"Cpusets provide a mechanism for assigning a set of CPUs and Memory
Nodes to a set of tasks."
"Control Groups provide a mechanism for aggregating/partitioning sets
of tasks, and all their future children, into hierarchical groups
with specialized behaviour."
To be clear, hwloc supports all of the above concepts. It is simply worth noting that they are different things.
Locations can be specified in multiple ways:
Indexes may also be specified as ranges.
x-y enumerates from index x to y.
x:y enumerates y objects starting from index x
(wrapping around the end of the index range if needed).
x- enumerates all objects starting from index x.
all, odd, and even are also supported for listing
all objects, or only those with odd or even indexes.
Chaining multiple tuples together in the more general form
object1:index[.object2:index2[...]]
is permissable. While the first tuple's object may appear anywhere in
the topology, the Nth tuple's object must have a shallower topology
depth than the (N+1)th tuple's object. Put simply: as you move right
in a tuple chain, objects must go deeper in the topology tree.
When using logical indexes (which is the default),
indexes specified in chained tuples are relative to the scope of the
parent object. For example, "package:0.core:1" refers to the second
core in the first package.
When using OS/physical indexes, the first object matching the given
index is used.
PCI and OS devices may also be designed using their identifier.
For example, "pci=02:03.1" is the PCI device with bus ID "02:03.1".
"os=eth0" is the network interface whose software name is "eth0".
PCI devices may also be filtered based on their vendor and/or device IDs,
for instance "pci[15b3:]:2" for the third Mellanox PCI device (vendor ID 0x15b3).
OS devices may also be filtered based on their subtype,
for instance "os[gpu]:all" for all GPU OS devices.
"all" and "root" are special locations consisting in the root object in tree. It contains the entire current topology.
Some tools directly operate on these objects (e.g. hwloc-info and hwloc-annotate). They do not support hexadecimal locations because each location may correspond to multiple objects. For instance, there can be exactly one L3 cache per package and NUMA node, which means it's the same location. If multiple locations are given on the command-line, these tools will operation on each location individually and consecutively.
Some other tools internally manipulate objects as sets (e.g. hwloc-calc and hwloc-bind). They translate each input location into a hexidecimal location. When I/O or Misc objects are used, they are translated into the set of processors (or NUMA nodes) that are close to the given object (because I/O or Misc objects do not contain processors or NUMA nodes).
If multiple locations are specified on the command-line (delimited by whitespace), they are combined (the overall location is wider). If prefixed with "~", the given location will be cleared instead of added to the current list of locations. If prefixed with "x", the given location will be and'ed instead of added to the current list. If prefixed with "^", the given location will be xor'ed.
More complex operations may be performed by using hwloc-calc to compute intermediate values.
Objects in tuples can be any of the following strings (listed from "biggest" to "smallest"):
osdev, pcidev, bridge, and misc may also be used to specify special devices although some of them have dedicated identification ways as explained in Location Specification.
Finally, note that an object can be denoted by its numeric "depth" in the topology graph.
Logical indexes are relative to the object order in the output from the lstopo command. They always start with 0 and increment by 1 for each successive object.
Physical indexes are how the operating system refers to objects. Note that while physical indexes are non-negative integer values, the hardware and/or operating system may choose arbitrary values -- they may not start with 0, and successive objects may not have consecutive values.
For example, if the first few lines of lstopo -p output are the following:
Machine (47GB)
NUMANode P#0 (24GB) + Package P#0 + L3 (12MB)
L2 (256KB) + L1 (32KB) + Core P#0 + PU P#0
L2 (256KB) + L1 (32KB) + Core P#1 + PU P#0
L2 (256KB) + L1 (32KB) + Core P#2 + PU P#0
L2 (256KB) + L1 (32KB) + Core P#8 + PU P#0
L2 (256KB) + L1 (32KB) + Core P#9 + PU P#0
L2 (256KB) + L1 (32KB) + Core P#10 + PU P#0
NUMANode P#1 (24GB) + Package P#1 + L3 (12MB)
L2 (256KB) + L1 (32KB) + Core P#0 + PU P#0
L2 (256KB) + L1 (32KB) + Core P#1 + PU P#0
L2 (256KB) + L1 (32KB) + Core P#2 + PU P#0
L2 (256KB) + L1 (32KB) + Core P#8 + PU P#0
L2 (256KB) + L1 (32KB) + Core P#9 + PU P#0
L2 (256KB) + L1 (32KB) + Core P#10 + PU P#0
In this example, the first core on the second package is logically number 6 (i.e., logically the 7th core, starting from 0). Its physical index is 0, but note that another core also has a physical index of 0. Hence, physical indexes may only be relevant within the scope of their parent (or set of ancestors). In this example, to uniquely identify logical core 6 with physical indexes, you must specify (at a minimum) both a package and a core: package 1, core 0.
Index values, regardless of whether they are logical or physical, can be expressed in several different forms (where X, Y, and N are positive integers):
REMEMBER: hwloc's command line tools accept logical indexes for location values by default. Use --physical and --logical to switch from one mode to another.
hwloc has many C API functions, each of which have their own man page. Some top-level man pages are also provided, grouping similar functions together. A few good places to start might include: hwlocality_objects(3), hwlocality_types(3), hwlocality_creation(3), hwlocality_cpuset(3), hwlocality_information(3), and hwlocality_binding(3).
For a listing of all available hwloc man pages, look at all "hwloc*" files in the man1 and man3 directories.