Page 1 of 1

GW0 with NCORE

Posted: Thu Aug 27, 2015 1:47 am
by mef362
I have found that you can do a single-shot GW0 calculation with NCORE set (e.g. to the number of cores per node), if you are using a 1x1x1 k-point grid; however, the VASP documentation states that NCORE must be set to 1 (default). Are the calculations being done correctly in this case? I have tried doing the calculation with NCORE=1 to compare the results but I run out of memory. My system is relatively large: 75 atoms, 258 electrons, cubic cell a=b=c=~15A. What I find a little odd is that the calculation with NCORE=16 (number of cores per node) and using 24 nodes, the calculation only takes about 15-20 minutes. So it seems strange that I can't do the calculation with NCORE=1 even if I use more nodes (I've tried 48 nodes and even tried reducing the number of cores per node to 4)? I suppose the answer has something to do with not having enough memory to do 1 frequencies per core. FYI, my computer system has 64GB per node (4GB per core).

Error when NCORE=1:

Code: Select all

energies w=
   0.00   0.00   0.26   0.00   0.52   0.00   0.77   0.00   1.03   0.00
   1.28   0.00   1.53   0.00   1.77   0.00   2.01   0.00   2.25   0.00
   2.48   0.00   2.71   0.00   2.93   0.00   3.16   0.00   3.38   0.00
   3.59   0.00   3.81   0.00   4.03   0.00   4.24   0.00   4.46   0.00
   4.67   0.00   4.89   0.00   5.11   0.00   5.34   0.00   5.57   0.00
   5.80   0.00   6.04   0.00   6.30   0.00   6.56   0.00   6.83   0.00
   7.11   0.00   7.42   0.00   7.74   0.00   8.08   0.00   8.45   0.00
   8.85   0.00   9.28   0.00   9.76   0.00  10.29   0.00  10.88   0.00
  11.55   0.00  12.31   0.00  13.19   0.00  14.21   0.00  15.43   0.00
  16.89   0.00  18.70   0.00  20.98   0.00  23.96   0.00  27.99   0.00
 responsefunction array rank=   40704
 LDA part: xc-table for Pade appr. of Perdew
 allocating   1 responsefunctions rank= 40704
 shmem allocating   1 responsefunctions rank= 40704
 response function shared by NCSHMEM nodes    1
 Doing            1  frequencies on each core in blocks of            1
NQ=   1    0.0000    0.0000    0.0000,
forrtl: error (78): process killed (SIGTERM)
Image              PC                Routine            Line        Source
vasp_std           00000000017DBE69  Unknown               Unknown  Unknown
vasp_std           00000000017DA73E  Unknown               Unknown  Unknown
vasp_std           0000000001788112  Unknown               Unknown  Unknown
vasp_std           00000000017351F3  Unknown               Unknown  Unknown
vasp_std           000000000173AEE9  Unknown               Unknown  Unknown
libpthread.so.0    00002AAAADE8F710  Unknown               Unknown  Unknown
vasp_std           0000000001198A07  Unknown               Unknown  Unknown
vasp_std           0000000001324166  Unknown               Unknown  Unknown
vasp_std           0000000001575256  Unknown               Unknown  Unknown
vasp_std           0000000000412326  Unknown               Unknown  Unknown
libc.so.6          00002AAAAECCBD5D  Unknown               Unknown  Unknown
vasp_std           0000000000412219  Unknown               Unknown  Unknown


Thanks for any help

Re: GW0 with NCORE

Posted: Thu Sep 12, 2024 8:46 am
by support_vasp

Hi,

We're sorry that we didn’t answer your question. This does not live up to the quality of support that we aim to provide. The team has since expanded. If we can still help with your problem, please ask again in a new post, linking to this one, and we will answer as quickly as possible.

Best wishes,

VASP