Post by thread:Multitasking without interrupt. Rather [OT]?

Peter van Hoof wrote: > Actually .... I disassembled the print.exe of dos 6.0 today, just to > satisfy my curiosity (there is nothing much on the net about old > technologie) and print seems to use one of the oldest multitasking > mechanisms around (in dos and windows) the msdos multitask interrupt. > and this is timer driven Just another small part of the mystery of the PC. The printer port was fitted with an interrupt mechanism so that it could be efficiently interrupt driven. The interrupt mechanism has a tri-state gate so the interrupt can be shared. That is why it *is* usually shared or perhaps, subverted by the sound card on IRQ7. Because ... they never *used* it to efficiently drive the printer (I too have disassembled print.exe!). More silliness. (Older) network cards tend to be set to IRQ3 by default. WIN95 assumes this which is *totally insane*, given that all machines since it was introduced are (quite properly) fitted with *two* serial ports, the second of which is hard-wired to IRQ3, generally cannot be disconnected from IRQ3 even if the serial port is disabled (but *why* would you relinquish a serial port anyway?). This is because the IBM version of the serial port, like the printer port, *had* a tri-state gate on the IRQ line but no-one (well, certainly not ...) could write OS code to utilize it to share the IRQ so - manufacturers decided it wasn't wanted and designed it out. Consequently, you have to set the network card to use IRQ5, which is *why* you must use IRQ7 for the sound card. It's not so bad if you use a 16-bit network card and can use the higher IRQs, but WIN95 was too dumb to figure this out itself. That's why setting up PCs is a tricky business. You think it's all sorted nowadays? <infernal laughter> Maybe? </infernal laughter> -- Cheers, Paul B.

I tri-state gate on a interrupt line is NOT what you'd want to use if the interrupt were sharable. Two systems that decide to DRIVE the interrupt line at the same time would still cause problems. You need something like open-collector lines instead. My understanding is that no one implemented shared interrupt drivers because the hardware was essentially broken... BillW

|I tri-state gate on a interrupt line is NOT what you'd want to use if the |interrupt were sharable. Two systems that decide to DRIVE the interrupt |line at the same time would still cause problems. You need something like |open-collector lines instead. IBM in their infinite wisdom decided to have the interrupts on the PC be active high. Thus, the simplest way of making a driver that can share the line is to use a 3-state buffer chip whose input is always high; the enable on the buffer then switches the output between floating and high. |My understanding is that no one implemented shared interrupt drivers because |the hardware was essentially broken... In another piece of their infinite wisdom, IBM decided to make interrupts edge-triggered rather than level-controlled. Sharing a level-controlled interrupt is easy: simply have the ISR check the state of any device which could cause the int- terrupt. Even if a device earlier in the testing sequence starts asserting IRQ while a later device is being checked, the device will keep IRQ asserted until it's acknowleged; this in turn will force the CPU to re-execute the ISR from the start, and detect the interrupting condition which it earlier had missed. Edge-triggered interrupts are much harder to share reliably: after any interrupting device is dealt with it's essential to re-poll *ALL* the devices that could assert the interrupt line. Returning from an interrupt when a device on the line has not yet been serviced will effectively put that IRQ line out of commission. Annoying...

----- > IBM in their infinite wisdom decided to have the interrupts > on the PC be active high. Thus, the simplest way of making > a driver that can share the line is to use a 3-state buffer > chip whose input is always high; the enable on the buffer > then switches the output between floating and high. My understanding was that the PC came about as a way to control the market, and help prevent erosion of their mini sales. Since the PC explosion was going to happen anyway, this was an opportunity to poison the well by introducing a functional but broken standard. Here we are, still bound by our peripherals. They blew it with the PS-2 line, when they apparently didn't understand how well bound we were. All you had to do was dump your entire in-box investment and buy a new one, to get on board with a much better hardware system.. Anyone running PS-2? Then we had the comical re-invention of the PS-1, as a face-saving fallback.

John Payson wrote: > IBM in their infinite wisdom decided to have the interrupts > on the PC be active high. Thus, the simplest way of making > a driver that can share the line is to use a 3-state buffer > chip whose input is always high; the enable on the buffer > then switches the output between floating and high. That's true, but *not* what was implemented. >> My understanding is that no one implemented shared interrupt drivers >> because the hardware was essentially broken... I agree, that is the long and short of it. > In another piece of their infinite wisdom, IBM decided to > make interrupts edge-triggered rather than level-controlled. If my previous comments were to convey the impression that the PC design was, and parts of it remain, anything other than an overall balls-up, then please forgive me. -- Cheers, Paul B.