Minutes of FIGGY BAR RT Conference.

Date: 10/04/90   Time: 22:28EDT

Attendees:
[[Len] NMORGENSTERN]
[GARY-S]
[[Wil] W.BADEN1]
  
There was not a lot of Forth discussed in this FIGGY, but there was some 
good discussion of computer architecture and systems level programming.
Concurrency, Interpreters and pipelines don't mix, parsing Forth's QUIT,
quantitative design, distributed parallel processing.

Minutes:
<[Wil] W.BADEN1> is here.
<[Wil] W.BADEN1> Hi Gary.
<[Wil] W.BADEN1> Welcome back.
<[Wil] W.BADEN1> Who's in room3
<GARY-S> thanks
<GARY-S> len
<GARY-S> what is happening these days with Mr. Apple ?
<[Wil] W.BADEN1> Evry one is waiting for Oct 15 announcements.
<GARY-S> which is ?
<[Wil] W.BADEN1> 3 new, lower priced, models.
<[Wil] W.BADEN1> The Mac RT here sez that audio digitizing will be built in.
<GARY-S> doesn't that coincide all too nicely with Raskin's conference here on 
the 17th ?
<NMORGENSTERN> is here.
<[Wil] W.BADEN1> Yeeaah.  He shud hav sumthin to say about that.
<GARY-S> have you had an opportunity to run A/UX ?
<[Wil] W.BADEN1> No, I'm sorry to say.
<GARY-S> If you do I would welcome your comments
<[Wil] W.BADEN1> MPW is a version of unix that has been around for 4 years.
<[Wil] W.BADEN1> Mac Prog'rs Workshop.
<[Wil] W.BADEN1> It is much better than unix.
<GARY-S> It is closer akin to version 7 than more modern Unix platforms though
<[Wil] W.BADEN1> Yes, it is restricted.
<[Wil] W.BADEN1> A/UX 2. shd be even better.
<GARY-S> MPW was a near-clone of the PW run on vaxen
<[Wil] W.BADEN1> Cud u scroll back, select text, and execute it, with stdout 
appearing where u are?
<GARY-S> Can't really say with certainty wil
<GARY-S> Stuff Forth, anything break or happen in my brief absence ?
<[Len] NMORGENSTERN> Nothing special, Gary, as far as I know.
<GARY-S> len , were you able to finish your discussion on Sunday or does it 
merit rvival ?
<[Len] NMORGENSTERN> Sunday's disc can be taken up next Sun.
<[Wil] W.BADEN1> I hav bought two great books in the last 2 weeks.
<GARY-S> which are ?
<[Wil] W.BADEN1> Interpreters and concurrent processor execution.
<[Wil] W.BADEN1> This was mentioned the week after I bought it in a message by 
Spackman.
<GARY-S> titles/authors/underbelly used
<[Wil] W.BADEN1> I'll hav to get the book from the car.
<GARY-S> save it for next week
<[Wil] W.BADEN1> Forth and Modula are considered languages of similar nature.
<[Wil] W.BADEN1> The price, not known in the message, is $30.
<[Len] NMORGENSTERN> (is surprised) ==(:-)   (his hair is standing on end)
<[Len] NMORGENSTERN> How are Modula and Forth similar?
<[Wil] W.BADEN1> The other is Hennessey&Patterson, reviewed by Ray Duncan in 
DDJ.
<GARY-S> Wil, Modula 2 I have seen seems to devour mysterious chunks of memory 
- how is that similar to Forth ?
<[Wil] W.BADEN1> "conservative syntactic content"
<[Wil] W.BADEN1> Anyhow, it treats Forth seriously and sympathetically.
<[Wil] W.BADEN1> What was the discussion?  I logged on at 6:15 & there was no 
one there.
<GARY-S> Concurrency - to what degree and by what method ?
<[Wil] W.BADEN1> Separate programs, or instructions in a single program?
<[Len] NMORGENSTERN> Jax was on with a brief report on the Embedded systems 
conference
<[Len] NMORGENSTERN> Then I was on with Mike Christoff. We ended about 6:20 
because
<[Len] NMORGENSTERN> I had to leave slightly early.
