Debian Codegraph @

doc/doxygen-doc/man/man3/hwlocality_memattrs.3

raw
1 .TH "hwlocality_memattrs" 3 "Version 2.14.0" "Hardware Locality (hwloc)" \" -*- nroff -*-
2 .ad l
3 .nh
4 .SH NAME
5 hwlocality_memattrs \- Comparing memory node attributes for finding where to allocate on
6 .SH SYNOPSIS
7 .br
8 .PP
9 .SS "Data Structures"
10
11 .in +1c
12 .ti -1c
13 .RI "struct \fBhwloc_location\fP"
14 .br
15 .in -1c
16 .SS "Typedefs"
17
18 .in +1c
19 .ti -1c
20 .RI "typedef unsigned \fBhwloc_memattr_id_t\fP"
21 .br
22 .in -1c
23 .SS "Enumerations"
24
25 .in +1c
26 .ti -1c
27 .RI "enum \fBhwloc_memattr_id_e\fP { \fBHWLOC_MEMATTR_ID_CAPACITY\fP, \fBHWLOC_MEMATTR_ID_LOCALITY\fP, \fBHWLOC_MEMATTR_ID_BANDWIDTH\fP, \fBHWLOC_MEMATTR_ID_READ_BANDWIDTH\fP, \fBHWLOC_MEMATTR_ID_WRITE_BANDWIDTH\fP, \fBHWLOC_MEMATTR_ID_LATENCY\fP, \fBHWLOC_MEMATTR_ID_READ_LATENCY\fP, \fBHWLOC_MEMATTR_ID_WRITE_LATENCY\fP, \fBHWLOC_MEMATTR_ID_MAX\fP }"
28 .br
29 .ti -1c
30 .RI "enum \fBhwloc_location_type_e\fP { \fBHWLOC_LOCATION_TYPE_CPUSET\fP, \fBHWLOC_LOCATION_TYPE_OBJECT\fP }"
31 .br
32 .ti -1c
33 .RI "enum \fBhwloc_local_numanode_flag_e\fP { \fBHWLOC_LOCAL_NUMANODE_FLAG_LARGER_LOCALITY\fP, \fBHWLOC_LOCAL_NUMANODE_FLAG_SMALLER_LOCALITY\fP, \fBHWLOC_LOCAL_NUMANODE_FLAG_INTERSECT_LOCALITY\fP, \fBHWLOC_LOCAL_NUMANODE_FLAG_ALL\fP }"
34 .br
35 .in -1c
36 .SS "Functions"
37
38 .in +1c
39 .ti -1c
40 .RI "int \fBhwloc_memattr_get_by_name\fP (\fBhwloc_topology_t\fP topology, const char *name, \fBhwloc_memattr_id_t\fP *id)"
41 .br
42 .ti -1c
43 .RI "int \fBhwloc_get_local_numanode_objs\fP (\fBhwloc_topology_t\fP topology, struct \fBhwloc_location\fP *location, unsigned *nr, \fBhwloc_obj_t\fP *nodes, unsigned long flags)"
44 .br
45 .ti -1c
46 .RI "int \fBhwloc_topology_get_default_nodeset\fP (\fBhwloc_topology_t\fP topology, \fBhwloc_nodeset_t\fP nodeset, unsigned long flags)"
47 .br
48 .ti -1c
49 .RI "int \fBhwloc_memattr_get_value\fP (\fBhwloc_topology_t\fP topology, \fBhwloc_memattr_id_t\fP attribute, \fBhwloc_obj_t\fP target_node, struct \fBhwloc_location\fP *initiator, unsigned long flags, hwloc_uint64_t *value)"
50 .br
51 .ti -1c
52 .RI "int \fBhwloc_memattr_get_best_target\fP (\fBhwloc_topology_t\fP topology, \fBhwloc_memattr_id_t\fP attribute, struct \fBhwloc_location\fP *initiator, unsigned long flags, \fBhwloc_obj_t\fP *best_target, hwloc_uint64_t *value)"
53 .br
54 .ti -1c
55 .RI "int \fBhwloc_memattr_get_best_initiator\fP (\fBhwloc_topology_t\fP topology, \fBhwloc_memattr_id_t\fP attribute, \fBhwloc_obj_t\fP target_node, unsigned long flags, struct \fBhwloc_location\fP *best_initiator, hwloc_uint64_t *value)"
56 .br
57 .ti -1c
58 .RI "int \fBhwloc_memattr_get_targets\fP (\fBhwloc_topology_t\fP topology, \fBhwloc_memattr_id_t\fP attribute, struct \fBhwloc_location\fP *initiator, unsigned long flags, unsigned *nr, \fBhwloc_obj_t\fP *targets, hwloc_uint64_t *values)"
59 .br
60 .ti -1c
61 .RI "int \fBhwloc_memattr_get_initiators\fP (\fBhwloc_topology_t\fP topology, \fBhwloc_memattr_id_t\fP attribute, \fBhwloc_obj_t\fP target_node, unsigned long flags, unsigned *nr, struct \fBhwloc_location\fP *initiators, hwloc_uint64_t *values)"
62 .br
63 .in -1c
64 .SH "Detailed Description"
65 .PP
66 Platforms with heterogeneous memory require ways to decide whether a buffer should be allocated on "fast" memory (such as HBM), "normal" memory (DDR) or even "slow" but large-capacity memory (non-volatile memory)\&. These memory nodes are called "Targets" while the CPU accessing them is called the "Initiator"\&. Access performance depends on their locality (NUMA platforms) as well as the intrinsic performance of the targets (heterogeneous platforms)\&.
