CREATECLOUD 2000 1
CLOUDPARTS proton 100.
CLOUDCOORD 0.005 0.005 0.005 0.0 0.0 -0.144
TEMP 0.025
BUFFER 0 1e-4
ODE 5 0 1e-10
COULOMB 1 1
OUTPUT PCloudMirrorN2000 1e-8 1
REALTRAP 2 ../../fieldmaps/E_fieldmap_nested.txt ../../fieldmaps/B_fieldmap1000x1000_80Amps_zoom.txt
NE 1e-4
To be precise:
N=2005 works
N=1999-2004: terminal hangs just before Cunbody header should be outputted
N=1900: segmentation fault (core dumped)
N=1800: segmentation fault (core dumped)
N=1799: segmentation fault (core dumped)
N=1790: segmentation fault (core dumped)
N=1785: segmentation fault (core dumped)
N=1782: segmentation fault (core dumped)
N=1781: segmentation fault (core dumped)
N=1780 works
N=1750 works
N=1700 works
Number in between not yet teseted
(works means, that it shows the Cunbody header and progression bar in terminal)
the Idealtrap part is actually commented out (the hash seems to format the text...)
Last edit: Michael Bartel 2013-09-03
Hi Michael,
It works well with g++ + CUNBODY.
the OUTPUT line could be the problem, because each particle has its own output file (third term = 1). try with 0.
Cheers
Pierre
Hi Pierre.
Tried again, with the newest revision (210) for N=2000 output in one file, it (looks like) it runs fine with the CPUNBODY option compiled (shows the progression bar), but gets stuck with CUNBODY (again before the Cunbody header output). No other change made to the Makefile.
What machine did you test it on? If i am looking at the numbers that do not work, it might have something to do how CUNBODY is allocating memory (on the GPU) or something like that. So it might be machine (or GPU) dependent. Or its something entirely different :).
Cheers,
Michael
Last edit: Michael Bartel 2013-09-04
Hi Micheal,
Which CUDA version and gpu do you have ?
Cheers
Pierre
Hi Pierre.
Running on the sake cluster, so GTX 680 and Cuda Driver 5.0/ Runtime 4.2.
Michael