cavazzon
c1aedba74c
- clean up and increased robustness against the possibility that
...
different processors give different results on the same FP operation
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@5135 c92efa57-630b-4861-b058-cf58834340f0
2008-08-26 15:21:07 +00:00
cavazzon
3a248b5a0e
- new interfaces required to replace reduce with mp_sum
...
- renumbering of communication error codes
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@4823 c92efa57-630b-4861-b058-cf58834340f0
2008-04-19 10:12:09 +00:00
giannozz
4dc4266f0b
Misc small changes for Alpha compilation warning, by G. Cantele
...
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@4731 c92efa57-630b-4861-b058-cf58834340f0
2008-03-06 20:02:06 +00:00
cavazzon
8649944110
- Bug fix, wrong behaviour of some driveris for parallel code executed
...
using only one proc.
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@4547 c92efa57-630b-4861-b058-cf58834340f0
2007-12-21 09:24:15 +00:00
cavazzon
1567da896f
- further mpi wrappers (parallel min and max)
...
- use of DCOPY in fft_scatter is inefficient
when using a large number of processors
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@4444 c92efa57-630b-4861-b058-cf58834340f0
2007-11-19 13:19:42 +00:00
cavazzon
bf74ac09d1
- new file mp_base.f90 containing mpi wrapper for block communications,
...
like in it was in reduce sub of PW. This is to avoid large temporary
array allocations in module mp.f90 (mp.f90 calls subs in mp_base.f90).
On the other hands mp.f90 provide the general interfaces for different
kind of data and data size checking.
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@4441 c92efa57-630b-4861-b058-cf58834340f0
2007-11-17 14:32:10 +00:00