Path: utzoo!mnetor!uunet!husc6!bloom-beacon!gatech!purdue!i.cc.purdue.edu!k.cc.purdue.edu!l.cc.purdue.edu!cik
From: cik@l.cc.purdue.edu (Herman Rubin)
Newsgroups: comp.arch
Subject: Sparse vectors
Message-ID: <730@l.cc.purdue.edu>
Date: 28 Mar 88 12:05:29 GMT
References: <1135@pembina.UUCP> <28200126@ccvaxa> <1162@pembina.UUCP>
Organization: Purdue University Statistics Department
Lines: 15
Summary: It depends on the vector architecture


On most vector machines, a vector is a rigid object.  There are very
many situations in which this gives bad performance.  On a few, there
is vector hardware which permits the direct use of sparse vectors.  
The other flexible vector operations needed are collapse and merge
of vectors--this is needed to get efficient evaluation of functions
where different algorithms are needed for different parts of the domain.

The problem is even worse on parallel processors.  I suggest that some
of these non-rigid vector manipulations be added to vector and parallel
processors, even if it requires adding special hardware for those purposes.
-- 
Herman Rubin, Dept. of Statistics, Purdue Univ., West Lafayette IN47907
Phone: (317)494-6054
hrubin@l.cc.purdue.edu (ARPA or UUCP) or hrubin@purccvm.bitnet