Difference between revisions of "Applications/Matlab"

From HPC
Jump to: navigation , search
(Change Log)
(32 intermediate revisions by 2 users not shown)
Line 1: Line 1:
 
__TOC__
 
__TOC__
 
=== Application Details ===
 
=== Application Details ===
* Versions: 2016a
+
* Description: MATLAB  is a numerical computing environment and programming language developed by MathWorks, MATLAB allows matrix manipulations, plotting of functions and data, and implementation of algorithms. A  range of toolboxes are available to extend functionality.
* Module names: matlab/2016a
+
* Versions: 2016a, 2016b
* License: University of Hull Total Academic Headcount license
+
* Module names: matlab/2016a matlab/2016b
 +
* License: University of Hull Total Academic Headcount license, available to members of the university for research use
  
 
+
==== Toolboxes ====
=== Toolboxes ===
+
{| class="wikitable"  style="background-color:#f5f5dc;"
{| class="wikitable"
 
 
|  MATLAB
 
|  MATLAB
 
|  Filter Design HDL Coder
 
|  Filter Design HDL Coder
Line 72: Line 72:
 
| SimEvents
 
| SimEvents
 
| Wavelet Toolbox
 
| Wavelet Toolbox
 +
|}
 +
 +
== Usage Examples ==
 +
=== Executables ===
 +
{| class="wikitable" style="background-color:#f5f5dc;"
 +
! Executable
 +
! Description
 +
|-
 +
| matlab
 +
| By default Matlab will make use of multithreading for certain functions, with the maximum number of computational threads equal to the number of computational cores on the node. If a job or session hasn't requested appropriate resource, this multithreading can cause contention for CPU resource and negatively impact other users. For this reason the standard '''matlab''' command has been aliased so it runs with -singleCompThread option and will only use a single CPU core. See [[Applications/Matlab#Parallel_Matlab | Parallel Matlab]]
 +
|-
 +
| matlab-mt
 +
| To access standard matlab that supports multithreading, please use the '''matlab-mt''' command and request appropriate resource by either using an exclusive session with '''interactive --exclusive''' or submiting a task with the '''#SBATCH --exclusive'''. See [[Applications/Matlab#Parallel_Matlab | Parallel Matlab]]
 
|}
 
|}
  
 
=== Interactive ===
 
=== Interactive ===
 +
Interactive with full graphical interface:
 +
<pre style="background-color: #000000; color: white; border: 2px solid black; font-family: monospace, sans-serif;">
 +
[username@login01 ~]$ interactive
 +
salloc: Granted job allocation 289663
 +
Job ID 289663 connecting to c170, please wait...
 +
Last login: Thu Jan 26 12:59:11 2017 from 10.254.5.246
 +
[username@c170 ~]$ module add matlab/2016a
 +
[username@c170 ~]$ matlab
 +
</pre>
 +
 +
Interactive with command line only:
 +
<pre style="background-color: #000000; color: white; border: 2px solid black; font-family: monospace, sans-serif;">
 +
[username@c170 ~]$ matlab -nodisplay
 +
 +
                                                    < M A T L A B (R) >
 +
                                          Copyright 1984-2016 The MathWorks, Inc.
 +
                                          R2016a (9.0.0.341360) 64-bit (glnxa64)
 +
                                                    February 11, 2016
 +
 +
To get started, type one of these: helpwin, helpdesk, or demo.
 +
For product information, visit www.mathworks.com.
 +
 +
        Academic License
  
 +
>>
 +
</pre>
  
=== Job Submission Script ===
+
=== Batch Submission ===
<pre style="background-color: #C8C8C8; color: black; border: 2px solid blue; font-family: monospace, sans-serif;">
+
<pre style="background-color: #C8C8C8; color: black; font-family: monospace, sans-serif;">
 
#!/bin/bash
 
#!/bin/bash
#SBATCH -J MATLAB
+
#SBATCH -J MATLAB           # Job name, you can change it to whatever you want
#SBATCH -N 1
+
#SBATCH -N 1               # Number of nodes (for Matlab should be always one)
#SBATCH -o %N.%j.out
+
#SBATCH -o %N.%j.out       # Standard output will be written here
#SBATCH -e %N.%j.err
+
#SBATCH -e %N.%j.err       # Standard error will be written here
#SBATCH -p compute
+
#SBATCH -p compute         # Slurm partition, where you want the job to be queued
#SBATCH --exclusive
 
