Speaking of spam ...

Byron Davies davies@pobox.com
Wed, 1 Apr 1998 20:32:34 -0700

Pobox's automated spam filter appended "[may be junkmail -pobox]" to the
subject of the following email, as shown.  Perhaps it's been seeing too
many emails with the word "FREE" in capital letters lately.



Resent-Date: Wed, 1 Apr 1998 16:20:13 -0500 (EST)
Subject: Re: [may be junkmail -pobox] Re: Make LispM code FREE (fwd)
Date: Wed, 1 Apr 1998 16:19:41 -0500
x-sender: ptw@spoon-hill.callitrope.thecia.net
From: "P. T. Withington" <ptw@pobox.com>
To: "Kragen" <kragen@pobox.com>,
        "William A. Barnett-Lewis" <wlewis@mailbag.com>
cc: <lispos@math.gatech.edu>
Mime-Version: 1.0
Resent-From: lispos@math.gatech.edu
X-Mailing-List: <lispos@math.gatech.edu> archive/latest/2234
X-Loop: lispos@math.gatech.edu
Precedence: list
Resent-Sender: lispos-request@math.gatech.edu

On 4/1/98 15:53, Kragen wrote:

>On Wed, 1 Apr 1998, William A. Barnett-Lewis wrote:
>> Someone wrote:
>> > OTOH, the portable C prototype of the emulator ran like a dog...
>Hmm.  Is this because the things you want to do for writing a Lisp VM
>are harder to do in C than in assembler?

Possibly not today.  4 years ago, scheduling the instructions by hand was
the only way to crank every last cycle out of the Alpha.  Today, I
understand there is an optimization mode for gcc that will use annealing
to find the optimal instruction order.

[Our job was made easier by doing all the writing in Lisp and using Lisp
to translate into assembly.  Thus we could use Lisp macros, etc. to make
the job of writing assembly much easier, and editor commands such as m-x
Show Compiled Code would show a cycle-count annotated assembly,
highlighting the instruction dependencies and stalls.]