1PARALLELCPU(1) User Contributed Perl Documentation PARALLELCPU(1)
2
6 PDL::ParallelCPU - Parallel Processor MultiThreading Support in PDL
7 (Experimental)
8
10 PDL has support (currently experimental) for splitting up numerical
11 processing between multiple parallel processor threads (or pthreads)
12 using the set_autopthread_targ and set_autopthread_size functions.
13 This can improve processing performance (by greater than 2-4X in most
14 cases) by taking advantage of multi-core and/or multi-processor
15 machines.
16
18 use PDL;
19 # Set target of 4 parallel pthreads to create, with a lower limit of
21 # 5Meg elements for splitting processing into parallel pthreads.
22 set_autopthread_targ(4);
23 set_autopthread_size(5);
24 $a = zeroes(5000,5000); # Create 25Meg element array
26 $b = $a + 5; # Processing will be split up into multiple pthreads
28 # Get the actual number of pthreads for the last
30 # processing operation.
31 $actualPthreads = get_autopthread_actual();
32
34 The use of the term threading can be confusing with PDL, because it can
35 refer to PDL threading, as defined in the PDL::Threading docs, or to
36 processor multi-threading.
37 To reduce confusion with the existing PDL threading terminology, this
39 document uses pthreading to refer to processor multi-threading, which
40 is the use of multiple processor threads to split up numerical
41 processing into parallel operations.
42
44 This is a brief listing and description of the PDL pthreading
45 functions, see the PDL::Core docs for detailed information.
46 set_autopthread_targ
48 Set the target number of processor-threads (pthreads) for multi-
49 threaded processing. Setting auto_pthread_targ to 0 means that no
50 pthreading will occur.
51 See PDL::Core for details.
53 set_autopthread_size
55 Set the minimum size (in Meg-elements or 2**20 elements) of the
56 largest PDL involved in a function where auto-pthreading will be
57 performed. For small PDLs, it probably isn't worth starting
58 multiple pthreads, so this function is used to define a minimum
59 threshold where auto-pthreading won't be attempted.
60 See PDL::Core for details.
62 get_autopthread_actual
64 Get the actual number of pthreads executed for the last pdl
65 processing function.
66 See PDL::get_autopthread_actual for details.
68
70 PDL PThreading can be globally turned on, without modifying existing
71 code by setting environment variables PDL_AUTOPTHREAD_TARG and
72 PDL_AUTOPTHREAD_SIZE before running a PDL script. These environment
73 variables are checked when PDL starts up and calls to
74 set_autopthread_targ and set_autopthread_size functions made with the
75 environment variable's values.
76 For example, if the environment var PDL_AUTOPTHREAD_TARG is set to 3,
78 and PDL_AUTOPTHREAD_SIZE is set to 10, then any pdl script will run as
79 if the following lines were at the top of the file:
80 set_autopthread_targ(3);
82 set_autopthread_size(10);
83
85 The auto-pthreading process works by analyzing threaded array
86 dimensions in PDL operations and splitting up processing based on the
87 thread dimension sizes and desired number of pthreads (i.e. the pthread
88 target or pthread_targ). The offsets and increments that PDL uses to
89 step thru the data in memory are modified for each pthread so each one
90 sees a different set of data when performing processing.
91 Example
93 $a = sequence(20,4,3); # Small 3-D Array, size 20,4,3
95 # Setup auto-pthreading:
97 set_autopthread_targ(2); # Target of 2 pthreads
98 set_autopthread_size(0); # Zero so that the small PDLs in this example will be pthreaded
99 # This will be split up into 2 pthreads
101 $c = maximum($a);
102 For the above example, the maximum function has a signature of "(a(n);
104 [o]c())", which means that the first dimension of $a (size 20) is a
105 Core dimension of the maximum function. The other dimensions of $a
106 (size 4,3) are threaded dimensions (i.e. will be threaded-over in the
107 maximum function.
108 The auto-pthreading algorithm examines the threaded dims of size (4,3)
110 and picks the 4 dimension, since it is evenly divisible by the
111 autopthread_targ of 2. The processing of the maximum function is then
112 split into two pthreads on the size-4 dimension, with dim indexes 0,2
113 processed by one pthread
114 and dim indexes 1,3 processed by the other pthread.
115
117 Must have POSIX Threads Enabled
118 Auto-PThreading only works if your PDL installation was compiled with
119 POSIX threads enabled. This is normally the case if you are running on
120 linux, or other unix variants.
121 Non-Threadsafe Code
123 Not all the libraries that PDL intefaces to are thread-safe, i.e. they
124 aren't written to operate in a multi-threaded environment without
125 crashing or causing side-effects. Some examples in the PDL core is the
126 fft function and the pnmout functions.
127 To operate properly with these types of functions, the PPCode flag
129 NoPthread has been introduced to indicate a function as not being
130 pthread-safe. See PDL::PP docs for details.
131 Size of PDL Dimensions and PThread Target
133 Due to the way a PDL is split-up for operation using multiple pthreads,
134 the size of a dimension must be evenly divisible by the pthread target.
135 For example, if a PDL has threaded dimension sizes of (4,3,3) and the
136 auto_pthread_targ has been set to 2, then the first threaded dimension
137 (size 4) will be picked to be split up into two pthreads of size 2 and
138 2. However, if the threaded dimension sizes are (3,3,3) and the
139 auto_pthread_targ is still 2, then pthreading won't occur, because no
140 threaded dimensions are divisible by 2.
141 The algorithm that picks the actual number of pthreads has some smarts
143 (but could probably be improved) to adjust down from the
144 auto_pthread_targ to get a number of pthreads that can evenly divide
145 one of the threaded dimensions. For example, if a PDL has threaded
146 dimension sizes of (9,2,2) and the auto_pthread_targ is 4, the
147 algorithm will see that no dimension is divisible by 4, then adjust
148 down the target to 3, resulting in splitting up the first threaded
149 dimension (size 9) into 3 pthreads.
150 Speed improvement might be less than you expect.
152 If you have a 8 core machine and call auto_pthread_targ with 8 to
153 generate 8 parallel pthreads, you probably won't get a 8X improvement
154 in speed, due to memory bandwidth issues. Even though you have 8
155 separate CPUs crunching away on data, you will have (for most common
156 machine architectures) common RAM that now becomes your bottleneck. For
157 simple calculations (e.g simple additions) you can run into a
158 performance limit at about
159 4 pthreads. For more complex calculations the limit will be higher.
160
162 Copyright 2011 John Cerney. You can distribute and/or modify this
163 document under the same terms as the current Perl license.
164 See: http://dev.perl.org/licenses/
166perl v5.30.0 2019-09-05 PARALLELCPU(1)