MathLink

Controlling Kernels

Start a local kernel (the command to do so may depend on the operating system):

Start a remote kernel with rsh or ssh:

remote=LinkLaunch["ssh urania math -MathLink -noinit"]

You can also start the remote kernel manually with math -MathLink -linkcreate and then connect to the resulting port using LinkConnect.

Swallowing initial line-number packet.

Avoid blocking:

Sending expressions for evaluation:

Polling for result

Dealing with other packets

Ending session

Lining up Workers

Higher-level routines for starting/closing MathLink connections (OS dependent).

LaunchSlave["host", "cmdTemplate", options...]

ConnectSlave["port@host", options...]

cmdTemplate is a string used in StringForm, `1` will be replaced by the hostname. The second form connects to a waiting kernel; this form can be used if security prevents you from executing a remote command directly.

Start local kernel:

LaunchSlave["localhost", $mathkernel]

List of all available remote kernels

Sending and Receiving Expressions

Basic remote evaluation (blocking):

SetAttributes[RemoteEvaluate, HoldRest]

RemoteEvaluate[expr_, link_LinkObject]:=
    Module[{},
        Send[link, expr];
        Receive[link]
    ]

RemoteEvaluate[cmd] sends cmd to all kernels in parallel, gives list of all results.

Here we gather information about the remote kernels currently running.

Document converted by Mathematica of Wolfram Research