[Up][Tutorial main page]

First example: running chroma

Running, output files and stripping the data

To run chroma use the command

where, chroma_install_dir, denotes the directory where you put the chroma executable. The input file is specified using the "-i" option. A lot of output in the form of XML is generated by the program. By default, the XML output goes into a file named XMLDAT. Alternatively can specify a file name (as above) using the "-o" option.

Chroma also outputs a lot of information to the screen (standard output). You can redirect the standard output to a file using

You have now generated the following files:

• log_hadspec - the standard output.
• hadspec.out.xml - the xml output.
• hadspec.dat.xml - the correlator measurements (the file specified in the HADRON_SPECTRUM inline measurement).

Let us look at the output first, an example of which is given here: log_hadspec. There are several things to note:

• Input XML: most of the input xml from the input file is written out at the beginning.
• Timings: the time taken to perform each inline measurement is given, as is the total time taken for chroma to run.
• Inversions: there are 12 for each spin-colour of the source. For each solve, information on the performance of the inverter is given, along with the number of iterations required to reach convergence, the final residual and the time taken for the inversion.

The output file hadspec.out.xml contains a lot of (often repeated) information that can be difficult to digest. However, it is useful as a log of the calculation and does contain some important information. Don't spend much time looking into this at the moment, we comment on it's contents in section reference 1 below.

There is no information about the HADRON_SPECTRUM measurement in hadspec.out.xml as this is all in the file hadspec.dat.xml. This file contains the correlator data we are interested in, however, it also contains a lot of other output. To make it easier to read the correlator data into an analysis routine, for example to calculate effective masses, we will strip the data from the file hadspec.dat.xml and put it into a file with a simple text format. How to find the correlator data within hadspec.dat.xml is detailed below in section reference 2.

Stripping the correlator data from hadspec.dat.xml using ADAT

The ADAT software package contains many programs for stripping data from the various xml or binary files that chroma produces. It also contains programs for analysing the data. You will need the program strip_hadspec. Many files will be produced by the program so it is best to change to a new directory. Run the program with the command

where strip_hadspec_install_dir is the directory where you have the strip_hadspec executable.

The program strips the correlator data into separate files for each hadron for every momentum and smearing combination used (here we just have one smearing combination). You will have generated the files shown in this list. The file names have a straightforward format:

For example

• channel - this is the name of the particle, for example, pion, proton etc
• _[momenta] - if the momentum is non-zero then it is given in the form pxX_pyY_pzZ, where (X,Y,Z) is the momentum. If we had chosen not to average over equivalent momenta there would be one file for each permutation of (X,Y,Z). For zero momentum _[momenta] is omitted from the file name.
• MassString - this is related to the mass of the quark. The Mass parameter, which can be used as input instead of Kappa in the PROPAGATOR inline measurement is defined in chroma as Mass=0.5*(1/Kappa - 1/Kappa_c), where Kappa_c=1/8 is the kappa value corresponding to zero quark mass in the free field case. We chose Kappa=0.11 which gives Mass=0.54545. The value is rounded and the 0. is removed to give the MassString=5455. Note that in the interacting case this parameter can be negative.
• [SourceSmear_Channel] - this string indicates the smearing and operator used at the source. Point smearing is denoted by a P. There are often several different operators used for the same particle. For the pion one can use q &Gamma q where &Gamma = &gamma₅ (Channel = 1) or &Gamma = &gamma₄&gamma₅ (Channel = 2). The channels for the baryons are more complicated and are discussed in section Looking into the chroma code.
• [SinkSmear_Channel] - this follows the same conventions as for the source.
• [Smearing] - this re-iterates the smearing at the source and sink (in that order).

Now let us look into the file: pion.D5455.P_1.P_1.PP,

1 8 1 4 1
0 0.695079547680507 0
1 0.0371189174074971 0
2 0.00801121179028996 0
3 0.00306581884433399 0
4 0.00206155294108612 0
5 0.00306324860321183 0
6 0.0080019051238196 0
7 0.0371176435437519 0

The first line states, in order, the number of configurations (1), the temporal extent of the lattice (8), that the correlator is complex (1) (0 if it is real) and the spatial extent of the lattice (4). The last number is no longer used.

The zero momentum pion correlator is then listed as a function of the timeslice.

At the moment you only have one data file (with results from one configuration). Later, in the section Mini spectrum calculation, you will use strip_hadspec with multiple files and calculate the effective mass of a particle.

Playing around

Now that you have successfully run chroma and stripped the output into a manageable format it is instructive to play around with the input file. Here are a few things to try.

