Dear Michael,
I doubt you can do anything much to help you as given your input that is the best you can get out as speed form the DL_POLY machinery.
The low efficiency warning is just a warning. Take a look at the user manual about performance and what are the engineering limitations on performance if you'd like to find out more.
The only way to get better performance is to load more - larger system or the same one but with much shorter ranged interactions.
Regards,
Ilian
-----Original Message-----
From: DL_POLY Mailing List [mailto:[log in to unmask]] On Behalf Of Michael Ferguson
Sent: 18 August 2017 16:13
To: [log in to unmask]
Subject: DLPOLY 4.08 Low Efficiency Mode
Dear Developers,
I am currently studying a liquid system of approximately 40,000 atoms in a 50x50x200 cell using DLPOLY v4.08.
The computer to which I have access has nodes of 8, 16 or 32 processor cores. I have therefore attempted some scaling tests on each node to evaluate my most efficient option.
However, in the OUTPUT file of each simulation, I receive the following message;
8 CORE:
node/domain decomposition (x,y,z): 1 1 8
cutoffs driven limit on largest possible decomposition: 144 nodes/domains (3,3,16)
cutoffs driven limit on largest balanced decomposition: 8 nodes/domains (1,1,8)
link-cell decomposition 1 (x,y,z): 3 3 2
warning issued 100
*** warning - primary link cell algorithm has a link cell dimension that is < 3 !!! ***
*** DL_POLY_4 RUNNING IN LOW EFFICIENCY MODE !!! ***
16 CORE:
node/domain decomposition (x,y,z): 1 2 8
cutoffs driven limit on largest possible decomposition: 144 nodes/domains (3,3,16)
cutoffs driven limit on largest balanced decomposition: 8 nodes/domains (1,1,8)
link-cell decomposition 1 (x,y,z): 3 1 2
warning issued 100
*** warning - primary link cell algorithm has a link cell dimension that is < 3 !!! ***
*** DL_POLY_4 RUNNING IN LOW EFFICIENCY MODE !!! ***
32 CORE:
node/domain decomposition (x,y,z): 2 2 8
cutoffs driven limit on largest possible decomposition: 144 nodes/domains (3,3,16)
cutoffs driven limit on largest balanced decomposition: 8 nodes/domains (1,1,8)
link-cell decomposition 1 (x,y,z): 1 1 2
warning issued 100
*** warning - primary link cell algorithm has a link cell dimension that is < 3 !!! ***
*** DL_POLY_4 RUNNING IN LOW EFFICIENCY MODE !!! ***
Could you please explain to what I can do to increase the efficiency of the DLPOLY simulation.
Kind Regards
Michael
|