Skip to content
Snippets Groups Projects

addition to README

    • Open Issue created by Aart Heijboer @aheijboer

    Hi,

    In the interest of transparency, and preventing people getting confused, I would request that the following is added to the README:

    For the Users of ROOT, the equivalent functionality can be accomplished without additional packages in a bare ROOT6 session as follows.

    import ROOT
jppd = '/pbs/throng/km3net/src/Jpp/master/software'
ROOT.gInterpreter.AddIncludePath( jppd )
ROOT.gInterpreter.ProcessLine('#include "'+jppd+'/JPhysics/JPDF_t.hh"')

pdf = ROOT.JShowerPDF_t("/pbs/throng/km3net/src/Jpp/v12.1.0/data/J%p.dat", 1.5 )

result = pdf.calculate(1000, 100, 0.5, 0.1, 0.1 , 100 )
print ( result.f, result.V )

    btw: Tom: do you see warning when compiling JPDF_t.hh with clang?

    Designs

      Child items ...

      2. Activity

        • Tamas Gal
          Tamas Gal @tgal ·
          Owner

          I can certainly add such references with a warning that this method is neither tested, nor supported by this package. One of the issues is for example that PyROOT is extremely slow in this case. It is so slow that it falls definitely into the time range where people think it's frozen and they will probably bug is with this.

          A quick check in Lyon yields 8 minutes to create the pdf object and 8µs per lookup.

          In case of jppy it's only 1 minute for creating the pdf and 1.85µs per lookup.

          Here is the full output, I guess you are referring to those warnings? I don't see any warnings from clang while compiling with pybind11 but those might just be suppressed, I'm not sure...