 
   
 
   
 +
module purge
 
module add matlab/2016a
 
module add matlab/2016a
 
   
 
   
 
matlab -nodisplay -nojvm -nodesktop -nosplash -r my_matlab_m_file
 
matlab -nodisplay -nojvm -nodesktop -nosplash -r my_matlab_m_file
 
</pre>
 
</pre>
 
+
This submission script can be found at /path/to/sample/script
 
Matlab  
 
Matlab  
  
 
+
<pre style="background-color: #000000; color: white; border: 2px solid black; font-family: monospace, sans-serif;">
=== Job Submission ===
 
<pre style="background-color: #C8C8C8; color: black; border: 2px solid black; font-family: monospace, sans-serif;">
 
 
[username@login01 ~]$ sbatch MATLABtest.job
 
[username@login01 ~]$ sbatch MATLABtest.job
 
Submitted batch job 289522
 
Submitted batch job 289522
 
</pre>
 
</pre>
  
 +
== Parallel Matlab ==
 +
By default Matlab on Viper will run on a single core as mutli-threading conflicts with the schedulers resource management, however there are three ways in which Matlab can utilise multiple processors:
 +
 +
=== Multi-threading ===
 +
Matlab can use multiple threads for certain built in functions for example linear algebra and numerical functions such as fft, \ (mldivide), eig, svd, and sort. These functions automatically execute on multiple computational threads in a single MATLAB session, allowing them to execute faster on multicore-enabled machines. Additionally, many functions in Image Processing Toolbox are multithreaded.
 +
 +
Using the standard '''matlab''' command, on Viper this is aliased to ''matlab -singleCompThread'' to limit to one thread:
 +
<pre style="background-color: #000000; color: white; border: 2px solid black; font-family: monospace, sans-serif;">
 +
[username@c170 ~]$ matlab -nodisplay -nojvm -nodesktop -nosplash -r "maxNumCompThreads , exit"
 +
 +
                                                    < M A T L A B (R) >
 +
                                          Copyright 1984-2016 The MathWorks, Inc.
 +
                                          R2016a (9.0.0.341360) 64-bit (glnxa64)
 +
                                                    February 11, 2016
 +
 +
For online documentation, see http://www.mathworks.com/support
 +
For product information, visit www.mathworks.com.
 +
 +
        Academic License
 +
 +
Warning: maxNumCompThreads will be removed in a future release. Please remove any instances of this function from your code.
 +
> In maxNumCompThreads (line 26)
 +
 +
ans =
 +
 +
    1
 +
</pre>
 +
 +
Using the '''matlab-mt''' command, on Viper this starts Matlab with multithreading support enabled:
 +
<pre style="background-color: #000000; color: white; border: 2px solid black; font-family: monospace, sans-serif;">
 +
[username@c170 ~]$ matlab-mt -nodisplay -nojvm -nodesktop -nosplash -r "maxNumCompThreads , exit"
 +
 +
                                                    < M A T L A B (R) >
 +
                                          Copyright 1984-2016 The MathWorks, Inc.
 +
                                          R2016a (9.0.0.341360) 64-bit (glnxa64)
 +
                                                    February 11, 2016
 +
 +
For online documentation, see http://www.mathworks.com/support
 +
For product information, visit www.mathworks.com.
 +
 +
        Academic License
 +
 +
Warning: maxNumCompThreads will be removed in a future release. Please remove any instances of this function from your code.
 +
> In maxNumCompThreads (line 26)
 +
 +
ans =
 +
 +
    28
 +
</pre>
 +
 +
When using Matlab with multicore enabled (using the '''matlab-mt''' command), please ensure you have exclusive access to the node, see [[General/Interactive#Exclusive_interactive_session | Exclusive interactive]] or [[General/Batch#Exclusive_batch_job|Exclusive batch jobs]].
  
 
=== Parallel Computing Toolbox ===
 
=== Parallel Computing Toolbox ===
 +
The Parallel Computing Toolbox allows you to open a number of workers (separate Matlab engines) on the local node, up to the number of cores on that node (28 for standard nodes in an exclusive session).
  
