BYTE%mit-mc@sri-unix.UUCP (06/05/83)
From: Roger L. Long <BYTE@mit-mc> Put simply, you know not of what you speak! You should really take a look at the specs for the 68451 MMU before you try to tell the rest of us how it doesn't work. I think I can speak with a bit more authority since I just finished porting UNIX to my companies 68000 product which uses two 68451 MMUs. First of all, the page size may be any power of two, so the typical process might use 4 or 5 segments. You can get along easily with only one 68451 - our company chose to use two in order to reduce swapping of descriptors in systems with LOTS of main memory. Running in single-user mode, you will typically not use more descriptors than will fit on one MMU unless you have a lot going on in background. In comparing the 68000 with the National 16032, I think a main consideration to keep in mind is not the raw power of the CPU. If you look at benchmarks comparing the 68000 to a VAX in terms of raw CPU power, I believe the 68000 running at 8MHz is .8 the power of a VAX 780. However, you are not going to see a 68000 UNIX system outperform a VAX UNIX system. The bottleneck is the disk! UNIX is such a disk-intensive OS and most implementations, including ours, use winchesters, which are fast compared to floppies, but not compared to the disks that hang off a VAX. What started this all off was a query about putting UNIX on an S100 box in what I would assume would be a single-user environment. Whether it be a 68000 or a 16032, I think you're going to have more than enough CPU to keep the average hacker happy in that kind of environment. -roger
goldfarb.ucf-cs@rand-relay@sri-unix.UUCP (06/06/83)
Well, Andy, the page size on your beloved Vax is 1024.
norskog@fortune.UUCP (06/09/83)
#R:sri-arpa:-179700:fortune:25500001:000:310 fortune!norskog Jun 8 21:12:00 1983 It should be noted that the Berkeley paging algorithm, (global demand paging), is >20 years out of date. It has been known since the early sixties that paging a process against itself between low and high water marks (and swapping against other processes) is much more efficient than global demand paging.
POURNE@mit-mc@sri-unix.UUCP (08/28/83)
From: Jerry E. Pournelle <POURNE@mit-mc> alas, I know less than either of you, and thus cannot comprehend what is happening here. I know the Fortune computer uses UNIX on the 68000 and has for years, so I fear I find the discussion strange. Date: 5 June 1983 05:40 EDT From: Roger L. Long <BYTE> To: SJOBRG.ANDY at MIT-OZ cc: POURNE, info-cpm at BRL, fylstra.tsca at SRI-UNIX Re: UNIX on the 68000 Put simply, you know not of what you speak! You should really take a look at the specs for the 68451 MMU before you try to tell the rest of us how it doesn't work. I think I can speak with a bit more authority since I just finished porting UNIX to my companies 68000 product which uses two 68451 MMUs. First of all, the page size may be any power of two, so the typical process might use 4 or 5 segments. You can get along easily with only one 68451 - our company chose to use two in order to reduce swapping of descriptors in systems with LOTS of main memory. Running in single-user mode, you will typically not use more descriptors than will fit on one MMU unless you have a lot going on in background. In comparing the 68000 with the National 16032, I think a main consideration to keep in mind is not the raw power of the CPU. If you look at benchmarks comparing the 68000 to a VAX in terms of raw CPU power, I believe the 68000 running at 8MHz is .8 the power of a VAX 780. However, you are not going to see a 68000 UNIX system outperform a VAX UNIX system. The bottleneck is the disk! UNIX is such a disk-intensive OS and most implementations, including ours, use winchesters, which are fast compared to floppies, but not compared to the disks that hang off a VAX. What started this all off was a query about putting UNIX on an S100 box in what I would assume would be a single-user environment. Whether it be a 68000 or a 16032, I think you're going to have more than enough CPU to keep the average hacker happy in that kind of environment. -roger