update doc pages

This commit is contained in:
Steve Plimpton
2022-06-17 12:11:13 -06:00
parent 378511345a
commit 06ffed1965
3 changed files with 99 additions and 37 deletions

View File

@ -5,9 +5,9 @@ Client/server coupling of two (or more) codes is where one code is the
"client" and sends request messages (data) to one (or more) "server"
code(s). A server responds to each request with a reply message
(data). This enables two (or more) codes to work in tandem to perform
a simulation. LAMMPS can act as either a client or server code; it
does this by using the `MolSSI Driver Interface (MDI) library
<https://molssi-mdi.github.io/MDI_Library/html/index.html>`_,
a simulation. In this context, LAMMPS can act as either a client or
server code. It does this by using the `MolSSI Driver Interface (MDI)
library <https://molssi-mdi.github.io/MDI_Library/html/index.html>`_,
developed by the `Molecular Sciences Software Institute (MolSSI)
<https://molssi.org>`_, which is supported by the :ref:`MDI <PKG-MDI>`
package.
@ -80,11 +80,22 @@ scripts for all of these use cases.
The examples/mdi directory contains Python scripts and LAMMPS input
script which use LAMMPS as either an MDI driver or engine or both.
Three example use cases are provided:
Currently, 5 example use cases are provided:
* Run ab initio MD (AIMD) using 2 instances of LAMMPS, one as driver
and one as an engine. As an engine, LAMMPS is a surrogate for a
quantum code.
* Run ab initio MD (AIMD) using 2 instances of LAMMPS. As a driver
LAMMPS performs the timestepping in either NVE or NPT mode. As an
engine, LAMMPS computes forces and is a surrogate for a quantum
code.
* As a driver, LAMMPS runs an MD simulation. Every N steps it passes
the current snapshot to an MDI engine to evaluate the energy,
virial, and peratom forces. As the engine LAMMPS is a surrogate for
a quantum code.
* As a driver, LAMMPS loops over a series of data files and passes the
configuration to an MDI engine to evaluate the energy, virial, and
peratom forces. As the engine LAMMPS is a surrogate for a quantum
code.
* A Python script driver invokes a sequence of unrelated LAMMPS
calculations. Calculations can be single-point energy/force
@ -97,20 +108,22 @@ Three example use cases are provided:
Note that in any of these example where LAMMPS is used as an engine,
an actual QM code (which supports MDI) could be used in its place,
without modifying other code or scripts, except to specify the name of
the QM code.
without modifying the input scripts or launch commands, except to
specify the name of the QM code.
The examples/mdi/README file explains how to launch both driver and
The examples/mdi/Run.sh file illustrates how to launch both driver and
engine codes so that they communicate using the MDI library via either
MPI or sockets.
MPI or sockets. Or using the engine as a stand-alone code or plugin
library.
-------------
Currently there are two quantum DFT codes which have direct MDI
support, `Quantum ESPRESSO (QE) <https://www.quantum-espresso.org/>`_
and `INQ <https://qsg.llnl.gov/node/101.html>`_. There are also
several QM codes which have indirect support through QCEngine or i-PI.
The former means they require a wrapper program (QCEngine) with MDI
Currently there are at least two quantum DFT codes which have direct
MDI support, `Quantum ESPRESSO (QE)
<https://www.quantum-espresso.org/>`_ and `INQ
<https://qsg.llnl.gov/node/101.html>`_. There are also several QM
codes which have indirect support through QCEngine or i-PI. The
former means they require a wrapper program (QCEngine) with MDI
support which writes/read files to pass data to the quantum code
itself. The list of QCEngine-supported and i-PI-supported quantum
codes is on the `MDI webpage

View File

@ -13,7 +13,7 @@ Syntax
* ID, group-ID are documented in :doc:`fix <fix>` command
* mdi/qm = style name of this fix command
* zero or more keyword/value pairs may be appended
* keyword = *virial* or *add* or *every*
* keyword = *virial* or *add* or *every* or *connect*
.. parsed-literal::
@ -25,6 +25,9 @@ Syntax
no = do not add returned values to LAMMPS quantities
*every* args = Nevery
Nevery = request values from server code once every Nevery steps
*connect* args = *yes* or *no*
yes = perform a one-time connection to the MDI engine code
no = do not perform the connection operation
Examples
""""""""
@ -66,7 +69,7 @@ information about how LAMMPS can operate as either an MDI driver or
engine.
The examples/mdi directory contains input scripts using this fix in
the various use cases listed above. In each case, two instances of
the various use cases dicussed below. In each case, two instances of
LAMMPS are used, once as an MDI driver, once as an MDI engine
(surrogate for a QM code). The examples/mdi/README file explains how
to launch two codes so that they communicate via the MDI library using
@ -104,9 +107,21 @@ during a dynamics run with the current LAMMPS simulation box and
configuration of atoms. The QM code will be called once every
*Nevery* timesteps.
The *connect* keyword determines whether this fix performs a one-time
connection to the QM code. The default is *yes*. The only time a
*no* is needed is if this command is used multiple times in an input
script. E.g. if it used inside a loop which also uses the :doc:`clear
<clear>` command to destroy the system (including any defined fixes).
See the examples/mdi/in.series.driver script as an example of this,
where LAMMPS is using the QM code to compute energy and forces for a
series of system configurations. In this use case *connect no*
is used along with the :doc:`mdi connect and exit <mdi>` command
to one-time initiate/terminate the connection outside the loop.
----------
3 example use cases:
The following 3 example use cases are illustrated in the examples/mdi
directory. See its README file for more details.
(1) To run an ab initio MD (AIMD) dynamics simulation, or an energy
minimization with QM forces, or a multi-replica NEB calculation, use
@ -243,4 +258,4 @@ Default
"""""""
The default for the optional keywords are virial = no, add = yes,
every = 1.
every = 1, connect = yes.

