🚀 go-pugleaf

In article <1069@goanna.oz> wjb@goanna.oz (Warwick Bolam) writes:
>
>I think the offensiveness of the Byte article stems from the fact that
>it is heades "68020 vs 80386" when it should have been headed "Machine X
>using 68020 vs. 80386".  If you want to compare raw processor speed,
>then don't do what BYTE did.
>

And if you want to compare commonly used systems, you still don't want
to do what Byte did.  Byte's benchmarks for the 80386 systems were
compiled with a 80386-specific compiler, whereas 95+% of the 80386
systems in this world are spending their time running software developed
for the 8086.  This makes a big difference, since running old software
reduces the 80386 to more or less the performance of an 80286 at equivalent
clock speed.  By contrast, in the 680x0 world there is not much penalty
for running 68000 software.  In fact, the September Byte article pointed
out that a 68000 compiler (Lightspeed C) produced code that was 10-15%
*faster* than that produced by Consulair's 68020 compiler.

So Byte's approach would only be useful to someone who is using special
software or developing their own, and those sorts of people are likely
to run their own, more specific benchmarks.

On the subject of compilers, has anyone benchmarked MPW Pascal 2.0 versus
version 1.0?  It would be interesting to see how much difference 2.0's
peephole optimization makes on the standard benchmarks (sieve, dhrystones,
etc.).

Jim Matthews
matthews@dartvax.uucp