          In [1]: import ROOT
   ...: jppd = '/pbs/throng/km3net/src/Jpp/master/software'
   ...: ROOT.gInterpreter.AddIncludePath( jppd )
   ...: ROOT.gInterpreter.ProcessLine('#include "'+jppd+'/JPhysics/JPDF_t.hh"')
   ...:
In file included from input_line_20:1:
In file included from /pbs/throng/km3net/src/Jpp/master/software/JPhysics/JPDF_t.hh:6:
In file included from /pbs/throng/km3net/src/Jpp/master/software/JTools/JSpline.hh:11:
In file included from /pbs/throng/km3net/src/Jpp/master/software/JTools/JResult.hh:8:
/pbs/throng/km3net/src/Jpp/master/software/JLang/JAssert.hh:29:9: warning: 'STATIC_CHECK' macro redefined [-Wmacro-redefined]
#define STATIC_CHECK(expr) { JLANG::JAssert<expr>(); }
        ^
/pbs/throng/km3net/software/root/6.18.04/etc/../include/Math/StaticCheck.h:56:9: note: previous definition is here
#define STATIC_CHECK(expr, msg) \
        ^
In file included from input_line_20:1:
In file included from /pbs/throng/km3net/src/Jpp/master/software/JPhysics/JPDF_t.hh:3:
In file included from /pbs/throng/km3net/src/Jpp/master/software/JTools/JMap.hh:6:
/pbs/throng/km3net/src/Jpp/master/software/JTools/JElement.hh:44:10: warning: defaulted move assignment operator of
      'JSplineFunction1D<JTOOLS::JSplineElement2S<double, double>, JCollection,
      JTOOLS::JResultPDF<double>, JTOOLS::JDistance<double> >' will move assign
      virtual base class 'JFunctional<double, JTOOLS::JResultPDF<double> >'
      multiple times [-Wmultiple-move-vbase]
  struct JElement2D {
         ^
/usr/lib/gcc/x86_64-redhat-linux/4.8.5/../../../../include/c++/4.8.5/bits/vector.tcc:337:16: note: in implicit move assignment operator for 'JTOOLS::JElement2D<double,
      JTOOLS::JSplineFunction1D<JTOOLS::JSplineElement2S<double, double>,
      JCollection, JTOOLS::JResultPDF<double>, JTOOLS::JDistance<double> > >'
      first required here
          *__position = _Tp(std::forward<_Args>(__args)...);
                      ^
/usr/lib/gcc/x86_64-redhat-linux/4.8.5/../../../../include/c++/4.8.5/bits/vector.tcc:306:4: note: in instantiation of function template specialization
      'std::vector<JTOOLS::JElement2D<double,
      JTOOLS::JSplineFunction1D<JTOOLS::JSplineElement2S<double, double>,
      JCollection, JTOOLS::JResultPDF<double>, JTOOLS::JDistance<double> > >,
      std::allocator<JTOOLS::JElement2D<double,
      JTOOLS::JSplineFunction1D<JTOOLS::JSplineElement2S<double, double>,
      JCollection, JTOOLS::JResultPDF<double>, JTOOLS::JDistance<double> > > >
      >::_M_insert_aux<JTOOLS::JElement2D<double,
      JTOOLS::JSplineFunction1D<JTOOLS::JSplineElement2S<double, double>,
      JCollection, JTOOLS::JResultPDF<double>, JTOOLS::JDistance<double> > > >'
      requested here
          _M_insert_aux(__position, std::forward<_Args>(__args)...);
          ^
/usr/lib/gcc/x86_64-redhat-linux/4.8.5/../../../../include/c++/4.8.5/bits/stl_vector.h:989:16: note: in instantiation of function template specialization
      'std::vector<JTOOLS::JElement2D<double,
      JTOOLS::JSplineFunction1D<JTOOLS::JSplineElement2S<double, double>,
      JCollection, JTOOLS::JResultPDF<double>, JTOOLS::JDistance<double> > >,
      std::allocator<JTOOLS::JElement2D<double,
      JTOOLS::JSplineFunction1D<JTOOLS::JSplineElement2S<double, double>,
      JCollection, JTOOLS::JResultPDF<double>, JTOOLS::JDistance<double> > > >
      >::emplace<JTOOLS::JElement2D<double,
      JTOOLS::JSplineFunction1D<JTOOLS::JSplineElement2S<double, double>,
      JCollection, JTOOLS::JResultPDF<double>, JTOOLS::JDistance<double> > > >'
      requested here
      { return emplace(__position, std::move(__x)); }
               ^
/pbs/throng/km3net/src/Jpp/master/software/JTools/JCollection.hh:165:29: note: in instantiation of member function 'std::vector<JTOOLS::JElement2D<double,
      JTOOLS::JSplineFunction1D<JTOOLS::JSplineElement2S<double, double>,
      JCollection, JTOOLS::JResultPDF<double>, JTOOLS::JDistance<double> > >,
      std::allocator<JTOOLS::JElement2D<double,
      JTOOLS::JSplineFunction1D<JTOOLS::JSplineElement2S<double, double>,
      JCollection, JTOOLS::JResultPDF<double>, JTOOLS::JDistance<double> > > >
      >::insert' requested here
        i = container_type::insert(i, value_type(x, ordinate_type()));
                            ^
/pbs/throng/km3net/src/Jpp/master/software/JTools/JElement.hh:44:10: note: in instantiation of member function
      'JTOOLS::JCollection<JTOOLS::JElement2D<double,
      JTOOLS::JSplineFunction1D<JTOOLS::JSplineElement2S<double, double>,
      JCollection, JTOOLS::JResultPDF<double>, JTOOLS::JDistance<double> > >,
      JTOOLS::JDistance<double> >::get' requested here
  struct JElement2D {
         ^
/usr/lib/gcc/x86_64-redhat-linux/4.8.5/../../../../include/c++/4.8.5/bits/stl_construct.h:103:11: note: (skipping 5 contexts in backtrace; use -ftemplate-backtrace-limit=0 to see all)