67
68 .PP
69 The following attributes describe the performance of memory accesses from an Initiator to a memory Target, for instance their latency or bandwidth\&. Initiators performing these memory accesses are usually some PUs or Cores (described as a CPU set)\&. Hence a Core may choose where to allocate a memory buffer by comparing the attributes of different target memory nodes nearby\&.
70
71 .PP
72 There are also some attributes that are system-wide\&. Their value does not depend on a specific initiator performing an access\&. The memory node Capacity is an example of such attribute without initiator\&.
73
74 .PP
75 One way to use this API is to start with a cpuset describing the Cores where a program is bound\&. The best target NUMA node for allocating memory in this program on these Cores may be obtained by passing this cpuset as an initiator to \fBhwloc_memattr_get_best_target()\fP with the relevant memory attribute\&. For instance, if the code is latency limited, use the Latency attribute\&.
76
77 .PP
78 A more flexible approach consists in getting the list of local NUMA nodes by passing this cpuset to \fBhwloc_get_local_numanode_objs()\fP\&. Attribute values for these nodes, if any, may then be obtained with \fBhwloc_memattr_get_value()\fP and manually compared with the desired criteria\&.
79
80 .PP
81 Memory attributes are also used internally to build Memory Tiers which provide an easy way to distinguish NUMA nodes of different kinds, as explained in \fBHeterogeneous Memory\fP\&.
82
83 .PP
84 Beside tiers, hwloc defines a set of "default" nodes where normal memory allocations should be made from (see \fBhwloc_topology_get_default_nodeset()\fP)\&. This is also useful for dividing the machine into a set of non-overlapping NUMA domains, for instance for binding tasks per domain\&.
85
86 .PP
87 \fBSee also\fP
88 .RS 4
89 An example is available in doc/examples/memory-attributes\&.c in the source tree\&.
90 .RE
91 .PP
92 \fBNote\fP
93 .RS 4
94 The API also supports specific objects as initiator, but it is currently not used internally by hwloc\&. Users may for instance use it to provide custom performance values for host memory accesses performed by GPUs\&.
95
96 .PP
97 The interface actually also accepts targets that are not NUMA nodes\&.
98 .RE
99 .PP
100
101 .SH "Typedef Documentation"
102 .PP
103 .SS "typedef unsigned \fBhwloc_memattr_id_t\fP"
104
105 .PP
106 A memory attribute identifier\&. hwloc predefines some commonly-used attributes in \fBhwloc_memattr_id_e\fP\&. One may then dynamically register custom ones with \fBhwloc_memattr_register()\fP, they will be assigned IDs immediately after the predefined ones\&. See \fBManaging memory attributes\fP for more information about existing attribute IDs\&.
107 .SH "Enumeration Type Documentation"
108 .PP
109 .SS "enum \fBhwloc_local_numanode_flag_e\fP"
110
111 .PP
112 Flags for selecting target NUMA nodes\&.
113 .PP
114 \fBEnumerator\fP
115 .in +1c
116 .TP
117 \f(BIHWLOC_LOCAL_NUMANODE_FLAG_LARGER_LOCALITY \fP
118 Select NUMA nodes whose locality is larger than the given cpuset\&. For instance, if a single PU (or its cpuset) is given in \fRinitiator\fP, select all nodes close to the package that contains this PU\&.
