broehl@wateng.UUCP (Bernie Roehl) (05/18/84)
All this concern with benchmarks is fine and good, but one important factor
is being overlooked: Very few applications for 16-bit processors are written
in assembler. This means that the practical performance of most of these
processors is determined not so much by their architecture and design, but
rather by the efficiency of the high-level languages available for them.
Now, it can be argued that the two are interelated (i.e. that processors with
more "advanced" architectures will be easier to write good compilers for).
However, experience has shown that this isn't always true; far (FAR!) more
effort has been expended to date developing software for the 8086-family
processors that for the more powerful M68000 or NS16000 chips. Thus for
most everyday applications, the practical performance of the 8086 systems
tends to be greater (which is how Intel and others can get such good results
in most benchmarks; they just use programs written in Basic or Pascal, for
which there are some quite good 8086 implementations).
--
-Bernie Roehl (University of Waterloo)nathan@orstcs.UUCP (05/19/84)
RE: An Architectural Contrast 68000 vs x86 There can be no rebuttal. I have spoken. Call up your Motorola rep today; they're quite enthusiastic about sending out that particular document. It is very enlightening, and contains only one blooper (that I could find). The blooper is the non-mention of their MMU (which is considered by one and most to be an abortion...but still better than the one in a '286). They also carefully skirted mention of the NS16000. The discussion of software/operating system portability, if read carefully, should be matter for hearty laughter (unless, of course, you just designed in some intel junk...) -orstcs!nathan