🚀 go-pugleaf

In article <3670@cit-vax.Caltech.Edu> newton@cit-vax.UUCP (Mike Newton) writes:
>Look at the code and the timings for some of the
>PC compilers (say, Microsoft 5.0).  They are MUCH better!  And for
>such a (ok folks, here it starts again, get your k-keys (rn) ready)
>SHITTY microprocessor from the compiler writers point of view.

Sigh.  Although compiler writers have many reasons to dislike previous
intel kludges, the 386 is actually reasonably pleasant to work on.  There
are no more segment register hassles (set 'em and ferget 'em, 32 bit
offsets remove the worry--for now), and the general registers actually are
reasonably general (ok, string move and shift counts still have to go
into ecx, but that's nearly the only restriction).  The chip is very
fast;  they originally had bit string insert and extract instrcutions,
but they turned out to be slower in every case to the equivalent shift
and mask code, so they took them out.  The only place they are still
weak is in the area of the coprocessor--their coprpocessor interface is
less flexible, general, and has lower performance than the 68020's.  In
particular, the 386 is so complicated that the coprocessor no longer
runs asynchronously from the 386 itself...

Now, I realize that I haven't supplied you with any hard numbers, but
these are my objective observations after having lived with the chip for
nearly 2 years, part of which involved the design and implementation of
a C compiler.

	    --Darryl Richman, INTERACTIVE Systems Corp.
	    ...!cca!ima\
			>-!ism780c!darryl
	    ...!sdcrdcf/
	    The views expressed above are my opinions only.