ERMOD 0.2 beta release
Shun Sakuraba
ERMOD version 0.2-beta1 is released. This release is in "Beta" stage:
http://en.wikipedia.org/wiki/Software_release_life_cycle#Beta
This program is still not suitable for production run and only recommended to experienced users.
The program is released to test whether program works correctly or not, with the users' system.
The software can be obtained from:
http://kaimen.kuicr.kyoto-u.ac.jp/archive/er-solvation/0.2(experimental)/enerepr-0.2.beta1.tar.gz
The cumulative list of changes is as follows.
New features:
* Introduced a new parallelization mechanism. ERMOD can now be run with 100+ cores without significant I/O contention.
Related tickets: #23
* The programs for NAMD and GROMACS are now merged into a single program.
Related tickets: #7, #24
* In the reference system mode, progress.tt is generated, which indicates the number of solvent trajectory snapshots processed by the program.
Related tickets: #49
Improvements from 0.1:
* Scripts under tools/ are rewritten to support wider range of inputs.
Related tickets: #40, #51, #58
* The memory footprint is significantly decreased.
* In PME, FFT is now slightly faster and consumes much less memory.
* Reworked real-space computational kernel and improved performance. (up to 15% faster)
* For solution system with a fixed solute, self-energy is calculated only once for each process.
This optimization is applied even for NPT system.
Related tickets: #1
* Filename is more compiler-friendly.
Related tickets: #24, #50
Changes that breaks compatibility to 0.1:
* Users for NAMD version should also set ERMOD_PLUGINS from this version. See http://kaimen.kuicr.kyoto-u.ac.jp/redmine/projects/ersolv/wiki/Quick_Start_Guide#Obtaining-Software for details.
* Users for GROMACS version should run new version of gen_structure and gen_input to update input files to the latest format.
* Values in weight_soln may be different from that obtained in ver. 0.1.
* Changed random number generator to RANDOM_NUMBER() provided by Fortran compiler.
Several values in reference system may be different from 0.1 because of this change (and changes in parallelization scheme).
* "corref" file is now output in binary format and is incompatible with old style format.
Thanks for all users' feedback.
--
Shun SAKURABA, Ph.D.
Postdoc @ Prof. Matubayasi's Lab in KUICR, Kyoto Univ