Open MPI logo

Portable Hardware Locality (hwloc) Documentation: v2.7.0

  |   Home   |   Support   |   FAQ   |  
rsmi.h
1 /*
2  * Copyright © 2012-2021 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
29 extern "C" {
30 #endif
31 
32 
59 static __hwloc_inline int
60 hwloc_rsmi_get_device_cpuset(hwloc_topology_t topology __hwloc_attribute_unused,
61  uint32_t dv_ind, hwloc_cpuset_t set)
62 {
63 #ifdef HWLOC_LINUX_SYS
64  /* If we're on Linux, use the sysfs mechanism to get the local cpus */
65 #define HWLOC_RSMI_DEVICE_SYSFS_PATH_MAX 128
66  char path[HWLOC_RSMI_DEVICE_SYSFS_PATH_MAX];
67  rsmi_status_t ret;
68  uint64_t bdfid = 0;
69  unsigned domain, device, bus;
70 
71  if (!hwloc_topology_is_thissystem(topology)) {
72  errno = EINVAL;
73  return -1;
74  }
75 
76  ret = rsmi_dev_pci_id_get(dv_ind, &bdfid);
77  if (RSMI_STATUS_SUCCESS != ret) {
78  errno = EINVAL;
79  return -1;
80  }
81  domain = (bdfid>>32) & 0xffffffff;
82  bus = ((bdfid & 0xffff)>>8) & 0xff;
83  device = ((bdfid & 0xff)>>3) & 0x1f;
84 
85  sprintf(path, "/sys/bus/pci/devices/%04x:%02x:%02x.0/local_cpus", domain, bus, device);
86  if (hwloc_linux_read_path_as_cpumask(path, set) < 0
87  || hwloc_bitmap_iszero(set))
89 #else
90  /* Non-Linux systems simply get a full cpuset */
92 #endif
93  return 0;
94 }
95 
111 static __hwloc_inline hwloc_obj_t
113 {
114  hwloc_obj_t osdev = NULL;
115  while ((osdev = hwloc_get_next_osdev(topology, osdev)) != NULL) {
116  if (HWLOC_OBJ_OSDEV_GPU == osdev->attr->osdev.type
117  && osdev->name
118  && !strncmp("rsmi", osdev->name, 4)
119  && atoi(osdev->name + 4) == (int) dv_ind)
120  return osdev;
121  }
122  return NULL;
123 }
124 
140 static __hwloc_inline hwloc_obj_t
142 {
143  hwloc_obj_t osdev;
144  rsmi_status_t ret;
145  uint64_t bdfid = 0;
146  unsigned domain, device, bus, func;
147  uint64_t id;
148  char uuid[64];
149 
150  if (!hwloc_topology_is_thissystem(topology)) {
151  errno = EINVAL;
152  return NULL;
153  }
154 
155  ret = rsmi_dev_pci_id_get(dv_ind, &bdfid);
156  if (RSMI_STATUS_SUCCESS != ret) {
157  errno = EINVAL;
158  return NULL;
159  }
160  domain = (bdfid>>32) & 0xffffffff;
161  bus = ((bdfid & 0xffff)>>8) & 0xff;
162  device = ((bdfid & 0xff)>>3) & 0x1f;
163  func = bdfid & 0x7;
164 
165  ret = rsmi_dev_unique_id_get(dv_ind, &id);
166  if (RSMI_STATUS_SUCCESS != ret)
167  uuid[0] = '\0';
168  else
169  sprintf(uuid, "%lx", id);
170 
171  osdev = NULL;
172  while ((osdev = hwloc_get_next_osdev(topology, osdev)) != NULL) {
173  hwloc_obj_t pcidev = osdev->parent;
174  const char *info;
175 
176  if (strncmp(osdev->name, "rsmi", 4))
177  continue;
178 
179  if (pcidev
180  && pcidev->type == HWLOC_OBJ_PCI_DEVICE
181  && pcidev->attr->pcidev.domain == domain
182  && pcidev->attr->pcidev.bus == bus
183  && pcidev->attr->pcidev.dev == device
184  && pcidev->attr->pcidev.func == func)
185  return osdev;
186 
187  info = hwloc_obj_get_info_by_name(osdev, "AMDUUID");
188  if (info && !strcmp(info, uuid))
189  return osdev;
190  }
191 
192  return NULL;
193 }
194 
198 #ifdef __cplusplus
199 } /* extern "C" */
200 #endif
201 
202 
203 #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:140
@ 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:342
@ HWLOC_OBJ_PCI_DEVICE
PCI device (filtered out by default).
Definition: hwloc.h:269
struct hwloc_topology * hwloc_topology_t
Topology context.
Definition: hwloc.h:692
static const char * hwloc_obj_get_info_by_name(hwloc_obj_t obj, const char *name)
Search the given key 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:1194
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:112
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:141
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:60
Structure of a topology object.
Definition: hwloc.h:396
char * name
Object-specific name if any. Mostly used for identifying OS devices and Misc objects where a name str...
Definition: hwloc.h:408
hwloc_obj_type_t type
Type of object.
Definition: hwloc.h:398
union hwloc_obj_attr_u * attr
Object type-specific Attributes, may be NULL if no attribute value was found.
Definition: hwloc.h:415
struct hwloc_obj * parent
Parent, NULL if root (Machine object)
Definition: hwloc.h:446
struct hwloc_obj_attr_u::hwloc_pcidev_attr_s pcidev
struct hwloc_obj_attr_u::hwloc_osdev_attr_s osdev
unsigned char dev
Definition: hwloc.h:639
unsigned char func
Definition: hwloc.h:639
unsigned short domain
Definition: hwloc.h:635
unsigned char bus
Definition: hwloc.h:639
hwloc_obj_osdev_type_t type
Definition: hwloc.h:666