WorkS Digest V1 #44
ucbvax!works
Works Digest Monday, 14 Dec 1981 Volume 1 : Issue 44
Today's Topics: Limiting 68000 To Permit Demand Paging
(Very Short Digest)
----------------------------------------------------------------------
Date: 9 Dec 1981 20:06:22-PST
From: decvax!watmath!idallen at Berkeley
To: decvax!ucbvax!works@Berkeley
Subject: Limiting 68000 to permit demand paging.
Re: Ian Merritt's comment on a "real implementation" of paging.
The observations of hardware deficiencies published in this digest
lead me to believe that a single 68000 can not be involved in a
non-restartable instruction when a page fault occurs, since there is
no way to preserve the internal state of the chip, service the page
fault, and resume. I am led to think that a "real implementation"
must involve more than one processor, essentially leaving the 68000
stuck in mid-instruction while the second processor brings in the
missing page(s). Some comments seemed to suggest that, even with this
dual-processor kludge, it was still necessary to restrict the use of
certain instructions. This is the point I am unsure of. I can not
see this restriction as being a good thing if it means that one must
avoid auto[increment|decrement] modes. I guess I am asking, What
precisely is a "real implementation"? -IAN!
------------------------------
End of WorkS Digest
*******************
-------