Top Page > Browsing

Keep getting Segmentation fault for large NEGF calculations Date: 2015/07/18 02:11 Name: KZ Hi, Dear OpenMX developers, I am using OpenMX to calculate the transmission of some large MoS2 systems. My system is 42 atoms as left lead and 42 as right lead, and 42 atoms device. I am using supercomputers like Stampede at TACC for the calculations. My running environment consists of nodes with 1T memory. But I still keep getting segmentation fault for the calculations. Now I have little clue about what is wrong with the calculations. Is it because the communication between nodes are too heavy or the needed memory is larger than 1T? Has anyone met similar problems or could anyone provide some suggestions about it ? Thanks! KZ

Page: [1]

Re: Keep getting Segmentation fault for large NEGF calculations ( No.1 ) Date: 2015/08/06 10:36 Name: T. Ozaki Hi, You can confirm if this happens due to memory shortage by changing computational parameters such as a basis set and grid fineness so that memory requirement can be reduced. If you still get a similar problem even if you reduce memory requirement, this may happen due to the other problem such as improper structural setting. Regards, TO

Page: [1]