Path: utzoo!mnetor!uunet!husc6!bloom-beacon!think!ames!amdcad!decwrl!alverson
From: alverson@decwrl.dec.com (Robert Alverson)
Newsgroups: comp.arch
Subject: Re: Tasting of Dhrystone 2.0 Results
Message-ID: <329@bacchus.DEC.COM>
Date: 16 Mar 88 20:46:01 GMT
References: <503@pcrat.UUCP> <4074@vdsvax.steinmetz.ge.com> <11309@duke.cs.duke.edu>
Reply-To: alverson@decwrl.UUCP (Robert Alverson)
Organization: Digital Equipment Corporation
Lines: 14


In the postings so far, there have been very few pascal dhrystone numbers
given.  How about it folks?  When I try the pascal version here, one of
the answers doesn't match.  Is this a problem with the code?  Anyway,
it seems that Pascal could potentially run the benchmark must faster
since the strings in it are fixed length.  The string assignment need
only copy 30 bytes of data from one spot to another.  The compiler
may even be able to guarantee word alignment.  Similarly, the string
compare may be optimized knowing the length to compare is fixed and
that the two strings are word-aligned.

Is this fair?  Or is it just a builtin inefficiency of C strings?

Bob