Cuda package error

LAMMPS LAMMPS Mailing List Mirror
1

Hi lammps users,

I met an error when using cuda package. Do you have any suggestions? Thanks in advance.

CUDA: VerletCuda::setup: Allocate memory on device for maximum of 64644 atoms…

CUDA: Using precision: Global: 8 X: 8 V: 8 F: 8 PPPM: 8

Setting up run …

CUDA: VerletCuda::setup: Upload data…

CUDA: Total Device Memory useage post setup: 418.695312 MB

Memory usage per processor = 53.9738 Mbytes

Step Temp TotEng Dt Elapsed 4

0 299.58522 -268679.44 0.001 0 0

WARNING: # CUDA: You asked for a Verlet integration using Cuda, but several fixes have not yet been ported to Cuda.

This can cause a severe speed penalty due to frequent data synchronization between host and GPU. (verlet_cuda.cpp:553)

[mez508:27117] *** Process received signal ***

[mez508:27117] Signal: Segmentation fault (11)

[mez508:27117] Signal code: Invalid permissions (2)

[mez508:27117] Failing at address: 0x3ae1f0200

[mez508:27117] [ 0] /lib64/libpthread.so.0() [0x3c5760f4a0]

[mez508:27117] [ 1] lmp_linux(_ZN9LAMMPS_NS13ComputePEAtom19unpack_reverse_commEiPiPd+0x23) [0x6219e1]

[mez508:27117] [ 2] lmp_linux(_ZN9LAMMPS_NS4Comm20reverse_comm_computeEPNS_7ComputeE+0x132) [0x607f2a]

[mez508:27117] [ 3] lmp_linux(_ZN9LAMMPS_NS13ComputePEAtom15compute_peratomEv+0x2ed) [0x621d31]

[mez508:27117] [ 4] lmp_linux(_ZN9LAMMPS_NS10FixAveAtom11end_of_stepEv+0x27d) [0x6c00af]

[mez508:27117] [ 5] lmp_linux(_ZN9LAMMPS_NS10ModifyCuda11end_of_stepEv+0x153) [0x7e26dd]

[mez508:27117] [ 6] lmp_linux(_ZN9LAMMPS_NS10VerletCuda3runEi+0x1534) [0xb4c0e4]

[mez508:27117] [ 7] lmp_linux(_ZN9LAMMPS_NS3Run7commandEiPPc+0x752) [0xb25224]

[mez508:27117] [ 8] lmp_linux(_ZN9LAMMPS_NS5Input15execute_commandEv+0xe38) [0x7be24e]

[mez508:27117] [ 9] lmp_linux(_ZN9LAMMPS_NS5Input4fileEv+0x28b) [0x7be75f]

[mez508:27117] [10] lmp_linux(main+0x5b) [0x7c8fde]

[mez508:27117] [11] /lib64/libc.so.6(__libc_start_main+0xfd) [0x33a121ecdd]

[mez508:27117] [12] lmp_linux() [0x574089]

[mez508:27117] *** End of error message ***

Segmentation fault (core dumped)

Best Regards

Richard

2

If you post a (simple) input script that illustrates
the problem, Christian can take a look at it.

Steve

3

If you post a (simple) input script that illustrates the problem, Christian
can take a look at it.

Steve

4

Hi Steve,

Thanks for your suggestion. It would be very nice if Christian could help.
The problem is that the simulation can run although there is the WARNING (#
CUDA: You asked for a Verlet integration using Cuda, but several fixes have
not yet been ported to Cuda.) but it will stop after running sometime, maybe
1000000 timesteps or longer.
We use a -sf cuda command to start gpu. The input is as follow:

#initialization
units metal
boundary p p p
dimension 3
echo screen
newton on
atom_style atomic
log log.si_temp

#Atom definitions ####################################
read_data data.s4_238

#force field setting
pair_style sw
pair_coeff * * Si.sw Si
neighbor 1.0 bin
neigh_modify delay 0 every 20 check no
mass 1 28.0
thermo 700

#npt equillibration
velocity all create 300 429349 dist gaussian
fix 1 all npt temp 300 300 10 iso 0 0 100

timestep 0.001
run 300000

#nve equillibration
unfix 1
fix 2 all nve
restart 400000 tmp.restart
run 100000
restart 0

#muller-plathe thermal conductivity

#Temp gradient
compute ke all ke/atom
compute pe all pe/atom
compute cord all property/atom xu yu zu
variable temp atom c_ke/(1.5*8.617*10^-5)
reset_timestep 0

#Flux
fix 4 all thermal/conductivity 700 z 100 swap 1
#dump 1 all xyz 100 dump.xyz

thermo_style custom step temp etotal dt elapsed f_4

log log.si_flux

fix 5 all ave/spatial 10 100000 1000000 z lower 0.01 v_temp file
tmp2.profile units reduced

#Output every 1000000 average ke pe temp
fix pu1 all ave/atom 10 100000 1000000 c_cord[1] c_cord[2] c_cord[3]
fix ke1 all ave/atom 10 100000 1000000 c_ke
fix pe1 all ave/atom 10 100000 1000000 c_pe
fix temp1 all ave/atom 10 100000 1000000 v_temp
dump ke1 all custom 1000000 dump.ke1 id type f_pu1[1] f_pu1[2] f_pu1[3]
f_ke1 f_pe1 f_temp1
dump_modify ke1 sort id

restart 1000000 tmp.restart
run 4000000

5

I'm CCing Christian on this message, to see
if he has ideas.

Steve