<pre style="background-color: #C8C8C8; color: black; border: 2px solid green; font-family: monospace, sans-serif;">
+
<pre style="background-color: #f5f5dc; color: black; font-family: monospace, sans-serif;">
 
>> parpool('local',28)
 
>> parpool('local',28)
 
Starting parallel pool (parpool) using the 'local' profile ... connected to 28 workers.
 
Starting parallel pool (parpool) using the 'local' profile ... connected to 28 workers.
Line 119: Line 207:
 
           SpmdEnabled: true
 
           SpmdEnabled: true
 
</pre>
 
</pre>
 +
 +
When using Matlab with the Parallel Computing Toolbox, please ensure you have exclusive access to the node, see [[General/Interactive#Exclusive_interactive_session | Exclusive interactive]] or [[General/Batch#Exclusive_batch_job|Exclusive batch jobs]].
 +
 +
=== Matlab Distributed Computing Server ===
 +
The Matlab Distributed Computing Server is not currently supported on Viper.
 +
 +
== Change Log ==
 +
[2017-06-09] Matlab 2016b installed
 +
 +
[2017-02-22] Changed default Matlab launch command in module to matlab -singleCompThread to disable multi-threading
 +
 +
== Further Information ==
 +
* [https://www.hpc.hull.ac.uk/forum/viewforum.php?f=14 UoH Matlab forum]
 +
* [https://www.mathworks.com/products/matlab.html Matlab website]
 +
* [http://www.mathworks.com/academia/student_center/tutorials.html Useful Matlab tutorials]
 +
* [https://uk.mathworks.com/discovery/matlab-multicore.html Matlab multicore]
 +
* [[Main Page#General_Support|Wiki: General Support]]
 +
* [[Main Page|Wiki: Home]]

Revision as of 12:53, 9 June 2017

Application Details

  • Description: MATLAB is a numerical computing environment and programming language developed by MathWorks, MATLAB allows matrix manipulations, plotting of functions and data, and implementation of algorithms. A range of toolboxes are available to extend functionality.
  • Versions: 2016a, 2016b
  • Module names: matlab/2016a matlab/2016b
  • License: University of Hull Total Academic Headcount license, available to members of the university for research use

Toolboxes

MATLAB Filter Design HDL Coder MATLAB Report Generator SimRF
Simulink Financial Instruments Toolbox Model Predictive Control Toolbox Simscape
Aerospace Blockset Financial Toolbox Neural Network Toolbox Simscape Electronics
Bioinformatics Toolbox Fixed-Point Designer Optimization Toolbox Simscape Multibody
Communications System Toolbox Fuzzy Logic Toolbox Parallel Computing Toolbox Simscape Power Systems
Computer Vision System Toolbox Global Optimization Toolbox Partial Differential Equation Toolbox Simulink Coder
Control System Toolbox Image Acquisition Toolbox Phased Array System Toolbox Simulink Control Design
Curve Fitting Toolbox Image Processing Toolbox RF Toolbox Simulink Design Optimization
DSP System Toolbox Instrument Control Toolbox Robotics System Toolbox Stateflow
Database Toolbox LTE System Toolbox Robust Control Toolbox Statistics and Machine Learning Toolbox
Datafeed Toolbox MATLAB Coder Signal Processing Toolbox Symbolic Math Toolbox
Econometrics Toolbox MATLAB Compiler SimBiology System Identification Toolbox
Embedded Coder MATLAB Compiler SDK SimEvents Wavelet Toolbox

Usage Examples

Executables

Executable Description
matlab By default Matlab will make use of multithreading for certain functions, with the maximum number of computational threads equal to the number of computational cores on the node. If a job or session hasn't requested appropriate resource, this multithreading can cause contention for CPU resource and negatively impact other users. For this reason the standard matlab command has been aliased so it runs with -singleCompThread option and will only use a single CPU core. See Parallel Matlab
matlab-mt To access standard matlab that supports multithreading, please use the matlab-mt command and request appropriate resource by either using an exclusive session with interactive --exclusive or submiting a task with the #SBATCH --exclusive. See Parallel Matlab

Interactive

Interactive with full graphical interface:

[username@login01 ~]$ interactive
salloc: Granted job allocation 289663
Job ID 289663 connecting to c170, please wait...
Last login: Thu Jan 26 12:59:11 2017 from 10.254.5.246
[username@c170 ~]$ module add matlab/2016a
[username@c170 ~]$ matlab

Interactive with command line only:

[username@c170 ~]$ matlab -nodisplay

                                                    < M A T L A B (R) >
                                          Copyright 1984-2016 The MathWorks, Inc.
                                           R2016a (9.0.0.341360) 64-bit (glnxa64)
                                                     February 11, 2016

To get started, type one of these: helpwin, helpdesk, or demo.
For product information, visit www.mathworks.com.

