===== SPACK - a package manager for HPC systems =====
[[https://spack.readthedocs.io/en/latest/basic_usage.html|Official documentation of SPACK]]
==== New SPACK without environments ====
Having worked with spack environments for some time, we have encountered several severe issues which have convinced us that we need to find a more practical way of maintaining software packages at VSC.
There are now three separate spack installation trees corresponding to the CPU/GPU architectures on VSC:
* skylake - Intel CPUs; works on Intel Skylake and Cascadelake CPUs
* zen - AMD CPUs; works on Zen 2 and 3 CPUs
* cuda-zen - AMD CPUs + NVIDIA GPUs; works on all nodes equipped with graphics cards
By default the spack installation tree suitable for the current compute/login node is activated and will be indicated by a **prefix** on the command line, e.g.:
zen [user@l51 ~]$
Read more at:
[[doku:spack-transition | Transition to new SPACK without Environments]]
==== List available packages ====
Most software on VSC4/5 is installed via spack, to list available software packages use:
spack find
List all available packages compiled with a specific compiler:
spack find %gcc@12.2.0
List specific installed software packages use:
spack find openmpi
A specific version can be requested as well:
spack find openmpi@4.1.4
List a specific version compiled with a specific compiler:
spack find openmpi@4.1.4%gcc@12.2.0
The detail level of information can be specified with the following options:
spack find -l openmpi@4.1.4%gcc@12.2.0
spack find -lv openmpi@4.1.4%gcc@12.2.0
spack find -lvd openmpi@4.1.4%gcc@12.2.0
The latter lists all dependencies of the package.
**Dependencies**: dependencies of software packages can be specified with a ''^''.
e.g.:
spack find -lvd py-scipy %intel@19.0.5.281 ^python@3.7.4 ^intel-mkl
This will list all available versions of ''py-scipy'' which have been built with ''python@3.7.4'' and any version of ''intel-mkl''.
List all currently loaded packages:
spack find --loaded
==== Load modules ====
The ''spack'' command can be used to load software modules. The general form is:
spack load
''package'' has to be a uniquely defined software package where all dependencies have been resolved, e.g.:
spack load openmpi@4.1.4%gcc@12.2.0
This will load the corresponding module, even if ''module list'' will not show anything!
zen xyz@l51:~$ spack load openmpi@4.1.4%gcc@12.2.0 arch=linux-almalinux8-zen3 /2vqdnay
zen xyz@l51:~$ module list
No Modulefiles Currently Loaded.
zen xyz@l51:~$ spack find --loaded
-- linux-almalinux8-zen3 / gcc@12.2.0 ---------------------------
autoconf@2.69 curl@7.85.0 hwloc@2.8.0 libgpg-error@1.46 m4@1.4.19 openssh@9.1p1 py-setuptools@65.5.0 util-linux-uuid@2.38.1
autoconf-archive@2022.02.11 diffutils@3.8 json-c@0.16 libiconv@1.16 meson@0.63.3 openssl@1.1.1s py-wheel@0.37.1 util-macros@1.19.3
automake@1.16.5 expat@2.4.8 krb5@1.19.3 libmd@1.0.4 mpfr@4.1.0 pcre2@10.39 python@3.9.15 xz@5.2.7
berkeley-db@18.1.40 gawk@5.1.1 libbsd@0.11.5 libpciaccess@0.16 munge@0.5.15 perl@5.36.0 readline@8.1.2 zlib@1.2.13
bison@3.8.2 gdbm@1.23 libedit@3.1-20210216 libsigsegv@2.13 ncurses@6.3 pigz@2.7 slurm@22.05.2.1 zstd@1.5.2
bzip2@1.0.8 gettext@0.21.1 libevent@2.1.12 libtool@2.4.7 ninja@1.11.1 pkgconf@1.8.0 sqlite@3.39.4
ca-certificates-mozilla@2022-10-11 glib@2.74.1 libffi@3.4.2 libxml2@2.10.1 numactl@2.0.14 pmix@4.1.2 tar@1.34
cmake@3.24.3 gmp@6.2.1 libgcrypt@1.10.1 lz4@1.9.4 openmpi@4.1.4 py-pip@22.2.2 texinfo@6.5
==> 61 loaded packages
and is equivalent to loading the module as:
module load openmpi/4.1.4-gcc-12.2.0-2vqdnay
Alternatively a package can be loaded by using the unique hash of a package.
The option ''-l'' will display the short hash of a package:
zen xyz@l51:~$ spack find -l openmpi@4.1.4%gcc@12.2.0
-- linux-almalinux8-zen3 / gcc@12.2.0 ---------------------------
2vqdnay openmpi@4.1.4
==> 1 installed package
The hash ''2vqdnay'' uniquely identifies the package and can also be used to load the software module:
spack load /2vqdnay
Unload all currently loaded packages:
spack unload --all