root / www / HPL_plindx1.html
Historique | Voir | Annoter | Télécharger (4,48 ko)
| 1 |
<HTML>
|
|---|---|
| 2 |
<HEAD>
|
| 3 |
<TITLE>HPL_plindx1 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_plindx1</B> Compute local swapping index arrays. |
| 11 |
|
| 12 |
<H1>Synopsis</H1> |
| 13 |
<CODE>#include "hpl.h"</CODE><BR><BR> |
| 14 |
<CODE>void</CODE> |
| 15 |
<CODE>HPL_plindx1(</CODE> |
| 16 |
<CODE>HPL_T_panel *</CODE> |
| 17 |
<CODE>PANEL</CODE>, |
| 18 |
<CODE>const int</CODE> |
| 19 |
<CODE>K</CODE>, |
| 20 |
<CODE>const int *</CODE> |
| 21 |
<CODE>IPID</CODE>, |
| 22 |
<CODE>int *</CODE> |
| 23 |
<CODE>IPA</CODE>, |
| 24 |
<CODE>int *</CODE> |
| 25 |
<CODE>LINDXA</CODE>, |
| 26 |
<CODE>int *</CODE> |
| 27 |
<CODE>LINDXAU</CODE>, |
| 28 |
<CODE>int *</CODE> |
| 29 |
<CODE>IPLEN</CODE>, |
| 30 |
<CODE>int *</CODE> |
| 31 |
<CODE>IPMAP</CODE>, |
| 32 |
<CODE>int *</CODE> |
| 33 |
<CODE>IPMAPM1</CODE>, |
| 34 |
<CODE>int *</CODE> |
| 35 |
<CODE>PERMU</CODE>, |
| 36 |
<CODE>int *</CODE> |
| 37 |
<CODE>IWORK</CODE> |
| 38 |
<CODE>);</CODE> |
| 39 |
|
| 40 |
<H1>Description</H1> |
| 41 |
<B>HPL_plindx1</B> |
| 42 |
computes two local arrays LINDXA and LINDXAU containing |
| 43 |
the local source and final destination position resulting from the |
| 44 |
application of row interchanges. In addition, this function computes |
| 45 |
three arrays IPLEN, IPMAP and IPMAPM1 that contain the logarithmic |
| 46 |
mapping information for the spreading phase. |
| 47 |
|
| 48 |
<H1>Arguments</H1> |
| 49 |
<PRE>
|
| 50 |
PANEL (local input/output) HPL_T_panel * |
| 51 |
On entry, PANEL points to the data structure containing the |
| 52 |
panel information. |
| 53 |
</PRE>
|
| 54 |
<PRE>
|
| 55 |
K (global input) const int |
| 56 |
On entry, K specifies the number of entries in IPID. K is at |
| 57 |
least 2*N, and at most 4*N. |
| 58 |
</PRE>
|
| 59 |
<PRE>
|
| 60 |
IPID (global input) const int * |
| 61 |
On entry, IPID is an array of length K. The first K entries |
| 62 |
of that array contain the src and final destination resulting |
| 63 |
from the application of the interchanges. |
| 64 |
</PRE>
|
| 65 |
<PRE>
|
| 66 |
IPA (global output) int * |
| 67 |
On exit, IPA specifies the number of rows that the current |
| 68 |
process row has that either belong to U or should be swapped |
| 69 |
with remote rows of A. |
| 70 |
</PRE>
|
| 71 |
<PRE>
|
| 72 |
LINDXA (global output) int * |
| 73 |
On entry, LINDXA is an array of dimension 2*N. On exit, this |
| 74 |
array contains the local indexes of the rows of A I have that |
| 75 |
should be copied into U. |
| 76 |
</PRE>
|
| 77 |
<PRE>
|
| 78 |
LINDXAU (global output) int * |
| 79 |
On exit, LINDXAU is an array of dimension 2*N. On exit, this |
| 80 |
array contains the local destination information encoded as |
| 81 |
follows. If LINDXAU(k) >= 0, row LINDXA(k) of A is to be
|
| 82 |
copied in U at position LINDXAU(k). Otherwise, row LINDXA(k) |
| 83 |
of A should be locally copied into A(-LINDXAU(k),:). |
| 84 |
</PRE>
|
| 85 |
<PRE>
|
| 86 |
IPLEN (global output) int * |
| 87 |
On entry, IPLEN is an array of dimension NPROW + 1. On exit, |
| 88 |
this array is such that IPLEN[i] is the number of rows of A |
| 89 |
in the processes before process IPMAP[i] after the sort |
| 90 |
with the convention that IPLEN[nprow] is the total number of |
| 91 |
rows of the panel. In other words IPLEN[i+1]-IPLEN[i] is the |
| 92 |
local number of rows of A that should be moved to the process |
| 93 |
IPMAP[i]. IPLEN is such that the number of rows of the source |
| 94 |
process row can be computed as IPLEN[1] - IPLEN[0], and the |
| 95 |
remaining entries of this array are sorted so that the |
| 96 |
quantities IPLEN[i+1] - IPLEN[i] are logarithmically sorted. |
| 97 |
</PRE>
|
| 98 |
<PRE>
|
| 99 |
IPMAP (global output) int * |
| 100 |
On entry, IPMAP is an array of dimension NPROW. On exit, this |
| 101 |
array contains the logarithmic mapping of the processes. In |
| 102 |
other words, IPMAP[myrow] is the corresponding sorted process |
| 103 |
coordinate. |
| 104 |
</PRE>
|
| 105 |
<PRE>
|
| 106 |
IPMAPM1 (global output) int * |
| 107 |
On entry, IPMAPM1 is an array of dimension NPROW. On exit, |
| 108 |
this array contains the inverse of the logarithmic mapping |
| 109 |
contained in IPMAP: IPMAPM1[ IPMAP[i] ] = i, for all i in |
| 110 |
[0.. NPROCS) |
| 111 |
</PRE>
|
| 112 |
<PRE>
|
| 113 |
PERMU (global output) int * |
| 114 |
On entry, PERMU is an array of dimension JB. On exit, PERMU |
| 115 |
contains a sequence of permutations, that should be applied |
| 116 |
in increasing order to permute in place the row panel U. |
| 117 |
</PRE>
|
| 118 |
<PRE>
|
| 119 |
IWORK (workspace) int * |
| 120 |
On entry, IWORK is a workarray of dimension 2*JB. |
| 121 |
</PRE>
|
| 122 |
|
| 123 |
<H1>See Also</H1> |
| 124 |
<A HREF="HPL_pdlaswp00N.html">HPL_pdlaswp00N</A>, |
| 125 |
<A HREF="HPL_pdlaswp00T.html">HPL_pdlaswp00T</A>, |
| 126 |
<A HREF="HPL_pdlaswp01N.html">HPL_pdlaswp01N</A>, |
| 127 |
<A HREF="HPL_pdlaswp01T.html">HPL_pdlaswp01T</A>. |
| 128 |
|
| 129 |
</BODY>
|
| 130 |
</HTML>
|