Path: utzoo!utgpu!water!watmath!clyde!rutgers!sri-spam!ames!amdcad!sun!pitstop!sundc!seismo!uunet!steinmetz!sunray!oconnor
From: oconnor@sunray.steinmetz (Dennis M. O'Connor)
Newsgroups: comp.arch
Subject: Re: Performance increase - a suggestion
Message-ID: <9375@steinmetz.steinmetz.UUCP>
Date: 29 Jan 88 19:51:25 GMT
References: <235@unicom.UUCP>
Sender: news@steinmetz.steinmetz.UUCP
Reply-To: sunray!oconnor@steinmetz.UUCP
Organization: GE Corporate R&D Center
Lines: 48

An article by aglew@ccvaxa.UUCP says:
]
]>	I really don't think the real world really needs anything more
]>expansive than a 32 bit processor to get most jobs done.
]
]Probably, but... I wonder how much interest might be out there for
]a true double-precision floating point engine - one that did 64 bit
]floating point, or IEEE 80 bit extended floating point, or even 
]128 bit floating point, as its native floating point mode, as fast
]as single precision on nearly any other machine in its price range?
]
]I'm sure that most people wouldn't need this, but some might - and I'd 
]like to get a feel for the size of such a niche, if it exists.
]
]Andy "Krazy" Glew. Gould CSD-Urbana.    1101 E. University, Urbana, IL 61801   

There is indeed a market for this. Several in fact.

  1. Simulation of Physical Systems ( Chemistry, Aerodynamics, Mechanics )
     These people need all the accuracy they can get. Truncation
     error just grows to fast on 32-bit floats, and the dynamic range
     is too small.

  2. Real-time control systems ( aircraft engines, aircraft control surfaces )
     These systems can't afford to flake out because of truncation
     errors accumulating or dynamic range being exceeded. They also
     are so tightly pressed for performance that they can't afford
     ellaborate checks of their results. 

  3. Sonar and OTH Radar systems
     These are TOUGH signal processing problems, with very noisy
     data, needing long, complicated munging. And must be real time.

  4. Strategic Weapons Systems, including SDI
     Think about aiming at a 1 meter target thats 500,000 meters away.
     That's moving at 10,000 meters per second. How much error can
     you afford ? How much time have you got to do the job ?

Single precision is apparantly becoming obsolete. It's just too
much of a pain to deal with. Double precision is the only way to go.
Optimize for double, keep single only if you need "backwards"
compatability. "You'll thank me for it."

  
--
	Dennis O'Connor 	oconnor@sungoddess.steinmetz.UUCP ??
				ARPA: OCONNORDM@ge-crd.arpa
        "If I have an "s" in my name, am I a PHIL-OSS-IF-FER?"