- Posts: 56
-rdd
- syahidah
-
Topic Author
- Offline
- Senior Boarder
Less
More
11 years 10 months ago #1039
by syahidah
-rdd was created by syahidah
Dear all,
may i know what makes us need to add -rdd,
Because when i run the self-assembly of DSPC using gromacs 4.0, 4.5.5 and 4.5.4 in fedora 16,
i didn't add -rdd, the bilayer was formed.
However, when i run the same process (same input,same command) using gromacs 4.5.5 in ubuntu 12.04LTS, we were asked to add -rdd 0.14, then only the production run can be simulated...
may i know what makes we need to add -rdd?
anyone here have any idea?
your reply is greatly appreciated..
best,
syahida
may i know what makes us need to add -rdd,
Because when i run the self-assembly of DSPC using gromacs 4.0, 4.5.5 and 4.5.4 in fedora 16,
i didn't add -rdd, the bilayer was formed.
However, when i run the same process (same input,same command) using gromacs 4.5.5 in ubuntu 12.04LTS, we were asked to add -rdd 0.14, then only the production run can be simulated...
may i know what makes we need to add -rdd?
anyone here have any idea?
your reply is greatly appreciated..
best,
syahida
Please Log in or Create an account to join the conversation.
- djurre
-
- Offline
- Admin
Less
More
- Posts: 272
11 years 10 months ago #1043
by djurre
Replied by djurre on topic -rdd
-rdd concerns: "The maximum distance for bonded interactions with DD (nm), 0 is determine from initial coordinates".
Since in Martini topologies bonded interactions can stretch over a long distance, due to long bond lengths, mdrun sometimes estimates rdd to short. This means that if two atoms that are bonded (bond, angle, dihedral) move further apart then the estimated rdd and are on different compute nodes, mdrun can't find these atoms and the system crashes.
That is why it is better to set rdd a bit bigger, typically 0.14-0.16 should be enough. This will not influence your results. It could limit the number of compute nodes you can use, but since the number of beads per volume in Martini is much lower then atomistic, the scaling will go down much earlier.
Since in Martini topologies bonded interactions can stretch over a long distance, due to long bond lengths, mdrun sometimes estimates rdd to short. This means that if two atoms that are bonded (bond, angle, dihedral) move further apart then the estimated rdd and are on different compute nodes, mdrun can't find these atoms and the system crashes.
That is why it is better to set rdd a bit bigger, typically 0.14-0.16 should be enough. This will not influence your results. It could limit the number of compute nodes you can use, but since the number of beads per volume in Martini is much lower then atomistic, the scaling will go down much earlier.
Please Log in or Create an account to join the conversation.
- syahidah
-
Topic Author
- Offline
- Senior Boarder
Less
More
- Posts: 56
11 years 10 months ago #1044
by syahidah
Replied by syahidah on topic -rdd
Dear djurre,
Thank you so much for the explanation..
got it=)
Thank you so much for the explanation..
got it=)
Please Log in or Create an account to join the conversation.
Time to create page: 0.134 seconds