Open MPI logo

Portable Hardware Locality (hwloc) Documentation: v2.11.2

  |   Home   |   Support   |   FAQ   |  
rsmi.h
1/*
2 * Copyright © 2012-2023 Inria. All rights reserved.
3 * Copyright (c) 2020, Advanced Micro Devices, Inc. All rights reserved.
4 * Written by Advanced Micro Devices,
5 * See COPYING in top-level directory.
6 */
7
15#ifndef HWLOC_RSMI_H
16#define HWLOC_RSMI_H
17
18#include "hwloc.h"
19#include "hwloc/autogen/config.h"
20#include "hwloc/helper.h"
21#ifdef HWLOC_LINUX_SYS
22#include "hwloc/linux.h"
23#endif
24
25#include <rocm_smi/rocm_smi.h>
26
27
28#ifdef __cplusplus
29extern "C" {
30#endif
31
32
62static __hwloc_inline int
63hwloc_rsmi_get_device_cpuset(hwloc_topology_t topology __hwloc_attribute_unused,
64 uint32_t dv_ind, hwloc_cpuset_t set)
65{
66#ifdef HWLOC_LINUX_SYS
67 /* If we're on Linux, use the sysfs mechanism to get the local cpus */
68#define HWLOC_RSMI_DEVICE_SYSFS_PATH_MAX 128
69 char path[HWLOC_RSMI_DEVICE_SYSFS_PATH_MAX];
70 rsmi_status_t ret;
71 uint64_t bdfid = 0;
72 unsigned domain, device, bus;
73
74 if (!hwloc_topology_is_thissystem(topology)) {
75 errno = EINVAL;
76 return -1;
77 }
78
79 ret = rsmi_dev_pci_id_get(dv_ind, &bdfid);
80 if (RSMI_STATUS_SUCCESS != ret) {
81 errno = EINVAL;
82 return -1;
83 }
84 domain = (bdfid>>32) & 0xffffffff;
85 bus = ((bdfid & 0xffff)>>8) & 0xff;
86 device = ((bdfid & 0xff)>>3) & 0x1f;
87
88 sprintf(path, "/sys/bus/pci/devices/%04x:%02x:%02x.0/local_cpus", domain, bus, device);
89 if (hwloc_linux_read_path_as_cpumask(path, set) < 0
90 || hwloc_bitmap_iszero(set))
92#else
93 /* Non-Linux systems simply get a full cpuset */
95#endif
96 return 0;
97}
98
114static __hwloc_inline hwloc_obj_t
116{
117 hwloc_obj_t osdev = NULL;
118 while ((osdev = hwloc_get_next_osdev(topology, osdev)) != NULL) {
119 if (HWLOC_OBJ_OSDEV_GPU == osdev->attr->osdev.type
120 && osdev->name
121 && !strncmp("rsmi", osdev->name, 4)
122 && atoi(osdev->name + 4) == (int) dv_ind)
123 return osdev;
124 }
125 return NULL;
126}
127
143static __hwloc_inline hwloc_obj_t
145{
146 hwloc_obj_t osdev;
147 rsmi_status_t ret;
148 uint64_t bdfid = 0;
149 unsigned domain, device, bus, func;
150 uint64_t id;
151 char uuid[64];
152
153 if (!hwloc_topology_is_thissystem(topology)) {
154 errno = EINVAL;
155 return NULL;
156 }
157
158 ret = rsmi_dev_pci_id_get(dv_ind, &bdfid);
159 if (RSMI_STATUS_SUCCESS != ret) {
160 errno = EINVAL;
161 return NULL;
162 }
163 domain = (bdfid>>32) & 0xffffffff;
164 bus = ((bdfid & 0xffff)>>8) & 0xff;
165 device = ((bdfid & 0xff)>>3) & 0x1f;
166 func = bdfid & 0x7;
167
168 ret = rsmi_dev_unique_id_get(dv_ind, &id);
169 if (RSMI_STATUS_SUCCESS != ret)
170 uuid[0] = '\0';
171 else
172 sprintf(uuid, "%lx", id);
173
174 osdev = NULL;
175 while ((osdev = hwloc_get_next_osdev(topology, osdev)) != NULL) {
176 hwloc_obj_t pcidev = osdev->parent;
177 const char *info;
178
179 if (strncmp(osdev->name, "rsmi", 4))
180 continue;
181
182 if (pcidev
183 && pcidev->type == HWLOC_OBJ_PCI_DEVICE
184 && pcidev->attr->pcidev.domain == domain
185 && pcidev->attr->pcidev.bus == bus
186 && pcidev->attr->pcidev.dev == device
187 && pcidev->attr->pcidev.func == func)
188 return osdev;
189
190 info = hwloc_obj_get_info_by_name(osdev, "AMDUUID");
191 if (info && !strcmp(info, uuid))
192 return osdev;
193 }
194
195 return NULL;
196}
197
201#ifdef __cplusplus
202} /* extern "C" */
203#endif
204
205
206#endif /* HWLOC_RSMI_H */
hwloc_bitmap_t hwloc_cpuset_t
A CPU set is a bitmap whose bits are set according to CPU physical OS indexes.