119 .TP
120 \f(BIHWLOC_LOCAL_NUMANODE_FLAG_SMALLER_LOCALITY \fP
121 Select NUMA nodes whose locality is smaller than the given cpuset\&. For instance, if a package (or its cpuset) is given in \fRinitiator\fP, also select nodes that are attached to only a half of that package\&.
122 .TP
123 \f(BIHWLOC_LOCAL_NUMANODE_FLAG_INTERSECT_LOCALITY \fP
124 Select NUMA nodes whose locality intersects the given cpuset\&. This includes larger and smaller localities as well as localities that are partially included\&. For instance, if the locality is one core of both packages, a NUMA node local to one package is neither larger nor smaller than this locality, but it intersects it\&.
125 .TP
126 \f(BIHWLOC_LOCAL_NUMANODE_FLAG_ALL \fP
127 Select all NUMA nodes in the topology\&. The initiator \fRinitiator\fP is ignored\&.
128 .SS "enum \fBhwloc_location_type_e\fP"
129
130 .PP
131 Type of location\&.
132 .PP
133 \fBEnumerator\fP
134 .in +1c
135 .TP
136 \f(BIHWLOC_LOCATION_TYPE_CPUSET \fP
137 Location is given as a cpuset, in the location cpuset union field\&.
138 .TP
139 \f(BIHWLOC_LOCATION_TYPE_OBJECT \fP
140 Location is given as an object, in the location object union field\&.
141 .SS "enum \fBhwloc_memattr_id_e\fP"
142
143 .PP
144 Predefined memory attribute IDs\&. See \fBhwloc_memattr_id_t\fP for the generic definition of IDs for predefined or custom attributes\&.
145 .PP
146 \fBEnumerator\fP
147 .in +1c
148 .TP
149 \f(BIHWLOC_MEMATTR_ID_CAPACITY \fP
150 The "Capacity" is returned in bytes (local_memory attribute in objects)\&. Best capacity nodes are nodes with \fBhigher capacity\fP\&.
151
152 .PP
153 No initiator is involved when looking at this attribute\&. The corresponding attribute flags are \fBHWLOC_MEMATTR_FLAG_HIGHER_FIRST\fP\&.
154
155 .PP
156 Capacity values may not be modified using \fBhwloc_memattr_set_value()\fP\&.
157 .TP
158 \f(BIHWLOC_MEMATTR_ID_LOCALITY \fP
159 The "Locality" is returned as the number of PUs in that locality (e\&.g\&. the weight of its cpuset)\&. Best locality nodes are nodes with \fBsmaller locality\fP (nodes that are local to very few PUs)\&. Poor locality nodes are nodes with larger locality (nodes that are local to the entire machine)\&.
160
161 .PP
162 No initiator is involved when looking at this attribute\&. The corresponding attribute flags are \fBHWLOC_MEMATTR_FLAG_HIGHER_FIRST\fP\&.
163
164 .PP
165 Locality values may not be modified using \fBhwloc_memattr_set_value()\fP\&.
166 .TP
167 \f(BIHWLOC_MEMATTR_ID_BANDWIDTH \fP
168 The "Bandwidth" is returned in MiB/s, as seen from the given initiator location\&. Best bandwidth nodes are nodes with \fBhigher bandwidth\fP\&.
169
170 .PP
171 The corresponding attribute flags are \fBHWLOC_MEMATTR_FLAG_HIGHER_FIRST\fP and \fBHWLOC_MEMATTR_FLAG_NEED_INITIATOR\fP\&.
172
173 .PP
174 This is the average bandwidth for read and write accesses\&. If the platform provides individual read and write bandwidths but no explicit average value, hwloc computes and returns the average\&.
175 .TP
176 \f(BIHWLOC_MEMATTR_ID_READ_BANDWIDTH \fP
177 The "ReadBandwidth" is returned in MiB/s, as seen from the given initiator location\&. Best bandwidth nodes are nodes with \fBhigher bandwidth\fP\&.
