Mjöllnir 0.3-alpha is available
Tenebrosus Scriptor
I am glad to inform you, that I have actually finished the development
of the next Mjöllnir version, 0.3.
In this version a lot of new functionality was added to Mjöllnir, and
its code was extensively optimised in order to increase performance.
Short changelog:
* Implemented multipart systems treatment (dividing system into a set of
non-interacting parts).
* Implemented analytical equations usage.
* Initial parallelisation support (eigenvalues calculation may be
parallelised on SMP systems with OpenMP).
* The model parser is extremely optimised (the main point --- avoided
slow matrix by matrix multiplications) --- model parsing is a few times
faster now.
* Initial lanthanides support.
* Changed input files requirements and launching scheme.
* Fixed bug caused by using LAPACK's ``slamch'' function instead of
required ``dlamch'' (this lowered the precision of eigenvalues
calculation to maximum precision of single precision real numbers
instead of possible double precision ones).
Binaries for win32 are available here:
https://sites.google.com/site/mjollnirmagn/home/mjollnir-0.3-alpha.zip?attredirects=0
They are considered to be "alpha stage" (unstable) version, this code is
not free from bugs, mistakes and misunderstandings (including stupid
ones), so I encourage you to run as much tests on it as you are able to.
The manual also needs to be read and checked for mistakes and
obscurities. This code should be used very carefully. I hope to prepare
stable "release stage" version as soon, as possible.
!!!!!! Note! In this version the names of input files were changed! !!!!!!!
In order to treat your old models with the new version, you have to:
1. Create additional file named "fragments" in the working directory
(please check out the manual and examples);
2. Rename your "model.xml" file to "model0.xml".
In before your questions and feedbacks :)
Yours sincerely,
Anton S. Lytvynenko.