1HWLOC-BIND(1) hwloc HWLOC-BIND(1)
2
6 hwloc-bind - Launch a command that is bound to specific processors
7 and/or memory, or consult the binding of an existing program
8
10 hwloc-bind [options] <location1> [<location2> [...] ] [--] <command>
11 ...
12 Note that hwloc(7) provides a detailed explanation of the hwloc system
14 and of valid <location> formats; it should be read before reading this
15 man page.
16
18 --cpubind Use following arguments for CPU binding (default).
19 --membind Use following arguments for memory binding. If --mempolicy
21 is not also given, the default policy is bind.
22 --mempolicy <policy>
24 Change the memory binding policy. The available policies are
25 default, firsttouch, bind, interleave replicate and next‐
26 touch. This option is only meaningful when an actual binding
27 is also given with --membind. If --membind is given without
28 --mempolicy, the default policy is bind.
29 --get Report the current bindings. The output is an opaque bitmask
32 that may be translated into objects with hwloc-calc (see
33 EXAMPLES below).
34 When a command is given, the binding is displayed before exe‐
36 cuting the command. When no command is given, the program
37 exits after displaying the current binding.
38 When combined with --membind, report the memory binding
40 instead of CPU binding.
41 No location may be given since no binding is performed.
43 --nodeset Report binding as a NUMA memory node set instead of a CPU set
46 if --get was given. This is useful for manipulating CPU-less
47 NUMA nodes since their cpuset is empty while their nodeset is
48 correct.
49 Also parse input bitmasks as nodesets instead of cpusets.
51 -e --get-last-cpu-location
54 Report the last processors where the process ran. The output
55 is an opaque bitmask that may be translated into objects with
56 hwloc-calc (see EXAMPLES below).
57 Note that the result may already be outdated when reported
59 since the operating system may move the process to other pro‐
60 cessors at any time according to the binding.
61 When a command is given, the last processors is displayed
63 before executing the command. When no command is given, the
64 program exits after displaying the last processors.
65 This option cannot be combined with --membind.
67 No location may be given since no binding is performed.
69 --single Bind on a single CPU to prevent migration.
72 --strict Require strict binding.
74 --pid <pid>
76 Operate on pid <pid>
77 --tid <tid>
79 Operate on thread <tid> instead of on an entire process. The
80 feature is only supported on Linux for thread CPU binding, or
81 for reporting the last processor where the thread ran if -e
82 was also passed.
83 -p --physical
85 Interpret input locations with OS/physical indexes instead of
86 logical indexes. This option does not apply to the output,
87 see --get above.
88 -l --logical
90 Interpret input locations with logical indexes instead of
91 physical/OS indexes (default). This option does not apply to
92 the output, see --get above.
93 --taskset Display CPU set strings in the format recognized by the
95 taskset command-line program instead of hwloc-specific CPU
96 set string format. This option has no impact on the format
97 of input CPU set strings, both formats are always accepted.
98 --restrict <cpuset>
100 Restrict the topology to the given cpuset.
101 --whole-system
103 Do not consider administration limitations.
104 --hbm Only take high bandwidth memory nodes (such as KNL's MCDRAM)
106 in account when looking for NUMA nodes in the input loca‐
107 tions.
108 This option must be combined with NUMA node locations, such
110 as --hbm numa:1 for binding on the second HBM node. It may
111 also be written as hbm:1.
112 --no-hbm Ignore high bandwidth memory nodes (such as KNL's MCDRAM)
114 when looking for NUMA nodes in the input locations.
115 -f --force
117 Launch the executable even if binding failed.
118 -q --quiet
120 Hide non-fatal error messages. It includes locations point‐
121 ing to non-existing objects, as well as failure to bind.
122 This is usually useful in addition to --force.
123 -v --verbose
125 Verbose output.
126 --version Report version and exit.
128
130 hwloc-bind execs an executable (with optional command line arguments)
131 that is bound to the specified location (or list of locations). Upon
132 successful execution, hwloc-bind simply sets bindings and then execs
133 the executable over itself.
134 If multiple locations are given, they are combined in the sense that
136 the binding will be wider. The process will be allowed to run on every
137 location inside the combination.
138 The list of input locations may be explicitly ended with "--".
140 If binding fails, or if the binding set is empty, and --force was not
142 given, hwloc-bind returns with an error instead of launching the exe‐
143 cutable.