178
179 .PP
180 The corresponding attribute flags are \fBHWLOC_MEMATTR_FLAG_HIGHER_FIRST\fP and \fBHWLOC_MEMATTR_FLAG_NEED_INITIATOR\fP\&.
181 .TP
182 \f(BIHWLOC_MEMATTR_ID_WRITE_BANDWIDTH \fP
183 The "WriteBandwidth" is returned in MiB/s, as seen from the given initiator location\&. Best bandwidth nodes are nodes with \fBhigher bandwidth\fP\&.
184
185 .PP
186 The corresponding attribute flags are \fBHWLOC_MEMATTR_FLAG_HIGHER_FIRST\fP and \fBHWLOC_MEMATTR_FLAG_NEED_INITIATOR\fP\&.
187 .TP
188 \f(BIHWLOC_MEMATTR_ID_LATENCY \fP
189 The "Latency" is returned as nanoseconds, as seen from the given initiator location\&. Best latency nodes are nodes with \fBsmaller latency\fP\&.
190
191 .PP
192 The corresponding attribute flags are \fBHWLOC_MEMATTR_FLAG_LOWER_FIRST\fP and \fBHWLOC_MEMATTR_FLAG_NEED_INITIATOR\fP\&.
193
194 .PP
195 This is the average latency for read and write accesses\&. If the platform provides individual read and write latencies but no explicit average value, hwloc computes and returns the average\&.
196 .TP
197 \f(BIHWLOC_MEMATTR_ID_READ_LATENCY \fP
198 The "ReadLatency" is returned as nanoseconds, as seen from the given initiator location\&. Best latency nodes are nodes with \fBsmaller latency\fP\&.
199
200 .PP
201 The corresponding attribute flags are \fBHWLOC_MEMATTR_FLAG_LOWER_FIRST\fP and \fBHWLOC_MEMATTR_FLAG_NEED_INITIATOR\fP\&.
202 .TP
203 \f(BIHWLOC_MEMATTR_ID_WRITE_LATENCY \fP
204 The "WriteLatency" is returned as nanoseconds, as seen from the given initiator location\&. Best latency nodes are nodes with \fBsmaller latency\fP\&.
205
206 .PP
207 The corresponding attribute flags are \fBHWLOC_MEMATTR_FLAG_LOWER_FIRST\fP and \fBHWLOC_MEMATTR_FLAG_NEED_INITIATOR\fP\&.
208 .SH "Function Documentation"
209 .PP
210 .SS "int hwloc_get_local_numanode_objs (\fBhwloc_topology_t\fP topology, struct \fBhwloc_location\fP * location, unsigned * nr, \fBhwloc_obj_t\fP * nodes, unsigned long flags)"
211
212 .PP
213 Return an array of local NUMA nodes\&. By default only select the NUMA nodes whose locality is exactly the given \fRlocation\fP\&. More nodes may be selected if additional flags are given as a OR'ed set of \fBhwloc_local_numanode_flag_e\fP\&.
214
215 .PP
216 If \fRlocation\fP is given as an explicit object, its CPU set is used to find NUMA nodes with the corresponding locality\&. If the object does not have a CPU set (e\&.g\&. I/O object), the CPU parent (where the I/O object is attached) is used\&.
217
218 .PP
219 On input, \fRnr\fP points to the number of nodes that may be stored in the \fRnodes\fP array\&. On output, \fRnr\fP will be changed to the number of stored nodes, or the number of nodes that would have been stored if there were enough room\&.
220
221 .PP
222 \fBReturns\fP
223 .RS 4
224 0 on success or -1 on error\&.
225 .RE
226 .PP
227 \fBNote\fP
228 .RS 4
229 Some of these NUMA nodes may not have any memory attribute values and hence not be reported as actual targets in other functions\&.
230
231 .PP
232 The number of NUMA nodes in the topology (obtained by \fBhwloc_bitmap_weight()\fP on the root object nodeset) may be used to allocate the \fRnodes\fP array\&.
233
234 .PP
235 When an object CPU set is given as locality, for instance a Package, and when flags contain both \fBHWLOC_LOCAL_NUMANODE_FLAG_LARGER_LOCALITY\fP and \fBHWLOC_LOCAL_NUMANODE_FLAG_SMALLER_LOCALITY\fP, the returned array corresponds to the nodeset of that object\&.