<[Len] NMORGENSTERN> and there were only the 2 of us
<[Len] NMORGENSTERN> Jax thinkx that Forth got some good "press" at the Emb sys 
conf
<GARY-S> It takes a sophisticated interrupt manager to properly handle 
concurrency, does it not. Can this be done with an interpretive process ?
<[Wil] W.BADEN1> Yes, but interpretive processes defeat the purpose for 
pipelined instruction execution.
<GARY-S> If you use a CISC processor you almost have to pipeline
<GARY-S> and cache
<GARY-S> what is the alternative according to your author(s) ?
<[Wil] W.BADEN1> I wonder if you cud write a compiler for Forth that wud rename 
registers instead of DUP and SWAP.
<[Len] NMORGENSTERN> The key word is "compiler"
<GARY-S> why not - but entire registers ?
<[Len] NMORGENSTERN> Forth's compiler is , (comma)
<[Wil] W.BADEN1> Alternativ to what?  Pipelining is state-of-the-art for the 
90's.
<[Len] NMORGENSTERN> If you were willing to add a parsing step in the quit
<[Len] NMORGENSTERN> loop, you would have a true compiler & you could do what 
you wanted
<GARY-S> I was responding to your comment that interpreters defeat pipelining
<[Wil] W.BADEN1> It wud hav to recognize stack-growing operation sequences and 
turn this off.
<GARY-S> That would require an additional pointer
<[Wil] W.BADEN1> The purpose of pipelining is to increase the instructin 
thru-put.
<GARY-S> Not that big a thing reALLY
<[Wil] W.BADEN1> This means unrolling many functions.
<GARY-S> I understand that - your original premise was Interpreters and 
Concurrency unless I misunderstood
<[Wil] W.BADEN1> It also shud re-order sequence of operations, which is 
fundamentally opposed to Forth style programming.
<[Wil] W.BADEN1> Let me get the book.  afk.
<[Len] NMORGENSTERN> What does afk mean, Gary
<GARY-S> I don't know, but I still think my question was reasonable, or did I 
miss something ?
<GARY-S> The whole point of pipelining and caching is to predetermine 
instruction fetches
<[Len] NMORGENSTERN> My expertise in these things is about 3 on a scale
<[Len] NMORGENSTERN> of 100, and I just don't know, Gary
<[Len] NMORGENSTERN> I have been thinking about the idea of incorporating
<[Len] NMORGENSTERN> a parsing step in QUIT It definitely would change
<[Len] NMORGENSTERN> the nature of Forth!
<GARY-S> For what purpose ? Is this like the step routines I have seen in FD ?
<[Len] NMORGENSTERN> We do a lot of things in Forth becaue the input
<[Len] NMORGENSTERN> is interpreted one word at a time. If we parsed
<[Len] NMORGENSTERN> it woould give more flexibility
<[Len] NMORGENSTERN> (some would say chaos)
<GARY-S> You would also corrupt the stack, right ?
<[Len] NMORGENSTERN> Not unless you made a mistake. Example:
<GARY-S> We're safe - never make mistakes, do we ??? :-)
<[Len] NMORGENSTERN> an adventure game. If we make North South etc constants,
<[Len] NMORGENSTERN> then we could say North Go But this is clumsy
<[Len] NMORGENSTERN> We want to use English like syntax. Programming
<[Len] NMORGENSTERN> Go North is fairly complex. If we parsed, we could
<[Len] NMORGENSTERN> analyze the input, locat the verb, and relate the otthe
<[Len] NMORGENSTERN> sutff to it, and execute.
<[Wil] W.BADEN1> bak. Back at keyboard.
<[Wil] W.BADEN1> Afk is away from keyboard, (but still recording -- I haven't 
missed anything).
<[Len] NMORGENSTERN> Thanx
<[Wil] W.BADEN1> One book is Interpretation and Instruction Path Coprossing, by 
Debaere and Campenhout.
<GARY-S> What do you gain - I see a slowdown
<[Len] NMORGENSTERN> (Sound like folks from Arcturus)
<[Len] NMORGENSTERN> Was that for me Gary?
<[Wil] W.BADEN1> The second book is Computer Architecture -- a Quantitative 
Approach.
<[Len] NMORGENSTERN> If so, adding exta steps always slows things down. In this 
case
<[Len] NMORGENSTERN> it would only be during keyboard interpretation or 
compilation
<[Len] NMORGENSTERN> and would not be a significant penalty.
