atrip/README.org

133 lines
3.2 KiB
Org Mode

#+title: ATRIP: An MPI-asynchronous implementation of CCSD(T)
#+PROPERTY: header-args+ :noweb yes :comments noweb :mkdirp t
#+begin_export rst
.. toctree::
:maxdepth: 2
atrip
#+end_export
* About this document
You can find this project written as a literate program
- [[https://alejandrogallo.github.io/atrip/atrip.html][Literate program]]
or if you prefer reading the source code with Doxygen there is also a built
of doxygen available:
- [[https://alejandrogallo.github.io/atrip/doxygen/html/index.html][Doxygen]]
* How to cite
If you need to cite this algorithm
before the proper paper is released please contact me.
In the mean time the code has been used in
[[https://aip.scitation.org/doi/10.1063/5.0074936][this publication]] and can therefore been cited.
* Building
Atrip uses autotools to build the system.
Autotools works by first creating a =configure= script from
a =configure.ac= file.
Atrip should be built out of source, this means that
you have to create a build directory other that the root
directory, for instance in the =build/tutorial= directory
#+begin_src sh :exports code
mkdir -p build/tutorial/
cd build/tutorial
#+end_src
First you have to build the =configure= script by doing
#+begin_src sh :dir build/tutorial :exports code :results raw drawer
../../bootstrap.sh
#+end_src
#+RESULTS:
:results:
Creating configure script
Now you can build by doing
mkdir build
cd build
../configure
make extern
make all
:end:
And then you can see the =configure= options
#+begin_src sh :dir build/tutorial :results raw drawer :eval no
../../configure --help
#+end_src
** Benches
The script =tools/configure-benches.sh= can be used to create
a couple of configurations for benches:
#+begin_src sh :exports results :results verbatim org :results verbatim drawer replace output
awk '/begin +doc/,/end +doc/ { print $NL }' tools/configure-benches.sh |
grep -v -e "begin \+doc" -e "end \+doc" |
sed "s/^# //; s/^# *$//; /^$/d"
#+end_src
#+RESULTS:
:results:
- default ::
This configuration uses a CPU code with dgemm
and without computing slices.
- only-dgemm ::
This only runs the computation part that involves dgemms.
- slices-on-gpu-only-dgemm ::
This configuration tests that slices reside completely on the gpu
and it should use a CUDA aware MPI implementation.
It also only uses the routines that involve dgemm.
:end:
In order to generate the benches just create a suitable directory for it
#+begin_src sh :eval no
mkdir -p build/benches
cd buid/benches
../../tools/configure-benches.sh CXX=g++ ...
#+end_src
and you will get a Makefile together with several project folders.
You can either configure all projects with =make all= or
then go in each folder.
Notice that you can give a path for ctf for all of them by doing
#+begin_src sh :eval no
../../tools/configure-benches.sh --with-ctf=/absolute/path/to/ctf
#+end_src
* Running benches
** Main benchmark
The main benchmark gets built in =bench/main= and is used to run an
atrip run with random tensors.
A common run of this script will be the following
#+begin_src sh
bench/main \
--no 100 \
--nv 1000 \
--mod 1 \
--% 0 \
--dist group \
--nocheckpoint \
--max-iterations 1000
#+end_src