            std::_Destroy(std::__addressof(*__first));
                 ^
/usr/lib/gcc/x86_64-redhat-linux/4.8.5/../../../../include/c++/4.8.5/bits/stl_construct.h:103:11: note: in instantiation of function template specialization
      'std::_Destroy<JTOOLS::JElement2D<double, JTOOLS::JMultiMap<double,
      JTOOLS::JSplineFunction1D<JTOOLS::JSplineElement2S<double, double>,
      JCollection, JTOOLS::JResultPDF<double>, JTOOLS::JDistance<double> >,
      JTOOLS::JMapList<JPolint0FunctionalGridMap,
      JTOOLS::JMapList<JPolint0FunctionalGridMap, JLANG::JNullType> >,
      JTOOLS::JDistance<double> > > >' requested here
/usr/lib/gcc/x86_64-redhat-linux/4.8.5/../../../../include/c++/4.8.5/bits/stl_construct.h:127:2: note: in instantiation of function template specialization
      'std::_Destroy_aux<false>::__destroy<JTOOLS::JElement2D<double,
      JTOOLS::JMultiMap<double,
      JTOOLS::JSplineFunction1D<JTOOLS::JSplineElement2S<double, double>,
      JCollection, JTOOLS::JResultPDF<double>, JTOOLS::JDistance<double> >,
      JTOOLS::JMapList<JPolint0FunctionalGridMap,
      JTOOLS::JMapList<JPolint0FunctionalGridMap, JLANG::JNullType> >,
      JTOOLS::JDistance<double> > > *>' requested here
        __destroy(__first, __last);
        ^
/usr/lib/gcc/x86_64-redhat-linux/4.8.5/../../../../include/c++/4.8.5/bits/stl_construct.h:151:7: note: in instantiation of function template specialization
      'std::_Destroy<JTOOLS::JElement2D<double, JTOOLS::JMultiMap<double,
      JTOOLS::JSplineFunction1D<JTOOLS::JSplineElement2S<double, double>,
      JCollection, JTOOLS::JResultPDF<double>, JTOOLS::JDistance<double> >,
      JTOOLS::JMapList<JPolint0FunctionalGridMap,
      JTOOLS::JMapList<JPolint0FunctionalGridMap, JLANG::JNullType> >,
      JTOOLS::JDistance<double> > > *>' requested here
      _Destroy(__first, __last);
      ^
/usr/lib/gcc/x86_64-redhat-linux/4.8.5/../../../../include/c++/4.8.5/bits/stl_vector.h:415:14: note: in instantiation of function template specialization
      'std::_Destroy<JTOOLS::JElement2D<double, JTOOLS::JMultiMap<double,
      JTOOLS::JSplineFunction1D<JTOOLS::JSplineElement2S<double, double>,
      JCollection, JTOOLS::JResultPDF<double>, JTOOLS::JDistance<double> >,
      JTOOLS::JMapList<JPolint0FunctionalGridMap,
      JTOOLS::JMapList<JPolint0FunctionalGridMap, JLANG::JNullType> >,
      JTOOLS::JDistance<double> > > *, JTOOLS::JElement2D<double,
      JTOOLS::JMultiMap<double,
      JTOOLS::JSplineFunction1D<JTOOLS::JSplineElement2S<double, double>,
      JCollection, JTOOLS::JResultPDF<double>, JTOOLS::JDistance<double> >,
      JTOOLS::JMapList<JPolint0FunctionalGridMap,
      JTOOLS::JMapList<JPolint0FunctionalGridMap, JLANG::JNullType> >,
      JTOOLS::JDistance<double> > > >' requested here
      { std::_Destroy(this->_M_impl._M_start, this->_M_impl._M_finish,
             ^
/pbs/throng/km3net/src/Jpp/master/software/JTools/JCollection.hh:71:9: note: in instantiation of member function 'std::vector<JTOOLS::JElement2D<double,
      JTOOLS::JMultiMap<double,
      JTOOLS::JSplineFunction1D<JTOOLS::JSplineElement2S<double, double>,
      JCollection, JTOOLS::JResultPDF<double>, JTOOLS::JDistance<double> >,
      JTOOLS::JMapList<JPolint0FunctionalGridMap,
      JTOOLS::JMapList<JPolint0FunctionalGridMap, JLANG::JNullType> >,
      JTOOLS::JDistance<double> > >, std::allocator<JTOOLS::JElement2D<double,
      JTOOLS::JMultiMap<double,
      JTOOLS::JSplineFunction1D<JTOOLS::JSplineElement2S<double, double>,
      JCollection, JTOOLS::JResultPDF<double>, JTOOLS::JDistance<double> >,
      JTOOLS::JMapList<JPolint0FunctionalGridMap,
      JTOOLS::JMapList<JPolint0FunctionalGridMap, JLANG::JNullType> >,
      JTOOLS::JDistance<double> > > > >::~vector' requested here
  class JCollection :
        ^
/pbs/throng/km3net/src/Jpp/master/software/JTools/JSpline.hh:658:5: note: 'JFunctional<double, JTOOLS::JResultPDF<double> >' is a virtual base class of
      base class 'JSplineFunction<JTOOLS::JSplineElement2S<double, double>,
      JCollection, JTOOLS::JResultPDF<double>, JTOOLS::JDistance<double> >'
      declared here
    public JSplineFunction<JElement_t, JCollection_t, JResult_t, JDistance_t>,
    ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/pbs/throng/km3net/src/Jpp/master/software/JTools/JSpline.hh:659:5: note: 'JFunctional<double, JTOOLS::JResultPDF<double> >' is a virtual base class of
      base class 'JFunction1D<typename JSplineElement2S<double,
      double>::abscissa_type, JTOOLS::JResultPDF<double> >' (aka
      'JFunction1D<double, JTOOLS::JResultPDF<double> >') declared here
    public JFunction1D<typename JElement_t::abscissa_type, JResult_t>
    ^
Out[1]: 0

