Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!ncar!gatech!usenet.ins.cwru.edu!eagle!data.nas.nasa.gov!amelia!eugene
From: eugene@nas.nasa.gov (Eugene N. Miya)
Newsgroups: comp.graphics
Subject: Re: Digital Holography
Message-ID: <1991May13.045426.7871@nas.nasa.gov>
Date: 13 May 91 04:54:26 GMT
References: <261@rins.ryukoku.ac.jp> <1991May9.153446.21742@leland.Stanford.EDU> <1991May10.165256.12414@nas.nasa.gov> <1991May11.152915.6488@dartvax.dartmouth.edu>
Sender: news@nas.nasa.gov
Reply-To: eugene@amelia.nas.nasa.gov (Eugene N. Miya)
Organization: NASA-Ames Research Center
Lines: 48

In article <1991May11.152915.6488@dartvax.dartmouth.edu>
npw@eleazar.dartmouth.edu (Nicholas Wilt) writes:
>In article <1991May10.165256.12414@nas.nasa.gov>
uselton@nas.nasa.gov (Samuel P. Uselton) writes:
>>In article <1991May9.153446.21742@leland.Stanford.EDU>
rick@pangea.Stanford.EDU (Rick Ottolini) writes:
>>>With the computing speeds increasing an order of magnitude
>>>every five years and no end in sight, 
>>	          ^^^^^^^^^^^^^^^^^^^
>>	There is a growing number of "experts" pointing out limits to current
>>	hardware techniques that we ARE rapidly approaching.  We need 
>>	BREAK-THROUGH improvements in technology, not just incremental
>>	improvements in the technology we have now.
>
>What about massively parallel architectures?
>techniques are as trivially parallelizable as ray tracing, then you don't
>even need any bandwidth between nodes.
>
>Sure there are issues (load balancing and stuff).  That's just software.
>_Lots_ of people are working on better software for parallel architectures.
>The hardware guys specialize in disproving statements like this.

You have three things working here.
1) Do not trivialize the software problem.  If the problem was simple,
it would have been solved in 1968.  Similar to the "automatic programming"
problem of the 1950s.
2) Every body has all these graphics codes in sequential languages.
The future dusty decks.
We are either going to have to throw out all of this code, and rewrite,
or have some awfully good software.
3) Parallel architectures are an O(n) (or at best (n^2)) solution to
problems that many times have greater complexity.  This is why students
should take computing theory classes.

We can only rely on so many hardware improvements.  Got to rememebr we
compute in the physical (real, not virtual) world.  Start reading and
reasing issues in comp.arch (although few architects read that group any more).
There are some ends in sight.

Those components (that software) which runs fastest and most reliable
are those which aren't there.  --Gordon Bell

"It's the things that nobody knows anything about that we can discuss..."
--R.P. Feynman

--eugene miya, NASA Ames Research Center, eugene@orville.nas.nasa.gov
  Resident Cynic, Rock of Ages Home for Retired Hackers
  {uunet,mailrus,other gateways}!ames!eugene