Tartan C compiler summary

[William L. Sebok]

About a week and a half ago I asked if anyone had experience with the
Tartan C compiler for the Vax.  This is a summary of responses I received.
Nobody admitted to actually having experience with the compiler but some
of the comments were informative anyway.  There were, of course, lots of the
"tell me what you find out" responses.

As mentioned in my previous posting the compiler claims to be a good optimizing
compiler. It costs $2000 ($1700 for educational institutions).  I have since
learned that one can get it on 30 day approval.  As of this afternoon we have
decided to go ahead and order it.

I am quite aware that it would be dangerous to compile the kernel with
a heavily optimizing compiler like this.  Neverless the potential benefits
are so great that I think it is worth a try.  Perhaps a mixed solution would
work where the drivers and some of the more asynchronous parts of the
kernel continue to be compiled with the pcc.  Even if it doesn't work
on the kernel there are potential gains in recompiling the system utilities
with it.

The summary follows:

>From: adrvax!angel
To: astrovax!wls
Subject: Re: Anybody have any experience with (Vax) Tartan C compiler?

Bill,

I have not direct experience with the TC compiler,
but the company is quite serious, and responsable.
There are sugesting us to go with the modula-2 compiler
and the product looks very well.
	      Angel

>From ihnp4!verdix!rascal Wed Dec  5 22:08:41 1984
Date: Wed, 5 Dec 84 17:27:39 pst
To: ihnp4!astrovax!wls
Subject: Re: Anybody have any experience with (Vax) Tartan C compiler?

We ordered Tartan's C compiler for the VAX in August.  It still hasn't
been delivered.  We've occasionally phoned Tartan to ask their status.

At the end of August, Tartan reported that the compiler was in Beta
test with shipment scheduled for September first.

At the end of September, Tartan reported that the compiler had not
performed too well in Beta test.  They said that problems were with
"compatibility" but not the quality of the generated code.

At the end of November, Tartan reported that the compiler was due
for Beta test on November 15th.  They promised us a call.

No call yet.  I'm still optimistic but I'm not in a hurry.

	Steve Scalpone
	verdix!rascal

(Ed: it sounded to me when I called them that they were shipping.)

>From noao!seismo!rochester!ritcv!mjl Thu Dec  6 03:51:17 1984
Date: Wed, 5 Dec 84 21:53:06 est
To: carina!astrovax!wls
Subject: Re: Anybody have any experience with (Vax) Tartan C compiler?

If I'm not mistaken, many of the folks at Tartan Labs are the same
ones who worked at CMU on highly optimizing compilers (esp. BLISS-11,
which was the focus of the monograph by Wulf, et. al. call something
like "An Optimizing Compiler").  If I'm correct, then I'd bet they're
claims are valid.  In fact, now that I see the price is reasonable,
I may even convince our facilities folks to give Tartan a call.

	Mike Lutz
	{allegra,seismo}!rochester!ritcv!mjl

>From allegra!rocksvax!ritcv!odin!mjl Mon Dec 10 22:00:49 1984
Date: Mon, 10 Dec 84 11:39:16 est
To: ritcv!rocksvax!allegra!astrovax!wls
Subject: Re: Anybody have any experience with (Vax) Tartan C compiler?

Bill:
	If you do act as guinea pigs, please let me know what the
results are.  If nothing else, perhaps Tartan C could be used to
recompile PCC.  In our teaching environment, where compilation time
exceeds execution time in most cases, fast compilation is more important
than super optimal code.  So, if PCC could be made to run as fast
as possible, this might be a win for us.

Mike Lutz
Rochester Institute of Technology
seismo!rochester!ritcv!mjl

-- 
Bill Sebok			Princeton University, Astrophysics
{allegra,akgua,burl,cbosgd,decvax,ihnp4,noao,princeton,vax135}!astrovax!wls