Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site ubc-cs.UUCP
Path: utzoo!linus!decvax!harpo!whuxlm!whuxl!houxm!ihnp4!alberta!ubc-vision!ubc-cs!ludemann
From: ludemann@ubc-cs.UUCP (Peter Ludemann)
Newsgroups: net.lang.c
Subject: Re: An MS-Dos C compiler for $49.95 ? (really DeSmet C)
Message-ID: <1169@ubc-cs.UUCP>
Date: Wed, 17-Jul-85 23:22:34 EDT
Article-I.D.: ubc-cs.1169
Posted: Wed Jul 17 23:22:34 1985
Date-Received: Sun, 21-Jul-85 21:23:18 EDT
References: <11587@brl-tgr.ARPA>
Reply-To: ludemann@ubc-cs.UUCP (Peter Ludemann)
Organization: UBC Department of Computer Science, Vancouver, B.C., Canada
Lines: 55
Keywords: C compiler, IBM-PC
Summary: compiler review

In article <11587@brl-tgr.ARPA> jpm@BNL44.ARPA (John McNamee) writes:
>The DeSmet package is excellent if you only need a small model compiler.
>My personal opinion is that the 8086 line is so braindamaged when you go
>outside of 64K code/64K data that large model isn't worth it. For $109
>you get an awful lot of bang for the buck (I don't use the symbolic debugger,
>but it is probably excellent too). The best thing about the system is that
>it is a complete development environment, and is blindingly fast. You
>get the fastest C compiler, fastest assembler, and fastest linker that I
>have ever seen. On my Tandy 2000 (an 8Mhz 186 system), they are disk bound.
>I put the compiler temporary files on RAM disk, and it seems that 90% of
>the compile time is spent loading the compiler and source text off disk.
>If I had the memory to put the whole compiler/linker/library on RAM disk,
>I bet it would compile and link 50K source programs in under 30 seconds.
>I should also point out that Desmet includes an editor that is fantastic.
>It isn't EMACS, but it is fast and well suited to editing source code.
>
>"Turbo C" for Borland will not be able to touch this package unless Borland
>includes a good linking method. Remember that Turbo Pascal is based around
>idea of one source file for the entire program (I guess they have something
>like #include, but I don't think that counts). Borland has yet to prove they
>can produce a system as fast as Turbo Pascal when separate compilation is
>needed. Turbo Modula will tell (it should be out before Turbo C).
>--
>
>			 John McNamee
>		..!decvax!philabs!sbcs!bnl44!jpm
>			jpm@BNL44.ARPA
>
>		  "MS-DOS is a communist plot"

In article <11587@brl-tgr.ARPA> jpm@BNL44.ARPA (John McNamee) writes:
>The DeSmet package is excellent if you only need a small model compiler.

I heartily agree.  Apparently there is a package which will let
you use more than 64K of data - at any rate, the compiler has overlays
which will let you use more than 64K of code (the overlays
can be memory resident).

On my machine, compiling and linking are limited by disk access.
When I used a friend's machine with lots of RAM-disk, my compile
time went from 12 minutes to 2 minutes (about 5,000 lines with
lots of includes) (on a standard IBM-PC).

The version of the symbolic debugger that I have has a few flaws left
in it.  However, the deSmet people have been very good in responding
to my bug reports (when was the last time you got your bug list back
with anotations as to what they were going to about them?).  So, I
suspect that the next version of the debugger will be pretty good.

Their visual editor ("see") is pretty good (and fast) for a freebie.
And they include a RAM-disk as well.

I noticed an ad for the Macintosh deSmet compiler.  Does anyone have
any experience with it?