RIVET usage
- PYTHIA with RIVET via plugin
- Using PYTHIA with RIVET via HEPMC
- Compiling PYTHIA with RIVET
RIVET is a toolkit for
the validation of Monte Carlo event generators [Buc10]. It
contains the results of many experimental analyses, so that generator
output can easily be compared to data, as well as providing a framework to
implement your own analyses. Although using PYTHIA with RIVET is not
officially supported, some helpful hints are given below. The full RIVET
manual is available online.
PYTHIA with RIVET via plugin
A PYTHIA plugin is available, RivetHooks
defined in
Pythia8Plugins/RivetHooks.h
. This plugin can be loaded,
and analyses can then be easily added for processing. The plugin works with
PythiaParallel
, and RIVET analyses can be run asynchronously.
For the plugin to be available, PYTHIA must be configured with the
--with-rivet
flag before being built.
./configure --with-rivet
If successful, the plugin library lib/pythia8rivet.so
will be built. This requires the RIVET configuration to be available via the
rivet-config
command and is only available for RIVET 4.
There are a number of examples which show how to work with this
plugin: main144
, main164
,
main225
, and main421
. The general usage is
as follows. First, load the plugin.
Init:plugins = {libpythia8rivet.so::RivetHooks}
None of the options described below is available if the plugin is not
successfully loaded. Once the plugin is loaded, a number of additional
settings will be available. To begin, set the output YODA file name;
it must end in .yoda
(this is a RIVET requirement, not
PYTHIA).
Rivet:fileName = NAME
If directory structure is included in the YODA file name, an attempt
will be made to create that directory structure.
Next, define the analyses to run. This can be done with comma
separated lists, {ANALYSIS1,ANALYSIS2,ANALYSIS3,...}
. The
list of analyses can be appended with the +=
syntax.
Rivet:analyses = {ANALYSIS1,ANALYSIS2}
Rivet:analyses += {ANALYSIS3}
Parameters for each analysis can be modified using the following syntax.
Rivet:analyses = {ANALYSIS1:PARM1=VALUE1:PARM2=VALUE2:...,ANALYSIS2}
Finally, set any additional configuration for RIVET.
! This is a list of YODA files which can be preloaded.
Rivet:preloads = {YODA1,YODA2,YODA3,...}
! This is a temporary debugging output filename, which must end with .yoda.
Rivet:dumpName = NAME
! If NUMBER > 0, dump RIVET output to Rivet:dumpName every NUMBER events.
Rivet:dumpPeriod = NUMBER
! Flag to require RIVET to check the beam configuration.
Rivet:checkBeams = off
! By default, zero weight events are skipped for analysis.
Rivet:skipZeroWeights = on
Using PYTHIA with RIVET via HEPMC
The following assumes that you already have RIVET installed. Instructions
for this may be found
here.
Events are passed from PYTHIA to RIVET using the HepMC format. PYTHIA must
be compiled with HepMC support, using the same version of HepMC used when
compiling RIVET. This is setup through the PYTHIA configure
script e.g.
./configure --with-hepmc=/path/to/HepMC
The PYTHIA library itself does not need to be recompiled.
The examples/main132.cc
sample program can then be used to
generate events in HepMC format (which examples/main133.cc
extends by allowing subruns). When in the examples
directory,
the main program can be built and used as follows
make main132
./main132 main132.cmnd main132.hepmc
The first argument is the input file which provides the options for event
generation, while the second is the output file where the HepMC events
should be written.
This HepMC file may now be read and processed by RIVET
rivet --analysis=ANALYSIS_NAME main132.hepmc
where ANALYSIS_NAME
is a
built-in RIVET
analysis, or one you have created yourself. The output of RIVET is in
the form of .aida
files, containing the histograms for the
analysis, which can be processed further with RIVET (see the
RIVET documentation for more details).
The above examples requires that (potentially large) HepMC events are stored
to disk before being read by RIVET. It is possible, instead, to pass the
events directly to RIVET as they are produced by using a FIFO
pipe. This is done with the mkfifo
command
mkfifo my_fifo
./main132.exe main132.cmnd my_fifo &
rivet --analysis=ANALYSIS_NAME my_fifo
Note that main132
is run in the background.
Compiling PYTHIA with RIVET
It is also possible to compile a PYTHIA main program together with the
RIVET library. To facilitate this, there is a header file called
Pythia8Plugins/Pythia8Rivet.h
defining a helper class
called Pythia8::Pythia8Rivet
. This header class has been
deprecated in favor of the plugin structure above, and will be removed
in a future version of PYTHIA. To use this class, a main program needs
to be modified as follows:
#include "Pythia8/Pythia.h"
// Include the Pythia8Rivet header file.
#include "Pythia8Plugins/Pythia8Rivet.h"
int main() {
Pythia pythia;
// Setup the run by reading strings or a command file.
pythia.init();
// Create a Pythia8Rivet object and add (one or several) analyses.
Pythia8Rivet rivet(pythia, "outputfile.yoda");
rivet.addAnalysis("AnalysisName");
rivet.addAnalysis("AnotherAnalysisName");
for (int iEvent = 0; iEvent < 100; ++iEvent) {
if (!pythia.next()) continue;
// Push event to Rivet.
rivet();
// Maybe do other non-Rivet analysis.
}
// Tell Rivet to finalise the run.
rivet.done();
}
To compile the program, information about where Rivet and its
dependencies are installed is needed. This information is available
via the rivet-config
script via the following.
rivet-config --includedir --libdir --cppflags --libs