236 .RE
237 .PP
238
239 .SS "int hwloc_memattr_get_best_initiator (\fBhwloc_topology_t\fP topology, \fBhwloc_memattr_id_t\fP attribute, \fBhwloc_obj_t\fP target_node, unsigned long flags, struct \fBhwloc_location\fP * best_initiator, hwloc_uint64_t * value)"
240
241 .PP
242 Return the best initiator for the given attribute and target NUMA node\&. If \fRvalue\fP is non \fRNULL\fP, the corresponding value is returned there\&.
243
244 .PP
245 If multiple initiators have the same attribute values, only one is returned (and there is no way to clarify how that one is chosen)\&. Applications that want to detect initiators with identical/similar values, or that want to look at values for multiple attributes, should rather get all values using \fBhwloc_memattr_get_value()\fP and manually select the initiator they consider the best\&.
246
247 .PP
248 The returned initiator should not be modified or freed, it belongs to the topology\&.
249
250 .PP
251 \fRtarget_node\fP cannot be \fRNULL\fP\&.
252
253 .PP
254 \fRflags\fP must be \fR0\fP for now\&.
255
256 .PP
257 \fBReturns\fP
258 .RS 4
259 0 on success\&.
260
261 .PP
262 -1 with errno set to \fRENOENT\fP if there are no matching initiators\&.
263
264 .PP
265 -1 with errno set to \fREINVAL\fP if the attribute does not relate to a specific initiator (it does not have the flag \fBHWLOC_MEMATTR_FLAG_NEED_INITIATOR\fP)\&.
266 .RE
267 .PP
268
269 .SS "int hwloc_memattr_get_best_target (\fBhwloc_topology_t\fP topology, \fBhwloc_memattr_id_t\fP attribute, struct \fBhwloc_location\fP * initiator, unsigned long flags, \fBhwloc_obj_t\fP * best_target, hwloc_uint64_t * value)"
270
271 .PP
272 Return the best target NUMA node for the given attribute and initiator\&. If the attribute does not relate to a specific initiator (it does not have the flag \fBHWLOC_MEMATTR_FLAG_NEED_INITIATOR\fP), location \fRinitiator\fP is ignored and may be \fRNULL\fP\&.
273
274 .PP
275 If \fRvalue\fP is non \fRNULL\fP, the corresponding value is returned there\&.
276
277 .PP
278 If multiple targets have the same attribute values, only one is returned (and there is no way to clarify how that one is chosen)\&. Applications that want to detect targets with identical/similar values, or that want to look at values for multiple attributes, should rather get all values using \fBhwloc_memattr_get_value()\fP and manually select the target they consider the best\&.
279
280 .PP
281 \fRflags\fP must be \fR0\fP for now\&.
282
283 .PP
284 \fBReturns\fP
285 .RS 4
286 0 on success\&.
287
288 .PP
289 -1 with errno set to \fRENOENT\fP if there are no matching targets\&.
290
291 .PP
292 -1 with errno set to \fREINVAL\fP if flags are invalid, or no such attribute exists\&.
293 .RE
294 .PP
295 \fBNote\fP
296 .RS 4
297 The initiator \fRinitiator\fP should be of type \fBHWLOC_LOCATION_TYPE_CPUSET\fP when refering to accesses performed by CPU cores\&. \fBHWLOC_LOCATION_TYPE_OBJECT\fP is currently unused internally by hwloc, but users may for instance use it to provide custom information about host memory accesses performed by GPUs\&.
298 .RE
299 .PP
300
301 .SS "int hwloc_memattr_get_by_name (\fBhwloc_topology_t\fP topology, const char * name, \fBhwloc_memattr_id_t\fP * id)"
302
303 .PP
304 Return the identifier of the memory attribute with the given name\&.
305 .PP
306 \fBReturns\fP
307 .RS 4
308 0 on success\&.
309
310 .PP
311 -1 with errno set to \fREINVAL\fP if no such attribute exists\&.
312 .RE
313 .PP
314
315 .SS "int hwloc_memattr_get_initiators (\fBhwloc_topology_t\fP topology, \fBhwloc_memattr_id_t\fP attribute, \fBhwloc_obj_t\fP target_node, unsigned long flags, unsigned * nr, struct \fBhwloc_location\fP * initiators, hwloc_uint64_t * values)"
316
317 .PP
318 Return the initiators that have values for a given attribute for a specific target NUMA node\&. Return initiators for the given attribute and target node in the \fRinitiators\fP array\&. If \fRvalues\fP is not \fRNULL\fP, the corresponding attribute values are stored in the array it points to\&.
