Sender: crough45 AT amc DOT de
Message-Id: <97Sep12.202559gmt+0100.11649@internet01.amc.de>
Date: Fri, 12 Sep 1997 19:31:03 +0100
From: Chris Croughton <crough45 AT amc DOT de>
Mime-Version: 1.0
To: Klaus-Georg DOT Adams AT chemie DOT uni-karlsruhe DOT de
Cc: djgpp AT delorie DOT com
Subject: Re: The numer 1 compiler, DJGPP or MSVC Here's a good rating comparision
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Precedence: bulk

Klaus-Georg Adams wrote:

> Yeah, as you say. Every workstation vendor makes a compiler for their
> platform. And these native compilers usually beet the sh** out of gcc
> when it comes to optimization (with the exception of SUN's, if you
> believe the benchmarks)

Well, benchmarks mean about as much as IQ tests - they show the
score relative to that benchmark (or IQ test) but almost nothing
in real terms.

The only way in which the native compilers beat gss on the 
platforms with which I'm familiar is compile time.  Half of
the presupplied ones aren't even ANSI C (not just Sun) and
lots don't provide C++ at all, not even CFront.  The native
compilers, on pure K&R-1 C code, usually beat gcc on compile
time by a factor of 2 or more (but often the linkers and
make let them down in other parts of the build process).

The code generated is seldom much different.  The exact
instruction choices are slightly different, but usually
it's hard to say which is more efficient because they are
optimising for different things.

And bugs?  The native compilers are usually designed for
recompiling the drivers and relinking the kernel, so they
aren't tested much more rigorously than that.  Try them
with code at the limits of the standard and they often fail.

(OK, you mentioned error checking, but a lot of the native
compilers give not much more than "error somewhere in
program, probably before line <n>".  Some are a bit better.)

Chris C