Scroll to navigation

HWLOC-INFO(1) hwloc HWLOC-INFO(1)

NAME

hwloc-info - Show some information about some objects, a topology or supported features

SYNOPSIS

hwloc-info [ options ]... <object>...

hwloc-info [ options ] root hwloc-info [ options ] [ topology | levels | support ]

<object>... may be a space-separated list of target objects to query. The program reports all information about one object before looking at the next one:


$ hwloc-info core:2 package:1 pu:all
Core L#2
type = Core
...
Package L#1
...
PU L#0
...
PU L#1
...

The list may also contain special keywords such as:


root for the topology root object.


levels for information about topology levels (default if no target object is given, identical to --topology).


topology for topology info attributes stored in the root object (for forward compatibility with topology info attrs in hwloc 3.0).


support for information about supported features (identical to --support).

Real targets and special keywords may be combined:


$ hwloc-info core:2 topology pu:3 levels

Note that hwloc(7) provides a detailed explanation of the hwloc system and of valid <object> formats; it should be read before reading this man page.

OPTIONS

Report information specific objects. This is the default if some objects are given on the command-line.
Report a summary of the topology instead of about some specific objects. This is equivalent to passing levels as a target object on the command-line. This is the default if no object is given on the command-line.
Report the features that are supported by hwloc on the topology. The features are those available through the hwloc_topology_get_support() function. This is useful for verifying which CPU or memory binding options are supported by the current hwloc installation. This is equivalent to passing support as a target object on the command-line.
Read the topology from <path> instead of discovering the topology of the local machine.

If <path> is a file, it may be a XML file exported by a previous hwloc program. If <path> is "-", the standard input may be used as a XML file.

On Linux, <path> may be a directory containing the topology files gathered from another machine topology with hwloc-gather-topology.

On x86, <path> may be a directory containing a cpuid dump gathered with hwloc-gather-cpuid.

When the archivemount program is available, <path> may also be a tarball containing such Linux or x86 topology files.

Simulate a fake hierarchy (instead of discovering the topology on the local machine). If <specification> is "node:2 pu:3", the topology will contain two NUMA nodes with 3 processing units in each of them. The <specification> string must end with a number of PUs.
Enforce the input in the given format, among xml, fsroot, cpuid and synthetic.
Include additional detail.
Reduce the amount of details to show. A single summary line per object is displayed.
Only report the attribute of name <name> for each object (instead of all attributes). The name must exactly match what is usually reported by the program, for instance "complete cpuset" in "0.1: complete cpuset = %0x00ffff00".

Only the value is reported, any other prefix or object name is ignored, so that the output may easily be used by other tools.

This option also works on topology information but it is ignored for levels and support keywords.

Display information about the object as well as about all its ancestors up to the root of the topology. This is identical to --ancestor all
Only display the object ancestors that match the given type.

Some special values matching multiple types may also be given: kind=normal (CPU objects, including caches), kind=cpu (CPU objects, excluding caches), kind=cache (all caches, including memory-side caches), kind=memory (NUMA nodes or memory-side caches), kind=io (IO objects), kind=all (all objects). See also Object Kind in Terms and Definitions in the documentation. The prefix kind= may be omitted if there is no ambiguity.

If multiple ancestors match, they are reported from the deepest to the highest in the topology. Adding --first will only show the first one.

Display information about the object children.
Display information about the object descendants that match the given type.

Some special values matching multiple types may also be given: kind=normal (CPU objects, including caches), kind=cpu (CPU objects, excluding caches), kind=cache (all caches, including memory-side caches), kind=memory (NUMA nodes or memory-side caches), kind=io (IO objects), kind=all (all objects). See also Object Kind in Terms and Definitions in the documentation. The prefix kind= may be omitted if there is no ambiguity.

If multiple objects match, they are reported in a depth-first order (first child, then its children, etc., then second child, etc.). Adding --first will only show the first one.

Display information about the NUMA nodes that are local to the given object.
Change the flags used to select local NUMA nodes. Flags may be given as numeric values or as a comma-separated list of flag names that are passed to hwloc_get_local_numanode_objs(). Those names may be substrings of actual flag names as long as a single one matches. The default is 3 (or smaller,larger) which means NUMA nodes are displayed if their locality either contains or is contained in the locality of the given object.

This option enables --local-memory.

Enable the listing of local memory nodes with --local-memory, but only display the local nodes that have the best value for the memory attribute given by <name> (or as an index). If the memory attribute values depend on the initiator, the object given to hwloc-info is used as the initiator.

<name> may be suffixed with flags to tune the selection of best nodes, for instance as bandwidth,strict,default. default means that all local nodes are reported if no best could be found. strict means that nodes are selected only if their performance is the best for all the input CPUs. On a dual-socket machine with HBM in each socket, both HBMs are the best for their local socket, but not for the remote socket. Hence both HBM are also considered best for the entire machine by default, but none if strict.

