There are restrictions on the number of atoms per GPU when running -the gayberne/gpu pair style. Upon initialization of the gayberne/gpu -pair style, LAMMPS will reserve memory for 64K atoms per GPU or 70% -of each card's GPU memory, whichever value is limiting. If the GPU -library is compiled for double precision, the maximum number of -atoms per GPU is 32K. When running a periodic system and/or in -parallel, this maximum atom count includes ghost atoms. -
-The value of 70% can be changed by editing the PERCENT_GPU_MEMORY -definition in the appopriate lammps/lib/gpu source file. For the -gayberne/gpu pair style, the value of 64K cannot be increased and is -the maximum number of atoms allowed per GPU. By default, enough -memory to store at least the maximum number of neighbors per atom is -reserved on the GPU, which is set by the -neigh_modify one command. The default value of -2000 will be very high for many cases. If memory on the graphics card -is limiting, the number of atoms allowed can be increased by -decreasing the maximum number of neighbors. For example placing, -
-neigh_modify one 100 --
in the input script will decrease the maximum number of neighbors per -atom to 100, allowing more atoms to be run on the GPU. -
The newton pair setting must be off.
-You should use the neigh_modify one command and -set its value to something close (but slightly larger) than the number -of pairwise neighbors/atom you expect to have in your model. This is -a function of the pairwise cutoff. Note that the default for this -setting is 2000, which is much larger than most models need. Unlike -neighbor lists in LAMMPS itself, the GPU version of this pair style -uses that setting to allocate memory on the GPU for neighbor -information. If the setting is too large, it will limit the number of -atoms that can be stored on the GPU. -
Mixing, shift, table, tail correction, restart, rRESPA info: diff -Naur lammps-7Mar10/doc/pair_gayberne.txt lammps-8Mar10/doc/pair_gayberne.txt --- lammps-7Mar10/doc/pair_gayberne.txt 2009-10-29 09:41:27.000000000 -0600 +++ lammps-8Mar10/doc/pair_gayberne.txt 2010-03-05 11:42:10.000000000 -0700 @@ -166,16 +166,6 @@ The "newton pair"_newton.html setting must be {off}. -You should use the "neigh_modify one"_neigh_modify.html command and -set its value to something close (but slightly larger) than the number -of pairwise neighbors/atom you expect to have in your model. This is -a function of the pairwise cutoff. Note that the default for this -setting is 2000, which is much larger than most models need. Unlike -neighbor lists in LAMMPS itself, the GPU version of this pair style -uses that setting to allocate memory on the GPU for neighbor -information. If the setting is too large, it will limit the number of -atoms that can be stored on the GPU. - :line [Mixing, shift, table, tail correction, restart, rRESPA info]: diff -Naur lammps-7Mar10/doc/pair_lj.html lammps-8Mar10/doc/pair_lj.html --- lammps-7Mar10/doc/pair_lj.html 2009-08-13 11:31:27.000000000 -0600 +++ lammps-8Mar10/doc/pair_lj.html 2010-03-05 11:42:10.000000000 -0700 @@ -209,16 +209,6 @@
The newton pair setting must be off.
-You should use the neigh_modify one command and -set its value to something close (but slightly larger) than the number -of pairwise neighbors/atom you expect to have in your model. This is -a function of the pairwise cutoff. Note that the default for this -setting is 2000, which is much larger than most models need. Unlike -neighbor lists in LAMMPS itself, the GPU version of this pair style -uses that setting to allocate memory on the GPU for neighbor -information. If the setting is too large, it will limit the number of -atoms that can be stored on the GPU. -
Mixing, shift, table, tail correction, restart, rRESPA info: diff -Naur lammps-7Mar10/doc/pair_lj.txt lammps-8Mar10/doc/pair_lj.txt --- lammps-7Mar10/doc/pair_lj.txt 2009-08-13 11:31:27.000000000 -0600 +++ lammps-8Mar10/doc/pair_lj.txt 2010-03-05 11:42:10.000000000 -0700 @@ -200,16 +200,6 @@ The "newton pair"_newton.html setting must be {off}. -You should use the "neigh_modify one"_neigh_modify.html command and -set its value to something close (but slightly larger) than the number -of pairwise neighbors/atom you expect to have in your model. This is -a function of the pairwise cutoff. Note that the default for this -setting is 2000, which is much larger than most models need. Unlike -neighbor lists in LAMMPS itself, the GPU version of this pair style -uses that setting to allocate memory on the GPU for neighbor -information. If the setting is too large, it will limit the number of -atoms that can be stored on the GPU. - :line [Mixing, shift, table, tail correction, restart, rRESPA info]: diff -Naur lammps-7Mar10/src/fix_indent.cpp lammps-8Mar10/src/fix_indent.cpp --- lammps-7Mar10/src/fix_indent.cpp 2010-03-03 11:40:02.000000000 -0700 +++ lammps-8Mar10/src/fix_indent.cpp 2010-03-08 11:22:06.000000000 -0700 @@ -193,7 +193,7 @@ domain->remap(ctr); double radius; - if (rstr) radius = input->variable->compute_equal(zvar); + if (rstr) radius = input->variable->compute_equal(rvar); else radius = rvalue; double **x = atom->x; diff -Naur lammps-7Mar10/src/version.h lammps-8Mar10/src/version.h --- lammps-7Mar10/src/version.h 2010-03-05 07:43:24.000000000 -0700 +++ lammps-8Mar10/src/version.h 2010-03-08 11:23:36.000000000 -0700 @@ -1 +1 @@ -#define LAMMPS_VERSION "7 Mar 2010" +#define LAMMPS_VERSION "8 Mar 2010"