Date: Sun, 23 Feb 1997 11:52:13 +0200 (IST)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
To: Jesse Bennett <jbennett AT ti DOT com>
cc: djgpp AT delorie DOT com
Subject: Re: Netlib code [was Re: flops...]
In-Reply-To: <5egilh$k7g$1@superb.csc.ti.com>
Message-ID: <Pine.SUN.3.91.970223115026.1447C-100000@is>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII


On 20 Feb 1997, Jesse Bennett wrote:

> Since all of the array indexing is explicitly coded using pointers the
> issue of aliasing is avoided.  Standard F77 does not provide memory
> pointers and this level of "human optimization" is not possible.  I
> understand that F90 incorporates the concept of pointers, but I have
> no experience with it.  Quality F90 compilers are not currently
> available on a wide variety of common platforms at a reasonable cost
> (IMHO).
> 
> I have done some benchmarking of the above matrix multiply code on a
> DEC Alpha using the native DEC compilers.  The "pointer based" C
> implementation was fastest, followed by the F77 code and the "array
> based" C code.  Hence my comments.

Try it with gcc.  In most cases, it converts the array-based code to 
pointer-based automatically, as far as I could see, and in many cases it 
does that better than you would.