root / www / HPL_logsort.html
Historique | Voir | Annoter | Télécharger (2,88 ko)
| 1 |
<HTML>
|
|---|---|
| 2 |
<HEAD>
|
| 3 |
<TITLE>HPL_logsort HPL 2.0 Library Functions September 10, 2008</TITLE> |
| 4 |
</HEAD>
|
| 5 |
|
| 6 |
<BODY BGCOLOR="WHITE" TEXT = "#000000" LINK = "#0000ff" VLINK = "#000099" |
| 7 |
ALINK = "#ffff00"> |
| 8 |
|
| 9 |
<H1>Name</H1> |
| 10 |
<B>HPL_logsort</B> Sort the processes in logarithmic order. |
| 11 |
|
| 12 |
<H1>Synopsis</H1> |
| 13 |
<CODE>#include "hpl.h"</CODE><BR><BR> |
| 14 |
<CODE>void</CODE> |
| 15 |
<CODE>HPL_logsort(</CODE> |
| 16 |
<CODE>const int</CODE> |
| 17 |
<CODE>NPROCS</CODE>, |
| 18 |
<CODE>const int</CODE> |
| 19 |
<CODE>ICURROC</CODE>, |
| 20 |
<CODE>int *</CODE> |
| 21 |
<CODE>IPLEN</CODE>, |
| 22 |
<CODE>int *</CODE> |
| 23 |
<CODE>IPMAP</CODE>, |
| 24 |
<CODE>int *</CODE> |
| 25 |
<CODE>IPMAPM1</CODE> |
| 26 |
<CODE>);</CODE> |
| 27 |
|
| 28 |
<H1>Description</H1> |
| 29 |
<B>HPL_logsort</B> |
| 30 |
computes an array IPMAP and its inverse IPMAPM1 that |
| 31 |
contain the logarithmic sorted processes id with repect to the local |
| 32 |
number of rows of U that they own. This is necessary to ensure that |
| 33 |
the logarithmic spreading of U is optimal in terms of number of steps |
| 34 |
and communication volume as well. In other words, the larget pieces |
| 35 |
of U will be sent a minimal number of times. |
| 36 |
|
| 37 |
<H1>Arguments</H1> |
| 38 |
<PRE>
|
| 39 |
NPROCS (global input) const int |
| 40 |
On entry, NPROCS specifies the number of process rows in the |
| 41 |
process grid. NPROCS is at least one. |
| 42 |
</PRE>
|
| 43 |
<PRE>
|
| 44 |
ICURROC (global input) const int |
| 45 |
On entry, ICURROC is the source process row. |
| 46 |
</PRE>
|
| 47 |
<PRE>
|
| 48 |
IPLEN (global input/output) int * |
| 49 |
On entry, IPLEN is an array of dimension NPROCS+1, such that |
| 50 |
IPLEN[0] is 0, and IPLEN[i] contains the number of rows of U, |
| 51 |
that process i-1 has. On exit, IPLEN[i] is the number of |
| 52 |
rows of U in the processes before process IPMAP[i] after the |
| 53 |
sort, with the convention that IPLEN[NPROCS] is the total |
| 54 |
number of rows of the panel. In other words, IPLEN[i+1] - |
| 55 |
IPLEN[i] is the number of rows of A that should be moved to |
| 56 |
the process IPMAP[i]. IPLEN is such that the number of rows |
| 57 |
of the source process row is IPLEN[1] - IPLEN[0], and the |
| 58 |
remaining entries of this array are sorted so that the |
| 59 |
quantities IPLEN[i+1]-IPLEN[i] are logarithmically sorted. |
| 60 |
</PRE>
|
| 61 |
<PRE>
|
| 62 |
IPMAP (global output) int * |
| 63 |
On entry, IPMAP is an array of dimension NPROCS. On exit, |
| 64 |
array contains the logarithmic mapping of the processes. In |
| 65 |
other words, IPMAP[myroc] is the corresponding sorted process |
| 66 |
coordinate. |
| 67 |
</PRE>
|
| 68 |
<PRE>
|
| 69 |
IPMAPM1 (global output) int * |
| 70 |
On entry, IPMAPM1 is an array of dimension NPROCS. On exit, |
| 71 |
this array contains the inverse of the logarithmic mapping |
| 72 |
contained in IPMAP: IPMAPM1[ IPMAP[i] ] = i, for all i in |
| 73 |
[0.. NPROCS) |
| 74 |
</PRE>
|
| 75 |
|
| 76 |
<H1>See Also</H1> |
| 77 |
<A HREF="HPL_plindx1.html">HPL_plindx1</A>, |
| 78 |
<A HREF="HPL_plindx10.html">HPL_plindx10</A>, |
| 79 |
<A HREF="HPL_pdlaswp01N.html">HPL_pdlaswp01N</A>, |
| 80 |
<A HREF="HPL_pdlaswp01T.html">HPL_pdlaswp01T</A>. |
| 81 |
|
| 82 |
</BODY>
|
| 83 |
</HTML>
|