minor bug fix, version gadget-2.0.2

Hi,

There has been a bug in gadget-2.0.1 in the routine
force_treeupdate_pseudos(), file forcetree.c, line 759. Instead of the correct
        maxsofttype = (DomainMoment[i].bitflags >> 2) & 7;
        diffsoftflag = (DomainMoment[i].bitflags >> 5) & 1;

there was a

        maxsofttype = (Nodes[no].u.d.bitflags >> 2) & 7;
        diffsoftflag = (Nodes[no].u.d.bitflags >> 5) & 1;

This meant that there was a chance that internal top-level tree nodes may
not have contained the correct maximum gravitational softening length for
the represented particles in all cases when the UNEQUALSOFTENING option was
used. Since these nodes are rarely if ever used for force computations (and
if they are then they are at distances larger than the softening length), it
is unlikely that this ever affected the force computation in practice. The
error was also present in a slight different form when in addition
ADAPTIVE_GRAVSOFT_FORGAS was set.

Two other small changes were made:

* the FLEXSTEPS option has been improved, and may now in principle used
together with the PMGRID option
* when FLEXSTEPS is used, the PM-stepsize may now also shrink with time and
the PM computation is not restricted anymore to steps where all particles
are active

The updated version is available on the web-site as version 2.0.2. (Patched
files are forcetree.c, timestep.c, longrange.c, init.c, run.c, allvars.h)

Greetings,
Volker
Received on 2006-04-22 16:02:08