• Change some of the parameters:
  • the position of the source, for example, to timeslice 5, i.e. use <t_srce>0 0 0 5<t_srce>. You will find that when you strip the output file that the correlators look similar to the results for those generated with the source at timeslice 0 (i.e. the largest number, approximately 0.695, is at timeslice 0). This is because chroma automatically shifts the correlator (after it has been calculated) so that the source timeslice is always labelled by 0.
  • The kappa value to correspond to a lighter quark mass, for example Kappa=0.12. The number of iterations for the inversions should increase, along with the time taken. Decrease the residual: this will also increase the number of iterations. Decrease the maximum number of iterations so that it is lower than the number required for convergence, for example <MaxCG>5<MaxCG>. The program still runs through successfully but you now get a non-convergence warning (!!!).
  • The hadrons which you calculate, for example only calculate the mesons. Try setting avg_equiv_mom to false, and so on.
• Mistyping names in the input file: this is a very good way to find out all the values certain parameters can take. If you mistype any key (they are capitalized) for example MAKE_SOURCE or WILSON etc, chroma will give you a list of possible values before exiting. You can go through each key in turn mistyping it (but nothing else before it). Here are a few examples
  • MAKE_SOURCE - this is the name of an inline measurement. If you mistype this you will get a list of all the possible inline measurements. You can see the list here: list of inline measurements
  • POINT_SOURCE - mistyping this will give you all the source types.
  • PROPAGATOR - mistyping this will just give you the same list as for MAKE_SOURCE as it is also an inline measurement key.
  • Do the same for WILSON to get all the fermion actions and CG_INVERTER to get all the inverters.

You can now go on to the change the XML file in a significant way in the next part of the tutorial: Modifying the input file.

Reference 1: the output file hadspec.out.xml

• Input XML: this is given, in it's entirety, at the beginning of the output file.
• Plaquette: the plaquettes and Polyakov loops are calculated at the beginning of most inline measurements and so you will see this information, within the <Observables> tags, repeated 4 times in the output file.
• Inversion: the number of iterations and the residual for the inversion of each spin-colour component is given. Look below the <QuarkProp4> tag.
• Pion propagator: this is calculated and written out to this file in several inline measurements. Firstly, in PROPAGATOR, after the propagator has been calculated. Look for <Prop_correlator>, to find the pion correlator as a function of the timeslice. Secondly, in SINK_SMEAR, where the pion correlator is calculated before smearing (look for <Forward_prop_correlator>) and after smearing (look for <SinkSmearedProp_correlator>). Since we have used point smearing the propagator, and hence the pion correlator, is unchanged by smearing.

Reference 2: how to find the correlator data within hadspec.dat.xml

To find the correlator data within hadspec.dat.xml search for the tag <Point_Point_Wilson_Mesons>. The first few lines of XML that follow this tag are shown below.

      <Point_Point_Wilson_Mesons>
        <elem>
          <gamma_value>0</gamma_value>
          <momenta>
            <elem>
              <sink_mom_num>0</sink_mom_num>
              <sink_mom>0 0 0</sink_mom>
              <mesprop>
                <elem>
                  <re>0.559868925531191</re>
                  <im>0</im>
                </elem>
                <elem>
                  <re>-0.0119817844101817</re>
                  <im>0</im>
                </elem>
                <elem>
                  <re>-0.00286980391956604</re>
                  <im>0</im>
                </elem>
                <elem>
                  <re>-0.00197637835844944</re>
                  <im>0</im>
                </elem>
                <elem>
                  <re>-0.00187173219273973</re>
                  <im>0</im>
                </elem>
                <elem>
                  <re>-0.00197547156858491</re>
                  <im>0</im>
                </elem>
                <elem>
                  <re>-0.00286186982066283</re>
                  <im>0</im>
                </elem>
                <elem>
                  <re>-0.0119718946280045</re>
                  <im>0</im>
                </elem>
              </mesprop>
            </elem>

The correlators are listed in terms of the &Gamma combinations used in the meson operator. There are 16 combinations, which are labelled by the tag <gamma_value>. What value corresponds to what &Gamma combination is given here, chroma_gamma_matrices (n_&Gamma(dec) in the table).

The correlator is given for each momentum combination, labelled with the tag <sink_mom_num> and explicitly specified within the tag <sink_mom>.

The timeslice is not specified but the correlator is always displayed starting from the source timeslice.

After the mesons, the currents are listed, look for the <Point_Point_Meson_Currents> tag, and then follow the baryons (<Point_Point_Wilson_Baryons>). To find out what the values labelled by the <baryon_num> tag correspond to you will need to look in the chroma code itself, see section Looking in the chroma code.

[Up][Tutorial main page]