According to LinuxReviews.org, “in recent months, both AMD and Intel have proposed higher approaches to interrupt and exception handling.”
Then they percentaged this evaluation, which Linus Torvalds published at the Real World Technologies forum:
“The AMD model is essentially “Fix recognized insects within the exception dealing with definition.”
According to the Intel model, “the covered mode of 80286 exception handling changed to bad, then 386 made it even odder with the 32-bit extensions, then syscall and sysenter made everything worse, and the x86-64 extensions added even more complications.” So let’s upload a style bit in which all of the crap is going away.
In contrast, the AMD one is largely a bare-bones attempt to restore real-world issues with all that legacy-precipitated crap that is difficult to work with and has caused problems…
Both are legitimate on their own, and they’re pretty independent. To be honest, the AMD paper appears to be a short “we have not even finished questioning all of the information through, but we realize those elements have been broken, so we might as well launch this.”
I’m not sure how long it’s been brewing; however, based on the “TBD” items in that paper, I assume it’s an “early difficult draft.”
In the article (shared by longtime Slashdot reader xiando), LinuxReviews.org summarizes the kingdom of the communiqué today:
Torvalds went on to mention that even though AMD’s proposed “short restore” could be easier to put into force for him and other working gadget vendors, it is now no longer perfect in the long run. Intel’s inspiration throws the complete present interrupt descriptor table (IDT) transport gadget below the bus so it could get replaced with what they call a “FRED occasion transport” gadget. Torvalds believes that is a more lengthy-term solution…
While the pros and cons of Intel and AMD’s respective proposals for interrupt and event handling in future processors are debatable, AMD is frequently as aggressive as Intel. They are the larger and more effective corporation. It is much more likely now than it was a year ago that destiny processors from Intel will use their proposed Flexible Return and Event Delivery gadget. Their subsequent-generation processors will not; consumer CPUs will not enter the FRED era for years, not months.Remember, the above-noted technical record changed and was posted in advance this month [in March]. Things do now no longer magically move from the drawing-board to store-cabinets overnight.
Intel is not going to simply hand the FRED era over to AMD and assist them in forcing it. We will probably see each move ahead with their very own proposals. Intel may have FRED and AMD may have Supervisor Entry Extensions until AMD, inevitably, adopts FRED or some variation of it years down the line.
Linus Torvalds Discusses Intel and AMD’s New Proposals for Interrupt/Exception Handling
They are additionally aware that Torvalds took issue with a poster arguing that microkernel systems are more stable than monolithic ones like Linux.
“Bah, you’re just repeating the standard party line, which has absolutely no foundation in fact and, while you check out the data, would not absolutely hold up.”
It’s all ideas and handwaving and simply repeating the same old FUD that has evolved into something that is no longer absolutely relevant.