Frequently-asked questions

Q: Why is my code running so slowly?

A: If you see massive performance problems (slow down by huge factors like 100), then you might have an issue with OpenMP when using MKL for instance. Setting OMP_NUM_THREADS=1 solves this problem.

Q: How do I save the triqs and cthyb hash and script for debugging purposes?

A: Simply add this to your script:

from pytriqs.applications.impurity_solvers.cthyb import version
if mpi.is_master_node():
  with HDFArchive(filename+".h5",'a') as Results:
    if "log" not in Results: Results.create_group("log")
    log = Results["log"]
    log["version"] = version.version
    log["triqs_hash"] = version.triqs_hash
    log["cthyb_hash"] = version.cthyb_hash
    log["script"] = open(sys.argv[0]).read() # read myself !

Q: Why does my data look so noisy?

A: If you are running a parallel calculation, ensure that you are using a different random seed on each core, i.e., that it is a function of the MPI rank:

param['random_seed'] = 34788 + 928374 * mpi.rank   # Default random seed

Q: How do I use the segment picture?

A: This cthyb code is based on the matrix formulation and does not include optimisations for the segment picture (applicable only in cases with density-density interactions only).