taylor (04/02/83)
Eric, (re: 286 vs 68K continued) WRONG WRONG WRONG!!! I am not too sure why everyone is so positive that the 286 is only a 64K page machine...according to the specs that I have gotten from Intel Corp. there is also an optional 32-bit addressing scheme supported, where the first 16 (?) bits would be the page select and the remaining simple an address in the 64K page. Timing-wise, I suspect that the 286 is REAL close to the 68K in > 64K programs/data/etc. but I know for a fact that it beats heck out of the 68K in all smaller programs (except bit-map operations, like graphics processing, where the 68K bit-oriented instructions give it a fighting edge!). Besides, in all realism, given that the instruction set is exceptional (for a micro) on both machines, I suspect that the majority of code, other than the OS, will BE under 64K, and that the main cost of data storage will be the text...not the code, or the numeric data. Ah well, I like the 286 better, so I admit to being biased. -- Dave Taylor
guy (04/04/83)
I wouldn't assume that most programs other than the OS will fit into 64K...
give a look at the Lisa software, or the Fortran-77 compiler, or the DEC
BLISS-11 compiler, or TeX, or.... If, for example, you want to provide a
very sophisticated user interface, and make it possible to move data from a
spreadsheet to a graphics package to a freehand-drawing package to a word-
processing document, it's going to take code, and probably data space as well.
Our Office Power system has programs that take 170KB of code on a VAX (and
slightly more on a 68K), and that's because it's feature-rich, not because
the compiler or the programmers aren't any good. You may want to argue against
feature-rich software; on the other hand, it's nice to be able to move a block
of text containing a footnote and not even have to *know* that there's a
footnote attached to one word in that block (the footnote moves invisibly
with it) or have to renumber the footnotes (the numbers change automatically
as the block is cut and pasted). That takes code. You can make software fit
in a Procrustean bed, but you have to cut something off.
I have had my fill of machines with 16-bit pointers; if UNIX had an
overlaying facility worthy of the name, I might feel differently. But
even with an overlaying facility, your DATA space can get big, and you
have to write code by hand to play with the segment registers. Without
an overlaying facility, you're up the creek.
Guy Harris
RLG Corporation
{seismo,mcnc,we13}!rlgvax!guy