LAMMPS WWW Site - LAMMPS Documentation - LAMMPS Mailing List Archives
Re: [lammps-users] [EXTERNAL] Re: [Kokkos] inconsistency in rmass across pair/fix/compute
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [lammps-users] [EXTERNAL] Re: [Kokkos] inconsistency in rmass across pair/fix/compute


From: "Moore, Stan" <stamoor@...3...>
Date: Tue, 15 May 2018 22:14:44 +0000

I’ve noticed this too, but never investigated. I will take a look.

 

Thanks,

 

Stan

 

From: Steve Plimpton [mailto:sjplimp@...24...]
Sent: Monday, May 14, 2018 7:39 AM
To: Denis Taniguchi <Denis.Taniguchi@...7548...>; Moore, Stan <stamoor@...3...>
Cc: LAMMPS Users Mailing List <lammps-users@lists.sourceforge.net>
Subject: [EXTERNAL] Re: [lammps-users] [Kokkos] inconsistency in rmass across pair/fix/compute

 

This is a Q for Stan (CCd)

Steve

 

On Fri, May 11, 2018 at 4:51 AM, Denis Taniguchi <Denis.Taniguchi@...7548...> wrote:

Hi all,

 

I noticed that sometimes rmass is just treated as a double* and sometimes as a AT::t_float_1d_randomread.

When treated as double* it takes the content of atomKK->rmass which is a pointer to host data. In FixNVEKokkos it stores rmass as double* but uses it inside a functor (FixNVEKokkosInitialIntegrateFunctor calling initial_integrate_rmass_item) which is dispatched by parallel_for.

Will the device kernel automagically copy the content to device memory?

Regards,

 

Denis

 


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
lammps-users mailing list
lammps-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/lammps-users