Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!caen!news.cs.indiana.edu!news.nd.edu!mentor.cc.purdue.edu!pop.stat.purdue.edu!hrubin
From: hrubin@pop.stat.purdue.edu (Herman Rubin)
Newsgroups: comp.arch
Subject: Re: Vector vs Cache/Superscalar
Message-ID: <11875@mentor.cc.purdue.edu>
Date: 4 May 91 12:47:52 GMT
References: <1991May4.031835.7979@midway.uchicago.edu
Sender: news@mentor.cc.purdue.edu
Lines: 42

In article <1991May4.031835.7979@midway.uchicago.edu>, rtp1@quads.uchicago.edu (raymond thomas pierrehumbert) writes:
> McAlpin comments that he finds vectorization (even on the Cyber 205)
> simpler, more intuitive and more transportable than the optimization
> techniques used on cached machines like the RS/6000.
 
> I think this is partly because the vector model of parallelism is
> so rigid; optimization for the superscalars involves a bigger bag
> of tricks.  Still, I have found that there are fewer things they
> choke on, and that it is easier to localize optimization in a few
> reusable routines.  Two case-studies:

			...................

There are differences in the various types of vector machines; most
have highly rigid vectors, but the 205 has quite a bit of flexibility.

> (b) Tridiagonal solving.  Comes up in lots of codes, and it is
> a real vector-breaker.  In fact, vector machines choke on all
> sorts of recursion, whereas the superscalars just love them.
> On the RS/6000, the tridiag code basically vanished, whereas on
> the vector Stardent, it was a bottleneck.

There are tricky ways of doing this efficiently on vector machines,
especially flexible ones.  This uses partitioning.

> A third example that occurs to me is evaluation of transcendental
> functions.  Lots of recursion, and pretty efficient on the RISCS.
> On a vector machine, you have to keep iterating the vector until
> the slowest converging argument is done converging (unless you
> do a lot of reshuffling in memory)

But this reshuffling in memory on the 205 is dirt cheap!  It could
be improved by adding more instructions compatible with a CISC
vector machine.  In fact, this type of management is quite useful
in the most efficient type of generation of non-uniform random 
numbers, acceptance-rejection, and the vector methods are quite
competitive on scalar machines with the older scalar versions, and
only use more memory, no longer a bottleneck here.
-- 
Herman Rubin, Dept. of Statistics, Purdue Univ., West Lafayette IN47907-1399
Phone: (317)494-6054
hrubin@l.cc.purdue.edu (Internet, bitnet)   {purdue,pur-ee}!l.cc!hrubin(UUCP)