rhl@computing-maths.cardiff.ac.uk (Robert Hartill) (02/05/91)
Background:
I'm designing a real-time operating system for an Intel 80386, which will
be used as a front-end-processor for a PC.
I am interested in designing the OS to use the OBIOS standard. But need to
clarify a few points.
(i) My OS could well find itself connected to say 256 serial devices which
are identical (except for the port number). The code for a port could
typically range from 4k to 16k in size.
I don't want to replicate the code 256 times.
The OBIOS spec implies (to me) that
(a) I need 1 set of routines for each port, if they
are to *run* concurrently
---> too much code.
or
(b) 1 set of non-reentrant routines shared by all,
allowing just 1 port to be serviced at a time.
---> for 256 ports, real-time response is lost.
I do hope that I've misunderstood the spec...
(ii) Is there any reason why I can't define my *client* to break a few rules,
i.e. the OBIOS routines are supposed to be non-reentrant, if I write
OBIOS routines which *just happen* to be reentrant, and allow my client
to assume reentrancy ONLY for this particular set of OBIOS routines, is
there a problem ?. The routines would conform to the OBIOS wouldn't they ?
------
Reference : The OBIOS is a proposed interface standard for a basic I/O system,
specifically in real-time environments.
Contact buz@ready.com for more details (i hope he doesn't mind me doing this).
------
Thanks in advance,
Robert.
:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
:: Robert Hartill, Dept Of Computing Mathematics, UWCC, Cardiff, UK. ::
:: e-mail : rhl@cm.cf.ac.uk Fax : 0222 371921 ::
:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
A radical is a man with both feet firmly planted in the air.
-Franklin D. Roosevelt 1939.
-------------------------------------------------------------------------------