Difference between revisions of "Intel VTune"
(Created page with "The Intel VTune™ Amplifier can be used to identify and analyse various aspects in both serial and parallel programs. * Hotspot Analysis * Concurrency Analysis * Hardware Pe...") |
|||
| Line 7: | Line 7: | ||
* False Sharing | * False Sharing | ||
| − | One usually starts off with the hotspot analysis. It identifies | + | One usually starts off with the hotspot analysis. It identifies compute-intensive parts in the code and also evaluates the utilization of the available hardware. It is important not to neglect the serial part, as it can seriously weigh down the performance of the application no matter how efficiently parallelised the rest may be. Overhead and waiting time are collected for OpenMP applications. Other types of analysis as well as the option of a custom one are available for a finer tuning. |
| + | |||
| + | == Intel VTune on the RWTH Cluster == | ||
| + | |||
| + | Execute this command to print all available versions of the Intel VTune module (among other Intel modules): | ||
| + | |||
| + | module apropos intel | ||
| + | |||
| + | DO NOT use the versions intelvtune/XE2018-u03 and intelvtune/XE2019-u01 as they have bugs that influence! | ||
| + | It is recommended to load the latest update. | ||
| + | To load the GUI and run analyses from there, execute this command: | ||
| + | |||
| + | amplxe-gui | ||
| + | |||
== References == | == References == | ||
Revision as of 09:21, 9 January 2019
The Intel VTune™ Amplifier can be used to identify and analyse various aspects in both serial and parallel programs.
- Hotspot Analysis
- Concurrency Analysis
- Hardware Performance Counter Support
- IO waits
- False Sharing
One usually starts off with the hotspot analysis. It identifies compute-intensive parts in the code and also evaluates the utilization of the available hardware. It is important not to neglect the serial part, as it can seriously weigh down the performance of the application no matter how efficiently parallelised the rest may be. Overhead and waiting time are collected for OpenMP applications. Other types of analysis as well as the option of a custom one are available for a finer tuning.
Intel VTune on the RWTH Cluster
Execute this command to print all available versions of the Intel VTune module (among other Intel modules):
module apropos intel
DO NOT use the versions intelvtune/XE2018-u03 and intelvtune/XE2019-u01 as they have bugs that influence! It is recommended to load the latest update. To load the GUI and run analyses from there, execute this command:
amplxe-gui
References
Tutorials by Intel [1]
Intel VTune™ Amplifier Performance Analysis Cookbook [2]