wall clock sanity check

From: David Riethmiller <rieth_at_phy.ohiou.edu>
Date: Tue, 6 Nov 2012 12:22:53 -0500

Hello Gadget users,

I'm running a galaxy simulation, SPH particles only within an NFW
static potential (implemented in a way similarly described here in
other archived posts). Looking for a sanity check on the amount of
time that should required to run this sim:

1 million particles, running on 24 processors
MaxSizeTimestep 0.1
MinSizeTimestep 0.00
ErrTolTheta 0.5
ErrTolForceAcc 1.e-2
TreeDomainUpdateFrequency 0.1
PartAllocFactor 5.0
TreeAllocFactor 5.0
BufferSize 25 % in MByte
UnitLength_in_cm 3.085678e21 % 1.0 kpc
UnitMass_in_g 1.989e43 % 1.0e10 solar masses
UnitVelocity_in_cm_per_s 2.0735e7 % 207 km/sec
(this makes UnitTime_in_s = 1.488e14, or about 4.7 Myrs)
MinGasHsmlFractional 0.0001
SofteningGas 0.01

Roughly spherical system, within a radius of ~100, total gas mass of
~1, all in gadget units. No additional physics turned on, just
settling the Poisson noise in my initial conditions. So far, I've
reached a simulation time of ~200 (=940 Myrs) in a wall clock time of
about three days. Does this sound normal, given the parameters that
I've chosen, or have I likely bogged something down, perhaps requiring
an unreasonably high precision or something?

I should also note that I have changed
        (TOPNODEFACTOR * NTask * NTask)
        (TOPNODEFACTOR * NTask),
as per suggestions in other archived posts, and this has helped to
speed up a bit.


David A. Riethmiller
Ph.D. Candidate, Astrophysical Institute
Ohio University

Clippinger Labs 338
Received on 2012-11-06 18:22:59

This archive was generated by hypermail 2.3.0 : 2023-01-10 10:01:31 CET