Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Mon, 13 Aug 2001 17:40:40 -0500

Rolf Keitel wrote:
> 
> b) RT Linux
>   How viable will it be for EPICS?
>   On what time-scale?
>   Who has any experience?

The 3.14 work is not porting EPICS to RTLinux (which is a micro-kernel
that runs Linux in its idle task), but to standard linux, although KEK
were looking at developing a port to the L4-linux real-time microkernel
IIRC.  We should be able to run EPICS as a real-time scheduled process on
a normal Linux system, although I'm not sure that the task priority stuff
is making use of this yet.  I have seen some statistics that imply that
while the minimum interrupt response time of Linux is slightly larger than
vxWorks, the mean response time is actually lower because the vxWorks
network layer adds a lot of variability (disabling interrupts for
relatively long periods).

The main problems will be I/O as device support has to communicate to
hardware through the user/kernel space divide, although for some
I/O-mapped hardware without interrupts it is possible to avoid the extra
layer and do everything from user space.  APS has started developing some
experience in writing kernel drivers, but we haven't released anything yet
and they're always going to be harder to do than vxWorks ones.  There are
some devices that only do network I/O though (the HP LAN/GPIB box is the
best example for which I believe the 3.14 software is already available)
and for these Linux would seem to be ideal.

- Andrew
-- 
The world is such a cheerful place when viewed from upside-down
It makes a rise of every fall, a smile of every frown

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System