144 NOTE: It is highly recommended that you read the hwloc(7) overview page
146 before reading this man page. Most of the concepts described in
147 hwloc(7) directly apply to the hwloc-bind utility.
148
150 hwloc-bind's operation is best described through several examples.
151 More details about how locations are specified on the hwloc-bind com‐
152 mand line are described in hwloc(7).
153 To run the echo command on the first logical processor of the second
155 package:
156 $ hwloc-bind package:1.pu:0 -- echo hello
158 which is exactly equivalent to the following line as long as there is
160 no ambiguity between hwloc-bind option names and the executed command
161 name:
162 $ hwloc-bind package:1.pu:0 echo hello
164 To bind the "echo" command to the first core of the second package and
166 the second core of the first package:
167 $ hwloc-bind package:1.core:0 package:0.core:1 -- echo hello
169 To bind memory on the first high-bandwidth memory node:
171 $ hwloc-bind --membind hbm:0 -- echo hello
173 $ hwloc-bind --membind --hbm numa:0 -- echo hello
174 Note that binding the "echo" command to multiple processors is probably
176 meaningless (because "echo" is likely implemented as a single-threaded
177 application); these examples just serve to show what hwloc-bind can do.
178 To run on the first three packages on the second and third nodes:
180 $ hwloc-bind node:1-2.package:0:3 -- echo hello
182 which is also equivalent to:
184 $ hwloc-bind node:1-2.package:0-2 -- echo hello
186 Note that if you attempt to bind to objects that do not exist, hwloc-
188 bind will not warn unless -v was specified.
189 To run on processor with physical index 2 in package with physical
191 index 1:
192 $ hwloc-bind --physical package:1.core:2 -- echo hello
194 To run on odd cores within even packages:
196 $ hwloc-bind package:even.core:odd -- echo hello
198 To run on the first package, except on its second and fifth cores:
200 $ hwloc-bind package:0 ~package:0.core:1 ~package:0.core:4 -- echo
202 hello
203 To run anywhere except on the first package:
205 $ hwloc-bind all ~package:0 -- echo hello
207 To run on a core near the network interface named eth0:
209 $ hwloc-bind os=eth0 -- echo hello
211 To run on a core near the PCI device whose bus ID is 0000:01:02.0:
213 $ hwloc-bind pci=0000:01:02.0 -- echo hello
215 To bind memory on second memory node and run on first node (when sup‐
217 ported by the OS):
218 $ hwloc-bind --cpubind node:1 --membind node:0 -- echo hello
220 The --get option can report current bindings. This example shows nest‐
222 ing hwloc-bind invocations to set a binding and then report it:
223 $ hwloc-bind node:1.package:2 -- hwloc-bind --get
225 0x00004444,0x44000000
226 hwloc-calc may convert this output into actual objects, either with
228 logical or physical indexes:
229 $ hwloc-calc --physical -I pu `hwloc-bind --get`
231 26,30,34,38,42,46
232 $ hwloc-calc --logical -I pu `hwloc-bind --get` --sep " "
233 24 25 26 27 28 29
234 Locations may also be specified as a hex bit mask (typically generated
237 by hwloc-calc). For example:
238 $ hwloc-bind 0x00004444,0x44000000 -- echo hello
240 $ hwloc-bind `hwloc-calc node:1.package:2` -- echo hello
241 The current memory binding may also be reported:
243 $ hwloc-bind --membind node:1 --mempolicy interleave -- hwloc-bind
245 --get --membind
246 0x000000f0 (interleave)
247 Note that if the system is not NUMA, the reported string may indicate
249 that the process is bound to the entire system memory (e.g.,
250 "0xf...f").
251
254 If the graphics-enabled lstopo is available, use for instance
255 $ hwloc-bind core:2 -- lstopo --pid 0
257 to check what the result of your binding command actually is. lstopo
259 will graphically show where it is bound to by hwloc-bind.
260
262 Upon successful execution, hwloc-bind execs the command over itself.
263 The return value is therefore whatever the return value of the command
264 is.
265 hwloc-bind will return nonzero if any kind of error occurs, such as
267 (but not limited to): failure to parse the command line, failure to
268 retrieve process bindings, or lack of a command to execute.
269
271 hwloc(7), lstopo(1), hwloc-calc(1), hwloc-distrib(1)
2721.11.8 Sep 06, 2017 HWLOC-BIND(1)