For each input object, only report the first matching output object (first ancestor, first child, etc.).
When outputting object information, prefix each line with the index of the considered object within the input. For instance, if three cores were given in input, the output lines will be prefixed with "0: ", "1: " or "2: ". If --ancestor is also used, the prefix will be "X.Y: " where X is the index of the considered object within the input, and Y is the parent index (0 for the object itself, increasing towards the root of the topology).
Include objects disallowed by administrative limitations.
Restrict the topology to the given cpuset. This removes some PUs and their now-child-less parents.

Beware that restricting the PUs in a topology may change the logical indexes of many objects, including NUMA nodes.

Restrict the topology to the given nodeset (unless --restrict-flags specifies something different). This removes some NUMA nodes and their now-child-less parents.

Beware that restricting the NUMA nodes in a topology may change the logical indexes of many objects, including PUs.

Restrict the topology to the current process binding. This option requires the use of the actual current machine topology (or any other topology with --thissystem or with HWLOC_THISSYSTEM set to 1 in the environment).

Beware that restricting the topology may change the logical indexes of many objects, including PUs and NUMA nodes.

Enforce flags when restricting the topology. Flags may be given as numeric values or as a comma-separated list of flag names that are passed to hwloc_topology_restrict(). Those names may be substrings of actual flag names as long as a single one matches, for instance bynodeset,memless. The default is 0 (or none).
Filter objects of type <type>, or of any type if <type> is "all". "io", "cache" and "icache" are also supported.

<kind> specifies the filtering behavior. If "none" or not specified, all objects of the given type are removed. If "all", all objects are kept as usual. If "structure", objects are kept when they bring structure to the topology. If "important" (only applicable to I/O and Misc), only important objects are kept. See hwloc_topology_set_type_filter() for more details.

Do not show Instruction caches, only Data and Unified caches are considered. This is identical to --filter icache:none.
Do not show any I/O device or bridge. This is identical to --filter io:none. By default, common devices (GPUs, NICs, block devices, ...) and interesting bridges are shown.
Do not show any I/O bridge except hostbridges. This is identical to --filter bridge:none. By default, common devices (GPUs, NICs, block devices, ...) and interesting bridges are shown.
Show all I/O devices and bridges. This is identical to --filter io:all. By default, only common devices (GPUs, NICs, block devices, ...) and interesting bridges are shown.
Assume that the selected backend provides the topology for the system on which we are running. This is useful when using --restrict binding and loading a custom topology such as an XML file.
Detect topology as seen by process <pid>, i.e. as if process <pid> did the discovery itself. Note that this can for instance change the set of allowed processors. Also show this process current CPU binding by marking the corresponding PUs (in Green in the graphical output, see the COLORS section below, or by appending (binding) to the verbose text output). If 0 is given as pid, the current binding for the lstopo process will be shown.
Use OS/physical indexes instead of logical indexes for input.
Use logical indexes instead of physical/OS indexes for input (default).
Report version and exit.
Display help message and exit.

DESCRIPTION

hwloc-info displays information about the specified objects. It is intended to be used with tools such as grep for filtering certain attribute lines. When no object is specified, or when --topology is passed, hwloc-info prints a summary of the topology. When --support is passed, hwloc-info lists the supported features for the topology.

Objects may be specified as location tuples, as explained in hwloc(7). However hexadecimal bitmasks are not accepted since they may correspond to multiple objects.

NOTE: It is highly recommended that you read the hwloc(7) overview page before reading this man page. Most of the concepts described in hwloc(7) directly apply to the hwloc-calc utility.

EXAMPLES

To display information about each package:


$ hwloc-info package:all
Package L#0
logical index = 0
...

To display information about the core whose physical index is 2:


$ hwloc-info -p core:2
Core L#1
logical index = 1
os index = 2
...

To list the OS devices that are of subtype OpenCL:


$ hwloc-info -s "os[OpenCL]:all"
CoProc:6
CoProc:8

To find the PCI bus ID of PCI devices containing OpenCL devices:


$ hwloc-info --ancestor PCI --get-attr "attr PCI bus id" 'os[opencl]:all'
0000:05:00.0
0000:42:00.0

To list the NUMA nodes that are local a PU:


$ hwloc-info --local-memory pu:25
NUMANode L#6 = local memory #0 of PU L#25
type = NUMANode
...
NUMANode L#7 = local memory #1 of PU L#25
type = NUMANode
...

To show the best-bandwidth node(s) among NUMA nodes local to a PU:


$ hwloc-info --local-memory --best-memattr bandwidth pu:25
NUMANode L#7 = local memory #1 of PU L#25
type = NUMANode
...

to find where a NUMA node is attached in the hierarchy of CPU cores:


$ hwloc-info --ancestor kind=normal --first -s numa:1
Package:0

To see levels and topology info attributes stored in the root object:


$ hwloc-info levels topology
depth 0: 1 Machine (type #0)
depth 1: 1 Package (type #1)
depth 2: 2 Core (type #2)
depth 3: 4 PU (type #3)
Special depth -3: 1 NUMANode (type #13)
info Backend = Linux
info LinuxCgroup = /user/622
info Architecture = x86_64
info hwlocVersion = 3.0.0a1-git

SEE ALSO

hwloc(7), lstopo(1), hwloc-calc(1), hwloc-bind(1), hwloc-ps(1)

September 26, 2024 2.11.2