On the NAU Monsoon super-computer, I have been using the user-space conda data science package manager for a few years. Today I heard there is another user-space package manager, spack. What is the difference?

  • spack builds from source downloaded from github (slow), whereas conda downloads pre-built binaries (fast).
  • conda’s focus is on packages for data science (including on windows), whereas spack’s focus is on HPC (no windows support).

I tried Basic installation on cygwin and I got:

th798@cmp2986 ~/spack
$ . share/spack/setup-env.sh
-bash: $'\r': command not found
-bash: $'\r': command not found
-bash: $'\r': command not found
-bash: share/spack/setup-env.sh: line 49: syntax error near unexpected token `$'{\r''
'bash: share/spack/setup-env.sh: line 49: `_spack_shell_wrapper() {

Related: Using Spack to Replace Homebrew/Conda.

TODO: check to see if spack supports these packages that I had to build from source on Monsoon.