        Academic License

>>

Batch Submission

#!/bin/bash
#SBATCH -J MATLAB           # Job name, you can change it to whatever you want
#SBATCH -N 1                # Number of nodes (for Matlab should be always one)
#SBATCH -o %N.%j.out        # Standard output will be written here
#SBATCH -e %N.%j.err        # Standard error will be written here
#SBATCH -p compute          # Slurm partition, where you want the job to be queued 
 
module purge
module add matlab/2016a
 
matlab -nodisplay -nojvm -nodesktop -nosplash -r my_matlab_m_file

This submission script can be found at /path/to/sample/script Matlab

[username@login01 ~]$ sbatch MATLABtest.job
Submitted batch job 289522

Parallel Matlab

By default Matlab on Viper will run on a single core as mutli-threading conflicts with the schedulers resource management, however there are three ways in which Matlab can utilise multiple processors:

Multi-threading

Matlab can use multiple threads for certain built in functions for example linear algebra and numerical functions such as fft, \ (mldivide), eig, svd, and sort. These functions automatically execute on multiple computational threads in a single MATLAB session, allowing them to execute faster on multicore-enabled machines. Additionally, many functions in Image Processing Toolbox are multithreaded.

Using the standard matlab command, on Viper this is aliased to matlab -singleCompThread to limit to one thread:

[username@c170 ~]$ matlab -nodisplay -nojvm -nodesktop -nosplash -r "maxNumCompThreads , exit"

                                                    < M A T L A B (R) >
                                          Copyright 1984-2016 The MathWorks, Inc.
                                           R2016a (9.0.0.341360) 64-bit (glnxa64)
                                                     February 11, 2016

For online documentation, see http://www.mathworks.com/support
For product information, visit www.mathworks.com.

        Academic License

Warning: maxNumCompThreads will be removed in a future release. Please remove any instances of this function from your code.
> In maxNumCompThreads (line 26)

ans =

     1

Using the matlab-mt command, on Viper this starts Matlab with multithreading support enabled:

[username@c170 ~]$ matlab-mt -nodisplay -nojvm -nodesktop -nosplash -r "maxNumCompThreads , exit"

                                                    < M A T L A B (R) >
                                          Copyright 1984-2016 The MathWorks, Inc.
                                           R2016a (9.0.0.341360) 64-bit (glnxa64)
                                                     February 11, 2016

For online documentation, see http://www.mathworks.com/support
For product information, visit www.mathworks.com.

        Academic License

Warning: maxNumCompThreads will be removed in a future release. Please remove any instances of this function from your code.
> In maxNumCompThreads (line 26)

ans =

     28

When using Matlab with multicore enabled (using the matlab-mt command), please ensure you have exclusive access to the node, see Exclusive interactive or Exclusive batch jobs.

Parallel Computing Toolbox

The Parallel Computing Toolbox allows you to open a number of workers (separate Matlab engines) on the local node, up to the number of cores on that node (28 for standard nodes in an exclusive session).

>> parpool('local',28)
Starting parallel pool (parpool) using the 'local' profile ... connected to 28 workers.

ans = 

 Pool with properties: 

            Connected: true
           NumWorkers: 28
              Cluster: local
        AttachedFiles: {}
          IdleTimeout: 30 minute(s) (30 minutes remaining)
          SpmdEnabled: true

When using Matlab with the Parallel Computing Toolbox, please ensure you have exclusive access to the node, see Exclusive interactive or Exclusive batch jobs.

Matlab Distributed Computing Server

The Matlab Distributed Computing Server is not currently supported on Viper.

Change Log

[2017-06-09] Matlab 2016b installed

[2017-02-22] Changed default Matlab launch command in module to matlab -singleCompThread to disable multi-threading

Further Information