Changes between Version 12 and Version 13 of cypress/Matlab


Ignore:
Timestamp:
Aug 24, 2016 12:19:50 PM (5 years ago)
Author:
fuji
Comment:

Legend:

Unmodified
Added
Removed
Modified
  • cypress/Matlab

    v12 v13  
    181181
    182182See [https://wiki.hpc.tulane.edu/trac/wiki/cypress/Matlab#CompiledMatlab]
     183{{{#!bash
     184#!/bin/bash
     185#SBATCH --qos=normal            # Quality of Service
     186#SBATCH --job-name=matlabMT     # Job Name
     187#SBATCH --time=1:00:00          # WallTime
     188#SBATCH --nodes=1               # Number of Nodes
     189#SBATCH --ntasks-per-node=1     # Number of tasks (MPI processes)
     190#SBATCH --cpus-per-task=10      # Number of threads per task (OMP threads)
     191
     192module load matlab
     193mcc -m FuncTest.m
     194./FuncTest
     195}}}
    183196
    184197==== Explicit parallelism ====
     
    201214
    202215module load matlab
    203 matlab  -nodesktop -nodisplay -nosplash -r "CreateWorker; ParforTest; exit;"
     216matlab  -nodesktop -nodisplay -nosplash -r "ParforTest; exit;"
    204217}}}
    205218
     
    222235% parfor "ParforTest.m"
    223236%
     237CreateWorker;
    224238iter = 10000;
    225239sz = 50;
     
    245259See [https://wiki.hpc.tulane.edu/trac/wiki/cypress/Matlab#CompiledMatlab]
    246260
     261{{{#!bash
     262#!/bin/bash
     263#SBATCH --qos=normal            # Quality of Service
     264#SBATCH --job-name=matlabPool     # Job Name
     265#SBATCH --time=1:00:00          # WallTime
     266#SBATCH --nodes=1               # Number of Nodes
     267#SBATCH --ntasks-per-node=1     # Number of tasks (MPI processes)
     268#SBATCH --cpus-per-task=4       # Number of threads per task (OMP threads)
     269
     270module load matlab
     271mcc -m ParforTest.m
     272./ParforTest
     273}}}
     274
    247275=== Running MATLAB with Automatic Offload ===
    248276Internally MATLAB uses Intel MKL Basic Linear Algebra Subroutines (BLAS) and Linear Algebra package (LAPACK) routines to perform the underlying computations when running on Intel processors.
     
    294322See [https://wiki.hpc.tulane.edu/trac/wiki/cypress/Matlab#CompiledMatlab]
    295323
     324{{{#!bash
     325#!/bin/bash
     326#SBATCH --qos=normal            # Quality of Service
     327#SBATCH --job-name=matlabAO     # Job Name
     328#SBATCH --time=1:00:00          # WallTime
     329#SBATCH --nodes=1               # Number of Nodes
     330#SBATCH --ntasks-per-node=1     # Number of tasks (MPI processes)
     331#SBATCH --cpus-per-task=1       # Number of threads per task (OMP threads)
     332#SBATCH --gres=mic:1            # Number of Co-Processors
     333
     334module load matlab
     335module load intel-psxe
     336mcc -m MatTest.m
     337export MKL_MIC_ENABLE=1
     338./MatTest
     339}}}
     340
     341To generate the offload report at run time,
     342{{{
     343export OFFLOAD_REPORT=2
     344}}}
     345
     346* Setting OFFLOAD_REPORT to 0 (or not setting it) results in no offload report.
     347
     348* Setting OFFLOAD_REPORT to 1 results in a report including:
     349 * Name of function called
     350 * Effective Work Division
     351 * Time spent on Host during call
     352 * Time spent on each available Phi coprocessor during call
     353
     354* Setting OFFLOAD_REPORT to 2 results in a report including everything from 1, and in addition:
     355 * Amount of data transferred to and from each Phi during call
     356