries of the tree given in the first column are functions.In this figure showing the load balance for MPI_Allreduce, you can see that TSelf
for MPI_Allreduce is pretty unbalanced across processes:
Do This: | To Do This: |
|---|---|
Right-click on Group MPI and select Ungroup MPI from the
context menu | See the distribution of execution time over the individual MPI
routines. |
Right-click on the child of Group MPI and select Regroup
MPI from the context menu.Or go to Function Aggregation ( Advanced > Function Aggregation ,
or the  toolbar button) and select Major Function Groups | Regroup the children of MPI |
Select the Children of Group All_Processes entry from the combo box
at the top of the tab. | View the data for the children of each process. |
Press the arrows at the side of each process in the Children of Group
All_Processes view | Expand and collapse the processes of interest |
Right-click on a process and select Command line for VTune
Profiler/Advisor... | Open the Command line for Intel® VTune™ Profiler/Intel® Advisor
dialog box for selected process |
Switch between the list and pie charts by pressing the button in the top
right corner of the tab | Analyze the overall load balance pattern (for TSelf ). For example,
see the Pie Charts in the Load Balance Tab image.However, there may be a huge number of processes in a relatively confined
space. |
Use two spin buttons above the pie charts | Control the minimum radius of the pies (left button) and how many pie
charts appear in a row (right button) |
Pie Charts in the Load Balance Tab
