virtual memory 4490

Sorry. It was better that you looked it up for the precise description; I don't write well. :-) I saw a lot of pink on the PDP-10 I met first.

Sure they're dynamic. An OS has to be able to cope without insisting that it has unreasonable constraints at boot time. It's like trying to fit a 2" square peg in a 1 mm round hole.

I'm speaking pure TOPS-10ese because I don't know the biz these days. Note that an OS also has to do its memory management. It has to have a page table for itself. Each user (or process) also needs to have a page table buttigned to it.

For more info about these you might try looking at the back of our cheat sheet, DECsystem-10 System Reference Card The one I have here in front of my nose includes the KI10. Order No. DEC-10-XSRCA-B-D. It has a description of the formats of the User Process Table (UPT) and the Exec Process Table (EPT) as it was shipped at that time (1974). That may give you an idea about how DEC did its first implementation.

I don't know where I can point you for the KL equivalent of that document. Thei system reference card has more knowledge in it, explicit and implicit, than you'ld find in a library dedicated to computering.

Not if this "data" is designed correctly. The goal is to do one mark and have the results cascaded through all "users" of that page. You cannot, as an OS, know who is going to use that reference; especially in this age of network and on-demand coding.

HUH? If your hardware architecture is page-based, everything is pagable. It is the job of the OS how...emoticon grasps for word elastic the determinations of what goes in and out are.

I'll guarantee you those parts are the wrong parts. I may be showing bias, but it's an experienced bias. When an OS executes an user mode instruction, it has to do so in behalf of the user. IOW, the OS "becomes" a user mode program for that instance. It was not uncommon for JMF to call TW's disk service as a "user mode" program even though he was the exec. There are a lot of time when you want the monitor (core-resident OS) to run as a mere user rather than The God called Exec.

JMF was the guy who was DEC's CPU device driver wizard and became the guy in charge of memory management. TW was the controller and device driver and file system guy. These were the TOPS-10 development guys.

Another term often used was the exec page table page and the user page table page. And indirection through this page could set any bit. So the only data maintenance one had to for knowing about pages was making sure the page table pages' data were pristine and always correct. Note that I'm being extremely simplistic talking about this stuff. What I know I learned from listening to the bit gods; I never did the work so the knowledge isn't embedded in my bones.

One of the problems the OS has to deal with is how to manage when the data kept in these page table pages spills over a page boundary.

Thank you for being kind :-). I had noticed that That Other Newsgroup was listed. I try to keep my bits out of there. :-)

BAH