import glob
import logging
import os
from contextlib import contextmanager
from typing import Any, Dict, List, Optional
import psutil
import torch
def get_numa_nodes_cores() -> Dict[str, Any]:
""" Returns numa nodes info in format:
..code-block::
{<node_id>: [(<core_id>, [<sibling_thread_id_0>, <sibling_thread_id_1>
...]), ...], ...}
# For example:
{0: [(0, [0, 4]), (1, [1, 5])], 1: [(2, [2, 6]), (3, [3, 7])]}
If not available, returns an empty dictionary.
"""
numa_node_paths = glob.glob('/sys/devices/system/node/node[0-9]*')
if not numa_node_paths:
return {}
nodes = {}
try:
for node_path in numa_node_paths:
numa_node_id = int(os.path.basename(node_path)[4:])
thread_siblings = {}
for cpu_dir in glob.glob(os.path.join(node_path, 'cpu[0-9]*')):
cpu_id = int(os.path.basename(cpu_dir)[3:])
if cpu_id > 0:
with open(os.path.join(cpu_dir,
'online')) as core_online_file:
core_online = int(
core_online_file.read().splitlines()[0])
else:
core_online = 1 # cpu0 is always online (special case)
if core_online == 1:
with open(os.path.join(cpu_dir, 'topology',
'core_id')) as core_id_file:
core_id = int(core_id_file.read().strip())
if core_id in thread_siblings:
thread_siblings[core_id].append(cpu_id)
else:
thread_siblings[core_id] = [cpu_id]
nodes[numa_node_id] = sorted([(k, sorted(v))
for k, v in thread_siblings.items()])
except (OSError, ValueError, IndexError, IOError):
Warning('Failed to read NUMA info')
return {}
return nodes
class WorkerInitWrapper:
r"""Wraps the :attr:`worker_init_fn` argument for
:class:`torch.utils.data.DataLoader` workers."""
def __init__(self, func):
self.func = func
def __call__(self, worker_id):
if self.func is not None:
self.func(worker_id)
[docs]class AffinityMixin:
r"""A context manager to enable CPU affinity for data loader workers
(only used when running on CPU devices).
Affinitization places data loader workers threads on specific CPU cores.
In effect, it allows for more efficient local memory allocation and reduces
remote memory calls.
Every time a process or thread moves from one core to another, registers
and caches need to be flushed and reloaded.
This can become very costly if it happens often, and our threads may also
no longer be close to their data, or be able to share data in a cache.
See `here <https://pytorch-geometric.readthedocs.io/en/latest/advanced/
cpu_affinity.html>`__ for the accompanying tutorial.
.. warning::
To correctly affinitize compute threads (*i.e.* with OMP), please make
sure that you exclude :obj:`loader_cores` from the list of cores
available for the main process.
This will cause core oversubsription and exacerbate performance.
.. code-block:: python
loader = NeigborLoader(data, num_workers=3)
with loader.enable_cpu_affinity(loader_cores=[0, 1, 2]):
for batch in loader:
pass
"""
[docs] @contextmanager
def enable_cpu_affinity(self, loader_cores: Optional[List[int]] = None):
r"""Enables CPU affinity.
Args:
loader_cores ([int], optional): List of CPU cores to which data
loader workers should affinitize to.
By default, will affinitize to cores starting at :obj:`cpu0`.
(default: :obj:`node0_cores[:num_workers]`)
"""
if not self.num_workers > 0:
raise ValueError(
f"'enable_cpu_affinity' should be used with at least one "
f"worker (got {self.num_workers})")
if loader_cores and len(loader_cores) != self.num_workers:
raise ValueError(
f"The number of loader cores (got {len(loader_cores)}) "
f"in 'enable_cpu_affinity' should match with the number "
f"of workers (got {self.num_workers})")
worker_init_fn_old = WorkerInitWrapper(self.worker_init_fn)
affinity_old = psutil.Process().cpu_affinity()
nthreads_old = torch.get_num_threads()
loader_cores = loader_cores[:] if loader_cores else None
def init_fn(worker_id):
try:
psutil.Process().cpu_affinity([loader_cores[worker_id]])
except IndexError:
raise ValueError(f"Cannot use CPU affinity for worker ID "
f"{worker_id} on CPU {loader_cores}")
worker_init_fn_old(worker_id)
if loader_cores is None:
numa_info = get_numa_nodes_cores()
if numa_info and len(numa_info[0]) > self.num_workers:
# Take one thread per each node 0 core:
node0_cores = [cpus[0] for core_id, cpus in numa_info[0]]
else:
node0_cores = list(range(psutil.cpu_count(logical=False)))
if len(node0_cores) < self.num_workers:
raise ValueError(
f"More workers (got {self.num_workers}) than available "
f"cores (got {len(node0_cores)})")
# Set default loader core IDs:
loader_cores = node0_cores[:self.num_workers]
try:
# Set CPU affinity for dataloader:
self.worker_init_fn = init_fn
logging.debug(f"{self.num_workers} data loader workers are "
f"assigned to CPUs {loader_cores}")
yield
finally:
# Restore omp_num_threads and cpu affinity:
psutil.Process().cpu_affinity(affinity_old)
torch.set_num_threads(nthreads_old)
self.worker_init_fn = worker_init_fn_old
self.cpu_affinity_enabled = False