Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!ames!oliveb!sun!chiba!khb
From: khb%chiba@Sun.COM (Keith Bierman - Sun Tactical Engineering)
Newsgroups: comp.arch
Subject: Re: Multi-Processor Serializability
Keywords: data ordering, coherence, shared memory multiprocessing
Message-ID: <88088@sun.uucp>
Date: 2 Feb 89 19:17:05 GMT
References: <3492@cloud9.Stratus.COM> <1066@cmx.npac.syr.edu>
Sender: news@sun.uucp
Reply-To: khb@sun.UUCP (Keith Bierman - Sun Tactical Engineering)
Organization: Sun Microsystems, Mountain View
Lines: 19

In article <1066@cmx.npac.syr.edu> billo@snacpac.npac.syr.edu.UUCP (Bill O'Farrell) writes:
>means that an outer loop is having its iterations performed in parallel, while
>the inner loop has been "unwound"  into vector operations. Both the Alliant
>and Encore compiler also perform "good old" global optimization, and both
>never NEVER perform an optimization if it would change the meaning of a program.


This is not quite true. Most very highly optimizing compilers have
(often undocumented) unsafe modes which intentionally change the
semantics of the code. On a CDC system with the u level of
optimization (0,1,2,3,U) it produced correct results about 70% of the
time, which made it a very useful tool.

In addition, even modest levels of optimization can cause subtle
changes in numerical results. On ieee 64-bit mode this is seldom seen,
but in 32-bits it can often be observed.

Keith H. Bierman
It's Not My Fault ---- I Voted for Bill & Opus