OpenMP parallelization of CNS 1.2
I have worked out how to efficiently parallelize CNS. To use this,
you'll have to patch the source code. The patch leaves the source code
practically unchanged, because it only inserts lines that start with
and are therefore understood as comments by the compiler, unless a
specific compiler option (-openmp in the case of ifort) is used.
The patch is for CNS 1.2 (and also works with 1.1). It was tested quite
a bit by me and others, but I cannot take any responsibility if it does
not work for you.
The speedup by parallelization occurs only for crystallographic
calculations which involve electron density calculations and FFTs. The
result is a reduction of wall-clock time for a CNS job by a factor of
1/2 - 1/3 on a Quad-processor machine. A factor of 2/3 can be expected
on a Dual-core machine.
Axel Bruenger approved that I send out the patch to interested people.
Please send requests for the patch to me ([log in to unmask]),
and please also report bugs in the patch to me.
Kay Diederichs http://strucbio.biologie.uni-konstanz.de
email: [log in to unmask] Tel +49 7531 88 4049 Fax 3183
Fachbereich Biologie, Universitšt Konstanz, Box M647, D-78457 Konstanz