View File

@ -8,9 +8,9 @@ Syntax
.. parsed-literal::
mdi mode args
mdi option args
* mode = *engine* or *plugin*
* option = *engine* or *plugin* or *connect* or *exit*
.. parsed-literal::
@ -22,7 +22,8 @@ Syntax
*infile* value = filename the engine will read at start-up
*extra* value = aditional command-line args to pass to engine library when loaded
*command* value = a LAMMPS input script command to execute
*connect* args = none
*exit* args = none
Examples
""""""""
@ -33,23 +34,15 @@ Examples
mdi plugin lammps mdi "-role ENGINE -name lammps -method LINK" &
infile in.aimd.engine extra "-log log.aimd.engine.plugin" &
command "run 5"
mdi connect
mdi exit
Description
"""""""""""
This command implements two high-level operations within LAMMPS to use
the `MDI Library
<https://molssi-mdi.github.io/MDI_Library/html/index.html>` for
coupling to other codes in a client/server protocol.
The *engine* mode enables LAMMPS to act as an MDI engine (server),
responding to requests from an MDI driver (client) code.
The *plugin* mode enables LAMMPS to act as an MDI driver (client), and
load the MDI engine (server) code as a library plugin. In this case
the MDI engine is a library plugin. It can also be a stand-alone
code, launched separately from LAMMPS, in which case the mdi plugin
command is not used.
This command implements operations within LAMMPS to use the `MDI
Library <https://molssi-mdi.github.io/MDI_Library/html/index.html>`
for coupling to other codes in a client/server protocol.
See the Howto MDI doc page for a discussion of all the different ways
2 or more codes can interact via MDI.
@ -61,6 +54,22 @@ stand-alone code or as a plugin. The README file in that directory
shows how to launch and couple codes for all the 4 usage modes, and so
they communicate via the MDI library using either MPI or sockets.
The scripts in that directory illustrate the use of all the options
for this command.
The *engine* option enables LAMMPS to act as an MDI engine (server),
responding to requests from an MDI driver (client) code.
The *plugin* option enables LAMMPS to act as an MDI driver (client),
and load the MDI engine (server) code as a library plugin. In this
case the MDI engine is a library plugin. An MDI engine can also be a
stand-alone code, launched separately from LAMMPS, in which case the
mdi plugin command is not used.
The *connect* and *exit* options are only used when LAMMPS is acting
as an MDI driver. As explained below, these options are normally not
needed, except for a specific kind of use case.
----------
The *mdi engine* command is used to make LAMMPS operate as an MDI
@ -121,7 +130,7 @@ commands, which are described further below.
* - <PE
- Request potential energy of the system (1 value)
* - <STRESS
- Request stress tensor (virial) of the system (6 values)
- Request symmetric stress tensor (virial) of the system (9 values)
* - >TOLERANCE
- Send 4 tolerance parameters for next MD minimization via OPTG command
* - >TYPES or <TYPES
@ -284,6 +293,31 @@ could specify a filename with multiple LAMMPS commands.
"mdi plugin" command could then load the same library plugin or
a different one if desired.
----------
The *mdi connect* and *mdi exit* commands are only used when LAMMPS is
operating as an MDI driver. And when other LAMMPS command(s) which
send MDI commands and associated data to/from the MDI engine are not
able to initiate and terminate the connection to the engine code.
The only current MDI driver command in LAMMPS is the :doc:`fix mdi/qm
<fix_mdi_qm>` command. If it is only used once in an input script
then it can initiate and terminate the connection. But if it is being
issuedd multiple times, e.g. in a loop that issues a :doc:`clear
<clear>` command, then it cannot initiate/terminate the connection
multiple times. Instead, the *mdi connect* and *mdi exit* commands
should be used outside the loop to intiate/terminate the connection.
See the examples/mdi/in.series.driver script for an example of how
this is done. The LOOP in that script is reading a series of data
file configurations and passing them to an MDI engine (e.g. quantum
code) for enery and force evaluation. A *clear* command inside the
loop wipes out the current system so a new one can be defined. This
operation also destroys all fixes. So the :doc:`fix mdi/qm
<fix_mdi_qm>` command is issued once per loop iteration. Note that it
includes a "connect no" option which disables the initiate/terminate
logic within that fix.
Restrictions
""""""""""""