This chapter is intended to get you up and running with VMUSBSpecTcl and custom configured SpecTcl's that don't extend the SpecTcl command set. I'll describe:
What mpiSpecTcl is.
How to know if your version of SpecTcl supports MPI Parallelism
Environment variables you will need to set to run MPI mpiSpecTcl
How to run mpiSpecTcl with parallel event processing pipelines
Debugging mpiSpecTcl tailored code.
mpiSpecTcl can be used with SpecTcl version 7.0 or later. It supports scalable performance by parallelizing the event processing pipeline. When you run mpiSpecTcl, events from the data source are distributed between one or more worker processes. Each of these processes passes events through its copy of the event processing pipeline and sends the parameters it produces to another process running the event sink pipline (histogramer).
Since in many nuclear science experminents, each event can be analyzed independently of all other events, running more than one worker process will scale up the rate at which events can be processed until data transfer limits this scaling.
For a more detailed description of how mpiSpecTcl works, see: How mpiSpecTcl works in parallel mode.