🚀 go-pugleaf

In <c1.01.2Gw8XV$17q@cretin.summerfield.network>, summer@summerfield.network (John Summerfield) writes:
>In <5qvvfa$9h2@news.nternet.net>, ossi@iname.com (Doug Toltzman) writes:

>]You may be missing a critical point.  One of the reasons that Macs are friendlier
>]is that they don't have the 8086 baggage with 16 IRQ's (of which only about 5)

>The 8086 family has no such restriction: it's in the design of the 
>    motherboard and the use fof PICs and how they're cascaded.

Correct.  The MAC uses vectored interrupts.  Therefore, only 1 IRQ is
required to handle any number of devices (up to the bit size of the
vector table at any rate).

>In any event, 16 (15 actually) is plenty: the AT (ISA) bus design is the 
>    problem as each interrupt has to be handled before the signal can be 
>    used for another interrupt.

Incorrect.  The problem is that the people who build the software and
the hardware design it conditional upon the fact that two devices will
not share a single interrupt.

With properly written drivers and properly designed hardware, you can
have as many devices as you like interrupting away on the same IRQ all
at the same time, whether or not the previous interrupt on that line has
been serviced.

It simply requires that the interrupt handler initiate each driver in
turn which services a device attached to that IRQ so that the driver may
determine if one of its devices caused the interrupt.  This is a
management issue which must be handled in the driver interface.  Note
that the driver interface specifications for no PC operating system
(that is from IBM or Microsoft to run on Intel architecture) has a
strategy for determining the interrupting device, not the capability to
register multiple devices against a single interrupt.

This is an IBM design flaw in the original PC BIOS which has been
perpetuated in all PC systems since.  Some PC operating systems (Linux,
other unices, QNX, for example) can handle multiple interrupting devices
on the same IRQ line with no problem whatsoever.

It is 99.99% a software issue and only 0.01% a hardware problem.
-------------------------------------------------------------
Keith Medcalf      (416) 410-5791      http://www.dessus.com/
IBM OS/2, LAN Server, DB2/2, TCP/IP, DOS, Windows, Windows NT