[gmx-users] GROMACS 4.6 release is ready!
mark.j.abraham at gmail.com
Mon Jan 21 09:09:41 CET 2013
Hi GROMACS users,
The day is finally here - GROMACS 4.6 is out!
As you've probably heard by now, there are lots of wonderful new
performance features, including
* a native GPU implementation layer - thanks to some heroic work from
Szilard Pall and Berk Hess, with special thanks to Mark Berger & Duncan
Poole from NVIDIA for their excellent advice and support
* new implementation of Verlet kernels with guaranteed buffered
interactions, with very good energy conservation
* brand new classical nonbonded interaction kernels, supporting any of
SSE2, SSE4.1, AMD's 128-bit AVX with FMA support, or Intel's 256-bit AVX
SIMD acceleration, for 30-50% faster performance
* use of OpenMP for better intra-node scaling
* much improved automatic load balancing, including between direct-space
and PME nodes
* improvements to integration algorithms, including lots of new free energy
You can find the code, release notes, installation instructions and test
suite at the links below.
If you have downloaded a 4.6 tarball already, we encourage you to get the
latest one - a last-minute bug fix forced us to change plans.
PDF installation instructions, and an updated manual release will follow in
the coming week.
A 4.5.6 bug-fix release has been prepared, and all its fixes are present in
4.6. We're still preparing its release notes, and will announce the release
For those of you using our git repository, please be advised that that
last-minute bug fix required us to move the tag for the 4.6 release. Your
repo will not change that for you automatically.
The GROMACS development team
More information about the gromacs.org_gmx-users