mrush@csuchico.edu (Matt "C P." Rush) (09/15/90)
(OK, this is not DIRECTLY Amiga oriented, but....)
I have a question about an MMU on a 680x0 based computer:
On a system with Memory Mapped I/O, is it possible to set up the
MMU so that USER programs can have ALL addresses go to RAM, yet have the OS
still be able to access the magic I/O address? (sort of make ALL I/O through
OS calls, and give the USER a completely RAM address space; regardless of
whether or not this is PRACTICAL :-)
On a more Amiga-oriented note: There's been talk that we may get
Virtual Memory in later versions of the OS; will it be possible to have
Virtual CHIP Memory?
And finally, how do all you ENGINEER-types (like Dave H.) manage to
design NEW hardware, AND keep up on everybody ELSES (Intel, IBM, Apple, etc.)
new hardware AT THE SAME TIME???
-- Matt
*~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~*
% "Progress is an up-hill battle % mrush@csuchico.edu %
% against backwards compatibility." % mrush@cscihp.UUCP %
% -- me %
% Now: mrush@cscihp.ecst.csuchico.edu %
*~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~*
This is a SCHOOL! Do you think they even CARE about MY opinions?!al158305@mtecv2.mty.itesm.mx (Gustavo Cordova Avila) (09/15/90)
mrush@csuchico.edu (Matt "C P." Rush) writes: > On a more Amiga-oriented note: There's been talk that we may get >Virtual Memory in later versions of the OS; will it be possible to have >Virtual CHIP Memory? I really don't think so, the principal behind virtual mem is that when the cpu tries to execute or access any memory that isn't actually in the main core (er, memory) at the moment, the mmu generates an exception that is trapped by the OS, which re-loads the stuff from disk. Now, I do think that video access is MUCH too fast to be able to be swapped in and out from memory/disk, don't you think? :) Gustavo Cordova sig: REAL computers have keyboard garages! :)
daveh@cbmvax.commodore.com (Dave Haynie) (09/18/90)
In article <1990Sep14.185419.28320@ecst.csuchico.edu> mrush@csuchico.edu writes: > I have a question about an MMU on a 680x0 based computer: > On a system with Memory Mapped I/O, is it possible to set up the >MMU so that USER programs can have ALL addresses go to RAM, yet have the OS >still be able to access the magic I/O address? (sort of make ALL I/O through >OS calls, and give the USER a completely RAM address space; regardless of >whether or not this is PRACTICAL :-) Well, you really wouldn't want non-RAM accesses to go to RAM. However, the 68030/68851 MMU gives you the option of having either unified or separate tables for each kind of memory space (the important ones being User Data, User Instruction, Supervisor Data, and Supervisor Instruction). So, for example, it would be simple to allow only Supervisor Data space accesses to talk to I/O locations. Of course, you would have to modify the Amiga OS somewhat to do that -- lots of I/O is currently done in user space. But I'm sure things like UNIX already behave this way on Amiga hardware. > On a more Amiga-oriented note: There's been talk that we may get >Virtual Memory in later versions of the OS; will it be possible to have >Virtual CHIP Memory? Virtual chip memory can't be done. Virtual memory is possible as long as you have control over that virtual memory exclusively by one CPU. That's because only the 68030, for instance, see it's virtual addresses -- everything else in the system sees physical addresses. As Chip memory is addressed by both 68030 and Agnus, you have two problems. First of all, Agnus has no way to detect or generate a page fault, and secondly, no way to speak in terms of the 68030's virtual addresses. > And finally, how do all you ENGINEER-types (like Dave H.) manage to >design NEW hardware, AND keep up on everybody ELSES (Intel, IBM, Apple, etc.) >new hardware AT THE SAME TIME??? We don't sleep much. And we never rest. > -- Matt -- Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy Get that coffee outta my face, put a Margarita in its place!
valentin@cbmvax.commodore.com (Valentin Pepelea) (09/22/90)
In article <1990Sep14.185419.28320@ecst.csuchico.edu> mrush@csuchico.edu writes: > > I have a question about an MMU on a 680x0 based computer: > > On a system with Memory Mapped I/O, is it possible to set up the > MMU so that USER programs can have ALL addresses go to RAM, yet have the OS > still be able to access the magic I/O address? (sort of make ALL I/O through > OS calls, and give the USER a completely RAM address space; regardless of > whether or not this is PRACTICAL :-) Yes, it is possible to do that. The virtual address space on the 68000 is not just a huge 4 Gigabyte plane. While indeed the processor is limited to access 4 Gigabytes of physical memory, virtual memory is divided into seven address spaces. Four of them are - user data space (001) - user program space (010) - supervisor data space (101) - supervisor program space (110) A separate translation table can be made for each of these cases. Therefore when CPU is in supervisor mode, the supervisor mode translation table will be used. But this strategy would not work on the Amiga, where I/O space is often accessed from user mode. > On a more Amiga-oriented note: There's been talk that we may get > Virtual Memory in later versions of the OS; will it be possible to have > Virtual CHIP Memory? Definitely not. The MMU resides between the CPU and external memory. The graphic chip processors are part of the external world. > And finally, how do all you ENGINEER-types (like Dave H.) manage to > design NEW hardware, AND keep up on everybody ELSES (Intel, IBM, Apple, etc.) > new hardware AT THE SAME TIME??? Notice that there is only one Dave Haynie. Valentin -- The Goddess of democracy? "The tyrants Name: Valentin Pepelea may distroy a statue, but they cannot Phone: (215) 431-9327 kill a god." UseNet: cbmvax!valentin@uunet.uu.net - Ancient Chinese Proverb Claimer: I not Commodore spokesman be