All migration tools are primarily intended for developer/expert-user usage, and are accordingly not polished for user friendliness. We do not provide user support for these tools. To use them, you will need a checkout of the LHAPDF6 development code, as well as a non-overlapping installation of LHAPDF5; these scripts are not provided in the LHAPDF6 releease tarballs.

Set migration

To migrate a PDF, install LHAPDF5 (including the Python module) and call e.g.

./creategrids CT10.LHgrid

This will create the .info set metadata file and .lha grid data files in a (new) CT10 directory.

The tomigrate file is a list of the LHAPDF5 PDF names to be migrated, plus any arguments to creategrids that those PDFs require for correct/optimal migration. A convenient way to use this file is as follows (in this case to dump all the HERAPDF sets):

cat tomigrate | grep -E -v "^#" | grep HERA | while read l; do eval "./creategrids $l"; done

(The eval is needed to preserve quote marks in arguments in the tomigrate file.)

Set validation and plotting

To test the new PDFs, build the dumpv5 and/or dumpv6 programs. These are compiled from the same source (with the appropriate LHAPDF library version installed) e.g.:

(cd /path/to/lhapdf5 && make install)
g++ dumpv5v6.cc `lhapdf-config --cppflags -cxxflags --ldflags --libs` -o dumpv5

or

(cd /path/to/lhapdf6 && make install)
g++ dumpv5v6.cc `lhapdf-config --cppflags -cxxflags --ldflags --libs` -o dumpv6

These programs are run like (from a PDF-specific validation directory in e.g. migration/validation/myset):

../../dumpvX <setname> <member_num>

For .LHgrid sets with LHAPDF5, the setname can be given without the .LHgrid extension, i.e. the exact same arguments can be used for LHAPDF5 and 6 dumps. You will need to specify a .LHpdf extension if there is one, though: this ~only applies to the CTEQ6L1 set, though. You'll also need to specify the data path to the preliminary v6 PDF data files if they're not installed in $prefix/share/LHAPDF:

LHAPDF_DATA_PATH=$PWD ./dumpv6 CT10 0

To numerically compare the PDFs and make all the validation plots, by far the easiest thing to do is run the cmpplotv5v6 batch script which loops over PIDs and x & Q values and runs the comparison and plotting programs. From our hypothetical validation/myset directory this is run like:

../../cmpplotv5v6

(You can use a –top flag with cmpplotv5v6 to include parton IDs 6 and -6.)

What follows is a more detailed description of how to use the scripts called by this: probably you don't need to know this, at least at first!

Bulk validation

To dump validation grids for the first (or any) member of each set, these commands may be helpful, to be run from the migration/validation directory:

export LHAPDF_DATA_PATH=$HOME/proj/hep/lhapdf6/migration/sets
mem=0
for i in `ls ../sets`; do mkdir -p ${i}_validation_$mem; done
[install LHAPDF5]
for i in `ls ../sets`; do (cd ${i}_validation_$mem && ../../dumpv5 $i $mem); done
[install LHAPDF6]
for i in `ls ../sets`; do (cd ${i}_validation_$mem && ../../dumpv6 $i $mem); done
for i in `ls ../sets`; do (cd ${i}_validation_$mem && echo $i && ../../cmpplotv5v6); done

Detailed comparison and plotting scripts

The cmpv5v6 script is used to numerically compare dumped-out PDF values between versions 5 and 6 of LHAPDF, and writes out the fractional deviations (with an "epsilon" parameter to excuse very small absolute deviations):

./cmpv5v6 xf_v{5,6}_scanx100_0.dat

(the optional output file is written in the same format as the PDF dumps, to allow the values and the differences to be visualised with the same code)

The following commands (run in a subdir containing the dumped sample point files) are useful for batch running of the comparisons across all standard PID codes and for the available "discrete" x and Q values:

for pid in 0 -5 -4 -3 -2 -1 1 2 3 4 5; do for q in 10 50 100 200 500 1000 2000 5000; do ../../cmpv5v6 xf_v*_scanx_q${q}_${pid}.dat -o xfacc_q${q}_${pid}.dat; done; done
for pid in 0 -5 -4 -3 -2 -1 1 2 3 4 5; do for x in 1e-08 1e-06 0.0001 0.01 0.1 0.2 0.5 0.8; do ../../cmpv5v6 xf_v*_scanq_x${x}_${pid}.dat -o xfacc_x${x}_${pid}.dat; done; done

The plotv5v6 script is useful for 1D comparison plotting of data files, either absolute PDF xf values or fractional deviations and with either x or Q on the x-axis of the plot. It requires the Python matplotlib library. plotv5v6 can be usefully used for batch plotting of PDF values or accuracies as follows, again run in a subdir in this example:

## Plotting absolute PDF values (from the output of dumpv5v6) for all PIDs:
for pid in 0 -5 -4 -3 -2 -1 1 2 3 4 5; do ../../plotv5v6 xf_v*_scanx_q*_$pid.dat --type=xf --axis=x --log=x --ylim=0: -o cmpv5v6_xf_scanx_$pid.pdf; done
for pid in 0 -5 -4 -3 -2 -1 1 2 3 4 5; do ../../plotv5v6 xf_v*_scanq_x*_$pid.dat --type=xf --axis=q -o cmpv5v6_xf_scanq_$pid.pdf; done

## Plotting differences (from the output of cmpv5v6) for all PIDs:
for pid in 0 -5 -4 -3 -2 -1 1 2 3 4 5; do ../../plotv5v6 xfacc_q*_$pid.dat --axis=x -o cmpv5v6_xfacc_scanx_$pid.pdf; done
for pid in 0 -5 -4 -3 -2 -1 1 2 3 4 5; do ../../plotv5v6 xfacc_x*_$pid.dat --axis=q -o cmpv5v6_xfacc_scanq_$pid.pdf; done