GitLab

Mesh the whole domain; take into account the bunch offset; ippl to 3D mapping for multicore simulations are handled

In autophase we stash the bunch and create a single particle on node 0. Thereafter we call itsBunch->update(). When OPAL is run in parallel then it can happen that this single particle is moved from node 0 to some other node. This we have check and act accordingly.

- recalculate reference trajectory inside bends on node 0 after autophasing; fixes problems of BerlinPro guys

/gpfs/DDNgpfs1/bekas/YVES/local/bgclang/r209570-20140527/libc++/include/algorithm:2073:5: error: no matching function for call to '__fill'
    _VSTD::__fill(__first, __last, __value_, typename iterator_traits<_ForwardIterator>::iterator_category());
    ^~~~~~~~~~~~~
/gpfs/DDNgpfs1/bekas/YVES/local/bgclang/r209570-20140527/libc++/include/__config:348:15: note: expanded from macro '_VSTD'
#define _VSTD std::_LIBCPP_NAMESPACE
              ^
/gpfs/DDNgpfs1/bekas/YVES/work/OPT/opal/src/Algorithms/ParallelTTracker.cpp:2098:10: note: in instantiation of function template specialization 'std::__1::fill<ParticleAttribIterator<double>, double>' requested
      here
    std::fill(itsBunch->dt.begin(), itsBunch->dt.end(), itsBunch->getdT());
         ^
/gpfs/DDNgpfs1/bekas/YVES/local/bgclang/r209570-20140527/libc++/include/algorithm:2054:1: note: candidate function [with _ForwardIterator = ParticleAttribIterator<double>, _Tp = double] not viable: no known
      conversion from 'typename iterator_traits<ParticleAttribIterator<double> >::iterator_category' (aka 'std::__1::input_iterator_tag') to 'std::__1::forward_iterator_tag' for 4th argument
__fill(_ForwardIterator __first, _ForwardIterator __last, const _Tp& __value_, forward_iterator_tag)
^
/gpfs/DDNgpfs1/bekas/YVES/local/bgclang/r209570-20140527/libc++/include/algorithm:2063:1: note: candidate function [with _RandomAccessIterator = ParticleAttribIterator<double>, _Tp = double] not viable: no known
      conversion from 'typename iterator_traits<ParticleAttribIterator<double> >::iterator_category' (aka 'std::__1::input_iterator_tag') to 'std::__1::random_access_iterator_tag' for 4th argument
__fill(_RandomAccessIterator __first, _RandomAccessIterator __last, const _Tp& __value_, random_access_iterator_tag)
^

Works now parallel. Remaining problems: -- does not go out out scope propperly, -- number of created particles maybe not correct

In case this goes equal or below 1 we stop tracking the bunch and integrating 
in the material until we have in minimum more than one particle on all nodes.

Works for one core (obviously) and the regression test Degarder-1, degrades a 72 MeV beam to 50.200 MeV.
This is the current benchmark, to be confirmed by FLUKA simulations. 

There is a new material Mylar (My) avaidable. 

Problems/ToDo:

-- the increase of NP from 100000 to 100072 is certainly non physical
-- check for more than one core
-- validate the regression test
-- validate Mylar parametrization

- moving the following classes over from OPAL to CLASSIC
  o OpalField       -> ClassicField
  o OpalRing        -> Ring
  o OpalRingSection -> RingSection
  o LossDataSink
- clean up
- splitting Options into classic and opal options

- let unit test be run from anywhere,
- clean up

- putting own header file first
  o for reason see http://www.cplusplus.com/forum/articles/10627/#msg49679 and
                   http://www.cplusplus.com/forum/articles/10627/#msg49680

- remove 'using namespace' from header files
- remove 'using std::string' and 'using std::abs' from header files

For reason see:
o http://stackoverflow.com/questions/5849457/using-namespace-in-c-headers

- updating .gitignore

- improve check if sections are glued together

Need to discuss this:

     OPAL> * intersectTinyLineSegmentBoundary:  WARNING: pt=( 0.00017769 , -8.1542e-05 , 0.00086429 )  is outside the bbox  i=-7  j=176  k=0


The only boundary geometry MUST be at the first element !

- bugfix in orienting the triangles: the STL set-intersection doesn't work
  as expected with unordered sets.
- multi-parted meshes are now supported (but not yet tested)
- several variables and methods refactored to improve readability.