designate the mode of operation for Grid computing
string, one of "mpi", "hpc", or "local"
host = string
(optional) the hostname or IP address of the hpc mode master server
port = integer
(optional) the port number of the hpc mode master server
mpidll = string
(optional) the name of an alternate MPI dynamic library
mpiwrapper = string
(optional) the name of an alternate wrapper library
numnodes = posint
(optional) number of nodes in the local grid
The Setup command allows you to designate the mode of operation for parallel computation using the Grid package. There are three available modes:
"local" - This mode provides support for multi-process parallel computation on a single machine. This is the default setup, and is available as part of a regular Maple installation.
"hpc" - This mode provides support for multi-process parallel computations distributed across a heterogeneous network. Job scheduling and control is monitored by daemon processes running on each node in the grid. This mode can only be enabled if the Grid Computing Toolbox add-on has been installed.
"mpi" - This mode provides support for multi-process parallel computations on a cluster using standard Message Passing Interface (MPI) protocols. Job scheduling is batch oriented and controlled by the operating system. This mode can only be enabled if the Grid Computing Toolbox has been installed.
The API for distributed remote-machine computation in the Grid package is mirrored in all modes. This makes it easy to prototype your algorithm in one mode and deploy it in another.
Setup does not need to (and should not) be called within the Maple code that defines a parallel computation. Setup only needs to be called prior to interactively querying the grid status or launching a parallel job into the grid.
Only one mode can be chosen in a given session. Calling Setup will cause the specified mode to be initialized. When using either hpc or mpi modes, the Setup command must be called prior to using any other function in the Grid package. Otherwise the local mode will be initialized, causing the other modes to become unavailable.
The host and port parameters tell Maple where one of the running Grid servers is located. When launching a computation, the designated host will act as the master server. In mpi mode, host is only required when the mpi cluster does not include the local machine (that is, the machine launching the job is not one of the compute nodes). In hpc mode, both host and port parameters are required. Additionally, in hpc mode, if you are running on multiple CPUs of the same machine, port needs to point at the base port (i.e. the port used by CPU 0).
The interactive Grid worksheet provides a gui interface for configuring hpc mode and launching jobs. The Setup options are configured by filling in the default node and default port fields and clicking the "Verify Grid Status" button.
The mpidll option allows you to specify an alternate .dll or shared library when in "mpi" mode. This option is used in the form mpidll = "msmpi" (Windows) or mpidll="mpich" (Linux). If the file is in the system path, you can specify only the basename of the file. If the file is not in the system path, you must specify the path and basename of the file.
The mpiwrapper option allows you to specify the name of the Maple wrapper .dll file in "mpi" mode. By default, the maplempi.dll file that is included with the Grid Computing Toolbox is used. Specify a value for this parameter only if you want to use a .dll file other than maplempi.dll. This is the interface layer between the Maple and the external mpi implementation, that takes care of data structure translations. In most cases the provided mpiwrapper implementation will be compatible with any mpi library.
The numnodes option allows you to specify the number of nodes to be used in subsequent computations. This option is only available in "local" mode.
The Grid[Setup] command was introduced in Maple 15.
For more information on Maple 15 changes, see Updates in Maple 15.
The Grid[Setup] command was updated in Maple 2015.
The numnodes parameter was introduced in Maple 2015.
For more information on Maple 2015 changes, see Updates in Maple 2015.
Download Help Document