zbyerly

# Cluster Benchmarks
The purpose of this guide is to document how to set up and run popular benchmarking tools.

This guide was written with SLURM in mind, and `philip.hpc.lsu.edu` in particular. 

## Downloading and installing OSU Micro-Benchmarks on SLURM
* Download [OSU Micro-Benchmarks](http://mvapich.cse.ohio-state.edu/benchmarks/)
* After untarring the files (`tar -xzvf <filename>`) installation is straightforward. Follow the instructions in the `README`
* To figure out what paths to use for for "./config", load the module for the compiler you want to use and run `which mpicc` and `which mpicxx`
* On `philip`, for example, I used `CC=/usr/local/packages/mpich/3.1.4/INTEL-15.0.3/bin/mpicc CXX=/usr/local/packages/mpich/3.1.4/INTEL-15.0.3/bin/mpicxx`

zbyerly

Jan  4 15:10:47 zachtop org.gnome.Shell.desktop[2307]: [31554:31554:0104/151047.517265:ERROR:buffer_manager.cc(491)] [.BrowserC
ompositor-0x1eacc40b3300]GL ERROR :GL_INVALID_OPERATION : glBufferData: <- error from previous GL command

zbyerly

pip install label_maker
Collecting label_maker
  Using cached https://files.pythonhosted.org/packages/3c/41/fdbd8fb7dffd2f53748bd404a390962250b3e0243651316d23dca5e88b63/label-maker-0.3.1.tar.gz
Collecting Cerberus==1.1 (from label_maker)
  Using cached https://files.pythonhosted.org/packages/e0/7e/3949c86f4e60bc2b3d24ebc94af55ffaf9d62ad221f47c194edc9bd7fa94/Cerberus-1.1.tar.gz
Collecting Pillow==4.3.0 (from label_maker)
  Using cached https://files.pythonhosted.org/packages/40/45/cd1000f1c474136236c5105c882d8e1e40bd94ae939b5ca53bf724967514/Pillow-4.3.0-cp27-cp27mu-manylinux1_x86_64.whl
Collecting Shapely==1.6.3 (from label_maker)
  Using cached https://files.pythonhosted.org/packages/1b/1f/dfbfd3b77e0441040b6d91373f98bd87aa0c80b387ab6e7bc2cdbbb88d14/Shapely-1.6.3-cp27-cp27mu-manylinux1_x86_64.whl
Collecting click==6.7 (from label_maker)

zbyerly

$ pip install label_maker
Collecting label_maker
  Using cached https://files.pythonhosted.org/packages/3c/41/fdbd8fb7dffd2f53748bd404a390962250b3e0243651316d23dca5e88b63/label-maker-0.3.1.tar.gz
Collecting Cerberus==1.1 (from label_maker)
  Using cached https://files.pythonhosted.org/packages/e0/7e/3949c86f4e60bc2b3d24ebc94af55ffaf9d62ad221f47c194edc9bd7fa94/Cerberus-1.1.tar.gz
Collecting Pillow==4.3.0 (from label_maker)
  Using cached https://files.pythonhosted.org/packages/40/45/cd1000f1c474136236c5105c882d8e1e40bd94ae939b5ca53bf724967514/Pillow-4.3.0-cp27-cp27mu-manylinux1_x86_64.whl
Collecting Shapely==1.6.3 (from label_maker)
  Using cached https://files.pythonhosted.org/packages/1b/1f/dfbfd3b77e0441040b6d91373f98bd87aa0c80b387ab6e7bc2cdbbb88d14/Shapely-1.6.3-cp27-cp27mu-manylinux1_x86_64.whl
Collecting click==6.7 (from label_maker)

zbyerly

DGSWEM/HPX

This is intended to be an informal, preliminary presentation of performance comparisons between DGSWEM (pure Fortran with MPI) and DGSWEM/HPX (using C++, HPX, and LibGeoDecomp in addition to Fortran). Results and instructions for reproducing the results will be updated periodically.

Eventually, this will also document the methods used to create DGSWEM/HPX from DGSWEM.

---

Meshes

zbyerly

#!/bin/bash
#----------------------------------------------------
# Sample SLURM job script
#   for TACC Stampede2 KNL nodes
#
#   *** MPI Job on Normal Queue ***
# 
# Last revised: 27 Jun 2017
#
# Notes:

zbyerly

In file included from /home/zbyerly/work/local_install_RelWithDebInfo/include/hpx/runtime/applier/apply.hpp:16:0,
                 from /home/zbyerly/work/local_install_RelWithDebInfo/include/hpx/runtime/actions/continuation_impl.hpp:10,
                 from /home/zbyerly/work/local_install_RelWithDebInfo/include/hpx/runtime/actions/make_continuation.hpp:12,
                 from /home/zbyerly/work/local_install_RelWithDebInfo/include/hpx/include/actions.hpp:15,
                 from /home/zbyerly/work/local_install_RelWithDebInfo/include/hpx/include/lcos.hpp:11,
                 from /home/zbyerly/work/local_install_RelWithDebInfo/include/hpx/hpx.hpp:10,
                 from main.cpp:2:
/home/zbyerly/work/local_install_RelWithDebInfo/include/hpx/runtime/components/client_base.hpp: In instantiation of ‘static void hpx::components::client_base<Derived, Stub>::register_as_helper(Derived&&, const string&) [with Derived = hpx::lcos::channel<double>; Stub = hpx::lcos::server::channel<double, double>; std::__cxx1

zbyerly

#include <hpx/hpx_init.hpp>
#include <hpx/hpx.hpp>

int hpx_main(int argc, char **argv) {

  hpx::lcos::channel<double> buffer;
  std::string const channel_string = "my_channel";
  buffer.register_as(channel_string);

  return hpx::finalize();

zbyerly

/threads{locality#0/total}/time/cumulative,1,1541.327264,[s],2.71173e+12,[ns]
/threads{locality#1/total}/time/cumulative,1,1541.348146,[s],2.64973e+12,[ns]
/threads{locality#2/total}/time/cumulative,1,1541.350993,[s],2.63434e+12,[ns]
/threads{locality#3/total}/time/cumulative,1,1541.349655,[s],2.67611e+12,[ns]
/threads{locality#4/total}/time/cumulative,1,1541.351111,[s],2.77525e+12,[ns]
/threads{locality#5/total}/time/cumulative,1,1541.350193,[s],2.72339e+12,[ns]
/threads{locality#6/total}/time/cumulative,1,1541.352344,[s],2.71144e+12,[ns]
/threads{locality#7/total}/time/cumulative,1,1541.349753,[s],2.65579e+12,[ns]
/threads{locality#8/total}/time/cumulative,1,1541.350426,[s],2.79069e+12,[ns]
/threads{locality#9/total}/time/cumulative,1,1541.348605,[s],2.85424e+12,[ns]

zbyerly

%gen fort.14 
%
%help: http://adcirc.org/home/documentation/users-manual-v52/input-file-descriptions/adcirc-grid-and-boundary-information-file-fort-14/
%
dofs = @(p) (p+2)*(p+1)/2;
% script takes in reference compute weights for p = 1
ref.hght = 190;
ref.wdth = 190;
ref.comput = ref.hght*ref.wdth*dofs(1);
% and attempts to compute a rectangular inlet with an aspect ratio of 3