Hi Mateusz,
This warning comes when trying to avoid a strange problem that appears very very rarely and for which I have not found a reproducible simple test case yet. It appears when for some chance of 1 in 10^-16 there are two kinks at two neighboring sites at exactly the same imaginary time, without there being a hopping term between them. I have checked all points where I can see that this might happen but it still occurs very very rarely and only when you do long runs. As you said, a simple workaround is to just split up the simulation and you can then just restart those where it happens with different seeds. A better solution might though be to use the new worm code, which is faster and should not suffer from the problem. Please contact Tama Ma pingnang@phys.ethz.ch if you want to be an alpha tester for that code.
Matthias
On Dec 29, 2012, at 6:13 PM, Mateusz Łącki mateusz.lacki@gmail.com wrote:
Dear All: Another update: it seems that it is possible to avoid the problem I mentioned by partitioning list of input cases into several files, that is
instead
parm5a: {T=1} {T=2} {T=3} ...
I have parm5a_1: {T=1} {T=2} parm5a_2: {T=3} {T=4} parm5a_3: {T=5} {T=6} ...
and this works.
I have no idea why though... All in all it seems I am able to do the computation within worm 2.0.2, but nevertheless I think I am still reporting the issue before as annoying problem.
Regards, Mateusz Łącki
Le Dec 28, 2012 à 10:03 AM, Mateusz Łącki a écrit :
One more thing: I am using worm 2.02 - perhaps this is something fixed in 2.1.....