Intel® VTune™ Amplifier
After performing an analysis on the command line, you can view the result in the GUI to get full benefit of Intel® VTune™ Amplifier GUI tools, or you can generate a report from the command line. A variety of report types, filtering and formatting options are available.
You cannot use the collect and report actions in the same command, so reports must be generated from a previously collected result.
The analysis type used to collect a result determines which report types are compatible and can be used to generate reports from the result.
By default, a report is written in text format to stdout and is not saved to a file. Instructions for saving, filtering and formatting reports are provided in the following sections.
Report Command Syntax
Here is the basic command syntax for generating a report:
amplxe-cl -R | -report <report_type> -result-dir <PATH> [report_options]
where:
<report_type> is the type of report that you want to create. To get the list of available report types, enter: amplxe-cl -help report. To display help for a specific report type, enter: amplxe-cl -help report <report_type>.
<PATH> is a directory where your result file is located. If you do not specify a result directory, the VTune Amplifier displays a report for the latest collected result.
[report_options] are action options used to manage the selected report. To view a list of available report action options, enter: amplxe-cl -help report.
-R is the short form of the report action, and -r is the short form of the result-dir action-option. The command syntax for generating a report could also be written as: amplxe-cl -R <report-type> -r <PATH>
Report Types
The amplxe-cl command can generate the following types of reports:
Description
Report on the overall performance of your target.
Report CPU time and Wait time (if it was collected for the analysis) for the analyzed target.
Display the total number of hardware events.
Report full stack data for each hotspot function; identify the impact of each stack on the function CPU or Wait time. You can use the group-by or filter options to sort the data by:
callstack
function
function-callstack
Report call sequences (stacks) detected during collection phase, starting from the application root (usually, the main() function). Use this report to see the impact of program units together with their callees.
Report a call tree with the time (CPU and Wait time, if available) spent in each function and its children.
Display a list of computing tasks (OpenCL™ kernels) or their instances annotated with total and average time, instance count and average values of GPU hardware metrics (if applicable). You can group or filter the report by computing-task (default) or computing-instance.
Display the time spent in different frequency modes and frequency transitions for the frequency analysis type.
Display the time spent in each sleep state and get information on processor wake-ups: time and cause for the sleep analysis type.
Display statistics that helps identify code regions for tracing on a HW simulator.
Here are a few examples of the reports you can generate with the amplxe-cl tool. More examples are available in other topics in this chapter.
The analysis type used with the collect action during analysis determines which report types can be used. This command collects Hotspots data. Several different types of reports can be generated from a Hotspots result.
$ amplxe-cl -collect hotspots -result-dir r0001hs -- ./MyApplication
The Summary report displays after a collect action by default, but you can generate a Summary report at any time, and for any analysis type. This example displays summary information including elapsed and CPU time of the target for the r001hs Hotspots analysis result created in the previous command.
$ amplxe-cl -report summary -result-dir r001hs
This example displays a Hotspots report for the same result, presenting CPU time for the functions of the target in descending order - starting from the most time-consuming function - and the cumulative percent for modules.
$ amplxe-cl -R hotspots -r r001hs
This example uses a result from a Concurrency analysis. The Summary report displays the average concurrency, elapsed and CPU time of the target for the r002cc Concurrency analysis result.
$ amplxe-cl -report summary -r r002cc