This post had a lot of advice and some references, but all of them pretty much missed the point about why there are so many interrupt problems lately.
First, you need the definition of IRQ [which, apparently, even the chief engineer at Cytrix needs to study]:
Microprocessor designers have gotten the concept of interrupts wrong from the beginnings of microprocessor architecture. I can say that because I've designed interrupt busses.
When Dr. Wirth began the microprocessor design, he based it on existing mainframe design [Time Magazine, interview, Dec. 1996].
However, the good Dr. missed something, the systems were 4x4 and each had a 16-bit Interrupt buss, yielding a total of 64 physically unique and separate interrupt lines. IBM had little to do with it, since it was not IBM's architecture that the microprocessor was based on, it was Unisys System, at the time, a Burroughs System.
Although the 16-bit buss worked at the time, since microprocessors were basically 16-bit machines, at the advent of 32-bit architecture it started to go awry. Which is when the PIC came out; a simple Programmable Logic Controller to reroute signals. The Programmable Logic Array was added and the PCI interrupt controller was born.
Right before 64-bit microprocessors, the PIC's became APIC's and advanced to 24 interrupts, 8 of which are control functions.
The modern busses are PCI-X busses; still 16-bit with the additional 8 and saying it's a 24-bit interrupt buss, which it is not.
The problem is not going to go away with devices [peripherals] until Intel, AMD, Cytrix, and others get with it and design a full 64-bit wide interrupt buss that has 64 physically unique interrupts number 0 through 63.
Every other buss is 64 bits wide, why isn't the interrupt buss?
Ignorance and misunderstanding of the Rand, Los Alamos, Palo Alto, and Oak Ridge computer design projects.
First, you need the definition of IRQ [which, apparently, even the chief engineer at Cytrix needs to study]:
Microprocessor designers have gotten the concept of interrupts wrong from the beginnings of microprocessor architecture. I can say that because I've designed interrupt busses.
When Dr. Wirth began the microprocessor design, he based it on existing mainframe design [Time Magazine, interview, Dec. 1996].
However, the good Dr. missed something, the systems were 4x4 and each had a 16-bit Interrupt buss, yielding a total of 64 physically unique and separate interrupt lines. IBM had little to do with it, since it was not IBM's architecture that the microprocessor was based on, it was Unisys System, at the time, a Burroughs System.
Although the 16-bit buss worked at the time, since microprocessors were basically 16-bit machines, at the advent of 32-bit architecture it started to go awry. Which is when the PIC came out; a simple Programmable Logic Controller to reroute signals. The Programmable Logic Array was added and the PCI interrupt controller was born.
Right before 64-bit microprocessors, the PIC's became APIC's and advanced to 24 interrupts, 8 of which are control functions.
The modern busses are PCI-X busses; still 16-bit with the additional 8 and saying it's a 24-bit interrupt buss, which it is not.
The problem is not going to go away with devices [peripherals] until Intel, AMD, Cytrix, and others get with it and design a full 64-bit wide interrupt buss that has 64 physically unique interrupts number 0 through 63.
Every other buss is 64 bits wide, why isn't the interrupt buss?
Ignorance and misunderstanding of the Rand, Los Alamos, Palo Alto, and Oak Ridge computer design projects.