LAMMPS WWW Site - LAMMPS Documentation - LAMMPS Mailing List Archives
Re: [lammps-users] How to accelerate the calculation speed by using existing hardware ?
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [lammps-users] How to accelerate the calculation speed by using existing hardware ?


From: Axel Kohlmeyer <akohlmey@...24...>
Date: Tue, 11 Jul 2017 08:44:57 -0400

On Mon, Jul 10, 2017 at 11:09 PM, 王淑婷 <wangshuting@...6572...> wrote:
> Dear lammps-users,
>
> Thanks a lot to Axel Kohlmeyer. It's very kind of you to reply me. I agree
> with your explanation about slow calculation speed caused by hybrid pair
> potential. Your suggestion about how to accelerate the calculation speed is
> very useful.
>
> I have a new question now. I use the command "mpirun -n 1 lmp_mpi -sf gpu
> -pk gpu 1 < in.W " to calculate the thermal conductivity. But it indicates
> the ERROR : Cannot open EAM potential file eam/alloy. Here is my input
> script blew. Could anyone give me some suggestions? I don't know what's
> wrong with my input script. Thank you for your help in advance.

the error message is self-explanatory. think about it.

axel.


>
> Best wishes.
> Shuting Wang
>
> ############################################
> #-----------------------------Initialization
>
> variable                      L equal 200/4.07
> variable                      W equal 30/4.07
> variable                      H_0 equal 50/4.07
> variable                      H_1 equal 20/4.07
> variable                      H_2 equal 30/4.07
> variable HOT_R equal round($L*0.125)
> variable HOT_L equal ${HOT_R}-1
> variable COLD_L equal round($L*0.875)
> variable COLD_R equal ${COLD_L}+1
>
> units                metal
> dimension                     3
> boundary                      p p p
> atom_style                    atomic
>
> #------------------------------Atom Defination
>
> lattice                       fcc 3.157
> region                        W block 0 $L 0 $W ${H_1} ${H_2}
> region                        total block 0 $L 0 $W 0 ${H_0}
> create_box                    1 total
> create_atoms                  1 region W
>
> region hot block ${HOT_L} ${HOT_R} 0 $W 0 ${H_0}
> region cold block ${COLD_L} ${COLD_R} 0 $W 0 ${H_0}
> group hot region hot
> group cold region cold
>
> mass                          1 196.96654
>
> #Force-field parameters
> #B=1,N=1
>
> pair_style      eam/alloy
> pair_coeff      * * eam/alloy W.eam.alloy
>
> neighbor                      2.0 nsq
> neigh_modify                  delay 0 every 10
>
> #---------------------------------Define Setting
>
> variable                      ke equal ke
> variable                      pe equal pe
> variable                      press equal press
> variable                      vol equal vol
> variable                      etotal equal etotal
> variable                      temp equal temp
> compute                       tot_temp all temp
> compute                       myKE all ke/atom
> variable                      temp1 atom c_myKE/0.0001292355
> compute                       hot_temp all temp/region hot
> compute                       cold_temp all temp/region cold
>
> min_style cg
> minimize 1.0e-4 1.0e-6 100 1000
> reset_timestep 0
>
>
>> -----原始邮件-----
>> 发件人: "Axel Kohlmeyer" <akohlmey@...24...>
>> 发送时间: 2017年7月10日 星期一
>> 收件人: "王淑婷" <wangshuting@...6572...>
>> 抄送: "LAMMPS Users Mailing List" <lammps-users@...431...ceforge.net>
>> 主题: Re: [lammps-users] How to accelerate the calculation speed by using
>> existing hardware ?
>
>>
>> On Sun, Jul 9, 2017 at 10:40 PM, 王淑婷 <wangshuting@...6572...> wrote:
>> > Dear lammps-users,
>> >
>> > Now I have several intel xeon processors e5 v4 with 280 cores. I find
>> > that
>> > the calculation speed is the fastest by using 54 cores when I use hybrid
>> > force field.
>> >
>> > I still have a NVIDIA Tesla P100 GPU. When the force field is tersoff,
>> > the
>> > calculation speed tested by one GPU core and one CPU core is four times
>> > higher than the 28 CPU cores. But when the force field is hybrid, the
>> > calculation speed tested by one GPU and one CPU core is two times lower
>> > than
>> > the 28 CPU cores.
>>
>> that is because with a hybrid pair potential, you must compute the
>> neighbor lists on the CPU and then transfer it to the GPU, even if all
>> individual pair styles are GPU accelerated.
>> that will cause a significant slowdown, in your case, since you are
>> using only one CPU. you can try to minimize this by using multiple
>> CPUs with the same GPU, but the additional overhead of computing
>> neighbor lists on the CPU and transferring significant additional data
>> to the GPU cannot be completely avoided.
>>
>> > I don't know how to use existing hardware to accelerate the calculation
>> > speed when my force field is hybrid. It's OK if I should buy some more
>> > hardware. Could you please give me some suggestions? Thank you for your
>> > help
>> > in advance.
>>
>> the best way to get the most performance with the GPU package, is to
>> avoid a hybrid pair style.
>> there are also many other, simulation specific factors, that can
>> affect performance, where it is not possible to give general advice.
>>
>> axel.
>>
>> >
>> > Yours sincerely
>> > Shuting Wang
>> >
>> >
>> >
>> >
>> >
>> >
>> >
>> > ------------------------------------------------------------------------------
>> > 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
>> >
>>
>>
>>
>> --
>> Dr. Axel Kohlmeyer  akohlmey@...24...  http://goo.gl/1wk0
>> College of Science & Technology, Temple University, Philadelphia PA, USA
>> International Centre for Theoretical Physics, Trieste. Italy.
>
>
>
>
> ------------------------------------------------------------------------------
> 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
>



-- 
Dr. Axel Kohlmeyer  akohlmey@...24...  http://goo.gl/1wk0
College of Science & Technology, Temple University, Philadelphia PA, USA
International Centre for Theoretical Physics, Trieste. Italy.