1TIPS(1)               User Contributed Perl Documentation              TIPS(1)
2
3
4

NAME

6       PDL::Tips - Small tidbits of useful arcana. Programming tidbits and
7       such.
8

SYNOPSIS

10               use PDL;
11
12               # Whatever happens here.
13

DESCRIPTION

15       This page documents useful idioms, helpful hints and tips for using
16       Perl Data Language v2.0.
17
18   Help
19       Use "help help" within perldl or pdl2 or use the "pdldoc" program from
20       the command line for access to the PerlDL documentation.  HTML versions
21       of the pages should also be present, in the HtmlDocs/PDL directory of
22       the PDL distribution. To find this directory, try the following
23
24        pdl> foreach ( map{"$_/PDL/HtmlDocs"}@INC ) { p "$_\n" if -d $_ }
25
26   Indexing idioms
27       The following code normalizes a bunch of vectors in $x.  This works
28       regardless of the dimensionality of $x.
29
30               $x /= $x->sumover->dummy(0);
31
32   What is actually happening?
33       If you want to see what the code is actually doing, try the command
34
35               PDL::Core::set_debugging(1);
36
37       somewhere. This spews out a huge amount of debug info for PDL into
38       STDOUT. Plans for the future include making it possible to redirect the
39       output, and also making it possible to select messages with more
40       precision.
41
42       Many of the messages come from "Basic/Core/pdlapi.c" and you can look
43       at the source to see what is going on.
44
45       If you have any extra time to work on these mechanisms, inform the pdl-
46       porters mailing list.
47
48   Memory savings
49       If you are running recursively something that selects certain indices
50       of a large ndarray, like
51
52               while(1) {
53                       $inds = where($x>0);
54                       $x = $x->index($inds);
55                       $y = $y->index($inds);
56                       func($y,$x);
57               }
58
59       If you are not writing to $y, it saves a lot of memory to change this
60       to
61
62                       $y = $y->index($inds)->sever;
63
64       The new method "sever" is a causes the write-back relation to be
65       forgotten. It is like copy except it changes the original ndarray and
66       returns it).
67
68       Of course, the probably best way to do the above is
69
70               $inds = xvals ($x->long);
71               while(1) {
72                       $inds0 = where($x>0);
73                       $inds1 = $inds->index($inds)->sever;
74                       $x = $a0->index($inds1);
75                       $y = $y->index($inds1)->sever;
76                       func($y,$x);
77               }
78
79       which doesn't save all the temporary instances of $x in memory.  See
80       "mandel.pl" in the Demos subdirectory of the PerlDL distribution for an
81       example.
82
83   PP speed
84       If you really want to write speedy PP code, the first thing you need to
85       do is to make sure that your C compiler is allowed to do the necessary
86       optimizations.
87
88       What this means is that you have to allow as many variables as possible
89       to go into registers:
90
91               loop(a) %{
92                       $a() += $COMP(foo_member) * $b()
93               %}
94
95       expands to
96
97               for(i=0; i<10000; i++) {
98                       a[i] += __privtrans->foo_member * b[i];
99               }
100
101       is about the worst you can do, since your C compiler is not allowed to
102       assume that "a" doesn't clobber "foo_member" which completely inhibits
103       vectorization. Instead, do
104
105               float foo = $COMP(foo_member);
106               loop(a) %{
107                       $a() += foo * $b();
108               %}
109
110       This is not a restriction caused by PP but by ANSI C semantics.  Of
111       course, we could copy the struct into local variables and back but that
112       could cause very strange things sometimes.
113
114       There are many other issues on organizing loops.
115
116       We are currently planning to make PP able to do fixed-width things as
117       well as physical ndarrays (where looping over the first dimensions
118       would be cheaper as there are less distinct increments, which might
119       make a difference on machines with a small number of registers).
120

AUTHOR

122       Copyright (C) Tuomas J. Lukka 1997. All rights reserved.  Duplication
123       in the same form and printing a copy for yourself allowed.
124
125
126
127perl v5.36.0                      2022-07-22                           TIPS(1)
Impressum