319
320 .PP
321 On input, \fRnr\fP points to the number of initiators that may be stored in the array \fRinitiators\fP (and \fRvalues\fP)\&. On output, \fRnr\fP points to the number of initiators (and values) that were actually found, even if some of them couldn't be stored in the array\&. Initiators that couldn't be stored are ignored, but the function still returns success (\fR0\fP)\&. The caller may find out by comparing the value pointed by \fRnr\fP before and after the function call\&.
322
323 .PP
324 The returned initiators should not be modified or freed, they belong to the topology\&.
325
326 .PP
327 \fRtarget_node\fP cannot be \fRNULL\fP\&.
328
329 .PP
330 \fRflags\fP must be \fR0\fP for now\&.
331
332 .PP
333 If the attribute does not relate to a specific initiator (it does not have the flag \fBHWLOC_MEMATTR_FLAG_NEED_INITIATOR\fP), no initiator is returned\&.
334
335 .PP
336 \fBReturns\fP
337 .RS 4
338 0 on success or -1 on error\&.
339 .RE
340 .PP
341 \fBNote\fP
342 .RS 4
343 This function is meant for tools and debugging (listing internal information) rather than for application queries\&. Applications should rather select useful NUMA nodes with \fBhwloc_get_local_numanode_objs()\fP and then look at their attribute values for some relevant initiators\&.
344 .RE
345 .PP
346
347 .SS "int hwloc_memattr_get_targets (\fBhwloc_topology_t\fP topology, \fBhwloc_memattr_id_t\fP attribute, struct \fBhwloc_location\fP * initiator, unsigned long flags, unsigned * nr, \fBhwloc_obj_t\fP * targets, hwloc_uint64_t * values)"
348
349 .PP
350 Return the target NUMA nodes that have some values for a given attribute\&. Return targets for the given attribute in the \fRtargets\fP array (for the given initiator if any)\&. If \fRvalues\fP is not \fRNULL\fP, the corresponding attribute values are stored in the array it points to\&.
351
352 .PP
353 On input, \fRnr\fP points to the number of targets that may be stored in the array \fRtargets\fP (and \fRvalues\fP)\&. On output, \fRnr\fP points to the number of targets (and values) that were actually found, even if some of them couldn't be stored in the array\&. Targets that couldn't be stored are ignored, but the function still returns success (\fR0\fP)\&. The caller may find out by comparing the value pointed by \fRnr\fP before and after the function call\&.
354
355 .PP
356 The returned targets should not be modified or freed, they belong to the topology\&.
357
358 .PP
359 Argument \fRinitiator\fP is ignored if the attribute does not relate to a specific initiator (it does not have the flag \fBHWLOC_MEMATTR_FLAG_NEED_INITIATOR\fP)\&. Otherwise \fRinitiator\fP may be non \fRNULL\fP to report only targets that have a value for that initiator\&.
360
361 .PP
362 \fRflags\fP must be \fR0\fP for now\&.
363
364 .PP
365 \fBNote\fP
366 .RS 4
367 This function is meant for tools and debugging (listing internal information) rather than for application queries\&. Applications should rather select useful NUMA nodes with \fBhwloc_get_local_numanode_objs()\fP and then look at their attribute values\&.
368 .RE
369 .PP
370 \fBReturns\fP
371 .RS 4
372 0 on success or -1 on error\&.
373 .RE
374 .PP
375 \fBNote\fP
376 .RS 4
377 The initiator \fRinitiator\fP should be of type \fBHWLOC_LOCATION_TYPE_CPUSET\fP when referring to accesses performed by CPU cores\&. \fBHWLOC_LOCATION_TYPE_OBJECT\fP is currently unused internally by hwloc, but users may for instance use it to provide custom information about host memory accesses performed by GPUs\&.
