LAMMPS WWW Site - LAMMPS Documentation - LAMMPS Mailing List Archives
Re: [lammps-users] problems of compiling with multiple gpu types
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [lammps-users] problems of compiling with multiple gpu types


From: "Moore, Stan" <stamoor@...3...>
Date: Tue, 11 Jul 2017 19:26:09 +0000

Yes, just use separate Makefiles, one Makefile for each GPU type. That will give separate executables that are built for each GPU type. I use this method all the time.

 

Stan

 

From: Chiang Yi [mailto:ckpccl1@...24...]
Sent: Monday, July 10, 2017 11:42 PM
To: lammps-users@lists.sourceforge.net
Subject: [EXTERNAL] Re: [lammps-users] problems of compiling with multiple gpu types

 

Dear lammps-users:

 

Thanks for your prompt reply.

But it didn't work with mpirun command, too. (We still couldn't compile with K40c.)

Is there any way to compile just with one gpu?

 

Best,

Chiang, Yi

 

 

寄件者: Steve Plimpton
傳送時間: 2017710 下午 10:09
收件者: Chiang Yi
副本: lammps-users@lists.sourceforge.net
主旨: Re: [lammps-users] problems of compiling with multiple gpu types

 

I suggest you first focus

on compiling and running LAMMPS successfully

on each GPU.  Then you have 2 executables,

each optimized for a different GPU, or even
built with a different Cuda lib.

Commands like mpirun, mpiexec

have ways to launch different executables on

different nodes, to run a single calculation on

a heterogeneous machine.

Steve

 

On Mon, Jul 10, 2017 at 3:08 AM, Chiang Yi <ckpccl1@...24...> wrote:

Dear lammps-users:

 

There are two different gpu types in our super computer:

 

Tesla P100-PCIE-16GB (Compute capability: 6)

Tesla K40c (Compute capability: 3.5)

 

        Device 0: Tesla P100-PCIE-16GB

       Device 1: Tesla P100-PCIE-16GB

        Device 2: Tesla K40c

        Device 3: Tesla K40c

        Device 4: Tesla P100-PCIE-16GB

        Device 5: Tesla P100-PCIE-16GB

        Device 6: Tesla K40c

        Device 7: Tesla K40c

        Device 8: Tesla K40c

 

First time I chose CUDA_arch = sm_60. I found that compiling didnt work with gpu 3 (Tesla K40c), but Its OK with gpu 0 (Tesla P100).

However, If I changed CUDA_arch = sm_35, It couldnt work with both P100 and K40c. The error messages are like this:

 

$ ./lmp_gpu -sf gpu -pk gpu 1 gpuID 3 3 -in in.lj

LAMMPS (31 Mar 2017)

ERROR: GPU library not compiled for this accelerator (../gpu_extra.h:40)

Last command: package gpu 1

Cuda driver error 4 in call at file 'geryon/nvd_device.h' in line 124.

 

So, I would like to ask:

  1. Does lammps not support different gpu types in one computer?
  2. Is it possible to pass the gpu 0, just compile with gpu 2?
  3. Or If there is any possible mistake I didnt notice?

 

Thank you.

 

Best,

Chiang, Yi


------------------------------------------------------------------------------
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