<[Len] NMORGENSTERN> (Might be hard to prgram, though.
<GARY-S> The purpose for adding these to your library Wil ? That might get my 
head in tune with yours.
<[Wil] W.BADEN1> Keeping up with the craft.
<GARY-S> Software engineering at the System level ?
<[Wil] W.BADEN1> BTW, "state of the art" means "ordinary".
<[Wil] W.BADEN1> Yes.
<[Len] NMORGENSTERN> The cutting edge is where you bleed!
<GARY-S> ok - I still find the premise of the first book at odds with current 
CISC design
<[Wil] W.BADEN1> Check out the message.  The writer of the message reviewed the 
book in SIGPLan.
<GARY-S> 'and this is the original point where we became talking heads earlier
<[Wil] W.BADEN1> Thatz wher I found out about it.  I call Chicago 555-1212 in 
hope to get his
<[Wil] W.BADEN1> phone number, but it was unlisted.  I was going to call him 
and ask him to
<[Wil] W.BADEN1> upload his review here.
<[Wil] W.BADEN1> A week later he posted his message.
<[Wil] W.BADEN1> He has, like me, mixed feelings about the book.
<GARY-S> That would be interesting
<[Wil] W.BADEN1> Spackman is the name of the messager.
<[Wil] W.BADEN1> Computer Architecture is a landmark book for the 90's. $55 and 
worth it.
<GARY-S> Have you ever read the 2 volume "Parallel Distributed processing"
<[Wil] W.BADEN1> No. By whom?
<GARY-S> Rumelhart,McClelland et al at MIT
<GARY-S> dry and difficult read but full of good insight
<[Wil] W.BADEN1> Did u know that Kogge, who wrote that great article on Forth's 
architectural trail, published a book on pipelining in 1981?
<GARY-S> no
<[Wil] W.BADEN1> No, I'll keep my eyes peeled.
<[Len] NMORGENSTERN> Which article is that, Wil?
<GARY-S> That predates actual incorporation - in Nova's anyway
<[Wil] W.BADEN1> "An architetural trail to threaded-code systems," IEEE 
Computer, March 1982.
<[Wil] W.BADEN1> disconnected.
<[Len] NMORGENSTERN> Thanx Wil
<GARY-S> For the information or for the logoff ? :-)))
<[Len] NMORGENSTERN> I read an article in Sci. Am on parallel processing
<[Len] NMORGENSTERN> The authors claimed that the radix sort is one of the 
fastest
<[Len] NMORGENSTERN> with a parallel processor because you can start processing 
a
<[Len] NMORGENSTERN> heap as soon as it is started insted of waiting until it 
is finished!
<GARY-S> Parallel Processing throws a curve into a lot of preconcieved notions 
simply because of the vector factor you just referred to
<[Len] NMORGENSTERN> This has been interesting, GAry. I am always interested
<[Len] NMORGENSTERN> in things I have never even known ABOUT much less known!
<GARY-S> It wasn't exactly Forthish, but maybe someone will find the transcript 
a keeper.
<GARY-S> Thanks for covering for me Len
<[Len] NMORGENSTERN> Right. No problem.
<GARY-S> I presume you read my e-mail
<[Len] NMORGENSTERN> I am usually free. How is Dennis doing? He sounds depressed
<[Len] NMORGENSTERN> but hopeful.
<GARY-S> You know as much as I do
<[Len] NMORGENSTERN> I hope things go well for you, Gary.
<GARY-S> It isn't me , but thanks
<[Len] NMORGENSTERN> True, but it is not easy.
<GARY-S> Think we should roll the sidewalk up - don't you
<[Len] NMORGENSTERN> Yes, it is that time. See you next week Will you be able to
<[Len] NMORGENSTERN> take Thursday?
<GARY-S> All I can tell you is if you notice a vacuum you might want to check 
up on me.
<GARY-S> You can tell if the ports stop fairly easily
<[Len] NMORGENSTERN> I will plan to be here about 15 minutes early & get on
<[Len] NMORGENSTERN> If you show up we will transfer, as we did today.
<GARY-S> Youre a class act Leonard Morgenstern, and I appreciate you
<GARY-S> g'nite
=== End of Steno notes. ===