Profiler: VTune

What is VTune?

VTune is a performance analysis tool.
It can be used to identify and analyse various aspects in both serial and parallel programs and can be used for both OpenMP and MPI applications.
It can be used with a command line interface (CLI) or a graphical user interface (GUI).

Where to find more resources on VTune?

• Main page is at VTune Profiler
• Tutorials are available at VTune Tutorials.

How to use VTune?

VTune is available on Mahuika by loading the VTune module.

module spider VTune

Versions:
VTune/2019_update4
VTune/2019_update8

First login to Mahuika and then run:

module load VTune
module load Qt5/5.12.3-GCCcore-9.2.0

To use the VTune amplifier command line tool help run:

amplxe-cl -help

How do I profile an application with VTune?

The hot-spot analysis is the most commonly used analysis and generally the first approach to optimizing an application.

• Example on Mahuika with the matrix sample.
  The matrix sample is composed of a pre-built matrix in C++ for matrix multiplication.

module load VTune/2019_update8
cp -r /opt/nesi/mahuika/VTune/2019_update8/samples_2019/en/vtune_amplifier/C++/matrix .
cd matrix
amplxe-cl -collect hotspots ./matrix

The amplxe-cl command collects hotspots data.

The option -collect specifies the collection experiment to run.
The option hotspots is to collect basic hotspots to have a general performance overview.

This is the type of output you are going to get:

amplxe-cl -collect hotspots ./matrix

amplxe: Collection started.
To stop the collection, either press CTRL-C or enter from another console window: amplxe-cl -r /scale_wlg_persistent/filesets/home/asav179/o/matrix/r000hs -command stop.
Addr of buf1 = 0x2aaab8e08010
Offs of buf1 = 0x2aaab8e08180
Addr of buf2 = 0x2aaabae09010
Offs of buf2 = 0x2aaabae091c0
Addr of buf3 = 0x2aaabce0a010
Offs of buf3 = 0x2aaabce0a100
Addr of buf4 = 0x2aaabee0b010
Offs of buf4 = 0x2aaabee0b140
Threads #: 16 Pthreads
Matrix size: 2048
Using multiply kernel: multiply1
Freq = 2.101000 GHz
Execution time = 2.515 seconds
amplxe: Collection stopped.
amplxe: Using result path `/scale_wlg_persistent/filesets/home/asav179/o/matrix/r000hs'
amplxe: Executing actions 19 % Resolving information for `libpthread.so.0'
amplxe: Warning: Cannot locate debugging information for file `/lib64/libpthread.so.0'.
amplxe: Executing actions 19 % Resolving information for `matrix'
amplxe: Warning: Cannot locate debugging information for file `/lib64/libc.so.6'.
amplxe: Executing actions 75 % Generating a report Elapsed Time: 2.552s
CPU Time: 36.440s
Effective Time: 36.440s
Idle: 0s
Poor: 36.440s
Ok: 0s
Ideal: 0s
Over: 0s
Spin Time: 0s
Overhead Time: 0s
Total Thread Count: 17
Paused Time: 0s
Top Hotspots
Function Module CPU Time
--------- ------ --------
multiply1 matrix 36.420s
init_arr matrix 0.010s
init_arr matrix 0.010s
Effective Physical Core Utilization: 85.4% (30.750 out of 36)
Effective Logical Core Utilization: 46.1% (33.194 out of 72)
| The metric value is low, which may signal a poor utilization of logical
| CPU cores while the utilization of physical cores is acceptable. Consider
| using logical cores, which in some cases can improve processor throughput
| and overall performance of multi-threaded applications.
Collection and Platform Info
Application Command Line: ./matrix
Operating System: 3.10.0-693.2.2.el7.x86_64 NAME="CentOS Linux" VERSION="7 (Core)" ID="centos" ID_LIKE="rhel fedora" VERSION_ID="7" PRETTY_NAME="CentOS Linux 7 (Core)" ANSI_COLOR="0;31" CPE_NAME="cpe:/o:centos:centos:7" HOME_URL="https://www.centos.org/" BUG_REPORT_URL="https://bugs.centos.org/" CENTOS_MANTISBT_PROJECT="CentOS-7" CENTOS_MANTISBT_PROJECT_VERSION="7" REDHAT_SUPPORT_PRODUCT="centos" REDHAT_SUPPORT_PRODUCT_VERSION="7"
Computer Name: mahuika01
Result Size: 3 MB
Collection start time: 05:35:42 15/09/2021 UTC
Collection stop time: 05:35:45 15/09/2021 UTC
Collector Type: Event-based counting driver,User-mode sampling and tracing
CPU
Name: Intel(R) Xeon(R) Processor code named Broadwell
Frequency: 2.095 GHz
Logical CPU Count: 72
amplxe: Executing actions 100 % done

The output one receives the overall elapsed and idle times as well as the CPU times of the individual functions in descending order (list of hotspots).
The utilization of the CPUs is also analyzed and judged.

Note

You can run the VTune GUI via Jupyter+VDT to visualise profiling results.

1. First launch VDT then open a terminal and cd to directory.
2. Then run:

module load VTune
amplxe-gui --path-to-open <vtune_result_dir>