378 .RE
379 .PP
380
381 .SS "int hwloc_memattr_get_value (\fBhwloc_topology_t\fP topology, \fBhwloc_memattr_id_t\fP attribute, \fBhwloc_obj_t\fP target_node, struct \fBhwloc_location\fP * initiator, unsigned long flags, hwloc_uint64_t * value)"
382
383 .PP
384 Return an attribute value for a specific target NUMA node\&. If the attribute does not relate to a specific initiator (it does not have the flag \fBHWLOC_MEMATTR_FLAG_NEED_INITIATOR\fP), location \fRinitiator\fP is ignored and may be \fRNULL\fP\&.
385
386 .PP
387 \fRtarget_node\fP cannot be \fRNULL\fP\&. If \fRattribute\fP is \fBHWLOC_MEMATTR_ID_CAPACITY\fP, \fRtarget_node\fP must be a NUMA node\&. If it is \fBHWLOC_MEMATTR_ID_LOCALITY\fP, \fRtarget_node\fP must have a CPU set\&.
388
389 .PP
390 \fRflags\fP must be \fR0\fP for now\&.
391
392 .PP
393 \fBReturns\fP
394 .RS 4
395 0 on success\&.
396
397 .PP
398 -1 on error, for instance with errno set to \fREINVAL\fP if flags are invalid or no such attribute exists\&.
399 .RE
400 .PP
401 \fBNote\fP
402 .RS 4
403 The initiator \fRinitiator\fP should be of type \fBHWLOC_LOCATION_TYPE_CPUSET\fP when refering to accesses performed by CPU cores\&. \fBHWLOC_LOCATION_TYPE_OBJECT\fP is currently unused internally by hwloc, but users may for instance use it to provide custom information about host memory accesses performed by GPUs\&.
404 .RE
405 .PP
406
407 .SS "int hwloc_topology_get_default_nodeset (\fBhwloc_topology_t\fP topology, \fBhwloc_nodeset_t\fP nodeset, unsigned long flags)"
408
409 .PP
410 Return the set of default NUMA nodes\&. In machines with heterogeneous memory, some NUMA nodes are considered the default ones, i\&.e\&. where basic allocations should be made from\&. These are usually DRAM nodes\&.
411
412 .PP
413 Other nodes may be reserved for specific use (I/O device memory, e\&.g\&. GPU memory), small but high performance (HBM), large but slow memory (NVM), etc\&. Buffers should usually not be allocated from there unless explicitly required\&.
414
415 .PP
416 This function fills \fRnodeset\fP with the bits of NUMA nodes considered default\&.
417
418 .PP
419 It is guaranteed that these nodes have non-intersecting CPU sets, i\&.e\&. cores may not have multiple local NUMA nodes anymore\&. Hence this may be used to iterate over the platform divided into separate NUMA localities, for instance for binding one task per NUMA domain\&.
420
421 .PP
422 Any core that had some local NUMA node(s) in the initial topology should still have one in the default nodeset\&. Corner cases where this would be wrong consist in asymmetric platforms with missing DRAM nodes, or topologies that were already restricted to less NUMA nodes\&.
423
424 .PP
425 The returned nodeset may be passed to \fBhwloc_topology_restrict()\fP with \fBHWLOC_RESTRICT_FLAG_BYNODESET\fP to remove all non-default nodes from the topology\&. The resulting topology will be easier to use when iterating over (now homogeneous) NUMA nodes\&.
426
427 .PP
428 The heuristics for finding default nodes relies on memory tiers and subtypes (see \fBHeterogeneous Memory\fP) as well as the assumption that hardware vendors list default nodes first in hardware tables\&.
429
430 .PP
431 \fRflags\fP must be \fR0\fP for now\&.
432
433 .PP
434 \fBReturns\fP
435 .RS 4
436 0 on success\&.
437
438 .PP
439 -1 on error\&.
440 .RE
441 .PP
442 \fBNote\fP
443 .RS 4
444 The returned nodeset usually contains all nodes from a single memory tier, likely the DRAM one\&.
445
446 .PP
447 The returned nodeset is included in the list of available nodes returned by \fBhwloc_topology_get_topology_nodeset()\fP\&. It is strictly smaller if the machine has heterogeneous memory\&.
448
449 .PP
450 The heuristics may return a suboptimal set of nodes if hwloc could not guess memory types and/or if some default nodes were removed earlier from the topology (e\&.g\&. with \fBhwloc_topology_restrict()\fP)\&.
451 .RE
452 .PP
453
454 .SH "Author"
455 .PP
456 Generated automatically by Doxygen for Hardware Locality (hwloc) from the source code\&.