Skip to content

Latest commit

 

History

History
105 lines (82 loc) · 4.02 KB

README.md

File metadata and controls

105 lines (82 loc) · 4.02 KB
Raw

MSIAC: A Julia package for SIAC filtering

Version: 1.0
For bugs / issues, email: [email protected]

Installing Julia (skip if you already have it)

  1. Download & install Julia
  2. Set up an alias to the Julia folder:
    • Linux / bash:
    vi ~/.bash_aliases 
alias julia="/etc/julia-1.7.2/bin/julia"  (change /etc/julia-1.7.2/  to your path to Julia folder)
    • iOS / zsh:
    vi ~/.zshrc 
alias julia="/etc/julia-1.7.2/bin/julia"  (change /etc/julia-1.7.2/  to your path to Julia folder)
  3. Check that it works. Source the file (or open a new terminal) and type $julia on terminal

Installing the MSIAC package

  1. Set up an environment variable This is optional. Useful to call MSIAC from anywhere in your machine but you can skip this and go to step 2.
    • Linux / bash
    vi ~/.bashrc
export MSIAC=$your_root_to_MSIAC_folder
    • iOS / zsh
    vi ~/.zshrc
export MSIAC=$your_root_to_this_folder
  2. Load & test the MSIAC Package 
    (1) open julia session inside your MSIAC folder
   $cd MSIAC
   $julia

   NOTE: if you are in an editor, make sure the REPL is indeed in MSIAC:
   julia> cd(ENV["MSIAC"])  #(ENV[""] looks for environment variables)

(2) Enter Pkg environment
   julia> ]
      (you should now see something like @(v1.7) pkg>)

(3) Activate MSIAC package
   (@v1.6) pkg> activate .   # The dot (.) tells Julia to look in local folder
      (you should see that (@v1.7) pkg> became (MSIAC) pkg> )

(4)	Load package
      (MSIAC) pkg> instantiate
      ** If you get errors, try running
                  pkg> resolve ; pkg> update

(5) Test the package (Optional)
   (MSIAC) pkg>  test
      (If all went well, the screen should show 4 (6) test PASSED )
***NOTE****: WriteVTK and VTKDataIO packages show some warnings during pre-compilation. However, they still work and allow reading .vtu & .vtk files correctly.
  1. Known issue For some configurations with iOS, building the package using the Project.toml may not work. If the installation failed, checkout README_pkg.md file

Running the MSIAC tool on Distributed Memory

Julia is more efficient using DistributedArrays than the default parallelization used by MSIAC: Threads.@threads

1. The file MSIAC_HPC.jl assumes ther is an environment variable with the MSIAC path
	@everywhere push!(LOAD_PATH,ENV["MSIAC"])

***NOTE****:  the everywhere command tells julia to include MSIAC module in all workers. Otherwise, it will only be loaded to the main worker ( # 1 )

   If you haven't set it, manually change ENV["MSIAC"] to your path to MSIAC:
	@everywhere push!(LOAD_PATH,absolute_path_to_MSIAC)


2. Start a session with several workers (CPUs) 
	$ julia --p number_of_CPUS

3. Include the MSIAC_HPC.jl file at the top of your script 
	using DelimitedFiles, Printf, Distributed
	@everywhere include(ENV["MSIAC"]*"/src/MSIAC_HPC.jl") 	

4. Run filter_data_hpc(args...) 
5. This [file](https://github.com/XuliaDS/MSIAC/tree/main/tutorials/msiac_hpc.jl) is an example of a filter + distributed

A quick guide on using the REPL

  1. Running a script

       open Julia session
   $julia
   julia> include("myfile.jl")  (you should be at the folder where myfile lives)
      Leaving a session open means that successive runs on the script are faster as it does not recompile:

   julia> include("myfile.jl")   #1st time: slower
   julia> include("myfile.jl")   #2nd time: faster

  2. Checkout the tutorials folder to see examples of how to run the MSIAC tool