ICODE-ADC/lammps
4
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

git-svn-id: svn://svn.icms.temple.edu/lammps-ro/trunk@8396 f3b2605a-c512-4ea7-a41b-209d697bcdaa

Browse Source
This commit is contained in:
sjplimp
2012-06-25 22:22:46 +00:00
parent 66c5617afb
commit 05e044abf4
10 changed files with 104 additions and 79 deletions
Download Patch File Download Diff File Expand all files Collapse all files

70
doc/Section_start.html
View File

@ -223,11 +223,14 @@ Restrictions section of the <A HREF = "dump.html">dump</A> command for details.
</P>
<P>Use at most one of the -DLAMMPS_SMALLBIG, -DLAMMPS_BIGBIG,
-D-DLAMMPS_SMALLSMALL settings. The default is -DLAMMPS_SMALLBIG.
These refer to use of 4-byte (small) vs 8-byte (big) integers within
LAMMPS, as described in src/lmptype.h. The only reason to use the
BIGBIG setting is to enable simulation of huge molecular systems with
more than 2 billion atoms. The only reason to use the SMALLSMALL
setting is if your machine does not support 64-bit integers.
These settings refer to use of 4-byte (small) vs 8-byte (big) integers
within LAMMPS, as specified in src/lmptype.h. The only reason to use
the BIGBIG setting is to enable simulation of huge molecular systems
with more than 2 billion atoms or to allow moving atoms to wrap back
through a periodic box more than 512 times. The only reason to use
the SMALLSMALL setting is if your machine does not support 64-bit
integers. See the <A HREF = "#start_2_4">Additional build tips</A> section below
for more details.
</P>
<P>The -DLAMMPS_LONGLONG_TO_LONG setting may be needed if your system or
MPI version does not recognize "long long" data types. In this case a
@ -434,41 +437,46 @@ Obj_target where it stores the system-specific *.o files.
files created when LAMMPS is built, for either all builds or for a
particular machine.
</P>
<P>(3) Changing the size limits in src/lmptype.h
<P>(3) Changing the LAMMPS size limits via -DLAMMPS_SMALLBIG or
-DLAMMPS_BIBIG or -DLAMMPS_SMALLSMALL
</P>
<P>If you are running a very large problem (billions of atoms or more)
and get a run-time error about the system being too big, either on a
per-processor basis or in total size, then you may need to change one
or more settings in src/lmptype.h and re-compile LAMMPS.
<P>As explained above, any of these 3 settings can be specified on the
LMP_INC line in your low-level src/MAKE/Makefile.foo.
</P>
<P>As the documentation in that file explains, you have basically
two choices to make:
<P>The default is -DLAMMPS_SMALLBIG which allows for systems with up to
2^63 atoms and timesteps (about 9 billion billion). The atom limit is
for atomic systems that do not require atom IDs. For molecular
models, which require atom IDs, the limit is 2^31 atoms (about 2
billion). With this setting, image flags are stored in 32-bit
integers, which means for 3 dimensions that atoms can only wrap around
a periodic box at most 512 times. If atoms move through the periodic
box more than this limit, the image flags will "roll over", e.g. from
511 to -512, which can cause diagnostics like the mean-squared
displacement, as calculated by the <A HREF = "compute_msd.html">compute msd</A>
command, to be faulty.
</P>
<UL><LI>set the data type size of integer atom IDs to 4 or 8 bytes
<LI>set the data type size of integers that store the total system size to 4 or 8 bytes
</UL>
<P>The default for atom IDs is 4-byte integers since there is a memory
and communication cost for 8-byte integers. Non-molecular problems do
not need atom IDs so this does not restrict their size. Molecular
problems (which use IDs to define molecular topology), are limited to
about 2 billion atoms (2^31) with 4-byte IDs. With 8-byte IDs they
are effectively unlimited in size (2^63).
<P>To allow for larger molecular systems or larger image flags, compile
with -DLAMMPS_BIGBIG. This enables molecular systems with up to 2^63
atoms (about 9 billion billion). And image flags will not "roll over"
until they reach 2^20 = 1048576.
</P>
<P>The default for total system size quantities (like the number of atoms
or timesteps) is 8-byte integers by default which is effectively
unlimited in size (2^63). If your system or MPI implementation does
not support 8-byte integers, an error will be generated, and you will
need to set "bigint" to 4-byte integers. This restricts your total
system size to about 2 billion atoms or timesteps (2^31).
<P>IMPORTANT NOTE: As of 6/2012, the BIGBIG setting does not yet enable
molecular systems to grow as large as 2^63. Only the image flag roll
over is currently affected by this compile option.
</P>
<P>If your system does not support 8-byte integers, you will need to
compile with the -DLAMMPS_SMALLSMALL setting. This will restrict your
total number of atoms (for atomic or molecular models) and timesteps
to 2^31 (about 2 billion). Image flags will roll over at 2^9 = 512.
</P>
<P>Note that in src/lmptype.h there are also settings for the MPI data
types associated with the integers that store atom IDs and total
system sizes, which need to be set consistent with the associated C
data types.
system sizes. These need to be consistent with the associated C data
types, or else LAMMPS will generate a run-time error.
</P>
<P>In all cases, the size of problem that can be run on a per-processor
basis is limited by 4-byte integer storage to about 2 billion atoms
per processor (2^31), which should not normally be a restriction since
basis is limited by 4-byte integer storage to 2^31 atoms per processor
(about 2 billion). This should not normally be a restriction since
such a problem would have a huge per-processor memory footprint due to
neighbor lists and would run very slowly in terms of CPU
secs/timestep.