You may have git already. To find out whether you do, type:
type git
On Mac or Linux, this will print the location of the executable git
file. If there is no such file, you’ll get an error message saying
type: git: not found. In that case, install git:
brew install git
Use git to clone Legofit onto your machine. This step will create a
subdirectory called “legofit”, which will contain the source
code. Before executing the command below, use the cd command to move
into the directory where you keep source code that other people have
written. I keep such code in a directory called distrib, which I
originally created by typing mkdir distrib. Having created such a
directory, move into it by typing
cd distrib
Then clone legofit by typing:
git clone git@github.com:alanrogers/legofit.git legofit
This will create a directory called legofit. Use cd to move into it.
To see whether you have a C compiler already, type type cc, or type
gcc, or type clang. You should get output like
cc is /usr/bin/cc
If you need to install a C compiler, there are several alternatives. On a Mac, you can install Xcode from the App Store. Or you can use Homebrew to install clang or gcc:
brew install clang
or
brew install gcc
On Linux, the command would be
sudo apt-get install clang
or
sudo apt-get install gcc
You will also need the program “make”, so type “type make”. If you need to install, then
brew install make
On Unix-like operating systems (MacOS and Linux) it is conventional for each user to maintain a directory named “bin”, just below the home directory, which contains executable files. This directory must also be added to the system’s PATH variable, which is used for finding executable files.
First create a “bin” directory, if you don’t already have one. To do so, first type the command
cd
at the shell prompt. This moves you into your home directory. Then type
mkdir bin
This creates a new directory called “bin”.
You now need to add this to your shell’s PATH variable. I’ll assume
you’re using the bash shell. In your home directory, look for a file
called either “.bash_profile” or “.profile”. Because the name begins
with “.”, it will not show up if you type ls. However, it will
appear if you type ls .bash_profile or ls .profile. If only one of
these files exists, open it with a text editor (not a word
processor). If they both exist, edit “.bash_profile”. If neither
exists, use the editor to create a new file called “.bash_profile”.
Within this file, you may find existing definitions of the PATH variable. Add the following after the last line that changes this variable:
export PATH=$HOME/bin:$PATH
Save this change and exit the editor.
The contents of this file are executed by the shell each time you log into your computer. Since you have just created the file, your PATH has not yet been reset. Log out and then log back in again, and your PATH should be set. To check it, type
echo $PATH
This will print your PATH.
The details are in the Legofit documentation.
In the sections that follow, I will introduce you to several directories and files on the CHPC servers. Each one is described by its “path name” relative to your own “home directory”, which is represented by the symbol “~”. Before these path names will work, you’ll need to define two “soft links” within your own home directory, which point to group storage devices owned by the Rogers lab. After you define these soft links, “~/grp1/” will refer to one of these group storage devices and “~/grp2” to the other. As I explained in class, you can establish these links by typing
ln -s /uufs/chpc.utah.edu/common/home/rogersa-group1 ~/grp1
ln -s /uufs/chpc.utah.edu/common/home/rogersa-group2 ~/grp2
If everyone in the lab defines these soft links in the same way, then we can all use path names of the form “~/grp1/rogers/data/whatever”.
Legofit uses data in “.raf” format. The following data sets are already available on the server at Utah’s CHPC.
~/grp1/rogers/data/altai/orig2/altai.raf.gz~/grp1/rogers/data/vindija/orig/vindija.raf.gz~/grp1/rogers/data/denisova/orig2/denisova.raf.gz~/grp1/rogers/data/simons/raf/weur.raf.gzYou can use any or all of these in your projects. But you’ll also need to make at least one more. To do so, copy the following script into a directory of your own:
~/grp1/rogers/data/simons/raf/weur.slr
Change its name (because “weur” stands for “western European”) and edit it to reflect the population or populations that you want to study. It’s set up to run on the “notchpeak” cluster, where I have only one node.
If that node is free, you can run it there by using “cd” to move to the directory that contains the script and then typing
sbatch <your script name>
where “<your script name>” is the name of your version of the script. This script will take many hours to complete. You can check on its status at any time by typing
squeue -u <your user id>
where “<your user id>” is the id you use to log onto the CHPC cluster. As a shorthand, I often type this as
squeue -u `whoami`
which uses Linux’s “whoami” command to fill in your user id.
If the notchpeak node is busy, try kingspeak instead. I have six nodes there, so your odds are better. Before doing so, edit your slurm script. The lines that read
#SBATCH --account=rogersa-np
#SBATCH --partition=rogersa-np
work on notchpeak but not on kingspeak. For kingspeak, they should read
#SBATCH --account=rogersa-kp
#SBATCH --partition=rogersa-kp
For each project, I create a directory tree with the same format. The top-level directory contains the following files:
Eventually, the top-level directory will contain other files, with names like “all.bootci” and “all.bma”, which contain the results of analyses.
Within the top-level directory, there should be a directory for each model that is fit to the data in “data.opf” and “boot/boot*.opf”. Name these as you wish, but it’s a good idea to keep the names short. Mine tend to have names like “a”, “b”, “ab”, etc. The “a” directory contains a model in which there is only one episode of admixture, labeled “alpha”. The “ab” directory is for a model that has episodes “alpha” and “beta”.
To be continued.