Definition hwloc.h:161
@ HWLOC_OBJ_OSDEV_GPU
Operating system GPU device. For instance ":0.0" for a GL display, "card0" for a Linux DRM device.
Definition hwloc.h:374
@ HWLOC_OBJ_PCI_DEVICE
PCI device (filtered out by default).
Definition hwloc.h:295
struct hwloc_topology * hwloc_topology_t
Topology context.
Definition hwloc.h:742
static const char * hwloc_obj_get_info_by_name(hwloc_obj_t obj, const char *name)
Search the given name in object infos and return the corresponding value.
int hwloc_topology_is_thissystem(hwloc_topology_t restrict topology)
Does the topology context come from this system?
hwloc_const_cpuset_t hwloc_topology_get_complete_cpuset(hwloc_topology_t topology)
Get complete CPU set.
static hwloc_obj_t hwloc_get_next_osdev(hwloc_topology_t topology, hwloc_obj_t prev)
Get the next OS device in the system.
Definition helper.h:1290
int hwloc_bitmap_iszero(hwloc_const_bitmap_t bitmap)
Test whether bitmap bitmap is empty.
int hwloc_bitmap_copy(hwloc_bitmap_t dst, hwloc_const_bitmap_t src)
Copy the contents of bitmap src into the already allocated bitmap dst.
int hwloc_linux_read_path_as_cpumask(const char *path, hwloc_bitmap_t set)
Convert a linux kernel cpumask file path into a hwloc bitmap set.
static hwloc_obj_t hwloc_rsmi_get_device_osdev_by_index(hwloc_topology_t topology, uint32_t dv_ind)
Get the hwloc OS device object corresponding to the AMD GPU device whose index is dv_ind.
Definition rsmi.h:115
static hwloc_obj_t hwloc_rsmi_get_device_osdev(hwloc_topology_t topology, uint32_t dv_ind)
Get the hwloc OS device object corresponding to AMD GPU device, whose index is dv_ind.
Definition rsmi.h:144
static int hwloc_rsmi_get_device_cpuset(hwloc_topology_t topology, uint32_t dv_ind, hwloc_cpuset_t set)
Get the CPU set of logical processors that are physically close to AMD GPU device whose index is dv_i...
Definition rsmi.h:63
Structure of a topology object.
Definition hwloc.h:431
char * name
Object-specific name if any. Mostly used for identifying OS devices and Misc objects where a name str...
Definition hwloc.h:443
hwloc_obj_type_t type
Type of object.
Definition hwloc.h:433
union hwloc_obj_attr_u * attr
Object type-specific Attributes, may be NULL if no attribute value was found.
Definition hwloc.h:450
struct hwloc_obj * parent
Parent, NULL if root (Machine object)
Definition hwloc.h:481
struct hwloc_obj_attr_u::hwloc_pcidev_attr_s pcidev
struct hwloc_obj_attr_u::hwloc_osdev_attr_s osdev
unsigned char dev
Device number (zz in the PCI BDF notation xxxx:yy:zz.t).
Definition hwloc.h:677
unsigned char func
Function number (t in the PCI BDF notation xxxx:yy:zz.t).
Definition hwloc.h:678
unsigned short domain
Domain number (xxxx in the PCI BDF notation xxxx:yy:zz.t). Only 16bits PCI domains are supported by d...
Definition hwloc.h:670
unsigned char bus
Bus number (yy in the PCI BDF notation xxxx:yy:zz.t).
Definition hwloc.h:676
hwloc_obj_osdev_type_t type
Definition hwloc.h:716