Dear All, I am using the new dwa code to compute sample test case (exactly like in tutorial05.py). The goal is to compute density of atoms in a harmonic trap for the BH model, canonical ensemble.
I would like two ask a few questions:
1. does dwa support canonical ensemble sampling like old worm code? I set RESTRICT_MEASUREMENTS[N]=16, but still I get non-integer “Total Particle Number” like 15.986.
2. I find, that the dwa produces lots of output, which amounts to several tens of MB per second:
Checkpoint Sweep 145857 ... Probability : 0.985054 / 0.447055 ... Measuring ... N = 16 ... speed = 0 Checkpoint Sweep 145858 ... Probability : 0.985054 / 0.447055 ... Measuring ... N = 16 ... speed = 0 Checkpoint Sweep 145859 ... Probability : 0.985054 / 0.447055 ... Measuring ... N = 16 ... speed = 0 Checkpoint Sweep 145860 ... Probability : 0.985054 / 0.447055 ... Measuring ... N = 16 ... speed = 0 Checkpoint Sweep 145861 ... Probability : 0.985054 / 0.447055 ... Measuring ... N = 16 ... speed = 0
in addition to that thre are “old checkpoint” lines: Checking if Simulation 3 is finished: not yet, next check in 5 seconds ( 0% done).
So my question is: what is the difference between “new checkpoint lines” and “old checkpoint lines”. Does in both cases actual checkpointing take place (meaning writing “current state to disk”)? I suspect that “new checkpoing lines” (the ones above) appear much to often to allow for any significant disk “i/o” (thousands of times per second).
3. is is possible to checkpoint every X secs like in worm core, or at least every X sweeps? I can set “SKIP=100” to get output every 100th iteration, but also to measure every 100 iterations. dwa —help shows : " --checkpoint-time arg (=1800) time between checkpoints” but definitely checkpointing is indicated several times a second, not every 30 minutes.
I run dwa as: mpiexec -np 3 ~/alps22/bin/dwa --mpi --Tmin=5 --Tmax=100 conf.in.xml >out 2>out2
My sample program outputs: Simulation ==========
Sweeps : 100001 Thermalization sweeps : 10000 Skip (Sweeps per measurement) : 500
Best, Mateusz Łącki