threads.9

CNC 的开放码,EMC2 V2.2.8版

.TH THREADS "9" "2007-01-16" "EMC Documentation" "HAL Component"
.de TQ
.br
.ns
.TP \\$1
..

.SH NAME
threads \- creates hard realtime HAL threads
.SH SYNOPSIS
\fBloadrt threads name1=\fIname\fB period1=\fIperiod\fR [\fBfp1=\fR<\fB0\fR|\fB1\fR>] [<thread-2-info>] [<thread-3-info>]

.SH DESCRIPTION
\fBthreads\fR is used to create hard realtime threads which can execute
HAL functions at specific intervals.  It is not a true HAL component, in
that it does not export any functions, pins, or parameters of its own.
Once it has created one or more threads, the threads stand alone, and
the \fBthreads\fR component can be unloaded without affecting them.  In
fact, it can be unloaded and then reloaded to create additional threads,
as many times as needed.

.P
\fBthreads\fR can create up to three realtime threads.  Threads must be
created in order, from fastest to slowest.  Each thread is specified by
three arguments.  \fBname1\fR is used to specify the name of the first
thread (thread 1).  \fBperiod1\fR is used to specify the period of thread
1 in nanoseconds.  Both \fIname\fR and \fIperiod\fR are required.  The 
third argument, \fBfp1\fR is optional, and is used to specify if thread
1 will be used to execute floating  point code.  If not specified, it
defaults to \fB1\fR, which means that the thread will support floating
point.  Specify \fB0\fR to disable floating point support, which saves
a small amount of execution time by not saving the FPU context.  For
additional threads, \fBname2\fR, \fBperiod2\fR, \fBfp2\fR, \fBname3\fR,
\fBperiod3\fR, and \fBfp3\fR work exactly the same.  If more than three
threads are needed, unload threads, then reload it to create more threads.

.SH FUNCTIONS
.P
None

.SH PINS
.P
None

.SH PARAMETERS
.P
None

.SH BUGS
.P
The existence of \fBthreads\fR might be considered a bug.  Ideally, creation
and deletion of threads would be done directly with \fBhalcmd\fR commands,
such as "\fBnewthread \fIname period\fR", "\fBdelthread \fIname\fR", or similar.
However, limitations in the current HAL implementation require thread creation
to take place in kernel space, and loading a component is the most straightforward
way to do that.