In [2]: %time pdf = ROOT.JShowerPDF_t("/pbs/throng/km3net/src/Jpp/v12.1.0/data/J
   ...: %p.dat", 1.5 )
loading input from file /pbs/throng/km3net/src/Jpp/v12.1.0/data/J14p.dat... OK
loading input from file /pbs/throng/km3net/src/Jpp/v12.1.0/data/J13p.dat... OK
bluring PDFs... OK
CPU times: user 7min 50s, sys: 6.78 s, total: 7min 57s
Wall time: 7min 52s

In [3]: %timeit result = pdf.calculate(1000, 100, 0.5, 0.1, 0.1 , 100 )
8.19 µs ± 987 ns per loop (mean ± std. dev. of 7 runs, 100000 loops each)
          In [1]: import jppy

In [2]: %time pdf = jppy.pdf.JShowerPDF("/pbs/throng/km3net/src/Jpp/v12.1.0/data
   ...: /J%p.dat", 1.5)
loading input from file /pbs/throng/km3net/src/Jpp/v12.1.0/data/J14p.dat... OK
loading input from file /pbs/throng/km3net/src/Jpp/v12.1.0/data/J13p.dat... OK
bluring PDFs... OK
CPU times: user 1min 7s, sys: 2.56 s, total: 1min 10s
Wall time: 1min 9s

In [3]: %timeit result = pdf.calculate(1000, 100, 0.5, 0.1, 0.1 , 100)
1.85 µs ± 232 ns per loop (mean ± std. dev. of 7 runs, 100000 loops each)
        • Aart Heijboer
          Aart Heijboer @aheijboer ·
          Author

          Thanks for including it. I think it will help to create clarity.

          I think the comment that this is not validated and tested by this package is certainly appropriate and correct. I also added this here: https://git.km3net.de/common/aanet/-/blob/master/examples/use_jpp_pdf_direct.py so maybe just refer to that, in stead of including the code.

          In this cases, the speed difference is not due to PyROOT (at least not the one for initializing the PDFs). It may be because code included in this way is compiled in ROOT without optimization. I'll have a quick look if something can be done about that.

          I asked about the warnings since -- of course I saw them, but I'm not sure if they point to genuine issues of the code. It seems they could be valid warnings..

          Edited by Aart Heijboer
        • Please register or sign in to reply
      • Tamas Gal added docs label

        added docs label

      • Aart Heijboer
        Aart Heijboer @aheijboer ·
        Author

        In case you care, it's indeed the ROOT JIT (cling) that is not optimizing (mucb). The following (unpolished) example uses a normal compiler and has theretofore similar speed as native code. Pyroot is still inbetween the call to calculate(), but its at 2.4 us, so not bad..

        import ROOT
ROOT.gInterpreter.AddIncludePath("/pbs/throng/km3net/src/Jpp/master/software/")

open('pdf.cc','w').write("""
#include "/pbs/throng/km3net/src/Jpp/master/software/JPhysics/JPDF_t.hh"
JShowerPDF_t getpdf() {
	static JShowerPDF_t r("/pbs/throng/km3net/src/Jpp/v12.1.0/data/J%p.dat", 1.5 );
	return r;
}
""")

ROOT.gInterpreter.ProcessLine('.L pdf.cc+')
pdf = ROOT.getpdf()

#etc
        Edited by Aart Heijboer
      • Tamas Gal
        Tamas Gal @tgal ·
        Owner

        Ah I see, interesting, thanks!

      • Tamas Gal created merge request !3 to address this issue

        created merge request !3 to address this issue

      • Tamas Gal mentioned in merge request !3

        mentioned in merge request !3

      Please register or sign in to reply