New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

comm_modify cutoff/multi option #8

Closed
herrmartinwagner opened this Issue Aug 19, 2015 · 6 comments

Comments

Projects
None yet
2 participants
@herrmartinwagner

herrmartinwagner commented Aug 19, 2015

Adds a cutoff/multi option to comm_modify to specify type-dependent communication cutoffs when in mode multi. Also adds explanation about default behaviour of option cutoff when in mode single to doc.

@akohlmey akohlmey self-assigned this Aug 24, 2015

@akohlmey

This comment has been minimized.

Show comment
Hide comment
@akohlmey

akohlmey Aug 25, 2015

Member

It is a bit annoying that github doesn't support attachments that are not images. One way to workaround this is to upload the patch to a service like dropbox or google docs and then create a readonly link for sharing. For the sake of testing and archiving, am uploading the patch sent via personal e-mail below.
https://www.dropbox.com/s/ahijw8bn7n91nk0/comm_modify_cutoff_multi.tar.gz?dl=0

Member

akohlmey commented Aug 25, 2015

It is a bit annoying that github doesn't support attachments that are not images. One way to workaround this is to upload the patch to a service like dropbox or google docs and then create a readonly link for sharing. For the sake of testing and archiving, am uploading the patch sent via personal e-mail below.
https://www.dropbox.com/s/ahijw8bn7n91nk0/comm_modify_cutoff_multi.tar.gz?dl=0

@herrmartinwagner

This comment has been minimized.

Show comment
Hide comment
@herrmartinwagner

herrmartinwagner Aug 25, 2015

I found an issue with the option I added, in that somehow the normal cutoff option needs to still be specified, else (I think) the communication patterns are not set up correctly and simulations crash. I will (hopefully) get around to modifying and documenting this properly next week.

herrmartinwagner commented Aug 25, 2015

I found an issue with the option I added, in that somehow the normal cutoff option needs to still be specified, else (I think) the communication patterns are not set up correctly and simulations crash. I will (hopefully) get around to modifying and documenting this properly next week.

@akohlmey

This comment has been minimized.

Show comment
Hide comment
@akohlmey

akohlmey Aug 25, 2015

Member

there is another issue. you cannot access atom->ntypes before the simulation box is defined. and your code would cause a memory leak, if comm_modify multi is specified multiple times.
i am thus not directly applying your patch but am currently transferring/reimplementing it piece by piece.
i'll keep an eye out for the user cutoff.

Member

akohlmey commented Aug 25, 2015

there is another issue. you cannot access atom->ntypes before the simulation box is defined. and your code would cause a memory leak, if comm_modify multi is specified multiple times.
i am thus not directly applying your patch but am currently transferring/reimplementing it piece by piece.
i'll keep an eye out for the user cutoff.

@akohlmey

This comment has been minimized.

Show comment
Hide comment
@akohlmey

akohlmey Aug 25, 2015

Member

(re-)implemented in commit 0989c9d in the lammps-icms branch.

can you please (attach as shown above or) send me one or more (simple) test cases, so i can check the implementation before i send it to steve for inclusion into the upstream version. thanks

Member

akohlmey commented Aug 25, 2015

(re-)implemented in commit 0989c9d in the lammps-icms branch.

can you please (attach as shown above or) send me one or more (simple) test cases, so i can check the implementation before i send it to steve for inclusion into the upstream version. thanks

@akohlmey

This comment has been minimized.

Show comment
Hide comment
@akohlmey

akohlmey Aug 26, 2015

Member

i also ran into the single comm cutoff issue in some of my regression tests and think i have found a clean way to handle this now in commit 3748ec5

Member

akohlmey commented Aug 26, 2015

i also ran into the single comm cutoff issue in some of my regression tests and think i have found a clean way to handle this now in commit 3748ec5

@akohlmey

This comment has been minimized.

Show comment
Hide comment
@akohlmey

akohlmey Aug 26, 2015

Member

as of commit 5576ecf this feature is completely implemented into LAMMPS-ICMS and the code forward to steve for inclusion into the upstream version. it passes the integration testing as well https://travis-ci.org/lammps/lammps/builds/77402493 and i've added code to the info command to query the communication mode and cutoff status. closing this issue now.

Member

akohlmey commented Aug 26, 2015

as of commit 5576ecf this feature is completely implemented into LAMMPS-ICMS and the code forward to steve for inclusion into the upstream version. it passes the integration testing as well https://travis-ci.org/lammps/lammps/builds/77402493 and i've added code to the info command to query the communication mode and cutoff status. closing this issue now.

@akohlmey akohlmey closed this Aug 26, 2015

ibaned pushed a commit to ibaned/lammps that referenced this issue Jan 10, 2017

sjplimp pushed a commit that referenced this issue May 7, 2018

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment