Dear ALPS users/developers,
I have a quick question about the NONLOCAL parameter used in the worm algorithm. The ALPS-Wiki mentions it to be a compile time option that one must "undefine to speed up the code for local interactions". However when I run a simulation (linear chain of hardcore bosons with finite chemical potential) for a given chain size with and without this parameter un-defined, the results are quite different. Secondly, the results for the runs with NONLOCAL=0 for different chain sizes are quite comparable to each other (though not exactly equal) and to the exact result for a single-site calculation (again though not exactly equal).
I am trying to understand what this parameter does and why the results change when it is set versus unset; any ideas will be appreciated.
Thanks, Vipin