-
Notifications
You must be signed in to change notification settings - Fork 1
/
INSTALL
83 lines (59 loc) · 3.32 KB
/
INSTALL
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
================================================================================
OpenAtom Massively Parallel Code v1.5a
================================================================================
INSTALLATION
1. To compile OpenAtom, you need to specify the following things:
i. Charm++ Installation Path
ii. FFTW-2 and FFTW-3 Library Path
iii. Optimized Math Libraries
iv. Charm-FFT Library
2. First, copy a machine specific configuration file (config.MACHINE.mk) from
the makefiles/ directory to the OpenAtom base directory and rename it to
config.mk.
Linux Workstations config.mk
Blue Gene/Q config.mk.bgq
Cray XE6/XK7/XC30 config.mk.xk6
Use config.mk for systems that do not belong to any of the above options.
3. To install Charm++, download a copy from http://charm.cs.illinois.edu
Depending on the architecture of the machine, choose the right build options
and build Charm++. If you have trouble building Charm++, e-mail:
Next, edit the CHARMBASE variable in config.mk to point to the Charm++
installation.
4. OpenAtom requires double precision FFTW-2 and FFTW-3 installed in some location.
Edit the FFT2_HOME and FFT3_HOME variables in config.mk to point to the FFTW
installations. You might also need to edit the DUAL_FFTW variable depending
on how FFTW was compiled. It takes two values: DUAL_FFTW_OFF and DUAL_FFTW
5. You will also need to compile Charm-FFT library separately before compiling
OpenAtom. This library can be obtained using the following command:
git clone https://charm.cs.illinois.edu/gerrit/libs/fft
Follow the instructions in the downloaded code to compile it. Set the
variable CHARMFFT_HOME to the directory with compiled version of Charm-FFT.
6. To start the build, simply type make in the top-level directory along with
any other options that you require. The binary 'OpenAtom' is placed in the
build directory whose name is indicated at the start and end of the build
process.
OPTIONAL CONFIGURATION
1. Check to make sure the optimization options are the ones that you desire.
These are set as the value of the variable OPT in the config.mk file. These
flags are also suffixed to the name of the build directory giving you a
separate directory for each different set of flags.
2. Be sure to type 'make help' to review the different options you have when
invoking the build. For e.g., parallel builds can be much faster if your
machine supports it.
TEST YOUR BUILD
make test should run a minimal set of regression tests and report failures. To
use it, you will have to obtain the water (32 molecules, 10Ry cutoff) dataset
from our git repositories.
Anonymous access:
git clone http://charm.cs.illinois.edu/gerrit/datasets/openatom/water_32M_10Ry.git
Place the dataset in the data/ directory of the openatom tree. Alternatively,
you can place this anywhere and modify the 'w3210' variable in your make config
file (config.mk) to point to the location.
Note:
- This test setup is still quite nascent and in development.
- It currently directly invokes the OpenAtom binary for simple sequential
runs. If your machine has a job queue or a headnode, this may not be
desirable or possible.
- The setup still is not ready to be used in any environment where job
submission is required.