1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
|
<!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML 2.0//EN">
<!--Converted with LaTeX2HTML 96.1-h (September 30, 1996) by Nikos Drakos (nikos@cbl.leeds.ac.uk), CBLU, University of Leeds -->
<HTML>
<HEAD>
<TITLE>In-core Dense Matrices</TITLE>
<META NAME="description" CONTENT="In-core Dense Matrices">
<META NAME="keywords" CONTENT="slug">
<META NAME="resource-type" CONTENT="document">
<META NAME="distribution" CONTENT="global">
<LINK REL=STYLESHEET HREF="slug.css">
</HEAD>
<BODY LANG="EN" >
<A NAME="tex2html3104" HREF="node75.html"><IMG WIDTH=37 HEIGHT=24 ALIGN=BOTTOM ALT="next" SRC="http://www.netlib.org/utk/icons/next_motif.gif"></A> <A NAME="tex2html3102" HREF="node68.html"><IMG WIDTH=26 HEIGHT=24 ALIGN=BOTTOM ALT="up" SRC="http://www.netlib.org/utk/icons/up_motif.gif"></A> <A NAME="tex2html3096" HREF="node73.html"><IMG WIDTH=63 HEIGHT=24 ALIGN=BOTTOM ALT="previous" SRC="http://www.netlib.org/utk/icons/previous_motif.gif"></A> <A NAME="tex2html3106" HREF="node1.html"><IMG WIDTH=65 HEIGHT=24 ALIGN=BOTTOM ALT="contents" SRC="http://www.netlib.org/utk/icons/contents_motif.gif"></A> <A NAME="tex2html3107" HREF="node190.html"><IMG WIDTH=43 HEIGHT=24 ALIGN=BOTTOM ALT="index" SRC="http://www.netlib.org/utk/icons/index_motif.gif"></A> <BR>
<B> Next:</B> <A NAME="tex2html3105" HREF="node75.html">The Two-dimensional Block-Cyclic Distribution</A>
<B>Up:</B> <A NAME="tex2html3103" HREF="node68.html">Data Distributions and Software </A>
<B> Previous:</B> <A NAME="tex2html3097" HREF="node73.html">Array Descriptors</A>
<BR> <P>
<H1><A NAME="SECTION04430000000000000000">In-core Dense Matrices</A></H1>
<A NAME="secincore"> </A>
<P>
The choice of an appropriate
data distribution heavily
depends on the characteristics
or flow of the computation in
the algorithm. For dense matrix
computations, ScaLAPACK assumes
the data to be distributed
according to the <EM>two-dimensional
block-cyclic</EM> data layout scheme.
This section presents this
distribution and demonstrates
how the ScaLAPACK software
encodes this essential
information as well as
the related software
conventions.
<P>
Dense matrix computations feature
a large amount of parallelism, so
that a wide variety of distribution
schemes have the potential for
achieving high performance. The
<EM>block-cyclic</EM> data layout
has been selected for the dense
algorithms implemented in ScaLAPACK
principally because of its
scalability [<A HREF="node189.html#lawn43">51</A>],
load balance, and communication
[<A HREF="node189.html#hendrickson94a">76</A>] properties.
The block-partitioned computation
proceeds in consecutive order
just like a conventional serial
algorithm. This essential property
of the block cyclic data layout
explains why the ScaLAPACK design
has been able to reuse the numerical
and software expertise of the
sequential LAPACK library.
<P>
<BR> <HR>
<UL><A NAME="CHILD_LINKS"> </A>
<LI> <A NAME="tex2html3108" HREF="node75.html#SECTION04431000000000000000">The Two-dimensional Block-Cyclic Distribution</A>
<LI> <A NAME="tex2html3109" HREF="node76.html#SECTION04432000000000000000">Local Storage Scheme and Block-Cyclic Mapping</A>
<LI> <A NAME="tex2html3110" HREF="node77.html#SECTION04433000000000000000">Array Descriptor for In-core Dense Matrices</A>
<LI> <A NAME="tex2html3111" HREF="node78.html#SECTION04434000000000000000">Example</A>
<LI> <A NAME="tex2html3112" HREF="node79.html#SECTION04435000000000000000">Submatrix Argument Descriptions</A>
<LI> <A NAME="tex2html3113" HREF="node80.html#SECTION04436000000000000000">Matrix and Vector Storage Conventions</A>
</UL>
<BR> <HR>
<P><ADDRESS>
<I>Susan Blackford <BR>
Tue May 13 09:21:01 EDT 1997</I>
</ADDRESS>
</BODY>
</HTML>
|
