Re: Why Virtual Machines? (was: C++ -> Java VM compiler)

p.froehlich@link-m.de wrote:


: The first VM I can remember is the pCode-Interpreter for Pascal,
: and AFAIK it was dumped for efficiency reasons.


I'd say that the "1401 emulator" systems for the early IBM
360's beat the P-System :-) and I can't speak for anybody else, but I
"dumped" the UCSD P-System for the very specific reason that they
cashed my check and never delivered the product (UCSD Pascal for the
6502). About that time they "sold" the assets of the project (but
apparently not the obligations) and folded the shell company, so legal
recourse was pretty much out of the question.


Another early "virtual machine" was the heart of Tom Pittman's
"tiny Basic" systems, published in Dr Dobbs in the early 70's (After
ETH Pascal, but before UCSD Pascal, AFAIK) This was a quite
interesting system. The VM was targeted at efficiently representing
recursive-descent parsers, essentially. Efficient is in the eye of
the beholder, but allowing tasks to be coded quickly and run fairly
quickly on minimal systems was a big plus. The alternatives at the
time typically involved using an expensive minicomputer to
cross-assemble code that ran maybe 20x faster, but at considerable
"capital" cost in both money and development time.


: Okay, one can say
: "Today's machines are powerful enough to support VMs on a large
: scale.", but really *why* should one do that when a more efficient and
: easier to understand method is already present, namely in the guise of
: semantic dictionary encoding (SDE) and on-the-fly compilation.


My take is that JVM, ANDF, and Active-X are all aiming at the
same goal: Delivering "portable" computation _without_ transferring
enough information to the user to allow reconstruction of useful
source. I'm enough of an old codger that I use no "mission critical"
programs for which I do not have source, so the attraction is lost to
me, but for a reasonably efficient way to waste time with sparkly
web-pages, why not... :-)


Mike
--

Return to the comp.compilers page.
Search the comp.compilers archives again.