biod.pnpi.spb.ru / alexxy / gromacs.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 31e6835) | patch
added nbnxn analytical Ewald correction kernels
authorBerk Hess <hess@kth.se>
Thu, 22 Nov 2012 12:15:07 +0000 (13:15 +0100)
committerBerk Hess <hess@kth.se>
Mon, 26 Nov 2012 08:40:07 +0000 (09:40 +0100)
commiteb1403520bdaf058e9de289a96773e535e765a93
tree8606af1361fc8c4f6d92711e15029e17d612a51dtree | snapshot
parent31e68351b028af5d2618969685938510ed40a455commit | diff
added nbnxn analytical Ewald correction kernels

In single precision on Bulldozer these kernels are 15% faster than
the quadratic-spline table correction, on Sandy Bridge 2%.
These kernels are only selected with AVX in single precision.

Change-Id: I1c3eb3b24b5570107fe3fa0c4b03fe5e215c79c8
12 files changed:
include/gmx_x86_simd_macros.h diff | blob | history
include/types/nb_verlet.h diff | blob | history
include/types/nrnb.h diff | blob | history
src/gmxlib/nrnb.c diff | blob | history
src/mdlib/forcerec.c diff | blob | history
src/mdlib/nbnxn_kernels/nbnxn_kernel_x86_simd128.c diff | blob | history
src/mdlib/nbnxn_kernels/nbnxn_kernel_x86_simd128.h diff | blob | history
src/mdlib/nbnxn_kernels/nbnxn_kernel_x86_simd256.c diff | blob | history
src/mdlib/nbnxn_kernels/nbnxn_kernel_x86_simd256.h diff | blob | history
src/mdlib/nbnxn_kernels/nbnxn_kernel_x86_simd_inner.h diff | blob | history
src/mdlib/nbnxn_kernels/nbnxn_kernel_x86_simd_outer.h diff | blob | history
src/mdlib/sim_util.c diff | blob | history
Local GROMACS mirror with custom stuff