Running Multiple Analyses Automatically
Intel® Advisor
provides the
Batch Mode
feature that enables you to select several Advisor
collectors to run automatically in a queue. You can specify several options to
reduce the amount of collected data. Using this mode might be useful if you
want to fully analyze with
Intel® Advisor
an application that takes a lot of time to execute.
To enable the
Batch Mode
, you must be in the
Intel® Advisor
Vectorization Workflow. Use the
ON/OFF
selector to enable/disable the
Batch Mode
.
Use This
| To Do This
|
|---|---|
| To switch to the batch mode and back.
|
Check boxes
| To select the collectors that you want to run in the batch
mode.
|
Manual Selection | Select loops for Memory Access Patterns and Dependencies
analysis. You can select loops in the Survey report view after running the
Survey analysis.
|
Automatic Selection | Let the
Intel® Advisor
automatically select loops for deeper analysis according to the specified
(check-marked) criteria.
|
Scalar Serial Loops Only (for
Dependencies analysis only)
| Intel® Advisor
will analyze scalar loops that are not included into any parallel region. The
loops that do not meet this criteria will be excluded from the analysis.
|
Innermost Loops Only (for Dependencies
analysis only)
| Intel® Advisor
will analyze innermost loops. All other loops will be excluded from analysis.
|
Diagnosed as "Vector Dependence Prevent
Vectorization" Only (for Dependencies analysis only)
| Intel® Advisor
will analyze scalar loops that the compiler cannot vectorize automatically and
reports the "Vector Dependence Prevent Vectorization" diagnostic. All other
loops that do not meet this criteria will be excluded from analysis.
|
With "Possible Inefficient Memory Access Pattern"
issue only (for Memory Access Patterns analysis only)
| Intel® Advisor
will analyze the loops with performance issue "Possible Inefficient Memory
Access Pattern" detected. The loops that do not meet this criteria will be
excluded from the analysis.
|
Loop hierarchy choice (for Memory Access Patterns analysis
only):
| Select the hierarchy level of nested loops for analysis. The
loops that do not meet this criteria will be excluded from the analysis.
"Innermost Loops" stands for loops that don't have nested loops. "Second-level
Loops" stands for the loops that include innermost loop(s) only. "Third-level
Loops" stands for the loops that include second-level loops only.
|
Exclude loops without source location
(for Memory Access Patterns analysis only)
| Intel® Advisor
will exclude loops with empty source location from the analysis.
|
Above 2% of Total CPU Time Only (for
Memory Access Patterns analysis only)
| Intel® Advisor
will analyze only the loops with total time more than 2% of the Total CPU Time.
The loops that do not meet this criteria will be excluded from the analysis.
|